This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
n . Hence the function V: S — > , 1 - P(P)>, for np we get V(np) = <1 - p(p>, p(p)>, which coincides with the result, e. g. , from [6]. Depending on the way of defining, of the operation
a
enters place
1
with a
list of
these E-tokens
6 which are transferred in the current
time-moment from input to output
E-transition places and a list of the corresponding characteristics of these E-tokens. For a more detailed description of this process, the place 1 mast be changed (by the hierarchical operator H with a GH 6 1 from the type of the GH from Theorems 9.1.1 of § 9. 1 in [«]. We do not « describe precisely this element of the net
E
because it
will
only
complicate its description. As a result of the activation of the transition Z , 3 a and p enter places
1 and 1 or 1 , respectively. 7 8 9
the tokens
In place 1 7
the
token a will not receive a characteristic. The token p will enter pla ce 1 , if in the current characteristic of a there exists at least one
a special E-token; otherwise p enters place 1 . In
the first case p re-
9 ceives as
a new
characteristic the
new description of
the SHGB E,
whose description is received as a result of applying operator(s) over E
(the type of this (these) operator(s) is (are) determined as a cha
racteristic of the token a. In the second case, new characteristic. Transition Z 4
p
is activated when the three
does not receive a tokens are in
its
input places. After this, token a enters place 1 or 1 without cha10 11 racteristics, token p enters place 1 or 1 without characteristics, 12 13
NEW RESULTS IN THE THEORY OF GENERALIZED NETS
27
token r enters place 1 or 1 without characteristic in 1 and with 14 15 14 a characteristic "the new value of the current time" in 1 , in rela15 tioin to the value of the current time. Finally, transition Z
is activated when the time for the func5
tioning of the SMGN E expires. The token a
(which has a higher pri
ority over the tokens from place 1 ) enters place 1 without a cha16 17 racteristic. The tokens from 1 enter sequentially place 1 and the16 re
they receive
1A
as characteristic the
list of
all
characteristics
which are related to them The list is received as a characteristic of K
the token a at the time of its transfer in the net E . Thus the E-toK
kens a ,..., a will leave the G N E with 1 n
the same
characteristics
as those tokens leaving the SMGN E. Therefore the new net will represent
the functioning
of
the
SMGN E. We shall denote for E
as follows: all its places and transiti
ons have equal priorities; the places
1 , 1 1
capacity; the others, capacities 1. The GN time-components can be (a) T ; 0; (a) T ; 0; o
and 16
1
have infinite 18
26
CHAPTER 1
where a € la, p, rii
a way represents
the functioning
of
each SHQB and therefore it is a UGH for £ SHGH Therefore Theorem 1. 1. 3. 2 rem 1. 1, 3. 1 is also given. THEOREM i. 1. 3. 3: £ H L SHGH
i.e.
is proved and another proof of Theo £ SHGH
i s a conservative
extension
of £. Proof: From Theorem 1. 1. 3. 2,
it follows that each SHGH
sented by the UGH, which is an element of of £
can be represented
£.
by an element of
can be repre
Therefore every element £ and hence
£ I- £
SHGH
SHGH
The other direction is obvious. Thus we give an answer of the Problem 19 in App, 1. REFEREHCES: [1] E. Atanassov
Methodological aspect
of the theory of generalized
nets. II, AHSE Review Vol. 20, Ho. 4, 1992, 53-64. [2] E. Atanassov, Universal generalized nets. II,
AHSE
Review, Vol.
21 (1992), Ho. 1, 59-64. [3] E. Atanassov, L. Atanassova, E. Dimitrov, ski, H. Harinov and S. Petkov,
G. Gargov, I. Kazalar-
Generalized nets and expert
sys
tems. Hethods of Operations Research, Vol. 59. , Proc. of the th Symposium on Operations Research of the
Gesellscbaft
thematiK, Okonomie und Operations Research, 1987, Passau,
12-
fur HaFrank
furt a.H: Athenaeum, 1969, 301-310. [4] E. Atanassov, L. Atanassova, E. Dimitrov,
G. Gargov, I. Kazalar-
ski, H. Harinov and S. Petkov,
Generalized nets and expert
tems.
"Hetworks Information
II,
Proc of Int. Conf.
sys
Processing
NEW RESULTS IN THE THEORY OF GENERALIZED NETS
29
Systems", Sofia, May, 1988, Vol. 2, 54-67. [5] K. Atanassov, L. Atanassova, E. Dimitrov,
G. Gargov, I. Kazalar-
sKi, M. Marinov, S. Petkov and M, Stefanova-Pavlova, nets and expert Vol. 63.
systems. III.
Proc. of the 14-th
Generalized
Methods of
Operations
Research,
Symposium on
Operations
Research.
Ulm, Sept. 1989, 417-423. [6] K. Atanassov, L. Atanassova, E. Dimitrov, ski, M. Marinov and S. Petkov, tems. IV.
Proc. of the
G. Gargov, I. Kazalar-
Generalized nets and expert
XIX Spring
Conf. of
sys
the Union of Bulg.
Math. , Sunny Beach, April 1990, 155-161. [7] L. Hadjyisky
and
K. Atanassov,
A generalized net, representing
the elements of one neuron network set, AMSE Review
Vol. 14, No.
4, 1990, 55-59. [8] L. Hadjyisky L. and K. Atanassov K. Theorem for representation of the neuronal networks by generalized nets, AMSE Review
Vol. 12,
No. 3, 1990, 47-54. [9] Handbook of operations research foundations and fundamentals Moder and
S. Elmaghraby,
(J.
Eds. ), Van Nostrand Reinhold Co. , New
York, 1978 (Vol. 1 and 2). [10] R. Ackoff,
The art of problem solving,
A Wiley-Interscience Pu
blication, New York, 1978. [11] J. Kleijnen, Statistical techniques in simulation, M
Dekker INC,
New York, 1974 (Volss. 1 and 2). [12] P. Fishburn, Utility theory for decision making, J. Wiley & Sons, INC, New York, 1970. [13] H
Wagner, Principles of operations research, Prectice-Hall, Inc.
1969. [14] J. Kanter,
Management-oriented
management
information systems,
Prentice-Hall Inc. , Englewood Clitts, New Jersay, 1977. [15] R. Ackoff
and M
Sasieni,
Fundamentals of operations research,
John Wiley and Sons, Inc. , New York, 1968.
30
CHAFFER 1
[16] Atanassov K. , Open problems in the theory of generalized nets, Preprint IM-MKAIS-1-91, Sofia, 1991. [17] Atanassov E., Universal generalized nets, AHSE Review, Vol. 7 (1988), Ho. 4, 29-35.
HEM RESULTS IH THE THEORY OF GENERALIZED HETS
31
§ 1.2: ON THE REPRESENTATION OF GENERALIZED NET TRANSITIONS E7 A FIXED SET OF TRANSITIONS Favlin Gyurov Let
and Krassimir Atanassov
a fixed GN and a set S with elements - fixed graphical
structures of transitions be given. We snail describe an algorithm for a representation
of the transitions of the given GN E by the elements
of the set S. Thus we shall solve the problem 12 of App. 1: "To construct
an algorithm which for every set
of transition
graphical structures (obviously they must satisfy some conditions) and for every given GN constructs a GN for which (a) all its transitions have forms which are elements of the given set; (b) both nets have and
as a result
equal tokens with equal initial characteristics, of their functioning,
the equal tokens
receive
equal final characteristics. " The mentioned conditions for the form of
transition can be as
follows:
1 |LJ I i IS I + 1 Z€S Z (i. e. it is necessary that at least one transition of S exists which
contains at least two input places) and 2 |L"| i |S| + 1 Z€S Z (i. e. it is necessary -that at least one transition of contains at least two output places),
S
exists which
where IXI is the cardinality of
the set X. Let the above the set S = 1Z , Z 1 2
two conditions be
satisfied for the elements of
Z ] and let the GN s
o « E = <,
32
CHAPTER 1
2 =
r = [L J , L", ||r
II]
( s e e App. 2 ) .
V, »
From
the Theorem for
the completeness of
(see § 5 of App. 1), it follows that every GH can
the GH
transitions
be represented by a
union, a composition and an iteration of its transitions, and therefo re it is sufficient to show a method for representation of an arbitra ry GH's transition and this can also be transformed in all other tran sitions. Hie new GH will have transitions with
S-fonns.
It is conveni
ent that this GH be a GH with a global memory (see § 3 in App. 1). Let this new component contain
as parameters the lists of current capaci
ties of the transition's arcs and of the numbers in the output places: for a transition
2
of the free location
these lists are:
H Z
= [L', Z
L", llm II], where m are the current capacities of the transition's Z i,j i, j arcs between places 1 and 1 ; and (<1, c(l)-c(l, TIME)>/l € L"], where i J Z c is a function giving the number of tokens in place
1 at the current
time-moment TIME. The GH transitions can be ordered according to
their prioriti
es. Let a transition Z of GH E be given. Our algorithm is as follows: By
"//...//" we shall denote some
comments. A01. Order the input and output places of Z by their priorities, where the order in the first case is descending and in the second
case
in ascending series. A02. Check the inequality |pr Z I i ipr 2| 1 i 1
(1)
for every i (1 i i i sj, where pr X is the j-th projection of the j
NEW RESULTS IN THE THEORY OF GENERALIZED NETS
33
n-dimensional set X (1 i J i n) and S - [Z , Z 1 If the set of the S-transitlons
Z1 2
s
which satisfy (1) is empty, 90
to A07. A03. Check the inequality ipr S 1 z ipr Z| ipr2S 1 1 z ipr2Z| 2 i 2
(2) (2)
for every i (l <, i i s). If the set of the S-transitions
which satisfy (2) is empty, go
to A06. K
A04. Determine this s-transition Z , which satisfies (1) and (2), for which K It
It K
ipr |pr Z 1 + Ipr Z I 1 2 is minimum
* A05. Mark the first Ipr Zl 1
input places of Z
with the corresponding
place's identifiers of the input places of Z and the first Ipr Zl 2 it
output places of Z with the corresponding place's identifiers of it
the output places of Z; (different) *
*
If
t
It
t
z 2
Z
z
Z = t , 1 1
z
=t , 2
- L") and define it
Z
Z Z z
identifiers - these places are
- L') V (L"
(L'
mark the other places of
Z
Z
with other
elements of the
set
34
CHAPTER i
r r
= rr K
+ r, r, Z Z
z z
where r = [L' « Z
- L', L" - L" , llfalsell] Z * Z Z
is a
p r e d i c a t i v e index
matrix, M M **
Z z
; : M M + M, M, Z Z
where M = [L'
z
matrix, D D
»» z
- L\ K
L"
Z
z
- L" , II 0 II] «
i s a natural number index
Z
=T T || D ,, V V ll l l ll//HH L L' ' - L')>>. L')>>. Z «« Z z z
z
z
z
GO t o A20. //Obviously !D
iff !D , Z
it
z because the tokens do not enter in the places of the set L' »
- h'. Z
z The method
of checking
the functioning of
a transition
structed in such a way is similar to the one shown below
and that
con is
why it will be emitted here. // R
A06. Mark the first Ipr Z| 1
input places of Z
with the corresponding
place's identifiers of the input places of Z, and the other input K
places of
Z
with other (different)
are elements of the set (L' «
- L'), Z
identifiers - these places
Go to A10.
Z A07. Check the inequality (2) and if it is not valid, go to
Aoa, else
« mark the first Ipr Zi output places of Z 1
with the corresponding
NEW RESULTS IN THE THEORY OF GENERALIZED NETS
35
place's identifiers of the output places of Z, and the other outK
put places of Z
with other (different)
ces are elements of the set (L"
* z
identifiers - these pla
- L"). Go to A08.
z
A08. Let Z ; Z. Let i ~- 1. 0 A09. Determine this transition of places whose number
is
S with minimal number of input pla-
greater than
|Z
1-1
or,
if such a
i-1 transition does not exist, the S-transltion with the maximm num ber of input places
(let it be
Z'). Construct a transition Z i i from a transition Z by removing its final Ipr Z'I - 1 from the i-1 1i number of places. Mark the first input place of Z' by y' (an 1 identifier
which is not found among the other GN's
i
identifiers)
and the other input places of Z' with the identifiers of the cori responding input places of Z The final output place of the i-1 transition Z' is marKed by y' (it is the first input place of i i-1 the transition Z' , if it exists). i-1
MarK all other output places
of Z' with other identifiers which are not found among the other i GN's identifiers: x' , x' i, 1 i, 2
x'
. Define i, IL' 1-1 Z' i
c(x' > - O, o, i, J j w <x' ) = 0 (for 1 i j i |L' I -1), L i, j Z' i c(y' ) = c(y' > = oo, > i-1 i
36
CHAPTER 1
n (y') = max |n (1) / (1 € pr 2') 8 (1 * y')]. L i L 2 i i If the -transition
Z
has no input places
(i. e. if the second
i case mentioned above is valid), is valid for Z,
then finish the procedure, and if (2)
determine the priorities
and capacities of the other
input places of Z'-transitions, in relation to the corresponding prio rities
and capacities of the places of
Z and
go to A12,
else go to
A10; else i:= i * 1 and execute A09 again. A10. Let Z = Z. Let i r 1. 0 All. Determine this transition of
S
which has the minimum
output places whose number is greater than |Z
number of
1-1
or, if such
S-transition with
the ""'nim
i-1 a transition does not exist,
the
number of output places (let it be Z"). Construct a transition Z i i from a transition Z by removing its final Ipr Z" I - 1 for the i-1 1i number of places. Hark the first output place of Z" by y" (an i identifier, which is not found among the other GH's
i identifiers)
and the other output places of Z" - with the identifiers of the i corresponding output places of Z . The final input place of the i-1 transition Z", is marked by y" (the first output place of the i i-1 transition Z" , if it exists). Hark all other input places of Z" i-1 i with other identifiers
which are not found among the
identifiers: x" , x" ,..., x" . Define i, 1 1,2 i, IL" |-1 Z" i c(x"
) = 0,
other GHJ s
NEW RESULTS IN THE THEORY OF GENERALIZED NETS
37
f (x" ) |L" I I --1), IT T (x" ) = 0 (for ( f o r 1 i<. j j i i |L" 1), L Z' ii,J ,J Z' i i c(y" ) ) ^ r c(y") c(y") = to, c(y" to, i-1 i-l 1 ir (y") ir (y") = max [u [u (1) (1) / / (1 (I € pr pr Z') Z') & & (1 (1 * y')j. y')i. L ii L 22 ii ii If the transition
Z
has no output places
(i. e. if the second
i case
mentioned above is valid),
then finish the procedure, determine
the priorities and capacities of the other output places of Z"-transi tions,
in relation to
the corresponding priorities and capacities of
the places of Z and go to A12; else i:= i + 1 and execute All again. A12. Identify places y' and y" by the operator O (see § 5 in App. 1). 0 0 //See structures
Fig.
1. 6.
Nore that
by these
12 steps
the graphical
of the transitions which represent the given transition
Z
are determined. Below we shall determine the other components of the above con structed transitions. // A13. Let the number of the transitions of the first (left) group be n 1 and that of the second (right) group be n . Determine a new tlme2 step t' for which o t t' =
. n
+n 1
+1 2
//More generally, n 1
= max pr Y, Yes l
2 2
= max pr Y, Y€S 2 Y€S 2
n
and these values can be determined before the application gorithm. // A14. Determine the following time-components
of this al
38
CHAPTER 1
Fig.
1. 6
NEW RESULTS IN THE THEORY OF GENERALIZES) NETS
39
Z' Z' i i Z t t :-. tt *+ ( (n n --i i). | . Vt' I(lt ii i is <. n n|,), 1 11 1 11 11 Z" i i ZZ t (n -- 11 + ii). (1 ii ii i.i nn ). t = tt + (n ) . tt' ' (1 ). I l l 2 I l l A15. Determine all elements of the M-components of
the transitions as
"eo". A16. Determine the transition types
of all transitions as
"v"
(i.e.
they will be activated when at least one token enters some of its places). A17. Define the characteristic function
$
in the different places as
follows $
gives as the current characteristic of the token the list of
y i the output Z-places in which the token mist enter, $ , $ , $ are not defined, x' x" y" i,j i, J i $ 1" i
coincides with § which is defined for the given GN. 1" i
Ala. Determine the transition condition predicates sition from the left group
of the
i-th tran
4©
CHAPTER 1
x'
false
. . .
1' i, 1
I false iI I I false I
false
...
false
W
1' 1,2
I false | .
false .
... .
false
W
I false I| I I
false
...
false
W
= i
y
x'
1,2
y' iI r
...
x'
1,11 1,
1' 1, IL' 1-1 i, i-l Z' 1
i, IL" 1-1 Z' 1 false
i-l i-1
true
where UV [1' IV , 1 '' i 1, 1 1, 2
1' ) -.-- L' , 1, IL' i-l i, 1-1 Z Z' 1
and W
-. " ( 3 T : 1 £ T i
|L' l)((f(r ) = 1) & (m > 0) & Z' p, T p, T 1
the transition
sition from the right group:
condition predicates of the j-th tran
NEW RESULTS IN THE THEORY OF GENERALIZED NETS
y' y i i
x' i, i, 1 1
r = ri = i
x' x' i,2 i,2
I false | I I I I| false | false . | .
x' x'
1" i" J, 1 1 J,
1" i" ... j, 2 2 j,
ii-i -l
1" i" i, |L" 1-1 i, |L" i-l Z" Z" i i
false
false
. . .
false
false false . .
false false . .
. . . .. . . .
false false
|| false false false false false false
... ...
false false
ii
... ...
w
ii, , IL' IL' | -1-1 1 II Z" II ii II
y" y"
41
w
w
w
iI
where U (1" (1" , , 11 "" J J jj, , 11 jj, , 22
1" 1" ) ) = L" |L" I -1-1 j j, , |L" l ZZ Z" Z" J J
and
W = "There exists at least one identifier of the place of a tran sition with a bigger number,
which is an element of the set
of the last toKen characteristic", W" - "The identifier
of the output
place is
an element
of the
set - the last token characteristic", A19. Define: n {Z') = if (Z") = n (Z) for every Z' and Z" - transitions A i A J A i J from the left and from the right groups, respectively. A20. Check the list of the transitions of GN E. If all transitions are processed - end of the procedure; else - go to A01. it
Let the newly constructed GN
be marked by
E . From the above K
construction, it follows that all transitions of the set S.
Moreover, the last characteristics
E
are elements
of
which the tokens will
receive in Z"-transltion places will coincide with the characteristics J which the tokens will receive in the corresponding Z-places.
42
CHAPTER 1
Therefore the following theorem is valid. THBQRHJ: For every set of transition tisfy the inequalities
graphical structures,
(1) and (2)
and for every
which sa given GH,
there exists a GN for vtfiich - all its
transitions
have forms nfliich are elements
of the
given set, - both nets have equal tokens with equal initial characteris tics and,
as a result of their functioning,
the equal to
kens receive equal final characteristics. This paper is based on [1].
HEFEREHCE: [1] F. Gyurov and X. Atanassov,
On the representation
net transitions by fixed set of transitions, 91, Sofia, 1991.
of generalized
Preprint IH-HFAIS-2-
HEM RESULTS IN THE THEORY OF GENERALIZED NETS
43
§ 1. 3: TWO HEW CONSERVATIVE EXTENSIONS OF GENERALIZED NETS Rumen S. Christov
Some problems
and
Krassimir T. Atanassov
related to the extensions of the concept
"gene
ralized net" (GH) were generated during the development of the program realization of GNs. Below we discuss two new types
of nets
and prove
that they are conservative extensions of the ordinary GHs. Let the object o « <,
and
let it have the following property:
place of the set
b>; L'>
when in
a certain
L' c L there exist at least two of its tokens, which
are generated by one
token "predecessor"
independently from the fact
that (a) there can exist other "kindred" tokens somewhere (i.e. tokens with the same "predecessor"), (b) the tokens can be generated
in different moments and
from diffe
rent, but kindred, "predecessors", then these tokens are united in one token by analogy with the procedu re from
§ 8 in App. 1. The new token
will also be a kindred token of
the tokens which are kindred tokens of its generated ones. Let us call this new net
"a GH allowing
a particular
token's
uniting" (GH-APTU). THEOREM 1.3. 1: I
c I. GH-APTU
Proof: Obviously, every GH-APTU for which
L' = $,
is not defined
i. e.
for which
X
i- £.
LJ Let
GH-APTU a GH-APTU E be given. Initially, we construct an ordinary
GH
E
with
the same graphical structure. After this, for each output place 1 € L' of transition Z of GH-APTU E (let 1 be a fixed place), we construct a new transition Z which corresponds to 1 (this transition contains 1
44
CHAPTER 1
place
1 as an input place).
This new transition
(see Fig, 1.7) has
a transition condition with the form 1 rr
= 11 1
II 11
Z II 1 1 II 2 1 2 1
Z 11 1
W W W W 11 22 W W
ffalse alse
44
ZZ 1 2 W W 3 W W 5
where Z = "the token has no kindred tokens in places 1 and 1 "; 2
W 1 W
Z = "in place 1 there is already a kindred token of the present 2 2
to-
ken"; = i W
W 3
1
8 1W, 2
W = "the token has no kindred tokens in place 1 and it has received 4 as a current characteristic the characteristics of the last toZ ken in the place 1 " (this is described below in relation to the 1 characteristic function), W
= nW . 5 4 The following inequalities for the new places are valid -Z ZZ n (1) (1) > n (1 (1 ) ) > TII (1 (1 ) ) L L II L 2
and Z n (1) T (1) > T *I (1 (1 )• ). L L 2 Let the GH-characteristic function that in place
Z 1 2
the token receives
as
$ be defined in such new characteristic all
a way the
NEW RESULTS IN THE THEORY OF GENERALIZED NETS
46
Fig. 1. 7 Z characteristics, which the token, last entered place 1 , has received 1 Z after the time-moment when it has split with the token from place 1 , 2 Z or with its corresponding "predecessor". In place 1 the tokens do not 1 receive new characteristics.
In place
1 the tokens receive the same
characteristics as the tokens of GN-APTU E in place 1. For transition Z from E (we denote it by Z ), let E L Z
= (1/1 € L' D pr Z ). 2 E
Following the above algorithm we can construct the set
of transitions
of E (Z /l € L ), and to transition Z of E we Juxtapose the transition 1 Z Z : Z V U Z. 1€L 1 Z From this construction it is obvious that the transition Z from E will function in the same way as the transition
Z,
By analogy with
the proof of Theorem 5. 3. 1 [*] (see | 5 in App. l) we can prove
that
46
CHAPTER 5
the GHs
E
and E
are equivalent in the results of their functioning.
Therefore E represents GH-APTU E and hence J c J
.0
GH-APTU Hie other extension of the concept GH is related to the charac teristic
functions of the nets.
How they are
related to the tokens.
Tokens receive new values from the characteristic functions associated with the places which the tokens enter.
These functions may be depen
dent on the entire information in the net. On the other band, the pla ces are also influenced by
the processes in the nets, but they do not
have their characteristics. It can be obtained by defining the charac teristic functions which will give characteristics to the places. Let the net E E
= =
<
i,
n , n , c, f, e , e >, nA, nL, c, f, e1, e2>, A b»
L
be given, for which
1
$
o « <E, n , e >,
2
K
<x, x, $, <x, x, $,
E
is a function determining
the places' current
characteristics. For this net the places have initial
characteristics
which are elements of the set X. we shall call this net a GH with pla ces' characteristic functions (GH-PCF), The proof of the following assertion is easy. THEOREM 1. 3.5: £ : J, GH-PCF Every GH with a global memory presented
by an ordinary
GH.
characteristic functions (related at
different moments
(see § 3 in App. 1)
From the fact
that all
can be re
values of all
to tokens or places) are calculated
of the GN's functioning,
it follows that these
results can be remembered in the GH "memory" and therefore the GH "me mory" can represent them Therefore al 1 GH-PCF can be represented by a GH with a global memory. On the other band, this type of extensions of
NEW RESULTS IN THE THEORY OF GENERALIZED NETS
47
a GN can be represented by an ordinary GN, from which, the validity of the Theorem follows,
because the opposite is obvious:
every ordinary
GN is a GN-PCF without the specific components. 0 From the construction
of the above proof
follows the validity
of THEOREM 1. 3. 3: Every ordinary GN (or a GN-PCF) can be represented by a reduced GN with
a global memory but without the set of
initial characteristics and ons.
the characteristic functi
Chapter
2:
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
In this chapter different results are included ted to
the application of
the GHs in some areas
applied mathematics, artificial intelligence and illustrations
ntfiicb are rela
of science
scientometrix.
of the possible applications of GHs in science
ven. Four of the papers
<§ 2. 1, §2.2, § 2.4
and § 2.6)
Thus
are gi
are revised
versions of old papers, and the other four contain new results.
46
such as
SCHE APPLICATIONS OP GENERALIZED HEXS IB SCIENCE
§ 2. 1:
49
GENERALIZED NETS REPRESENTING THE ELEMENTS OF NEURON NETWORKS Ljubomir HadjyisKy
and Krassimir Atanassov
The class of Neuron BetworKs
(NNs)
represented with GNs is a variety of the
(see [1-3])
which will be
"additive model" [4).
This
class of HNs is described by the equation Y(t 4 at) = g(I
T B y (t)> € R B
(R is the set of all real numbers,
T <. . . >
is the transposed vector
of the indicated one. ] is the state of the BB at the time-moment t; (e) y (t) is the state of the neuron "i" at the time-moment t; i B x H (f) T € R
i s a matrix whose elements
T (1 i i, i, j
synaptic strength of the connection between the neuron
j i B) are t h e " j " and the
neuron "i"; I(t) = =
(g)
T B I (t)> (t)> € € R R I B B
is the potential of the somna obtained by integration of the separa te input impacts; (i) (i i
B I (t) = B2 T .y (t) Ii |t| •■ j=l I I ii J .JJ (t) i j=l ii J J
i <■ H ) ;
T B (j) gd(t)) =
the neurons.
50
CHAPTER 2
N H X H (R is a linear N-dimensional vector space, R is a linear (NxH) -dimensional vector space.) y (t) describes the state of the neuron "i". Ibis state can i exited or not and, according to this
y (t), accepts i
be
corresponding 1y
the values 1 or 0. The determination of g (I (t)) is realised by the following rei i 1ations:
4
(\
, if I (t) > e
i
7 (t) = g (I (t)) = 1 0 , if y i (t) = g i (ii (t)) = { o , if i i i y (t), if y (t), if where 6
I (t) ii (t) Ii (t) I (t)
'
< G , < ei , = 6i = 6
is the threshold value for the neuron "i". i
This class of HHs has the possibility of self-training, sed by changes of the synaptic streingths T
reali
or the threshold values i, j
6 . In case some neurons are dead,
the network can reorganize
itself
i and the other neurons will accept the functions of the dead ones.
The
representation with GHs includes this property as well. Let the GH and
E
E
be composed of two
graphical incoherent GHs, E
(see Fig. 2.1). GB E is a modelled 2 1
1 I(t)
somatical potential
and a determined state of the neurons Tf(t + at) by g(I(t)). GH E
per2
forms self-training and reorganization When a neuron is dead. E
The
GH
is composed of nine places, three transitions and H tokens, where H 1
is
the number of neurons in the network.
1 is c|l ] -- I i i
for
i i i i
9,
The capacity of every place
i. e. at some time-moment all tokens
9C*E AFFLICATICHS OF GEHERALIZED HETS IH SCIBHCE
can fall into one of the places.
It is convenient
51
to use an extended
GH with a global nEmory (see § 3 in App. 1). In this case, in the com-
Fifi. 2.1
5a
CHAPTER 2
ponent B different data sets about the net-topology IT J, i, j
the SOEOB-
tical potential of neurons (I (t)l, their threshold values 16 }, their i i state ly (t)] can be disposed of as follows ii JJ (1) (1) JJ (2) <2) .. .. .. N i i i i (a) X+ x+ i=l T T i=l ii II T T .. .. .. I I JJ ((l),i l),i JJ ((2), 2 ) , ii II ii ii where
H £+ marks the sum of i= l
number of the neuron
"i",
N r
JJ ((r r ) ) i i ii
, T T
JJ (|r],i r ],i ii ii
index matrices (see App. 2), i
is the
is the number of the neurons which are i
connected with neuron "i", JJ (k) / 1 i k i r 1 is the set of the numi i bers of the neurons which are connected with the neuron "i", for 1 4 JJ <, r , T is the synaptic strength of p-th connections, i J (JJ), i i ii JJ B
(d) (d)
H JJt t ii=l =l
ii
ii i i e ,, II
ii ii
i i l |l j |t] (t) . || ii
The relation among
places
is given
by the conditions of
index matrices for the corresponding transitions: 1
r
=
1
1
22
33
1
I W 1 1 1
W 2
V
1
I W 5 | 1
W 2
W 3
I W 1 1
W 2
W
1 1
7
44 3
3
,
the
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
where 1
- "y (t) = 1" & nW , i 3
2
- "y (t) = 0" & 1W , 1 3
W W
- "T
W 3
= 0"; a 1 5 r
= 2
1 6
1 l W 2 1 4
W , 5
1 I W 6 | 4
W 5
where W =W v W , 4 2 3 W 5
=W ; 1 1 7 7 r
= 3
1 8 8
1
1 I W 3 | 6
W
W
1 I W 9 1 1 6 9 6
W
9 9 ,
7
8
7 7
W 9 9
where W
=W 6
1
7
zW, 2
W W
V W , 3
= "no connection exists", 8
w
= "C - O", 9
where (i, if a neuron a is a live, = V (t) = \ a. a I O, if a neuron a is a dead,
T
V (t) is the function determining the physical existence of the neua ron a.
53
54
CHAPTER 2 $
The characteristic functions
(1 i i i 9)
are not defined.
i Hote, that if we use an ordinary GH, B,
i.e. a GN
without global memory
then a part of the characteristic functions will exist.
Hence the
function i is not defined. The time-moments
for firing
and for activity duration
of the
transitions are t
1 1
= t , o
2 t = - tJ , 1 t
3 1
= t , o
1 t
2
- 00
,
2 t = t" , 2 3 t
2 2
= 0D 00 ,
where where t (= T) is the initial time-moment of the HH and hence of to (= T) is the initial time-moment of the HH and hence of o
(
t
+ t
i if t = t
and there are no tokens in the
place 1 , 6
tJ =
+ t
E 2
+ t", otherwise,
o n o place 1 , ] and is the time at which |I (t) 6 i t + t + t", otherwise, ted, calc E t is the time at2which the unique token tE is the time at which |I (t) ] and calc i 2 ted, t is the time at which the unique token E 2 t calc
the GH E, the GH E,
|y (t + at)] will be calculai
of the GH E does one cir|y (t + at)] 2will be calculai of the GH
E 2
does one cir-
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
55
cle in E , 2 t" is the time at which N tokens transfer from input places
t n
to output
places of the transition Z , 2 is the time at which a token can be transferred from the transition Z 1
to the transition Z . 2 The calculation of (I (t)) and (y (t + 3t)I begins at the timei i
moment t'. When t = t , the values o
(y (t)J i
of some neurons which are
inputs in the NN can be changed and this will be an external impact to the NN. The GN E
is composed
of five places, two transitions
and one
2 token. The capacity of every place is 1. In this GN, self-training and reorganization
are accomplished.
Bearing in mind that the previous s states of the NN are remembered as a matrix (y ) (1 i s i n), where n i
is the number of this state.
They are selected by determined criteria
in the component B of the GN E. The relations among the places are given by the conditions: 1 11 11
r = 4
where - "F - 1",
W 10 w 11
= nW , 10
1 12 12
l 1 i w 10 I 10 lO
W w 11
1 I W 13 I 10
W 11
l I w 14 I lO
w 11
56
CHAPTER 2
1 13 13 r
= 5 5
1 14 14
1 | W 13 I 12 13 I 12
W 13 13
1 I W 14 12 14 I I 12
W 13 13
where W
= "F = 1", 12 - IV
W
13
,
12
where the parameters R and F have the following values: 1, if C d, old
= C, d
d, old
E= 0, if C where C
is the number of dead neurons before the last cycle, d, old
C d
is the current number of dead neurons; f1, if non-self-training, F ---- \ i F \ 0, if self-training. The values C
, C and F d, old d
component B of the GH E.
are defined as
parameters for the
The value of F is defined at the initial ti
me-moment. The characteristic functions $
(10 & i i 14) are not defii
ned. The time-moments for firing and activity duration of the transi tions Z and Z are 4 5 4 t -- t 1 calc 4 t =t 2 tr 5 t = t (= T) , 1 o
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
57
5 tt
= = tt
2
tr tr
where t
tr
is the time when one token transfers from an input
place
to an
output place of one transition, tt = tt + tt ,, E E rr r 2 44 5 ft r 4
II reor // tt ,, iif f R = 0,0,
[
ft /t tt
r
,, iif f R R -: 1,1,
-- . 5
t r 4
\ tr tr
,, iif f P -. 1,I,
self self tt ,, iif f P = 0,0, E E * * 2 2 is the time when a token is transferred throughout the transition
Z, 4 t r
is the time when a token is transferred throughout the transition 5
Z, 5 reor t is the time when the MS reorganizes when a neuron is dead, E 2 self t is the time when the NN is self-trained. E 2 The calculation of new characteristics of the NN when the net work has been reorganized begins at the time-moment t
and when i t calc
58
CHAPTER 2
has been self-trained, at the time-moment t + t . calc r 4 Let T be the starting moment at which GH t
be toe running time
E begins functioning,
of the model, given by a time-variable, diffe
rent from the time-variables of the GH
E, defined for the component B
of the GH. At the moment T, all tokens are in the place 1 . The transition 1 1 1 Z is fired (t = T, t =
the?
Heurons of
1 are transferred in places 1 and 1 (transition Z is fired). 3 7 8 3
Heu-
rons which change their state from non-excited to excited or dead fall in place 1 ; in place 1 , neurons which keep their 7 8 ve. The conditions for tokens transfer from in in 1 , 8 neurons
state and are ali-
1 in 1 , or from 1 aga8 7 8
are the same as the ones from 1 to 1 or 1 . Only excited 3 8 7
can change
the state
this case tokens (neurons)
of other neurons from the network.
transfer through
the transition
In
Z . The 2
change in the neuron state will be described in detail. At the starting-moment, place 1 ; 6
the transition Z
when t = T = t , o
there
are tokens in
is fired at the time-moment t 2
+ t , i. e. o n
when a token appears in place 1 and the duration of the fired transi2
SC*E APPLICATIONS OF GENERALIZED NETS IN SCIENCE
tion is t".
At the
time-moment
59
t + t" the calculation of o
(I (t)), i
(y (t)) and E begins as follows: let a token with a number a be trani 2 sferred through Z . The somatica1 potentials 2
I of J
the neurons
connected with neuron a will be I (t) -. T .y (t), J (k) J (k),a a a a with K = 1, 2
r. a
I (t) is the change of the potential of neuron J (k) and it J (k) a a is added to
the old value of this potential defined as
an element of
the component B. In the same way, for all other tokens passing through Z , the potentials of the neurons are determined in the NN. 2
When this
calculation is finished for all tokens, potentials I (l < a i N) a
are
determined completely. Then the calculation continues with the functi on g (I (t)) for 1 £ a £ N and the new states ly (t + at)! are detera a i mined. The running time t
is equal to t .At this moment the trancalc
sition Z of E is fired and 4 2
the token moves to place 1 or 11
1 . If 12
the token falls in place 1 , a reorganization is not necessary, 11 the number 1 , 12
of neurons does not decrease.
a reorganization is necessary in
If a token
falls
i. e.
in place
the data of the neurons in the
NN and the connections among them At the time t = t + t + t", calc r 4 the reorganization is completed and the transition
Z
is fired. 5
The
60
CHAPTER 2
token falls in place 1 or 1 , where the corresponding self-training \1 13 is
either performed
transitions Z
or not at the time t . When GH r 5
and Z 1
E
is working, 2
are fired and the excited tokens (neurons)
are
3
moved correspondingly from place 1 to place 1 , and 1 . The dead neufl 7 2 rons are moved correspondingly from place 1 to place 8
1 and 1 . When 7 4
the running time t is equal to the following: t = t + t + t", calc E 2 the transition Z is fired and the running time 2 t . So excited tokens are removed from places 0
t
receives the value
1 and 2
1 to place 1 ; 6 6
the non excited and dead ones are entered in to place 1 . 5 the process continues. If all tokens fall in place 1 8 the not excited
In this way
this means that
neurons in the network are absent and al 1
tokens are
entered into place 1 and the GH E stops functioning. 9 By the representation of HNs with GUs, the distribution of neu rons by their functional state is obtained. This allows a better visu alization during the UN
functioning, because only excited neurons ta
ke part in the corresponding operations. On the other hand, it is pos sible
that some of
the operations for the neuron's
data change
are
performed simultaneously. These properties obtained by
the representation with
GH
as a
possibility for an external impact of neuron characteristics made pos sible
the effective modelling
with HHs
with processing of information.
»
of real processes
connected
S O E APPLICATIONS OF GENERALIZED NETS IN SCIENCE
61
Below we shall define a new type of NNs and construct
a GN
E
which will represent their functioning and the results of their worK. Each
new NN has the same elements (neurons) as the NNs from
above. The functions, which determine the functioning,
the state
and
the potential (Y, I, g, etc. ) of the new NNs are the same as the first NNs, described by the GN above. The basic difference between the two types of networks
is the
possibility of new neurons being born in the second type of NNs. These neurons possess the same functional possibilities as the others in the network. In the second type of NNs new functions exist
(which are de
ft fined in the global memory
B
of the GN E ):
(a) V
a ) - the set of born neurons; V (t) 1 (c) V (N, t) = N + V (t) - the new neuron number for NN (N is the num3
1
ber of neurons in the NN before time-moment t; (d) V (x, t) - T U T , where 1 i i, j, x S V (N, t), x € V (t) 4 x, J i, x 3 2 and T , T are as above; x, j i, x (e) V (x, t) = 6 for x 6 V
and V 1
to the network
are stochastic or they can
be
related
4
parameters.
strongly connected with V
The functions
V , 2
V , 3
V
and 5
V
are 6
and its type. 1
At time-moment t of being borning,
every new neuron
x 6 V (t) 2
62
CHAPTER 2
has a zero potential
(I (t) = 0) and a zero state (Y (t) = 0), where z x
Y is defined by g X
and I as from above. X X Q How the function g(I(t)} e E , where Q is the number of the na tural numbers.
* The GH E has the form of Fig. 2.8 ons which are identical in both nets E
(cf. Fig. 2.1). All notati-
and E
are marked identically.
a The new places in E
are m , m , m and m ; the new transitions are Z 1 2 3 4 6
and Z (a t r a n s i t i o n Z of GH E does not e x i s t h e r e ) . 7 5
«
Z 7
r e a l i z e s the c o n n e c t i o n between t h e subnets E 1
analogous t o E
and E 1
which
are
>
The t r a n s i t i o n c o n d i t i o n s i n t h e GN E
= 1
and E , 2
i n the above GN E. 2
r
The t r a n s i t i o n
«
are a s f o l l o w s :
1 2
1 3
1
1
I W 1 1 1
W 2
W 3
1
I 5 1
W 1
W 2
W
1 I W 6 | 1
W 2
W 3
m I 4 1 4 1
W 2 2
W
W 1 1
4
3
3 3
where W = "y (TIME) : I" 8 lV : 1 i 3 W = "y (TIME) = 0" S lV ; 2 i 3 W = "6 = 0 " (for 6 s e e above). 3 a a The t r a n s i t i o n c o n d i t i o n s r , r and r are t h e same 2 3 4
a s above
SCME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
53
t
Fig. 2. 2 The transition Z
does not exist and hence the same is valid for r . 5
a
64
CHAPTER 2
rr 6 6
m ID 11
m m 2 2
]] || 11 || 11
V V
W W 14 14
15 15
1 II 12 I 12 I
W W 14 14
W W 15
Kfoere W 14
= "V (TIME) > 0"; 1
15
= "V (TIME) = 0". 1
W
x In place
m
the control token (of the second subnet of
GH
E
1 which corresponds to the GH E
from above),
receives a characteristic
2 "V (TIHE)", vfoere TIME i s the current time-moment of the GH-model. The 1 condition m 3
r 7
m
l 4
l 13
14
m I II
W 16
W 16
W 17
V
= m I 2 I
W 16
W 16
W 17
W
m I 3 1 3 1
W 16 16
W 16 16
W 17 17
W
18 18 18 18
where W = "x > 0"; 16 last W = "x = 0" 8 W ; 17 last 12 W - "x - 0 " ft W 18 last 13 (W and W are d e f i n e d above). 12 13 In p l a c e
m 3
the token r e c e i v e s a s c h a r a c t e r i s t i c
"x
- 1" last
and i n p l a c e m : "<x, V (x, TIHE), V (x, TIHE), V (x, TIHE), I (TIHE), 4 4 5 6 x
SCWE APPLICATIONS OF GENERALIZED NETS IN SCIENCE
65
Y (TIME)>", where x € V (IDE). x 2 All components In the GN global memory
are changed to new ones
for which N: = N + V (TIME) at the tune-moment TIME. 1 In this GN, the boundary of N is
2 and N
is a function of the
global model-time. Here the numbers of new connections between neurons are added to the elements r , if they are generated for the corresponi ding neurons. In this GN t E
= t + t + t , r r r 2 4 6 7
where t is defined above, r 4 / t
6
t
, if V (TIME) > 0
, otherwise (see above),
I tr tr where t W
is the time for the calculation of the function
W 1
and
1 t t r r
M M
7
= (TIME), t =V V (TIME), t 1 1
+ t , , + t r r 5
K
where t is the time for the calculation of the characteristic toKen in place m (for t see above). 4 r 5
of the
» The GN E
functions in a similar way as GN E. The tokens of the
NN enter E* in place 1 . These neurons which are generated at the time 1 of the functioning of E* GN E
appear in place
m . A control token 4
from the above transfers from transitions 2
of the
Z , Z and Z and is 4 6 7
"a mother" of the born neurons. In the other places
it functions only
66
CHAPTER 2
as a control token for the modelling process.
« The place capacities of GH E The other models
are infinite.
for functioning of HHs
can be represented by
similar GHs. The graphical structures of these GHs will be the same or simpler than the above ones. For example,
the GH which represents the
HH-back propagation method [5] will not have the transition
Z
from 3
the
first GH
functioning form
described above.
of HHs
On the other
band some ideas for the
(e.g. [6-9]) will force
some corrections in the
of the transition condition predicates
tions in this GH.
Finally,
and characteristic func
some relations between the Petri nets and
some types of the HHs are discussed in [10-12]. This paper is based on [13, 14]. REFERENCES: [1] Jun-Vtei Wong,
Recognition of
general patterns using neural net
works, Biol. Cybern. , Vol. 58 (1988), Ho. 6, 361-372. [2] H. Cottrell, Stability and atractivity in associative memory net works, Biol. Cybern., Vol. 58 (1988), Ho. 2, 129-139. [3] A. Fukushima,
A historical
neural network model for associative
memory. Biol. Cybern., Vol. 50, 1984, Ho. 2, 105-113. [4] A. Guez, V. Protopopsescu and J. Brahnen, On the stability stora ge capacity and design of continuous neural networks, Trans. SMC, Vol. 18 (1988), Ho. 1, 80-90. [5] B. Rumelhart and J. HcClellond,
Parallel distributed processing,
explorations in the microstructure of cognitron, HIT Press,
Cam
bridge, Mass. , 1986. [6] J. Hopfield,
Heural networks phisical systems with emergent col
lective computational abilities, 2554-2558.
Proc.
Hat.
Acad. Sci. , 1982,
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
[7] T. Kohonen,
Self-organization and associative memory,
67
Springer-
Verlag, Berlin, 1984. [8] L. Hadjyisky, Parallel processing
in frequency
dependent neural
network, Parallel and Distributed Processing, Elsevier Sci. Publ. , 1991, 59-74. [9] C. Lee, Intelligent
control based
theory. Proc, of Internat. Conf.
on fuzzy logic and neural net
on Fuzzy Logic
and Neural Net
works, Japan, 1990, 759-764. [10] A. Negro, R. Tagliaferri and S. Tagliaferri, Some remarks on Petri nets and neuron networks. Proc. of the IASTE3) Int. Symp. App lied Informatics,
Grindelwald,
(M Hamza, Ed. ),
Anaheim,
Acta
Press, 1987, 28-29. [11] T. Kristian, J. Pavlasek and P. Sapaty,
Modeling of neuronal-ac-
tivity using modified Petri nets, Physiologia-Bohemoslovaca, Vol. 36, No. 6, 1987, 541-551. [12] 14 Zargham,
Parallel processing neural networks,
Proceedings of
the First Annual Meeting of the nternational Neural Network Soci ety, 1988, Boston; Neural Networks Vol. 1, No. 1, 1988, 558. [13] L. Hadjyisky and K. Atanassov,
Theorem for representation of the
neuronal networks by generalized nets. AMSE Review, Vol. 12, No. 3, 1990, 47-54. [14] L. Hadjyisky and K. Atanassov,
A generalized net,
representing
the elements of one neuron network set. AMSE Review Vol. 14, No, 4, 1990, 55-59.
68
CHAPTER 2
§ 2. 2: GENERALIZED HETS REPRESENTING THE TRAVELLING SALESMAN FROELEM Krassimir T. Atanassov
The Generalized Nets (GNs) have been used as a means for model ling real processes and theoretical problems.
For example the process
of solution of the transportation problem was described using a GH. Initially, ire shall construct a GH the process of
(see Fig. 2.3),
solution of a variant of
problem (see e.g. [1-7]).
After this,
describing
the Travelling Salesman (TS)
we shall define the TS problem
in a more extended form. The temporal components of this GN will be
o « T = 0, t = 0 , t =
1. The GN E consists of two subnets E 1 has only one transition Z
GH
"a formal description of the graph of connections
The predicate
one) is "true".
2
1
between points which must be visited by the TS exists in it. The transition Z is activated at every time-moment with 1 ration 1.
E
with one place 1 . One token a with an ini1
tial characteristic
and E . The second 2
of the transition condition
(there is
Toe token a receives, during its transfer from
a duonly 1 to 1
a 1 , the characteristics x , which we shall describe below. 1 i The token p enters GH E in place 1
with an
initial characte-
2 ristic "a name of vertex, where the TS has been initially". The transition condition
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
r 2 2
=
1 3 3
1 4 4
1
1 I 2 2 I I
W
W
W 3 3
1 I I 3 I 3 I
W
1 1
2 2 W
1 1
1 I W I 1 11 5 |
69
5 5
W 2 2
3 3
2
W 3
W
where W
= "From the vertex of the graph, described as
an initial characte-
1 ristic of a,
where
a
is the TS,
more than one path goes
out
which is not passed by the TS", W
= "From the vertex of the graph, described as an
initial characte-
2 ristic of a,
where
a
is the TS,
at least
one path goes out
which is not passed by the TS", W
= "From the vertex of the graph,
described as an initial characte-
3 ristic of or, where a is the TS,
no path goes out
which
is not
passed by the TS". If a token goes in the place 1
(i. e. the predicate W
5 lid), vertex,
then it receives as
the next characteristic:
where the TS is going",
is va-
3 "<"name of a new
"length of the path between the
old
and the present vertices", "cannon length of the TS transfer")-". "The length" can be changed with other coefficient. If the predicate W is also valid, each token from i splits into two tokens. The first enters the place 1
1 or 1 2 5
and the second,
the place 1 without a characteristic. 3 Thus all tokens which determine the different directions of the TS-transfer are collected in the place 1 . 5 When some of these tokens do not satisfy
the predicates W
and 1
70
CHAPTER 2
Fig. 2.3 W , the token enters the place 1 without a characteristic. 3 4 The condition 1 6 r 3
1 I W 4 | 4
1 7 W 5
where W = "There are no untravelled vertices in the graph described 4 initial characteristic of the token a " V
- "There are no untravelled vertices in the graph
as an
described as
an
5 initial characteristic of the token a,
but there is no
path to
it without repetition of the route". Each one of the tokens mon
length
of the path
which enters place 1 receives the coro6
as the current characteristic.
The
tokens,
SOME APPLICATIONS OF GENERALIZED HETS IN SCIENCE
nftuch enter place
1 leave tbe net with 7
71
a final characteristic
"the
token does not pass all graph vertices". the transition condition 1 8 8 r 4 4
1 | W 6 1 1 6 6 6 1 I a i 8 I
1 9 9 W 7 7
W 8
W 9
where W 6
6 a = (" £ c(l , TIHE) > 1" 8 " c(l , TIHE) = 0") v "x > x "; TIHE last i=3 i 8
W
6 a = " I c(l , TIHE) + c(l , TIHE) = 1" v "x > x "; 7 i=3 i 8 TIHE last
W
a = "x <, x "; 8 TIHE last
W 9
6 a = " I c(l , TIHE) = 0" v "X > X i=3 i TIHE last
where x
is the last characteristic of the corresponding token from last
a 1 and x is tbe current characteristic of tbe token a and 6 TIHE is the number of tokens in the place 1 at tbe time-moment ct racteristic x TIHE ristic of tbe token
c(l, t)
t. The ena
is determined by the value of tbe current charactewhich is in
tbe place
1 . 8
In some time-moments
a x can be "0". TIHE Tbe classical TS problem
can be described by sush a GN.
problem can be generalized in some directions
This
(see e.g. [1-7]). Below
we shall introduce its new generalization and construct a GN represen ting one of them.
72
CHAPTER 2
Let a time-scale be fixed and let T
and T i
final time-moments of the process, which possible that T
be the initial and f
we shall model below.
It is
= co. f
Let
T = (T , T 1
T ] be a set of travelling salesmen (TSs; 2
which is different from TS
t as an abriviation of
"a travelling sales
man") . Let C = |C , C C J be the set of al 1 products which the 1 2 c TSs delivered. Let V = IV , V V ] be the set of built-up areas 1 2 v (the graph-vertices] which can be visited by the TSs. Let us use everywhere the following indices: - i € 11, 2
t],
- j € II, 2
si,
- k 6 |1, 2
cl,
- TIHE € [T , T ]. i
f
Below we shall define some functions related to the elements of the above sets which will determine the choices of the TSs' routes: - »
C
1 being the set of the different i, c i
types of products, which the t-th TS delivered at the current moment TIHE, where t u »(T , TIHE) c C i=l i (We assume the possibility of absence of some products from the TSs' baggage at a given moment.); - Q(T , C , TIHE) is the quantity of the j-th i j i-th TS at the moment TIHE;
product
which has the
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
- P(T , C , TIME) i J
is the price of
73
the j-th product which the i-th TS
has at the moment TIME; - L(T , C , TIME) i J
is the
limit of fitness of the
j-th product which
the i-th TS has at the moment TIME; - M(V , TIME) =
different necessary
i 0 is the quantity
products in the built-up area V
at mo le
merit TIME; - D(V , V ) is the time for transfer from vertex K K' vertex V
V to its neighbour k
, the equality k' d(V , V ) = d(V , V ) k k' kJ k
is not necessary; - E(V , V , TIME) k k'
is the price of the transfer from vertex V to its k
neighbour vertex V
at the moment TIME. k'
- G(V , V , I D C ) € [0, i) k k' from a vertex V
determines the possibility
to its neighbour vertex k
V
of a
tranfer
at the moment
TIME.
k'
Thus, in the frames of a given time-interval,
the paths can be for
bidden for transfers. - R(T , V , TIME) is the value of the gain of T i k i
in V . k
The value of
this function can be determined e. g. as follows: R(T , V , TIME) = I Q(T , C , TIME). P(T , C , TIME). i k C €*(T .TIME) i j i k J i sg(L(T , C , TIME)), i j when i
K k, j
E Q(T , C , TIME). C €V(T ,TIME) i j j i
74
CHAPTER 2
Analogously, the equality can be defined in the case when X < £ Q(T , C , TIME). k, j C €<MT , TIHE) i j j i - H(T , V , TIHE) is the time of staying of T in V after time-moment i k i k TIME. Based on these definitions, we shall formulate an the Travelling Salesman Problem (ETSP).
Extension of
We shall also formulate other
problems. The ETSP will be denoted as ETSP1. Trace out the optimal gain") route of the which
(according to a
criterion HG:
t-th TS among the independent TSs
must move in the built-up areas
V , V 1 2
V , v
"maximum
T , T 1 2
T t
carrying some
of the products C , C , . . . , C . 1 2 c THEOREM 1: There is a GH in the frames of which the ETSP1 is solved. Proof:
Let the GH from
Fig. 2.4 be constructed. It has the following
transition conditions: 1 3
r
= 1
1 I true 1 I I 1 | false 2 I 1
1
1 6
14
false
false
false
false
true
false
3
I 1
W 1
2
1 9
|I 1
W 1
2
1 I false 11 I
1 4
W
false
W 3
W
false
W 3
false
true
false
where W
= "there exists
at least one more path along which
the token
1 move"; W
= "there exists a path along which the token can be moved"; 2
can
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
Fig.
2.4
W = iW ; 3 2 5
r
= 1 I 2 4 1
W 4
where
a W = "TIME i p r i ": 4 2 cu 1 7 r 3 where a W = "TIME i x "i 5 cu
= 1 I 5 1
W 5
75
76
CHAPTER 2
1 fi a r = 1 4 7
I 1
1
1
W
W 6
1 I false 6 |
1] 10
9
false
11 ii false
7 false
V 6
W 7
where
« = "TIME i T + t ",
W 6 W 7
= TV ; 6
] I S I1 8 r = I 5 1 I 12 i i 12 1 I 14 I
1 12
1
W 8
W 9
W
W a a
9 9 W 8
13
W 9
where W
= "c(l 8
a i i , TIME] = 0" V "x i x "; 12 cu cu
i where x is the current characteristic of the a -token, cu i
which is lo-
cated in place 1 ; 12 W 9
: iV . 8 Tokens
a , a , . . ., a 1 2 t
with initial characteristics: "T " eni
ter 1 at time-moment TINE = T and after this they enter place 1 1 3
im-
mediately. There they receive new characteristic "j)(T , TIHE), i
KC , Q(T , C , TIME), P(T , C , TIME), L(T , C , TIHE) > j i j i j i j / C € j
<MT , TIHE] J, i
wtoere p(T , TIME) is the vertex V i k
where the TS
T
is located. i
Si-
SCME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
multaneously with the above transfers,
the token
7T
p enters place 1 2
with an initial characteristic "(V , M(V , TIME!), [
neighbour vertices
of
the
vertex V . The a-tokens split in places 1 (if the TS T has some posk . 3 i sibilities for motion) and 1 (if the path for motion 4 unique).
of the TS T is i
In place 1 some tokens mist be generated which have one li4
near predecessor.
All tokens which have equal initial characteristics
"T " are different representatives of the i
token
a i
and there
is no
concurrent situation among them Note that concurrent processes do not exist in this GN model, but such processes exist in other GN-processes of the TS-problem. The a-tokens do not receive any new characteristics in place 1 , but in place 1 they receive the characteristic 3 4"V , TIME + D(V , V )", k' k k' k' k k' where V kJ
is the vertex to which the TS T will come. After the transi
fer from 1 to 1 , 4 5
token a receives the characteristic
a a "TIME + N(pr x , pr X , TIME)" 10 1 cu and in place 1 - the characteristic 7 a a a a "R(pr x , pr x , TIME) - E(pr x , pr x , TIME)", 1 0 1 cu 1 cu-2 1 cu-1 which is the net profit for the vertex. When the a -token enters place i
78
CHAPTER 2
1 , it receives as current characteristic the name of the vertex, whe9 re it is located at these
tokens
the current moment, i.e. p(T , TIME). In place 1 i fl
receive
as a characteristic
the value
of the
tokens
transfer in the net according to the criterion HG. Finally, in the frames of a transition Z 5
the a -token with the i
maximm value for HG is determined, The token
p
is transfered in the
transfers of the a-tokens. current characteristic
GH
simultaneously with the
It enters place 1 , where it receives as a 6
the new values of the parameters
which it has
in its previous characteristic. Thus the process of the transfer of
TSs is described.
For the
J
GH s definition the following data are also necessary: (a) the tokens transfer in a package, i.e.
in the frames
of one tact
all tokens which can be transfered from input to output transition places, do this. o
is the minimum time which
is related
to TS-stays or TS-transfers (in real situations, this can be e.g., 1 min. ). (c) the transitions are activated in every time-moment about the fixed 0
time scale with parameters T , t i
and
• t = T - T. i f
o (d) the duration of the transition's activities is t . Therefore o e (t) = t 4 t . 2 (e) the tokens transfer is determined by their time-characteristics. (f) the capacities of all arcs and places are oo, without
SC*E APPLICATIONS OF GENERALIZED NETS IN SCIENCE
79
C(l > = c(l ) = C(l ) = c(l ) = 1. 2 6 10 11 (g) the priorities of all places
and transitions
are equal,
without
Places 1 , 1 and 1 for which it is valid: 0 14 12 n (1 ) = n (1 ) > Tt (1 ). L 6 L 14 L 12 (h) o 1
= A(V(1 , 1 , 1 ), v(l , 1 >), 1 3 9 2 11
(i) D = A(1 , 1 ). 4 6 7 The other transitions are of a disjunctive type. Thus a GN in frames of which the ETTSP1
is solved
is construc
ted. This GN has some faults and the reason for them is
that the num
ber of oc-tokens can grow un-1 .unitedly. Below we problems and
shall introduce some other modifications
of the TS-
discuss the possibilities of their descriptions with the
GNs: (ETSP2) Trace out the locality-optimal (for a criterion LMG: ty-maximal gain") TSs T , T , . . . , T 1 2 t
route of the i-th TS
(l <, i i t> which mist move in the built-
up areas V , V . . . , V 1 2 v .. . , C c
when
"locali
among the independent
carrying some of the products
the different TSs
do not Know
C , C, 1 2
the plans of the
others. (ETSP3) Trace out the locality-optimal (for a criterion LMG): route of a in number groups of TSs:
(T ,T 1,1 1,2
T
), l,t
(T 2,1
1 T
, . . . , T 2, 2 2, t 2
),...,
move in the built-up areas
IT ,T a, 1 a, 2
T ) Which a, t a
V , V , . .. , V , 1 2 v
must
carrying some of
80
CHAPTER 2
the products
C , C ,..., C ; 1 2 c
the TSs
from d i f f e r e n t s e t s do
not concur. The last problem can also be defined for deterministic
and for
non-deterministic cases. Variants of the above problems are those where
the TS
into all the V-places or go to some of these places only, TS must
have exactly determined qualities
must go
where every
of products or where every
TS must travel certain exactly determined distance. The above TS problems can also be described by QHs as above. This paper is based on [8,9]•
REFERENCES: [1] E. Lawler,
J. Lenstra,
A. Rinnooy Kan and D. Sbmoys
(Eds.), The
traveling salesman problem. A guided tour of combinatorial optimi zation, H. Y. : Wiley X Sons, 1985. [2] I. Helamed, S. Sergeev and I. Sigal,
The traveling salesman prob
lem Issues in theory. Avtomatika i telemecbaniKa,
1989,
9, 3-33
(in Russian). [3] I. Helamed, S. Sergeev and I. Sigal,
The traveling salesman prob
lem Exact methods. Avtomatika i telemechanika, 1989, 10, 3-29 (in Russian). [4] I. Helamed, S. Sergeev and I. Sigal,
The traveling salesman prob
lem Approximate algorithms. Avtomatika i telemechanika, 1989, 11, 3-26 (in Russian). [5] J. Little, K. Hurty, D. Sweeney and C. Karel, An algorithm for the traveling salesman problem
Operations Research,
Vol. 11,
1963,
972-989. [6] T. Volgenant and R. Jonker, On some generalizations of the travel ling salesman problem,
Journal of Operational
Vol. 38, Ho. 11, 1987, 1073-1079.
Research
Society,
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
[7] R. Ackoff
and M. Sasieni,
Fundamentals of operations
81
research.
John Wiley and Sons, Inc. , New York, 1968. [8] K. Atanassov,
A generalized net, representing
the travelling sa
lesman problem. AMSE Review Vol. 14, No. 4, 1990, 61-64. [9] K. Atanassov, Generalized net and extensions of the travelling sa lesman problem press).
AMSE
Review
Vol. 21,
No. 2,
1992,
19-26
(in
82
CHAPTER 2
§ 2. 3: GENERALIZED NETS AND LABYRINTHS Erassimir Atanassov and Rumen Cbristov
Different researches on the methods for exist (see e.g., [1-3]).
motion
in a labyrinth
Here we shall describe one of them, which is
based on the GNs. Let the labyrinth L be given. It can have the form of
(m x n) -
dimensional (0, 1)-matrix, the elements "0" and "1" of which correspond to the labyrinth corridors and walls, respectively. We construct
the GN from Fig. 2. 5. It has two transitions with
transition conditions
l1
3 3
m I false II m |false 2 1 r = I 1 1 I W 1 1 3 1 2 1 55
where
|I 1 I ||
W 3 W 33
l1
l1
4 4
false false
5 5
mm
false false
m m
2 2
3
W
W 1 2 W 1 2
W
w 4
W w 5
false
false
w 4 W 44
w W 5 W 55
false
false
false
false
W = "c(l , TIME] + c ( l , TIHE] > 0" 8 " c ( l , TIHE) = 0"; 1 3 5 3 W
= nW ;
2 W
1 = "end of the (unique) path or a meeting with another
token on the
4 path or along this path another token has already passed"; - IV ;
W
3 W
4 = "there is more than one path
5
and along this path
has not passed";
and
1 66 r
1 2
1
22
| W W W W 3 | 6 7
another token
SCME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
83
where W
= "the token arrives to an exit"; 6
w = -m .
7
Fig. 2. 5 First,
toKens a and p enter the net through places 1 and m , 1
respectively and
p has an initial characteristic in the form
1 of the
above mentioned matrix. If there exists exactly one path for motion, it goes to place 1 with a characteristic "nurber of this path, nurabe3 red
according to the hour hand".
If the number of the paths
is more
84
CHAPTER 2
than one, the token a splits sequentially to the tokens a , a , . . 1 2
(s is the number
time one token enters place 1 3
with the above characteristic
one (which will generate the next tokens) enters place 1
s
Each
of the different paths in front of token a ) .
another
without cha-
5
racteristic.
when there are no paths for motion
a igiven
in front of
token, it goes into place 1 with the characteristic "unsuccessful end 4 of .a motion". place 1 6
If there is an exit in front of
with the characteristic
a token,
it goes into else
"successful end of a motion";
the token enters place 1 without a characteristic. 2
All 1-places have equal priorities,
which are bigger
than the
priorities (equal) of all m places. Both transitions have equal rities. All places have equal (oo) capacities,
prio-
the temporal components
are standard and thus not given. The problem of aspects.
motion in
a labyrinth can
be extended in some
We shall show one extension and its GH-representation:
the labyrinth has some floors GH, we shall
(i.e. it is not planar),
when
in the ;above
change only the characteristic function of place 1 3
"number of the path, numbered as first underneath - uphill and
with
as se-
cond along the hour hand" and the initial characteristic of token P is (m :K n x k) -dimensional (0, 1) -matrix. REFERENCES:
tl] F. George, The foundations of cybernetics,
Gordon and Breach Sci.
Publishers, London, 1977.
[2] H, Gardner,
Mathematical puzzles
and diversions,
Bell and ,Sons,
London, 1965.
[3] K. Shannon,
Research on
theories of informatics and cybernetics,
Inostrannoi Literatury, Moscow, 1963 (in Russian).
SCHB APPLICATIONS OF GENERALIZED HETS IN SCIENCE
85
5 2. 4: FORMAL COMMUNICATION IN SCIENCE: A MODEL BASED ON GENERALIZED HETS Fadosvet Todorov
§ 2. 4. 1
and
Krassimir Atanassov
Introduction
Various studies
of the
been carried out (see [1-5]). plain the informal and/or
connunication network in science have Different models have been used
formal connunication
processes.
to ex
Models of
forma] connunication in science are predominantly linear and emphasize various media (articles, journals, books, etc.), participants (indivi duals and institutions) offers two views
or functions
(activities).
In [1], Atherton
of the connunication network of scientific and tech
nical information, one of which stresses for example
the place of the
publisher. Garvey presents (see [2]) a schematic overview of
the sci
entific communication system which includes all events that occur from the time nal)
the scientist begins his research,
publishes it (in a jour
until its appearance in reviews, treatises, etc. Also Garvey de
scribes in [3] features of the current dissemination system in science and suggests some innovations in it. Some linear models of the formal connunication process are pre sented in a recent review article by Hills ([4]). The review "the scholarly
connunication process in terms
looks at
of the interaction and
concerns of the scholar, the learned society, the publisher, the pro duct, the librarian, and the new technology". King
et al. offer (see [5]) a schematic overview of the disse
mination of scientific and technical research results by different me ans and of the general flow of information among scientists and neers.
The authors also supply a stochastic model of the
engi
transfer of
article manuscripts between authors and publishers in order to descri be the system of journal publishing in terms of cost and flow of mate rials.
86
CHAPTER 2
Here we shall present at outline of a formal communication pro cess by stressing
some
for the traditional
specific functions and participants required
transfer of
(originator) to the reader
article manuscripts
from the author
(consumer). The purpose of the study is to
show a possible description of this process by means of a mathematical model with GHs.
§ 2. 4. 2 Schematic overview o± formal comami cation Here we choose
a simplified
part of
in
science
the formal communication
networK in order to apply in a comprehensible manner the GH (having in mind the
unacquainted reader).
under consideration,
The schematic overview of the process
namely the publishing of article manuscripts and
distribution of journal articles, preprints in Fig. 2.6,
or reprints, is presented
The author (A) submits an article manuscript to the edi
tor (s) of a given journal. "Hie flow of manuscripts between authors and editors is represented by arrow 1 reject obvious
a paper whose subject matter is inappropriate
editor then
This is
symbolized by arrow 2.
transmits the remaining manusrcipts
they believe to be experts this
or which is "an
non-starter" [6]. A recommendation may be made to submit the
manuscript elsewhere. the
(see Fig. 2.6). The editor (E) may
in the particular
activity is represented by arrow 3.
(F) returns the evaluated paper (arrow 4) commendations for acceptance not need re-examination),
"to people whom
topics involved"
The expert or
[6],
the referee
to the editor with his re
(without change),
improvements,
Traditionally,
corrections
modifications or
(that do
rejection.
Only in the first case (direct acceptance) is the manuscript transfer red to the
publisher
(arrow 5),
who in our diagram is to be related
with recording and reproducing activities rather than being corporated in the publishing process. If the manuscript is rejected or a revision is recommended, then the author is informed (arrow 2). The author then
SCME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
87
decides whether to revise the manuscript, to send it to another jour nal
(usually
without
modifications)
or
to give up
publication of
(withdraw) his paper. After manuscript recording, the publisher sends proof-copies to the author (arrow 6).
The corrected proofs are returned for reproduc
tion and issue preparation (arrow 7). The published articles reach the user (U) through the
distributor (D), who incorporates the functions
of such participants in the formal communication process as publisher, librarian, information services, etc. Fig. 2. 6
also shows
the direct contact between the author and
the user. The significance of arrow 9 is a reprint request and that of arrow 10 correspondingly the satisfaction of the request.
Fig. 2. 6 Schematic overview of the formal conmunication by means of journal articles (A: Author, D: Di stributor, E: Editor, P: Publisher, R: Referee, U: User)
88
CHAPTER 2
§2.4.3
A GS model Fig. 2.7 presents the graphical structure of
a GH which models
toe relations discussed above (see also Fig. 2.6). The circled capital letters are the places Where different activities are performed ting, reviewing, etc.). f
are "fictitious"
(edi
places which are introdu-
i ced
to conform to
the graphical
representation of the GH.
They are
used to determine the length of time needed hy the tokens to pass from a "real" to a "real" place. The f are "fictitious" because the toKens 1 do not receive new characteristics as they transit through the f . i The token (author)
(article manuscripts)
enter the net through place A
with the following initial characteristic "name,
the author,
manuscript's title".
Place A
address of
is the entry point for all
tokens which represent the manuscripts ready for a submission to a jo urnal. Transition condition r
has the trivial form 1 l ff 88
ff II true true 7 II rr
=
R R
II true true II A I I true true A
1
ff II true true 4 II Transition condition r
has the form 2
E r
= 2
f
| W 8 | 1 I R I true
f f 2
f f 33
W W
ff
W 2
W 3
W 4
W 5
false
false
false
false
4
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
89
Fig. 2. 7 where z "A takes the decision to submit a manuscript";
W i W
= "A is ready to return the proof-copies to P"; 2
W
= "A is willing to send a reprint to U"; 3
W = "A decides not to publish or to withdraw his manuscript" 4 (The place W is an output place of the model. This implies that A will take no further steps to publish the paper under conside ration);
90
W
CHAPTER 2
= "A agrees to make changes in order
to publish
the manuscript in
5 the same journal or decides to resufamit it without
improvements
to another journal". When the GH's token
(which models some manuscripts] arrives in
place E, it receives the characteristic "journal title". Transition condition r
r
:
E
has the form 3 R f i 6 1
I W 1 6
3
W 7
W 8
where W
= "The editor E takes the decision to inform the author A" 6 (The notice
could include
a rejection of
the submitted manu
script, a reconmendation for revision, etc. In other words, the token arriving in place A
(coming from E)
receives the charac
teristic "information from E to A"); - "E decides to send the manuscript to be assessed by an expert R";
W 7 W
e
= "E takes the decision to transfer the manuscript to the publisher F in order to be published (after peer reviewing]". Transition conditions r
and r 4 U
f
--
r 1
I true 3| _ I D I true f
I true 5 I
and P r
=
f
5 f
I true 1I
I true 2 I
have the trivial forms 5
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
91
Transition condition r has the form 6 b
r = e e
P U
f 7
D
| i i
W
W
I I I
W
f 5
9 9
to 10 W
12 12
W it li W
13 13
14 14
where W
= "P is ready to send manuscript copies to A" 9 (These are the proof-copies or the reprints according
to the mo
del time), = "P is ready to send the issue containing the article in question
W lO
to D", z "P is ready to send the Journal issue directly to U",
W 11 W
= "U wants to send a letter
(with a comment)
or a request for a
12 reprint to A", W
s "U is willing to receive information or the original from D", 13
W
= "U
is willing to be disseminator of
the article
for
another
14 user". Entering f
from P, the token has the characteristics
"proof-
7 copies" or "reprints". Arriving in D from P, it receives the characte ristic "bibliographic data". When the token enters f from U (through 6 f ) it has the characteristic "Kind of information or requested mate7 rial". K
Let the GN function between the tune -moments T and T + t . Ini tially (at the moment T) the tokens enter the net in ve for E. The predicate W
place A and lea
has the current value "true" and the remal1 ning predicates in the transition condition r have zero-values. From 2
9E
CHAPTER 2
place E
the tokens move to places A or F according to the editor's
judgement. Visiting places A and E the toKens receive the above menti oned characteristics. When a token returns to f , it could enter place f (modifica8 4 tion of the manuscript), W (withdraw) or f (sending a reprint). If 3 the token enters R coining from E it returns obligatorily to E, and ac cording to the characteristic
received in R (referee opinion) arrives
at A or P. From P it moves to A with the characteristic "proof-copies" or "reprints". The token with the characteristic "proof-copy" goes from f to P through f . With a new characteristic (bibliographic da8 2 ta) the token arrives in D or f coming from P. The ultimate aim of 5 the tokens is to get to X). notice that in our case, the place f sym8 bolizes the author's activity and U that of the reader, i.e. we do not consider
(for the sake of simplicity]
the possible manifestation of
the user as an author and correspondingly that of the author as a rea der of publications. § 2.4. 4
Discussion
The proposed diagram of formal conmunication 2.6)
does not show all possible
the process. as well,
in science
(Fig.
interactions among participants
in
In the scheme, there could be represented other contacts
such as a possible
direct link between the author and
the
referee: the author could send an accompanying note to the referee and the latter could openly publish (or answer individually) his recomendations for manuscript Improvement. All the simplifications as well as some artificial divisions are made here to perceive more easily the GH as a modelling tool appropriate
to conmunication network.
The model
based on the GHs is sufficiently powerful to be used for the descrip-
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
93
tion of the whole (ccnplex) structure and dynamic picture of knowledge generation, representation,
distribution and use, or for detailed re
presentation
part of
of a specific
minimum modifications
the conminication process.
With
(simplifications), the model in Fig. 2. 7 can be
used for example, for the description of booK publishing and distribu tion. Fig. 2. 6
can be also presented
by means
of
a block diagram,
which would reflect the interactions among participants lopment of the
and the deve
process under consideration. But a block diagram
can
not describe the simultaneous movement even of two or more manuscripts submitted at the same time to one, two or more journals. neous
transfer of two (or more) manuscripts in the network can be de
scribed by means of ons,
The simulta
e. g. , the Petri nets and their other modificati
but with the increase
in the number of manuscripts, it would be
impossible to follow the individual fate of the manuscripts at various places of the net. The transfer of a given number sented by
of manuscripts
can
be
repre
the stochastic model of King [5]. However, in this case one
cannot follow
the dynamic development of the process or "predict with
certainty the time
required to perform any of the article preparation
functions, nor whether the articles will be rejected, accepted, or mo dified" [5]. Only by means of GNs it is possible to follow manuscript (on the basis of the different ved by the
the history of a
GN's characteristics recei
tokens at each GN's place) and to trace its propagation in
a time-frameworks.
§ 2. 4. 5 Possible
applications
of the GNs
The simulation of a formal comninication process will on
the
values of the transition condition predicates,
be based
which can
be
94
CHAPTER 2
determined by the token's characteristics or on the basis
of experts'
assessment. Using the results of the program realization and simulati on,
a wide variety of questions can be answered and various conclusi
ons can be drawn. The quantitative results may ticles
include: published ar
(on a certain topic or in a given Journal)
all the submitted manuscripts; without revision]
as a percentage of
or directly accepted manuscripts (i.e.
as a percentage of all
the accepted article manu
scripts; the number of resubmitted manuscripts that are accepted; number
of
submitted
etc. The qualitative how
does
the number
manuscripts by authors with
of
submitted manuscripts
change; is there any relation characteristics
the same
answers are to questions such as on
is the flow of
a specific topic and initial
submitted manu
scripts
sufficient for the "safe"
Through
this model many time-related questions can also be
One could
appreciate
address,
the following:
between the rejection rate
of the manuscript;
the
existence of a given journal, etc. answered.
the mean model time needed for the transfer
the manuscripts from place E to place D
of
(publication time-lag) or the
time required to pass a particular place, etc. This paper is based on [7],
REFERENCES: [1] P. Atherton,
Views of the communication network of scientific and
technical information. International Forum on Information
and Do
cumentation, Vol. 1 (1975), Ho. 1, 10. [2] W. Garvey et al. , Research studies in patterns of scientific com munication, Inf. Stor. Retr. Vol. 8 (1972), 111-122, 159-169, 207221, 265-276. [3] W. Garvey, S.
Gottfredson, banging the system: Innovations in the
interactive social system of scientific conmunlcation. Information Processing and Management, Vol. 12 (1976), Bo. 3, 165-176. [4] P. Hills,
The scholarly conmuni cation process.
Annual Revies for
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
Information Science
95
and Technology, Vol. 18 (1903), 99.
[5] D. King, D. McDonald, N. Roderer, The journal in scientific conminication:
The roles of authors, publishers, libraries and readers
in a vital
systen.
National
Science Foundation Contract NSF-C-
DS175-0694-2. RocKville, Mary lands, King Research, Inc. , May 1979. [6] A. Meadows,
The problem of refereeing. The Scientific Journal (A.
Meadows, Ed. ), Aslib Reader Series, Vol. 12 (1979), London, ASLIB, 104-111. [7] Todorov R. , Atanassov K. Formal communication in Science: based on Generalized nets, 177-185.
A model
Scientometrics, Vol.9 (1986), No. 3-4,
96
CHAPTER H
§ 2. 5: GENERALIZED HETS AND EXPERT SYSTEMS. V Krassimir Atanassov,
Peter Georgiev
The main problem in designing, Systems (ESs) is to select lism (see e.g. [1,2]). i.e.
and Martin Tetev
Implementing and
using
Expert
the proper Knowledge representation forma
Host of the ESs apply the rule-based approach,
they represent the knowledge in "IF situation THEN action" man
ner. The papers [3-6] deal with another means of representation, main ly with a generalization of Petri nets
and discuss how the rule-based
approach can be viewed through the new formalism with all the ensuring consequences.
On the other hand the GHs are an
extension of the well
Known Petri nets allowing a detailed description of parallel
(or con
current) processes. The GN described in the paper can be considered as an extension of the nets from
[3-6).
All notations of [3-6] are saved here.
This
gives the possibility to trace the development of the idea for the GHmodelling of the ESs. On
the other hand
we introduce new idea related to
the ESs,
based on the GH-description possibilities. The main components of a production (DB)
containing
system are:
the Data Base
the facts about the problem which are
to be solved,
the Knowledge Base (KB)
containing the rules which are
to be used in
the reasoning process, and the inference engine which operates through the KB using the DB for proving or rejecting rules in form,
the KB
are supposed
to be just
a given hypothesis. in a positive
The
conjunctive
since otherwise the rule can be split into a number of
"subru-
les" of that type. We shall describe the ES J s facts with ties.
How we can compare
priorities.
new components - priori
two arbitrary facts on the basis
This possibility is very
useful
of
their
in the particular cases
when both facts coincide or when both facts have contrary senses.
SCHE APPLICATIONS OF GENERALIZED NETS IN SCIENCE
97
Let to every fact A of KB he Juxtaposed by a natural nuntoer u A which corresponds to tbe priority of A. Let tbe new fact B with a pri ority \i be generated by some way in a certain time-moment when tbe B ES functions.
If both
facts
are not related,
ters the DB. In the ordinary ESs, fact
A, when
function
the new fact
then the new fact en B substitutes the old
B coincides with, or contradicts to A. Now the
in another way, basing on tbe new component,
ES will
when tbe facts
A and B coincide, tbeir representative (in particularly - A or B) stay in tbe DB, but with a new priority - tbe maximm of \i and \> . On tbe A B
other band, tbe fact with tbe mcnrimm priority between \i and \> stays A
B
in tbe DB when tbe facts A and B are in a contradiction. The constructed here GN (see Fig. 2. 8) represents the ESs with data's priorities in tbe above sense. There have been several attempts to describe production systems by some Kinds of nets (see cf. [7-15]). Following [3-6] some definiti ons
will be given here in order to describe
different components
of
tbe GN appearing below. Let a token
a enters place 1 of the GN with initial cbaracte1 a a ristic x {- "a hypothesis") and let z denotes its current cbaracO last teristic. Let a token
b b enters place 1 with initial characteristic x (= 2 O
"DB"). Let tbe token
c enters tbe place 1 with initial cbaracterisric 3
c x
(= "R"), where 8 : IF , I 1 I is tbe list of tbe rules. Each 0 1 2 n rule E has the form
98
CHAPTER 2
ent and A
A 1,1
are Uhe elements of the conjunction
presen-
i, s i
ting the antecedent,
z
denotes the i-th characteristic of
the token
i with the highest priority in a given place. Finally, let toKens d , d , ..., d 1 E s
(s z 0) enter place 1 wuth 40
d cu initial characteristics x
(= <"new fact", "its priority")), where
0 "cu" symbolises the current number of the d-type token, Z
The transitions in the GU are the following. = 9 40 2 33 48 41 42 46 9 9
where 1 41 41 1 I 40 I r
= 9
W 30
1
1 42 42 W
1 43 43 W
44 44
1
1 45 45 W 34
false
46 46
31
32
W 33
1 | false 2 I
false
false
false
false
true
1 I false 33 I
false
false
false
false
true
1 I false 48 I
false
false
false
false
true
where d d d cu cu p cu cu cu W W = "pr "pr xx ee xx " S "pr "pr xx 22 Q(pr Q(pr xx )", )", 30 30 1 00 llast ast 2 0 10 where Q(A] is the priority of fact A, if A € x
B lasr
P « x ; last d d d cu cu p cu cu cu W :: "pr W "pr xx €€ xx " 88 "pr "pr xx < Q(pr Q(pr xx J"; J"; 31 31 1 00 1 llast ast 2 00 2 11 0
and Q(A) = 0 ,
if A
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
d cu W 32
z "ipr : "lpr x l o
d cu
p ex last
" & "pr x 2 0
99
d cu i Q(npr x )", 10
where TA is the negation of the fact A; d cu W 33
- "ipr x l 1 0
ex last
d cu w
and D 9 Z 10
" s & "pr x 2 0
d cu < Q(npr x )", 10
d cu p " & & "npr x i x )" last 2 0 last
p i x
: "pr x 34 1 0
d cu
p
= v(l , 1 , M l i , v(l , 1 , 1 )); 2 33 40 2 46 33
= <(1 J 1, , fl (1 , 1 ), K, *, r , K, v(l )> 46 47 46 lO 46
where
r 10
1 47 47
1 48 48
: 1 I W 46 I 35
W 36
where W
= "c(l 35
W
, T H E ) = 0"; 40
r "C<1 36
, TIME) > 0"
40
where c(l, TIME) is the number of the toKens in place 1 at the current time-moment TBE; Z : <(1 , 1 , 1 , 1 1 where
1
47
3
35
!, fl , 1 4
5
1 !, *, *, r , K, D > 9
(here we shall use the notation from
ding corrections)
1
1
[3-6] with the correspon
100
CHAPTER 2
1 4 1 1 r
= 1
1
1
1
5
I W 1 1
1
6
w 2
false
false
false
W
false
false
false
1
I false 1
false
false
false
W v W W vW 1 2
1 || ffalse alse 35 I 35 I
ffalse alse
ffalse alse
ffalse alse
W W vv W W W W vv W W 1 2 1 2
where the predicates W, W
and W 1
v W 1
9
false
1 I false 47 I 3
W
1 8
7
v W 2
have the form 2
W = "there are more tokens before place 1 "; 1 a b W W = x € x ;; 2 00 0 W W 1
= = nW ;; 2
and DD 1
= A(I A ( I ,, ll ,, v v(l ( l ,, 1 >); >); 11 47 47 33 35 K
Z ss 2 4 6 8 10 12 13 14 15 16 2 2 2 4 6 8 10 12 13 14 15 16 2 2 where 11
r
= 2
K K
lO lO W 5
11
12 12 W 6
11
13 13 false
11
11 14 14 false
11 15 15 false
16 16 false
false
false
1 4
I 1
1 6
| false 1
false
W v W 5
W v W 6
1 8
| false 1
false
false
false
where the p r e d i c a t e s W and W have the form 5 6 c a W = "mnt>(pr x , x ) i- 0"; 5 1 0 last c a W = "mnf>4pr x , x ) - 0" 6 1 0 last
W v W W vW 5 6
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
where nun*>(Y, y) the ordered set
101
is the number of appearances of the element y in the Y.
(In [3-6] there are other predicates related with K
places 1 , 1 and 1 4 10 12
Here place
the GN from [6]. This subnet
1 10
is an input to the subnet of
realizes the process of checking of the
hypothesis and it is aside of the present model); Z =
= 11 II W W 13 13 II 16 16 1 II W W 31 II 31 16 16
where the predicates W
and W 16
W
1 31 W W 17 17 W W 17
have the form 17
= "there are no tokens in the subnet" 16
(the form of this predicate in [6] is similar to the above); W
= "a new fact is generated in the subnet" & "there are some tokens 17 in the subnet";
Z = <(1 , 1 !, (1 , 1 , 1 , 1 i, *, «, r , *, D > 7 29 15 32 33 34 35 7 7 where
rr 7 7
l1 32 32
33 33
= 11 ||W W vvW W 29 II 18 18 29
W W vvW W 19 19
1 30 I 30 I where the predicates W
and W 1«
W
= "the 18
W = nW . 19 ia
ffalse alse
ii
ll
ll 35
3434-
ffalse alse
ffalse alse W W vv W W 18 18
false false W W vv W W 19 19
have the form 19
GN-modelled ES is not a self-addapted system";
102
CHAPTER 2
Fig. 2. 8
SC*E APPLICATIONS OF GENERALIZED NETS IN SCIENCE
103
The characteristic functions f of the tokens associated to the i corresponding place 1 are as follows: i $ 5 $
a a — > "!x " (the hypothesis x is valid); 0 0
a a — > ""Mix )" (the hypothesis x is not valid); 12 o 0
»
$ 31
— > "x U fx), where last which
(The place
x is the last characteristic of the token
has entered place 1 " 20
1 20
is an element of the subnet.
In it enter tokens with
current characteristic "new fact", which is received in the process of a checking of the given hypothesis x ); 0 o d d d / p cu cu cu / (x - (<pr x , Q(pr x )>) U {x i 0 last 10 10 , if the token d is in place 1 cu 41 I •* I
$
46
— —>>
(
\ I
, if the token d is in places 1 or 1 CU 42 44 d
d d cu cu cu (x - (<pr x , Q(npr x )>! V (x ) last 10 10 0 p
, if the token d is in place 1 cu 43 p
cu
x U (x V last 0
1 , if the token d is in place 1 cu 45
(the other functions are not defined). The GN considered in [3-6] ting production systems
shows the possibility
of represen
(their functioning and their results) by GNs,
This possibility follows also from the fact that a GN can describe the functioning of systems were
Predicate/Transition nets,
and in
[7-15]
production
represented by means of Petri nets, Predicate/Transition
10*
CHAPTER 2
nets and other Petrl net modifications. [3-6]
have the following
The GNs described here and in
universal property:
it is not dependent on
the particular modelled production system These GNs are not connected with the description of
the parti
cular ES. It should be mentioned that existing
(Known from the literatu
re) nets representing production systems are the particular system they describe and
rigidly
associated with
cannot be changed (deformed).
On the other hand the two GNs from [121, the second GN from [3, 4] and the GNs from
[5, 6]
they represent.
are maximally
independent of the particular
ES
Finally, the described here GN includes as particular
cases all previous constructed GNs describing ESs. REFERENCES: [1] Building
expert
systems,
F. Hayes,
D. Waterman
and
D. Lenat
(Eds. ) Addi son-Wesley Publ. Co. , Reading, Mass. , 1963. [2] D. Waterman, A guide to expert systems, Addison-Wesley Publ. Co. , Reading, Mass. , 1906. [3] K. Atanassov, L. Atanassova, E. Dimitrov, G. Gargov, ski, M
Marinov, and S. Petkov, Generalized nets
tems. I. 12-th
Methods of Operations Research,
Symposium on
Mathematik,
I. Kazalar-
and expert sys
Vol. 59.
Proc. of the
Operations Research of the Gesellschaft
OKonomie
und
Operations
Research,
1987,
fur
Passau,
Frankfurt a.M: Athenaeum, 1989, 301-310. [4] K. Atanassov, L. Atanassova, E. Dimitrov, G. Gargov, ski, M
Marinov, and S. Petkov, Generalized nets
tems. II. IFIP Symp.
"Network
Information
I. Kazalar-
and expert sys
Processing Systems",
Sofia, May 1988, Vol. 2, 54-67. [5] K. Atanassov, L. Atanassova, E. Dimitrov, G. Gargov, ski, M
Marinov, S. Petkov and M
nets and
expert systems. III.
Stefanova-Pavlova,
Methods of
Operations
I. KazalarGeneralized Research,
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
Vol. 63.
Proc. of the
14-th
105
Synposiim on Operations
Research,
Ulm, Sept. 1989, 417-423. [6] K. Atanassov, L. Atanassova, E. Dimitrov, G. Gargov, ski, M
Marinov, and S. Petkov, Generalized nets
tems. IV.
Proc. of the XIX
Spring Conf. of
I. Kazalar-
and expert sys
the Union of Bulg.
Math. , Sunny Beach, April 1990, 155-161. [7] M. Bonacina, Petri nets for Knowledge representation,
Petri Nets
Newsletter 27, 1987, 28-36. [8] J. Duggan and J. Browne, ESPNET: Expert-system-based simulator of Petri nets, IEE Proc. D
(Control Theory and Applications),
Vol.
135, No. 4, 1988, 239-247. [9] A. Giordana and L. Saitta,
Modelling production
rules by
means
of predicate/transition networks, Inf. Sciences 35, 1965, 1-41. [10] U. Mainz, Netztheoretische reprasentation pradikatenlogischer begriffe und methoden, Diplomarbeit, Univ. Bonn, 1985. [11] A. Sinachopoulos,
Derivation of
a contradiction
by
resolution
using Petri nets, Petri Nets Newsletter 26, 1987, 16-29. [12] S. Stoeva
and
K. Atanassov,
Generalized net
production systems interpreters,
Proc. of the
representation of Fifteenth
Spring
Conf. of the Union of Bulg. Math. , Sunny Beach, 1986, 456-464. [13] R. Valette,
Nets in production
systems,
Proc.
of an Advanced
Course "Petri nets: Applications and Relationships to dels
of Concurrency",
Bad Honnef,
Other Mo
in Lecture Notes in Computer
Science, 255, 1986, 191-217. [14] K. Voss, Nets in Data Bases, Proc. of an nets:
Applications and Relationships to
rency" , Bad Honnef,
in
Advanced Course
"Petri
Other Models of Concur
Lecture Notes in Computer Science, 255,
1986, 97-134. [15] D. Koutanis, R. Rasshidi, expert systems. 1987, 143-152.
Petri net representation of rule based
First Annual BSD/SMI
Expert Systems Conference,
106
CHAPTER 2
§ 2.6: GENERALIZED NETS RESEARCHING THE BASIC PROPERTIES GP KNOWLEDGE BASE Ewgeni Dimitrov,
U y a KazalarsKy
and Krasslmir Atanassov
One of the basic elements of every
Expert System
(ES)
is the
Data Base (KB). The structure and correctness of the DB determine the properties of the corresponding ES. The Knowledge in the ES
is repre
sented as Production Rules (FRs). The possibility
to check the correctness of
the KB
is a very
important condition for the functioning of the ES. We shall construct a reduced the existence
GN
of at least two rules in
(see App. 5) which determines a given EB,
having identical
antecedents (see Fig. 2.9). The capacities of its places and arcs are
Fig. 2.9 infinite. I t s transitions are as follows: Z =
SOJE APPLICATIONS OF GENERALIZED NETS IN SCIENCE
107
where 1l 3 1 II W W r = 11 II ll r 1 || 1 II W W 2 11 11
1 4 W W 2 W W 2
where W = "the place 1 i s empty" or "in place 1 there e x i s t s at least one 1
3
3
token with the characteristic - the same
antecedent
as in
the
present token",
w = -m ; 2
1 Z 2
= <(1 1, (1 , 1 ), r > 3 5 6 2
where 1 5 5 r 2 2
= 1 I W 3 1 3 3 1 3
1 6 6 W 4 4
where W
= "the transitions Z 3
and Z are not active" and "at the moment of 1 3 activation of the transition Z in place 1 there exists only 2 3 one token"
(this token enters place 1
and leaves the net;
if it is
the unique
5 token of the net, it receives as
a final characteristic
"there exist
no two PRs with identical antecedents"); W = "the transitions Z and Z are not active" and "at the moment of 4 1 3 activation of the transition Z in place 1 there exist at le2
3
ast two tokens" (these tokens enter place 1 6
and leave the net with a final characte-
CHAPTER 2
loa
ristic "the PRs which are Initial characteristics of these tokens have identical antecedents"); Z 3
= <(l !, (1 ), r > 4 2 3
where
1 2 r
: 1 3
I W 4 | 5
where W = 5
'the places 1 and 1 are empty". 1 2 All PRs of a given KB can be described as initial
characteris-
tics of the tokens of the above GN and thus an answer for tic question of the existence of identical antecedents in PRs can be obtained. The GN
which checks the set of al1
PRs
of a given
the existence of at least two rules frcni the form has the form as in Fig 2. 10.
KB
about
A I- C and A & B I- c
The transitions Z and 1
Z 3
in both nets
coincide with those from Fig. 2. 9 without
w - • the 1
place 1 is empty" or "in place 1 there exists at least one 3 3
token with the characteristic - the same
consequent,
as in the
present token". The transition
z =
3
where
I ), 9
a , i ),r 5 1 5
r
2
6
> 2
1 6
= 1 | W 3 1 6
W
1 1 W 9 1 6
W
7 7
where W
6
= ' the place 1 is empty" or "in place 1 there exists at least one 5 5
SOJE APPLICATIONS OF GENERALIZED NETS IN SCIENCE
109
Fig. 2. 10 token with
the characteristic - an antecedent,
being a subtenu
of the antecedent which is an initial characteristic of the present token",
w = nw . 7
6 The t r a n s i t i o n z <[1 1, !, 11 11 ,, 1l ). ), rr > Z -:
where
where W
3
and W
4
1 77
1
rr = 11 II W W 4 55 11 33 4
W W
are as above.
The transition
8 8 4 4
110
CHAPTER 2
Z 5
= <(1 !, (1 ), r >, 6 9 5
where 1 9 r
: 1 I W 6 1 8
5 where W 6
= "places 1 and 1 are enpty". 3 5 The result
of functioning of this GN is as follows: the tokens
corresponding to the rules of the type exist)
leave the net at place
1 . 8
A I- C and
ASB
K
If all tokens leave
(if they
the net from
place 1 , then in the KB there exist no such rules. 7 The GN checking a given KB for contradictions, i. e. , for simul taneous existence of rules of types A I- B and
A I- IB,
is similar to
the GN from Fig. 2. 9, but here the consequent is C = nB. The GN
which checks a given KB
for existence of
two rules of
types A I- B and B I- A also has the form as in Fig. 2. 9 and all notati ons are identical without the predicate W = "the place 1 is empty" or "in place 1 there exists a token 1 3 3 which the antecedent
for
(in its initial characteristic) is a con
sequent in the initial
characteristic of the present token and
the consequent (in its initial characteristic) is an antecedent in the initial characteristic of the present token". As the result of place 1 have initial 6
functioning of this GN the tokens leaving the characteristics - contradicting
rules.
If all
tokens leave the net at place 1 , then the KB is noncontradictlng. 5 Considering the GN, graphic tool
we conclude
for representing
that
the structure
such nets are and functioning
a useful of the
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
ESs. The representation of the ESs by GNs it
will probably become
111
has certain
advantages and
even more popular in future.
This will con
tribute to an easy transition
from present systems to systems of
the
fifth generation computers. This paper is based on [1]. REFERENCE: [1] E. Dimitrov,
I. KazalarsKi and K. Atanassov.
Generalized net re
presenting the basic properties of Knowledge base. fic Session of the ligence" Seminar, Sofia, 1989, 12-14.
First Scienti
"Mathematical Foundation of Artificial Sofia,
Intel
Oct. 10, 1989, Preprint IM-MFAIS-7-89,
112
CHAPTER 2
§ 2. 7: PARALLEL KNOWLEDGE PEOCESSIHG HOEELLIHG IN EXPERT SYSTEM USIHG GENERALIZED HETS Ivan Hrlstozov
The main task
and
Evgeni Tzolov
in designing expert systems (ES) for application
in fields such as decision support, image trol and others is to
processing,
real time con
achieve high-speed data processing.
The appli
cation of the parallel computation principles in the ES as a whole and in the inference mechanism in
particular, is a way to achieve a quick
response. A method for a formalized
description of a parallel
inference
mechanism in ES based on production rules using Generalized Hets (GHs) is proposed
in this paper.
The organization and the interference of
the parallel processes in a multiprocessor system
are being
studied.
The system is built up of specialized processors for parallel computa tions (SFPC) of the transputers type [1-3]. The production systems are a popular means for knowledge repre sentation.
Each production system consists of three basic components:
a data base, a set of rules (a knowledge
base] and a rule interpreter
(a logical inference mechanism]. The knowledge is represented
by pro
duction rules of the type IF "condition 1", "condition 2". . . "condition H", THEB "conclusion". The activation of each rule changes the data in the data base. Parallel
data processing in ES on
the basis
of a production
system will be considered as a simultaneous interpretation of a set of rules in a multiprocessor system of SFPC.
The rule evaluation in each
SFPC is sequential. A list of ldentiflcators (names) of the conditions and the con clusions
included in the rules
hypotheses to be proved
and a list
are entered in
of identificators of
the process of
the
the knowledge
base formation. The list of hypotheses is a subset of the
list of the
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
conditions' and conclusions' identificators. ledge
base only the rules
113
After forming
the know
necessary for proving the given hypotheses
are separated. The Knowledge base thus reduced is divided into groups (levels) of
rules, which
can be
activated parallely, i.e.
they
are not in
formational ly interconnected. Rules whose conditions (the variables on their left sides) are not conclusions of other rules, i. e.
rules that
can be directly evaluated, are referred to as the first processing le vel. Rules that depend on at least one rule of the
(i-l)-th level are
referred to as the i-th level (i * 2). Rules of the i-th level may de pend on rules of the lower levels.
Only the rules whose
are the chosen hypotheses are evaluated on
conclusions
the last level.
The rules
from a given level are distributed for parallel processing between the processors of the multiprocessor net. For
the case where the number of independent rules
level is less than or
equal to
the number
of a given
of processors in the net,
each rule can be evaluated on a separate processor. When the number of rules on a level is greater than the number of processors, some of the independent rules cannot be evaluated parallelly,
because of the lacK
of resources (processors). If
the net
processors),
of processors is homogeneous
one rule
will be evaluated at
(built up of similar
one and the same time on
the different processors. So balance processor loading in
time may be
chosen as a distribution criterion. Ihe distributed rules (the indivi dual knowledge bases) are loaded in the processors. When starting up the inference mechanism, the data base is sent to each processor of the net. When processing the rules of a given le vel, each processor changes a given part of it. In order to standardi ze the data base
at the end of the rule processing on a
information exchange
between the
SPPC is necessary
values received in each SPPC are necessary to
given level,
(the
conclusion
continue the processing
114
CHAPTER 2
in the other processors of the net). The evaluation process of the rules continues no more
levels. The
until
there are
values of the chosen hypotheses are sent to
the
user (the controlled object), new data (if any) are received from him, and the processing is repeated. When developing sing,
the multiprocessor
the question as to
system for
parallel proces
the evaluation of its effectiveness arises.
On the stage of designing when the system Is still not physically rea lized, this is possible by using the imitation modelling methods. The GNs are Introduced as an effective means for modelling
and
simulating. They permit a detailed investigation, a natural way of re presenting
parallel
and asynchronous processes,
receiving temporary
characteristics. Hie operation
of the multiprocessor system
performing
parallel
lnferece in the above stated way may be described by a reduced GN (see App. 1), given on Fig. 2. 11. It has three transitions. Transition Z 1 of the type Z 1
= <(1 , 1
1 , 2
1 , 1 , 1 , 1 , 3 4 7 6
1
), 14
[1 , 8
1 , 4
1 , 9
1 , 5
1 , 6
1
,
the place
1
1
lO
is
!, 13
r >. 1
In the beginning of the GN functioning,
contains
1 tokens,
symbolizing the rules of
the following characteristics:
the Knowledge base.
Each token has
"execution time, "priority". The prio
rity locates the level of a rule. The rules with an identical priority are processed for one realization of the transition. The tokens from place 1 with the lowest priority pass into place 1 with the charac1 13 teristics - "rule number", "identificator showing whether the operati on in the THEN part is performed or not". Tokens with a higher priori ty pass in place 1 to without changing their characteristics and 8
are
SCME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
115
Fig. 2. 11 subjected to a subsequent inference all into 1 13
processing.
After terminating
rules are evaluated and the tokens
the logical
from 1 have passed 1
The tokens from 1 go into place 1 , if a history tracing 13 15
of the inference is necessary, or into 1
for a cyclic repetition of
14 the inference. In the beginning of the GN functioning, place
1
(input place)
2 has tokens with initial characteristics "data base identificator" a
"list of
hypotheses we
are interested in".
and
Each SPPC has its own
116
CHAPTER 2
data base ldentiflcator. These toKens pass into place 1 with the cha4 racteristic "updated data base". the next processing level.
Thus 1 4
becomes the input place for
When there are no more levels,
the tokens
from place 1 and 1 pass into Place 1 with the characteristic "final 2 4 9 data base".
In the beginning of
kens whose characteristics
the GN functioning place
1 has to3
are: "number of the corresponding proces-
sor", "numbers of the processors directly connected with it", "identificator of the data basej which is the initial
characteristic of
the
the processing of rules with the same priority ends,
the
token from 1 ". 2 When
tokens from 1 , under the conditions specified below, pass into places 3 1 5
and 1 , with the respective characteristics: 6
sors from which data are received",
"numbers of
proces-
"numbers of processors, to which
data are sent", besides "time for conmunication between the corresponding
processors".
In the case where the number of rules
level is smaller than the number of
processors in
cessors do not take part in the processing with
characteristics
of
a given
the net, some pro-
on that level.
The tokens
"the numbers of such processors" pass
into 1 6
and
have
no characteristics.
operating processors only,
The tokens from 1 , 3
pass into place
"new facts (changes in the data base the corresponding processor
corresponding
1 with 5
to
characteristics
received during the operation of
with the rules of
the specified level)".
So, when passing the transition Z , the tokens from 1 are divided in1 3 to new tokens, the number of which is equal to the number of transmissions/receivings . In 1 7
all processors have the characteristic
"new
SO»E APPLICATIONS OF GENERALIZED NETS IN SCIENCE
117
facts, received during operation". At the end
of the operation the tokens from 1 pass into 1 and 3 lO have no characteristics. The condition r
is of the type 1 1
1 44 1
I
1
false
55
11 66
11 8«
99
1l lO lO
false
false
w 5
false
false
11 1 I 2 1
w
false
1
1
false
W 2
I 3 |
r 1
= 1 I W 4 | 1 11
false W 3
1i
false false
1l 13 13 w 4
w 4
false
false
false
W 4
false
false
false
false
W 4
false
false
false
false
1 I false 7 1
W 2
W 3
false
I I false 8 I
false
false
w 1
false
false
false
false
false
false
w
false
false
ffalse alse
w w 4 4
1 I II II II
w
W 4
1
4
1 || ffalse alse 14 I 14 I
ffalse alse
ffalse alse
w w 55
ffalse alse
where W
= "a new fact is found", 1
W
= "there are processors that will transfer data"; 2
W
: "there are processors that will receive data"; 3
W
- "all rules are evaluated"; 4
W
: "there are groups of rules for processing". 5 The appearance of a token in place 1 activates the transition 9 Z 3
: < [ l , 1 I, II , I , 1 , 1 !, r > , 9 13 11 12 14 15 3
which defines data
exchange with
an external object.
The token from
110
1 9
CHAPTER 2
is divided in two: one part passes into place
ristic
"input data
identificators" and
1 with 11
receives data,
the final user; the other part passes into place 1
characte-
if any, from
with the
charac-
12 teristic "hypotheses" and is sent to the user (controlled object). The condition r
is of the form 3
r
11 11 11 1 I W 9 1 6
3
11
12 12 true
1 II ffalse alse 13 13 II
11
11 14 14 false
ffalse alse
W W 7
15 15 false true true
where W
= "there are new data for receiving"; 6
W
= "a new inference execution is necessary". 7 The
of
transition
processors ready
kens with the
Z is activated if there is at least 2 for
data
same initial
exchange between them
characteristic
one pair
It unites to
"processor
number".
By
the end of the last exchange operation, all tokens from 1 and 1 have 5 6 passed into place 1 , which is 7
the initial place
for processing the
next level of rules with one and the same priorities. has the form Z = <(1 <(1 ,, 11 ), ), (1 (1 ]], , rr > 2 55 66 77 2 with condition 1 7 1
I W i a r = I 2 I 1 I W 6 I 9 5
Transition
Z 2
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
119
where W
= "the processor in 16 has received all data";
a - "the
W
processor in 15 has transferred the new facts
to the other
9 processors". If a
moment of time t is compared with the moment 1
ting the transition Z , 1
and a moment t
of activa-
to the moment of the
end of
2
exchange with the controlled object (deactivating the transition the difference t - t 2 1
Z ), 3
determines the time for data processing in the
expert system The model developed of
logical
permits on to evaluate the characteristics
inference parallel computers with
a different connection
topology, number of processors, rule distribution,
exchange organiza
tion (priority of one processor to another) aiming at finding an opti mal (according to a specified criterion) version.
REFERENCES: [1] D. Richard, C. Askew and D. Carpenter, Practical parallelism using transputer
arrays, Lecture Notes in Computer Science 258, 1987.
[2] Transputer development system 2.0, IMS D700C, INMDS, 1987 [3] K. Yanev, D. Todorov, N. Avramov and E. Elitzina,
Designing digi
tal devices with microprocessors, Sofia, Tehnika, 1987 (in Bulg. ).
120
CHAPTER 2
§ 2. 8: GENERALIZED SETS AND DISTRIBUTED DATABASE TECHNOLOGY Alexander Georgiev
§ R. 8. 1. (GN)
Introduction
This article is an introduction to present our
Generalized Net
approach to the Distributed Database Technology
(DDBT). The GNs
should be considered as a possible response to
the problem of tbe de
velopment of tools for parallel process modelling. The DDBT is a state-of-the-art technology in the database theo ry.
The DDBT consists of two basic domains:
Global Intelligence (GI)
with laws and rules governing the Distributed Database (DUB) [1]. The aim of this article is to introduce the foundations
of the
GN approach to the building of the Global Distributed Database Manage ment System (GDDBHS)
in the DDBT research area.
the specifications of the ANSI/SPARC wing
project proposed in [2]. Follo
the current status of the DDBT theory,
to have
Our approach follows
the GDDBHS is considered
two conceptual domains: the DDB as a set of new basic
tural concepts in the DDBT
struc
as well as a lower conceptual domain,
and
an upper conceptual domain - the GI as a wise manager in a distributed environment. All discussions are on the conceptual level only.
Final
ly, we propose the architecture of an abstract GDDBMS which integrates all features from the DDBT area on the
conmon basis of the GN
forma
lism « K
§ 8. 8. 2. Distributed
X
database - basic structural
In the following paper we shall and their consequences of the
discuss three
for organization of
information network.
concepts basic
concepts
the DDB logical structure
They define the data architecture in the
DDB on some layers or schemes, as shown in [1]. The first notion
is fragmentation.
Fragmentation
is a scheme
SCME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
which describes how tables
(relations)
are shared among
121
the network
nodes logically. There are three fragmentation types: horizontal, ver tical
and mixed one.
Fragmentation
allows one
to define
different
users' views on the data in the database. The second notion is replication. Replication is a scheme which describes the distribution of different whole tables as copies of data among
the network nodes.
Replication refers to fast local access
to
data as well as to a higher assurance of the network's working. The third notion is allocation. Allocation of
data is a combi
nation of fragmentation and replication. Allocation describes the fragments with their physical
is a scheme which
addresses on the network.
Allocation is a key element in a project of DDR Allocation has a gui ding principle: local availability. The three concepts of the DDB. for
shown above define the layer's architecture
The DDB architecture has
to follow
the basic principles
logical and physical independency of the data.
Therefore,
should exist mappings between the layers. Mapping is facts and
rules for a transformation from one logical aspect
ther logical or physical aspect
of the DDB.
there
a description of to ano
In that sense, a mapping
could be viewed as a global, fragment at ed, allocated, or local-mapping scheme. it
§ 2. S. 3, Distributed
database - basic
«
logical
concepts
The notion "distributed transparency" is closely related to the DDB architecture. Each layer from the architecture a
conveniently distributed transparent layer.
is associated with
A transparency
DDB means that the user should see that the database user's
node with
its whole entity.
resided
of the at the
The distributed transparency
is
orientated to ensure both logical and physical independency of the da-
122
CHAPTER 2
ta in the DDB. The fragmentation transparency has the highest transpa rency level. The Distributed Data Integrity (DDI)
is a feature that ensures
the consistency and security of the data. It is related
to the follo
wing problems of supporting the activities on the DDB: query optimiza tion, concurrency control and transaction handling.
* K
§ 2. 6. 4. Global
It
Intelligence
We consider
GI as
an entity which is most important
for the
the existence of the DDB. A node contains a Database Management System (DBMS). The DBMS represents Local Intelligence (LI) in the DDB system The GI
exists above the LI. GI could
be considered as a
DBMS (DDBMS). The GI manages and controls data, Knowledge about the data
(metadata)
Distributed
the data structure, the row and all
activities of the
overall DDB. The GI is responsible to ensure: (a) Distribution
Transparency
as a structural
aspect as well
as an
activity in the DDB; (b) a DDI
by means of supporting the query optimization,
concurrency
control and the transaction handling. All these activities require the capacity to maintain processing. Parallel processing is related
parallel
to the notion of real-time
synchronization. It is the biggest problem in a DDBMS. For solving the problem it is necessary to have a power modelling tool for behavioural research of a parallel processing
in the real-time situation. The GNs
are the ideal ones in this sense. The GN can behave as
a set of
pro
cesses at a certain moment of time, each one with its own history, lo cal time and constraints. Moreover, the GNs can forecast the behaviour of a stand-alone process, cesses
as a whole,
a subset
and then detect
of related processes and the pro deadlock or
wrong situations as
SOME APPLICATIONS OF GENERALIZED NETS IN SCIENCE
well as correct them We consider the
123
GNs a conceptual modelling tool
for a GI above the DDB. We distinguish
some sub-models,
for example:
GN-model for query decomposition, GN-model for query optimization, GNconstraint integrity model, GN-transaction handling model, and GN-conceptual model of the GI (GNGI-model), which is a general one.
« § 2. 0. 5.
«
Conclusions
In this paper we have presented the idea of the GN in the DDBT area. We have reason to thinK that this to
the building of a DDEM5 as a GI in
will
is a new approach
a distributed environment.
discuss the submodels of the GNGI-model precisely
works.
Moreover, the idea of using
DDB area could reflect
application
We
in the future
the GN as a modelling tool in the
the performance of intelligent and power data
base machines.
REFERENCES: [1] R. Davis, Sharing the Wealth, BYTE, September 1989. [2] The ANSI/SPARC DBMS framework, Information Systemsm, 3, 1978.
Vol. 3, No.
Chapter
3:
SOME A P P L I C A T I O N S OF GENERALIZED NETS IN ECONOMICS, INDUSTRY AND TRANSPORT
Several GH-models describing various processes in areas of eco nomics, industry and transport are introduced. Some of them are publi shed for the first time (§3.3, § 3.4, § 3.5, §3.7, § 3.8). These mo dels
describe the processes from the view of Bulgarian economics, in
dustry and transport. Some of them give only an idea about the form of the GH-models which can be constructed (e.g. in § 3.3), others descri be the processes in a broad outline
(§ 3. 1, § 3.4, § 3.5, §3.7, § 3.
fl). The model from §3.2 and those from §3.6
are more detailed. The
models from §3.7 reflect the first steps made on the global modelling of the NEFTOCH1H Petrochemical Combine in Bourgas (Bulgaria). Therefo re,
the GH-models
described here are not
universally
valid
or of
worldwide importance. On the other hand, they do illustrate the possi bility for a GH-description of such processes and also for other simi lar ones. The models are realized in the frames of the current version of
the program package for GHs and they can be given to those who are
interested in them.
124
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
135
§3.1: MODELLING THE ACTIVITIES OF IHTERHATIONAL TRANSPORT MARKET BY GENERALIZED NETS Antoaneta Kirova
The international
and Krassimir Atanassov
transport market
is a complicated dynamical
system developing under the influence of a number of factors. Its stu dy
is connected
aspect
with creating
different models which
give a visual
of the processes and afterwards make it possible
to determine
the ways of improving them in practise. The modelling of economic processes includes retical and
statistical data
The models give an idea about ging
the elements
tasks of
statical which
the interconnection of ways
of a complicated
economic system
the management at a given stage define
modelling:
statical,
modelling
dynamical, is
a number of theo
concerning their appearance in reality. of arraan-
The gears
a number
expert-imnitational and so on.
characterized
surround the economic process.
by firmly set
and
of ways of
out
The
conditions
An example most typical in this
aspect is the transportation task. The creation of dynamical models is possible with non-stopping processes or phenomena. of
computers
in practise
gives
processes and phenomena which only describe.
This concerns
The widespread use
the possibility of modelling
those
the expert-imnitationa] method can
a number of management processes,
parallel-
going processes, and so on. One of the ways of creating expert-inmitational and simulational models
is by using different types of nets.
The main quality of QNs is that they give not only surement of the tokens, place
but also
the conditions
a time mea
for going
from one
to another. This creates some prerequisites for simulating re
al processes and for optimising the existing ones. Here we intend to show two of the processes connected
with the
international market of transport services. The modelling concerns the
126
CHAPTER 3
activities of the forwarding agent and the transporting firm.
» « Let the GU
*
of the forwarding agent be
E
(see Fig. 3.1).
We
c shall assume means
that the tokens
conducted by the
represent the
forwarding agents.
different transportation These tokens possess
the
following characteristics: type of the transportation means, the (cur rent) locations, time for covering a certain distance and price of the transportation service. All these are changing in time and in the uni verse when passing from one place to another. The forwarding agent's (reduced] GN is as follows: o « E : <(t, 1 , f>, K,
Z ), 2
6
(b) the set K of tokens consists of all the trailers and other portation means which the forwarding agent might
use at
trans a moment
of time. (c) the time GH's components are as follows: T - a fixed start time-moment, o t - one day, t
- one year. The token's characteristics, which are related
a very
important part
of the generated
to the time are
information in the net.
initial token's characteristics have the form
<"type of the transpor
tation means", "location"). The characteristic function $ lustrated at
each place after
and for them, the function
$
The
will be il
its description.
The GH has 12 places
juxtaposes to the
corresponding tokens
SCBE APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
Fig. 3. 1 the following values: 1 - "location of the transportation means", 1 $ - <"new (current) location", "price of passage">, 1 1 - "transportation means sent for loading", 2 $ - <"kind of a good", "price of passage", " time for passage">, 2 1 - "transportation means of sending to the agent's warehouse", 3 $
- <"Kind of a good", "price of passage", " time for passage">, 3
1 - "transportations means of sendinf for re-loading", 4
12T
128
CHAFFER 3
$ - <"direction of transportation">, 4 ] - "transportation means for proceeding to the consumer", 5 $ - <"direction of transportation">, 5 1 - "transportation means for a mixed transportation", 6 $
- <"location", "time for loading operations", "price of passage"), 6
] - "transportation means sent to an intermediate's warehouse", 7 $
- <"location", "time for loading operations", "price of passage"), 7
1 - "transportation means is discharged", 8 $ - <"direction of transportation"), 8 1 - "trailer/road railer on mixed mode of transportation", 9 $
- <"direction of transportation"), 9
1 - "trailer/road railer at the final point", 10 % - <"location", "time for loading operations", "price of passage"), 10 1 - "emergency situation", 11 $
- <"emergency situation"), 11
1 - "goods in the consumer's warehouse", 12 $
- <"location", "time for loading operations", "price of passage"). 12 "Ehe above mentioned places and the characteristic functions re
lated to them, because he
correspond
to the activity
represents the goods'
of the forwarding
agent,
owners to the transportation firms.
He is concluding the transportation contracts, determines the mode
of
transportation and is used to differentiate other operations connected
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
with
the preparation,
running and
completion of
129
the transportation
process. Components which are anticipated to have an indirect participa tion in the model are emitted.
They are t -, t 1 2
and M-transitions'
components, it -, c-, © -, © -, u -, © -, b-GNsJ components. Therefore, L 1 2 K K the described ©? is one that is a reduced from the class A ,A ,A ,11 , c, © , © ,TI , © ,b 3 4 6 L 1 2 K K
E The capacities
of the places
and the arcs
are infinite.
The
transitions have the following forms: Z
=
r
= 1
1 I true 7 I I 1 I true 8 I 1 I true 12 I
The transition's type of Z
is "v" because it can be fired if a 1
token of
(a transition mean
its input
places
which is already discharged) appears in one
(i.e.
in 1 , 1 , 1 7
ponds to the "sending
8
). This transition corres-
12
of the transition means from
forwarding agent". Z = <[1 1, (1 , 1 , 1 !, r , v(l )>, 2 1 2 3 4 2 1 where
its owner
to the
130
CHAPTER 3
1 2 r = 2 2
1 1 1
I W 11 4 1
1 3
1
W
W 3 3
4
2 2
where W
= "the transportation means is with the producer", 1
W
= "the transportation means is within the agent's warehouse", 2
W
= "the transportation means is sent for loading"; 3
Z 3
=
where 1 5 r
= 3
1 2
I 1
W 4
1 3
I 1
W
1 6
1
W
W 6
7
5 W 7
false 8
where W
= "the transportation means is moving from the producer to the con4 sumer",
W
= "the transportation means is for re-loading", 5
W
= "the transportation means is in the intermediate's warehouse", 6
W
= "the transportation means is moving from the agent's warehouse to 7 the consumer",
W
= "the transportation means is for re-loading in the agent's ware0 house". Here
1 and 1 are input and 1 , 1 , 1 2
transition's type is
3
5
6
are output places. The 7
"A", because for its activation it is necessary
to have at least one token in each of the transition's Z means "first stage of transportation". 3
input places.
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
131
Z = <[1 ), tl , 1 ), r , v(l )>, 4 6 S 9 46 where
r = 4
1 6
I |
1 «
1 9
W
W 9
10
where W
= "the transportation means is discharged", 9
W
= "trailer/road railer continuing". 10 Z denotes "end of the first stage of transportation and begin4
ning of the second one. Z 5
= <(1 , 1 ), [I , 1 ), r , A(1 , l )>, 9 4 lO 11 5 9 4
where
r = 5
1 lO
1 11 W
1 9
I I
W
1 4
I 1
W
11
12 W
13
14
where W
= "the goods have arrived", 11 = ~\W
W
12 W
,
11 = "the transportation means have arrived",
13 W 14
= nW . 13 Here
1 and 9
1 are input, and 1 and 1 4 10 11
are output places.
The transition's type is "«" because for its activation it is necessa ry to have at least one token in each
of the transitions' input
ces. Z means "end of the second stage of transportation 5
pla
and beginning
132
CHAPTER 3
of the third one". Z 6
-- < U , 1 ), II 1, r , «|1, 1 )>, 5 10 IE 6 5 10
where 1 12 r
=
1 ] I true 5 I
6
1 | true 10 I Z expresses "completion of the transportation process 6
and de-
livery of goods to the consumer". The proposed agent's
model
shows only
those parts of
the forwarding
activity which require management in time and universe.
is particularly
This
important in the international field, where the deli
very term needs punctuality in the contacts between the goods and
the
transportation means.
the
The effectiveness is greater in organizing
multimodal transportation abroad, for the processing there is
divided
into different stages (land - river - land, land - see - land, etc. ). > •
«
The transporter's GH E is as follows: n o
■
E = <, K,
b>>,
where A = IZ , Z 1
Z ] 2
6
(same as above, but the new transitions are different from the above}; the elements of the set of toKens K also
represent the transportation
means in the transportation firm; the time components of the net are the same as with E ; the token's characteristics in E are analogical c n
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
to E ; c
the initial token's characteristics are from the type:
133
<"type
of the transportation means", "location">. The places of this net denote the following: 1 - "the transportation means is ready for loading", 1 1 - "the transportation means is sent for loading", 2 1 - "the transportation means meets 3 port",
goods from
a multimodal trans-
1 - "the transportation means has got the goods", 4 1 - "the transportation means has given 5 port",
the goods to another trans-
1 - "the transportation means is discharged", 6 1 - "the transportation means is moving loaded", 7 1
- "the transportation means has left the goods to the consumer",
a 1 9
- "an emergency situation",
1 - "end of the second stage of a multimodal transportation", 10 1 - "an emergency situation with a multimodal transportation", 11 1 - "continuing of the second stage of a multimodal transportation", 12 1 - "start of the third stage of a multimodal transportation", 13 1 - "repair/storage of the transportation means". 14 The function $ of the places is similar to the above: $
- <"location", "time", "price of passage")-, 1
$ - <"direction of transportation'^, 2
134-
$
CHAPTER 3
- <"direction of transportation"), 3
$ - <"direction of transportation"), 4 $
- <"direction of transportation">, 5
$
- <"direction of transportation">, 6
$
- <"direction of transportation",
"time", "price of the passage"),
7 $
- <"direction of transportation">,
a $
- <"eniergency s i t u a t i o n " ) , 9
$
- <"location",
"time", "price of passage"),
10 $
- <"emergency situation"), 11
$
- <"direction of transportation"), 12
$
- <"direction of transportation", "time", "price of the passage"). 13 As with E , components which do not participate directly in the c
model will be omitted. This GN is a reduced one from the same class of reduced GNs as the first one. The transition Z
=
where 1 1 1 I true 6 I r = I 1 1 | true a i 1 I true 14 |I denotes
"the transportation means is ready for taking up goods/out of
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
Fig.
3. 2
usage". Z - <(1 I, 2 1
(1 , 1 , 1 ), r , v ( l )>, 2 3 14 2 1
where 1 2 r
= 2
1 1
I 1
1
1 3
W w 1 2
where w = "the t r a n s p o r t a t i o n means i s working", 1
14 W w 3
135
136
W
CHAPTER 3
= "the transportation means is sent to meet the goods", 2
W
= "the transportation means is sent to be repaired". 3 Z
denotes "the transportation means is sent to meet the goods". 2
Z
= <(1 , 1
3
2
),
(1 . 1 , 1 !, r , v ( l , 1
12
4
5
6
3
2
)>, 12
where
rr
33
= =
11 22 1 12 12
II 11 II II
1 1 1 14 15 16 4 5 6 W W W W W W 44 55 66 ffalse a l s e true true false false
where W
s "the transportation means is moving down a road", 4
W
= "the transportation means is on water/railway", 5
W
= "leaving of goods and moving back for taking the new ones". 6
Z
= <(1 , 1 , 1
4
4 7
),
11 , 1 , 1 i, r , v ( l , 1 , 1
13
7
8
9
4
4 7
11 4 4
1 17 7 II W W 1 6 6 1 I W W 11 66
)>, 13
where
rr
44
= =
11 77
11
I 13 I
1 18 8 W W 7 7
1 19 9 W W 88
W W 77
W W 8
W W
W W
6
7
W W 8
,,
where W
= "the transportation means is moving on", 6
W
= "the transportation means is at a new location", 7
W
= "the transportation means is having an emergency". 8
SOME APPLICATIONS OF (3*5 IN ECONOMICS, INDUSTRY AND TRANSPORT
137
Z denotes "the goods are transported by land to the consumer". 4 That transition aims to model the movement of the transportation means by land in different countries. Z
= < U ), (1 , 1 , 1 J, r , v(l )>, 5 5 10 11 12 5 5 where
r
= 5
1 5
| I
1 10
1 11
1
W
W
W
9
12
10
11
where W
= "trailer/road railer at the end of the second stage of transpor9 tation";
W
= "trailer/road railer having emergency"; 10
W
= "trailer/road railer moving on". 11 Z
denotes
"movement of the transportation means in the second
5 stage of nultimodal transportation". Z = < U , 1 ), (1 I, r , M i , 1 )>. 6 9 10 13 5 9 10 where 1 13 r = 6
1 I true 9 I 1 I true 10 l
The transition Z 6
is of the »-type
because it shows the exis-
tence of at least one token at both its input places. It describes the "end of the second stage of multimodal
transportation and begining of
a third one". Unlike GN E , c
there is a cycle between the transitions
Z and 5
138
Z
CHAPTER 3
here, which corresponds to going in a certain direction. The notion 3
of the transportation means with verse
causes
a change of their locations in uni
increasing charges for passing to a definite stage and
leads to changes in the characteristics of the tokens. The GN E can n be used for (a) prognosing the expenses of the transportation firm for giving ser vice
when knowing the direction and the way of changing
the cha
racteristics; (b) if the time-parameters related to the tokens are known, the trans portation firm can follow their movement and send
them for dispo
sal of the forwarding agent for a further use and new contracts. The above models
show only a part of the possibilities of GNs.
These models could be used not only for simulation of concrete proces ses, but also for management (in combination
with expert systems) and
control in real time. Thi s paper i s pased on [1]. REFERENCE: [1] A. Kirova,
K. Kirov,
K. Atanassov,
Modelling the
transporting market a c t i v i t i e s with generalized nets I z t c h i s l i t e l n a Tehnika i Avtomatizirani sistemi", 66.
(in Bulgarian).
international "Avtomatiua,
1985, No. 6, 61-
SCHE APPLICATIONS OF GHs IH ECCNOHICS, INDUSTRY AHD TRAHSPQRT
139
§ 3. 2: GENERALIZED HETS FOR SIMULATION QF BOOKING SYSTEMS FOR PASSENGER TRANSPORT Sergej Nedev
and
Lllija Atanassova
The automated systems for booking and sale of tickets are ter ritorially distributed complexes having a set Toe
of parallel
building of such systems requires great investment.
models
are necessary for a detailed analysis
function of toe complexes,
of
and for correcting
itself. The result of this analysis can be
processes. Mathematical
the parameters
errors in
and
the project
used for choosing an opti
mal hardware and software for building of the system A
mathematical
instrument for analysing
such systems
is the
simulation based on GNs. A GH represents not only the structure of the process but also the dynamics of the process. In comparison with other net models, GNs are very compact and simple in structure. They provide very detailed information about the simulated processes. In 11] a model of a booking system is described. Here the topic of the simulated processes will be extended. Let us examine the GN mo del
of a fragment of a booking system.
The fragment consists of
the
following components: 1. Points for booking - 30; 2. Box-offices at every point, numbered from 1 to 8; 3. The main computer centre of the system (HOC); 1. Telephone channels
for comunication between the HOC and the points
(three logic channels, each containing 5 physical channels). The model gives the possibility of the information about a part of the vehicles to be stored at the points from which they start.
The
channel access is of the type CSHA/CD. The maximum length of a message is 20 bytes. The transmission speeds in the channels are: HCC-point
-
24O0 bit/s;
point-box-office
- 500 000 bit/s.
140
CHAPTER 3
MCC response to a request is 0. 5s. Our GN model is based on the following reduced GN (see § 1. 1> , <X, $>>, where A = !Z , . . . , Z ) is a set of transitions of GN, K is a set of 1 9 Every
system's request is represented by exactly one token in
tokens.
the GN
model. Every token has the initial characteristic of a couple of posi tion n
integers (m,n), where m is the number of the
the number
of the box-office
at
the booking
booking point point.
characteristics are exactly the elements of the set X. specifies for every token ct€K the moment
and
These token
The function 6
of time it will enter the GN
through the initial place 1 . It is convenient to considered 1
the cha-
racteristic function as 16 $ = U $, i=2 i v*iere $ is a characteristic function related to the place 1 (the toi i kens
enter
1 with their previous characteristics without receiving 1
such in 1 ). A variable can be specified in the model. Its values cor1 respond to the increment of time for the simulated process. ponent
is not related
global time components
to the global
This com
time components of the GN.
are not necessary for this
model and
The
that is
y/tiy they are omitted. Let T denote the current value of this variable, T is its initial value, t is the time for the stay of the token 0 a the GN and t i
a in
the time for the stay of a token in the place 1 (1 <. i i
i 16). The places
and
the characteristic
have the following meanings:
functions
related to them
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
1
141
- stands for box-offices. In the box-offices the users requests are are created, shown as tokens of the GN.
1 - stands for points of booking. All requests are collected 2 point. $ - specifies the value 2
X + t a 1
as a characteristic of
the
the tokens,
where X is the final a
characteristic of
0. 002 s .K,
is a random number for which 1 S K 5 9 (the
where K
number 0. 002s
the token
at
a
and
t = 1
is the duration of a single transmission between a
box-office and a booking point). 1 - stands for the queue of requests, waiting for sending through 3 the telephone channel between the booking point and the MCC. $ 3
- specifies the duration of the tokens' stay in place 1 . 3
For 4 i i i 8: 1 - stands for sending i channels
of
the request through
$ - specifies the value X +t + t i a 2 3 kens, where t
as
= 0. 1 . K, where K
one of the telephone
a characteristic of the
to-
is a random number, 0 i K <. 4.
2 It is assumed that in 70Z of the cases are in 0 i K i 1. The num ber 0. 1
is the duration in seconds of
the sending of
a request
through the telephone channel, t
1 9 5 9
= 0. 1 . (Q + 1), where Q is the 3 number of the simultaneously transmitted requests (4 <. i < 8). - stands for request reception in the MCC of the system - specifies the duration of the tokens' stay in the place 1 . 9
1 - stands for request processing in the MCC. 10 $ 10
- specifies the value X + t as a characteristic of the tokens, a 4 where t = 0. 5 s. (It is assumed that one request is processed 4
142
CHAPTER 3
In the average time of 0. 5 seconds). 1 - stands for the queue of requests, 11
processed in
the booking
point itself. $ 11
- specifies the duration of the token's stay in the place 1 11
1 - stands for request processing in the booking point. 12
$ 12
-- $ . 10
1 - stands for cashier reception of the system response. 13 $
- specifies the value
X +t +t as a characteristic of the tokens, a 1 5
13
■where t = L, L being a random number equally distributed in the 5 the interval 60 - 120s. 1 - stands for the end of request processing. 14 1 - is analogous to 1 15 13
$ 15
=$ . 13
1 - is analogous to 1 16 14 The capacities of the different for 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 2 3 9 11 13 14 15
places are:
for 1 - infinite, 1 - 240, for 1 - 30, for 1 16 12 10
1, for 1 , 1 , 1 , 1 , 1 - 24. The capacity of each arc of 4 5 6 7 8 sitions is 1,
the tran-
e. g. every element of the capacity matrix is equal to 1
and that is why this component will be omitted in the transition defi nition below.
Besides all the transitions are disjunctive and that is
why this component will be omitted too. The formal description of the GN transitions of aring in mind that this ON is reduced) is Z 1
=
Fig. 3. 3
(be
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
143
Fig. 3. 3 where 1 is the input place for Z and 1 is the output place for 1 1 2
Z, 1
the transition condition is 1 2 r 1
= 1 I II
W 1
where W
= "in the
GN
does not exist any other token with the same initial
1 characteristic".
144-
CHAPTER 3
Z = <(1 Z <(1 ,, 1 ), ), (1 (1 ,, 1 ), ), rr >, >, 2 22 10 10 33 11 11 2 where 1 33 rr
-- 11 22 22
II W W 11 22
1 11 11 W W 3 3
1 II W W 10 II 4 10
false false
where = "d = 0" & "T i X
W 2
= "3 = 1" & "T i X
W
+ t ",
a
3
a
1 + t ", 1
where d is an indication for branching at the transition Z . 2 W = "T i X + t ".
3
a
4 Z 3
=
where the r ' s p r e d i c a t e s c o n t a i n the e s s e n t i a l of t h e channel 3 type CSMA./ CD. Z = <[1, 4 4
I, 5
1 , 1 , 1 I, 6 7 8
|1, 9
where
r 4
1 9
1
1 I 4 |
W 4
W 5
1 I 5 | = | 1 I 6|
W 4
W 5
W 4
W 5
1 I 7 |
W 4
W 5
1 I 8 1 8 1
W 4 4
W 5 5
where
w 4
= "a - o" & "T i x a
+t
+ t ", 2
3
15
1 15
), r >, 4
access
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
= »S - ln
W
& "T i X
5
+ t a
2
+ t ", 3
where 0 is an indication for branching at the transition Z . 4 Z = <(1 ), (1 ], r >, 5 9 10 5 where 1 10 r
= 1 I true ; 5 9 5 9I I Z =
r = 1 I 6 11 I
1 12 12 W 6
where W
= "in 1 6
there are no tokens with the same characteristic m"; 12 Z
= <(1 7 7
), (1 12 12
13 13
), r >, 7 7
where
r = 1 I 7 12 I
1 13 13 W 7
where = "T i x
w 7
a
+ t ": 4 Z
= <(1 8 8
13 13
i, (1 ], r >, 14 8 14 8
where 1 14 r = 1 I 8 13 8 13 I I where
W 8 8
14-5
146
W
CHAPTER 3
= "T i t +t + X ". z fl 1 15 5 a Z = <(1
where 1 16 r = 1 I A 15 a is Ii
W . A a
The described model shows the basic points of the structure and the function of an automated system for booking and sale of tickets in the passenger transport. This model can be used for two principal aims: 1. When the main parameters of the system
are specified
productivity of the MCC, loading of the system, of the channels,
by this model the following
such as
transmission capacity characteristics
of the
system can be defined: a) average time for processing of a request; b) maximum time for processing of a request; c) the part of processed requests for a fixed period of time; d) the time for processing the requests in
every
element of the sys
tem; 2. When the loading of the system is specified,
the productivity
of the elements can be defined so that the average time for processing of a request does not overrun a certain value. 3. The number of the processed requests for
a defined
period of
time and others. The suggested model can be developed, if necessary, for getting extra information about the simulated processes. This paper is based on [2]. REFERENCES; [1] S. Nedev,
K. Atanassov,
An automated
booking system modeling by
generalized networks, Automation, Computer Engineering and Automa-
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
14-7
tion Systems, 1989, No, 6, 17-20 (in Bulgarian). [2] S. Nedev,
L. Atanassova,
zervation systems for
Generalized nets for simulation of re-
passenger transport, First
the Math. Found. AI Seminar,
Sci. Session of
Sofia, Oct. 10, 1989,
MFAIS-7-89, Sofia, 1989, 45-49.
Preprint
IM-
148
CHAPTER 3
§ 3. 3: GENERALIZED NETS IN MODELLING OF PUBLIC TRANSPORT Katja Stefanova
In this section,
public transport is modelled by means of the
Generalized Nets (GNs). "A transport system"
means
the set consisting
of a transport
net in a given urbanized region, all vehicles and passengers, her
active or passive
and ot
participants in the public traffic of the same
region, A transport net is characterized by
its geometry, topology and
organization. Its aim is to realise vehicle
or pedestrian
connection
between al1 points which are to be connected in a given region. We shall (buses,
describe transport nets
trams,
public
transport
trolley-buses etc.) only. Such a transport net can be
characterized by
a directed graph
stations. Each pair of vertices a given route
involving
G
whose vertices
represent
the
representing neighbouring stations of
is connected with
arcs directed according to the traf
fic. In the sequel we shall use the GN
for transport systems model
ling. Let V = |v, v 1 2 be the set of vertices, and
v ] P
A = |a, a 1 2
a ) q
be the set of arcs of the graph G. To each vertex
v € V
and its cor
responding input and output arcs a', a', . . , a'; a", a", . . . , a" 1 2 i 1 2 j 1 i H
p, l i j i q
where
(see Fig. 3.4), we shall assign the transitions
2'and Z" from the GN (see Fig. 3.5). In view pairs
of the results of
§ 5. 1 in App. l,
of transitions corresponding to
the vertices
the union of
all
of the transport
SCME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
Fig. 3. 4-
Fig. 3. 5
149
150
CHAPTER 3
net represents a GN. This GN contains transitions of the two types gi ven
above and two kinds of tokens corresponding
the passenger stream
to the vehicles
and
The tokens of the first kind will run over pla
ces of the type 1', 1" and 1 and the second kind will run over places 2 of types 1"', 1"", 1 , 1 , 1 . i 2 3 The GN will also contain global time parameters such as an inio T, an elementary time-step t <e. g. a minute, for "ro the time step may be larger), duration of the model func-
tial moment ugh" models tioning t
(hours up to twenty-four hours).
of the GN has hicles,
the following
The first kind of tokens
initial characteristics: type of the ve
registration numbers, numbers of the line.
kind of token,
the number of
the passengers
As for the second
boarding
a vehicle
is
its initial characteristic. The tokens of the first kind are in the GN at the moment T,
whereas the tokens of the second
kind are in the GN
during its functioning through the places of type 1 . In
the traffic,
4 the tokens of the first kind can be characterized necessary
also
by
"the time
for a transition from an ex-station to the current station"
(in places of type
1J)
as well as by
"the sojourn time in
the last
station" (in a place of type 1"). The tokens of the second kind can be characterized by
"the number of passengers
vehicle" (in a place of type 1"") and
who are currently
in the
"the number of passengers
get
ting off the vehicle" (in a place of type 1 ). The tokens of the se1 cond kind do not possess any characteristics in a place of type 1 . 3
A transition of type Z' has the following form: Z'
=
1' 2
v(A(i', i"), 1 1
1', i A(r, 2
1", 1
1" 2
i») 2
1"], i A (
r, i
11 , 1 , 1 2 !">)>, i
1 i, 3
»,
x,
r\
x,
SOME APPLICATIONS OF <2>!s IN ECONOMICS, INDUSTRY AND TRANSPORT
151
where a) 1', I',..., l' 1 2
and
i
1", 1",..., 1" 1 2
correspond to the arcs between
i
the vertices of the graph which are connected with the vertex v; b) 1 is related to passengers getting off the vehicle (the tokens re1 ceive in it the numrber of these passengers); c) 1 is related to the presence of the vehicle; 2 d) 1 is related to the presence of passengers in the vehicle (the to3 Kens receive in it the number of these passengers); 1 1
r'
=
1 2
1
1' | false 1l
true
false
1' I false iI I 1" | W 1 1 1
true
false
1" I j I
W
false
3
W 2
false
W
1
2
where - "there is at least one passenger getting off the vehicle";
W 1 W
= "there is at least one passenger continuing the trip". 2 A transitions of type Z" has the following form;
Z" = <[1 , 1 , 1 ), fl"\ 1"' 2 3 4 1 2 x, A(1 , v(l , 1 )>>, 2 3 4 where
1"', 1"". 1"" i 1 2
1""!, *. *, r", i
152
CHAPTER 3
a) place 1 is related to passengers boarding the vehicle 4
(the tokens
receive in it the number of these passengers); e) the transition condition is
I I r" =
1"' 1
...
1"' j
1"" 1
...
1"" J
w
...
W
false
...
false
2 1 3 I 1 I false 3 1
3
1 | false 4 |
...
false
W 3
...
W 3
...
false
W 3
...
W 3
■where W
= "the next route station of the vehicle". 3 All transitions have equal priorities. The GN described above is a reduced GN of the type
A,A,A,Tl,e,9,b 3 4 6 A 1 2 T. We conclude that within the framework of the model we can
spe
cify the following main characteristics: priority of the stations (for the places of type
1 ),
capacity of the places of type 1
2
(number of
2
vehicles stopping at a given station) and the priority of the vehicles (tokens of the first kind). The fornullation of an analytical model of the considered tran sport system, particularly the analytical
part of the given GN model,
is related to the influence of factors of various nature. Some of them are:
settings and movements of the population,
presence of different
social and age groups with the appropriate requirements
for security,
comfort, costs, and travel time. These factors affect the determinati on of the initial
parameters of the second kind of tokens
passengers waiting for a vehicle at a given station,
(number of
with preferences
regarding the type of vehicles, current number of passengers, etc. ).
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
153
Different characteristics of the transport system can be obtai ned
by simulating the model built by means of a (34.
It makes it pos
sible to use the GN model as a tool for transport process
management.
Furthermore, it enables optimization of the vehicle routes the
traffic process and passenger streams.
simulating
It is easy to obtain
the
average and maximum time duration of a given route. In the case of any changes, the time schedule and the correspondingly adjusted streams
can be fixed.
On the basis of the obtained results,
passenger strati
fication of the transport net and appropriate estimation of the travel costs can be determined
154
CHAPTER 3
§ 3. 4: GENERALIZED HEX MODEL OF ACTIVITIES OF A BUS-DEPOT Favlin Gyurov
An example for using tion
toe Generalized Hets (GHs) for a descrip
of some transport situations is introduced.
It illustrates
the
solving of the following problem Suppose a public coach agency uses seats are Known,
n
coaches whose numbers of
when the coaches are not in use, they are in a gara
ge. There is a service for maintenance work. The coaches leave for the routes they
from a bus station.
take their own daily
Every morning, before leaving the garage, time table.
At the end of
their workday,
they go back to the garage. This process
will be described below by means of a GH and
the
loading of the lines and coaches will be reported. The GH has a graphical structure as shown in Fig. 3.6. Every coach is represented by
a token in the GH.
when a coach
is on a route,
the token corresponding to it goes into the GH. At the
time -moment T,
all coaches are
bus-tokens BT i
in the garage,
so at that moment all
(1 <, i i n) are in place 1 . If any coach is out of or3
der, it has to be
repaired so
the corresponding token goes
to place
1 through place 1 and the transitions Z and Z . when the coach is 13 15 2 3 in
order again the corresponding token will
come back
to place
1 3
through place in
1 and 1
the transitions Z
and Z . If any coach located 4 1
the garage is in order and it is time
to go on a route,
the cor
responding token leaves place 1 and goes to place 1 through place 1 3 7 5 and the transitions Z 7
and Z . 5
If there is much time before
the next
travel of the coach, it goes to the garage and the corresponding token
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
155
Fig. 3. 6
goes to place 1 3
through place 1 and the transitions Z and Z 2 6 1
The
situation is the same when its travel is completed. However,
if a coach
is damaged and has to go on the route, or
it gets damaged while travelling,
it cannot do the next planned
tra
vel. So it has to go for servicing. Then the corresponding token
goes
to place 1 that
But the travelling will not be done. So it is necessary
another coach
takes up the next travel.
This is why there is a
156
CHAPTER 3
private token in the GN. At the time-moment T it is in place 1 . This 12 token (which we will mark below by OT) gets as characteristic the next travel which the damaged
coach would to do. Then it goes to place 1 . 4
In this case, there is a choice for an optimal coach from those in the garage. The corresponding token goes to place 1 and the goes to place 5 1 16
The bus-token goes to place 1 7
through place 1 8
and
and the transitions Z
that can accept the order,
5
OT comes back to place
and Z . 8
OT waits in place
If there is no coach
1 . It is possible that 4
at that time-moment another coach come out of order. the transition Z 1 9
9
1 12
and its copy comes back in place
So OT 1
12
splits in
through place
and the transition Z . It is useful if the transitions 8
Z
8
and
Z 9
work more often than the others. Bus-tokens have only one (current) characteristic with the form: "<seats, maintenance, travels>", where seats is a number of sites, maintenance =
it gets a
new characteristic whose seats component is the number of sites of the damaged coach, its waiting-time component travels
component is
is the previous one and the
the set of travels which
when OT comes into place 1 , 9
this coach
would do.
it doesn't change its characteristic. In
SCME APPLICATIONS OF GNs IN ECONOMICS , INDUSTRY AND TRANSPORT
place 1 it gets new characteristics - the previous seats 16 components
157
and travel
and the waiting-time characteristic is the duration of
is
stay in place 1 . 4 Every bus-token gets as
a first characteristic
its
number of
seats and its daily time table. The maintenance component of its first characteristic is
<0, 0>.
When it goes to place
characteristic. Travel component reduces to
1 it changes 13
its
the set of travel fulfil
led. The current characteristic of the token includes information aboit gets
ut it. When a token comes in place 1
a characteristic - the
1 previous last
of
The nunfcer of seats
travel component.
can be changed
the repair component, and the duration
the repairs join
The of
these repairs adds to the current durat ion of maintenance. When
a token comes into place
1 it changes 14
the passengers
(TIME) characteristic. A bus-token that comes into place 1 can append 3 new travels to the travel component of its characteristic. Z
1 1 =
where 1 3 r
= 1
1 l 11 l 1 l 21
true true
1 3 M
= 1
1 I 11 1 1 1 21
n
' n
15&
CHAPTER 3
2 2 Z = <(1 , 1 1, 11 , 1 , 1 ), t , t , r , M , v(l , A(1 , 1 ))>, 2 3 4 15 5 16 1 2 2 2 3 3 4 where
r = 2
1 3 1 4
1 15
1 5
1 16
I W I 1 I I false I
W 2
false
false
W 3
where W
= "the coach is out of order"; 1
W 2
= iW & ("the moment for going to the station has come" or 1 is an OT
in 1 " 4
and
("the coach has no other travel"
("there or
"the
travel of OT can be done before the travel of the bus-token") and "this coach is optimal about the number of seats"); W
= "the order is taken up by the bus-token"; 3 1 15 1 M
= 2
Z
= <(1 3
I n 3 | I 1 1 0 4 I
1 5 n 1
3 3 ,1 ,1 !, (1 ), t , t , r , M , v(l , 1 , 1 )>, 15 10 11 13 1 2 3 3 15 10 11
where 1 13 r r
3 3
= =
1 I 15 15 I I I I 1 I 10 |
true
1 I 111 1 I
true
true
SCME APPLICATIONS OF GNs IN ECONOMICS,
INDUSTRY AND TRANSPORT
1 13 1 M = 3
I 15 I I 1 I 10 I
n
1
n
n
I
11 I Z = <[1 ), |, 4 13
4 4 [1 ]!,, t , t , r , , M ,, v ( l >>, 1 1 2 4 4 13
where 1 r
=
1
4
| 13 I
1 true 1 13 M = 4. 4 Z = <[1 , 1 , 1 ), 5 5 6 16
1
I i13 n Ii
n
5 5 (1 , 1 ), t , t , r , M , v ( l , 1 , 1 )>, 7 8 1 2 5 5 5 6 16
where
1I r
= 5
| 5 I I 1 | 6 I 1 | 16 I
1 M = M 5
5 1 66
1
I 1 I I II
1 16 16 II
0
1 7
1
true
false
true
false
false
true
1 7
I
n
0
n
0
8
a
1
159
160
CHAPTER 3
6 6 Z : = c|) (|) I, II !1 , 1 , 1 ), t , t , r , M , v(l )>, 6 7 10 2 11 1 2 6 6 7 where
r = 6 6
1 7 7
I 1 1
1 10 10
1 14 14
1 2 2
W
W
W
1 1
4 4
5 5
where = iV
W 4 W 5
& "it is time corresponding to the token coach to travel", 1
= iW & 1
("there is nuch time before the next travel" ach's work day is finished" ),
M
: 6
-
Z 7
1 | 7 I
1 10
1 14
1 2
n
n
n ;
7 7 ), fl , 1 ), t , t , r , M , v(l )>, 14 11 6 1 2 7 7 14
where
r
= 7 7
1 14 14
l I I I
1 11 11
1 6 6
W 1 1
w W 6 6
1 11
1 6
n
n
where W 6 6
- nW & "it is the end of travel", 1 1
M
= 7
1
I 14 I
;
Z = <[1 <[1 ,, 11 ]], , fl fl ), ), tt ,, tt ,, rr ,, M M ,, vv(l ( l ,, 11 )>, )>, 12 8 88 88 9 9 12 11 22 88 88 8 9 9 where
or
"the co-
SCME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
161
1 12 1 r
=
a
a 1 9
I I i I I
true true
1 12 M, M, = = a a
1 I 8 I a i i i 1 I l i 9 I 9 I
1
l1
9 9 Z ---
=
1
9 9
l 12 I 12 I
1 9
W
true 7 7
where W
= "there is a coach which is damaged"
&
"this coach
7 vels",
M
= 9
1
l 12 I
1 2
1
1
1
9 .
For t h e GN-components t o be v a l i d , i A
|Z | = i f o r i
n(l) L 4 12
S 9,
= n ( l ) = l; for all other places: TT (1) = 2, 9 L
) - 2;
c(l
I $i
for all other places: c(l) - n,
has to tra-
162
CHAPTER 3
i t
o + 2. t , if i £ (1, 2, 3, 5, 6, 7)
1 i i 6 (t ) ) = s 1 1
i o t + C. t , if i = 4 (C = const 2* 2), 2), 1 i t
o + t , if i € [», Eft, 9)
1 It means that the private tokens go faster than others.
In or
der to maintain the works character it is possible that the transition Z works rarely. 4 i i i o © (t ) = t = t (1 <, i i 9). 9>. 2 i1 2 2 i The numbers t (1 i i £ 9) must be large enough so that all to2 kens can go through transitions K = (BT / 1 i i <, n| tl iOT), i IT (a) K
- 1, for every a € K,
0 (a) = T, for every a 6 K. K All tokens enter the GN at the starting time-moment of its work. T
is the time-moment of the start of workday,
o t = 0. 5 minutes
M
example); T + t
is the time-moment of end of a workday.
(for
SCHE APPLICATIQHS CF GNs IH ECCHCHICS, INDUSTRY AND TRANSPORT
163
§ 3. 5: GEHERALIZED HET MODEL CF THE ACTIVITY CF AN AIR-TRAFFIC CCKTROL CENTER Stanislav Fistroanov
As an example of the description of real processes with Genera lized Nets
(GHs) we snail examine a GN-model of
air-traffic control center consists of area fixes
the following:
for landing,
the
activity of an
In general,
this activity
when an airplane is coming to the airport
it comes under
the control of
a dispatcher,
who
an air-corridor where the airplane can fly a round until a run
way is free. airplanes area,
at an airport.
At the sane time,
which have
there must also
just flown
off until
we shall examine an airport with
n
they tracks
can be used for both landing and for flying off), fic is being North (1),
controlled
be air-corridors for leave
the airport
(each one of which where the air-traf
by 4 dispatchers - one for each direction:
west (2), South (3) and East (4). All tracks are used by
all dispatchers at the same time. Every dispatcher can control up to m airplanes at any one moment, while in special situations for short in tervals of time this number can be increased to m + k, where fixed natural number, i.e. airplanes
waiting for
k
is a
the normal capacity of the airport is
landing at any one moment,
when
4.m
the number of
waiting airplanes is less than 4. m, this means that there are free aircorridors. from
We shall
the same way
corridor
examine the airplanes
which have just flown
as those waiting for landing - each takes
from the moment of flying off till the moment of
off
one air-
leaving the
airport area. Let us examine a GH with 7 transitions and 23 places for move ment of airplanes,
and a separate transition Z with 0
tokens moving, carrying managing information
2 places,
whit
(see Fig. 3. 7).
The separate transition Z has two input places 1 and 1 which 0 0 1
164
CHAPTER 3
are also
the output places
for it. In each of them, there is a token
which passes the transition; when it is activated, it renews its tran sition and goes back to its starting place, i. e.
the index
matrix of
the transition Z is 0 1 0
1 1
1 l true 0 |
false
1 | false 1l
true
There is only one token
a with the current characteristic (and
a without the other ones) x
= "", 1 2 n
where
a
is the
number of current free air-corridors of the whole airport at the mo ment, a , a , . . , , a € 1 0 , 1), a = 0 means that the i-th track is oc1 2 n i cupied at the moment, and a = 1 means the opposite situation (1 <, i i i n). In the place 1 there is only one token - p, which has the cha1 P racteristic x - "", where b is a natural number for 1 2 3 4i which
-k <, b i
i m, and b i
is the number of
free air-corridors under
the control of the i-th dispatcher (1 s i $ 4|. The set S with elements s is the set of tokens which represent j the airplanes in the airport area. set (1,2
Their number is an element
4. m) and they are transferred to the other section of the
GN. Each of these tokens has as an initial characteristic its fying data 1 4 re:
of the
(type,
identi
nationality, number of course, etc. ). In the place
j j j receives as new characteristic "
(1 i i £ 4. m), whe-
SCME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT j - c e (1, 2, 3, 4)
determines the direction from which the airplane
comes; for airplanes which determines
have just flown off, this characteristic
the direction to which the airplane
inside the ®*,
165
for airplanes
waiting to land,
will be travelling; this characteristic
carries the appropriate information; J - c is a parameter corresponding to the time during which the airpla2 ne can fly with its fuel;
for the airplanes just taking off
it al
ways has the maximum value;
J - c € {0, I) and it is equal to 0 for an airplane waiting to land and 3 to 1 for an airplane just taking off. Token s (1 <, j £ 4. m) can enter the GN through three input j places 1 , 1 and 1 . Every token entering through 1 1 2 16 2
corresponds to
the airplane, coming to the airport
If the airport
capacity is not exceeding
(i.e.
area for landing.
a pr x i- 0), 1
then the token moves to
place 1 (where it receives the above characteristic) and (simultaneo4 a usly the token a receives as a new characteristic such one with pr x 1 decreased by i
(the airplane is accepted in the airport).
Otherwise,
the token moves to place 1 , which is an output place for tokens
cor-
3 responding to airplanes (the airplane is directed to another airport). The transition condition of Z
is 1
r
= 1
where
1 3 1 I W 2 1 1
1 4 W 2
166
W
CHAPTER 3
a = "pr x * 0", 2 1
y/
--
1
-iw .
2 A token
s j
which has entered the GN through the second
input
J
place (1 ), corresponds to an airplane just taking off 15
and has
c = 3
1. To move through the transition Z , it is necessary to satisfy the 3 a a condition "pr x * <0, 0, . . . , 0>" & "pr x * 0" (i.e. at 2, 3, . . . , n+1 1 least
one track mist b e free a n d also at least o n e air-corridor for
leaving the airport mist b e free).
B e s i d e s a token leaving place
1 15
can m o v e only to place row
1 16
f o r assignment to a dispatcher,
of the index m a t r i x of the transition Z
corresponding
i. e. the t o place
3 1
contains the value
"false" in every position, so does t h e c o l u m
15 corresponding t o place
1 , w i t h only 16
o n e exception:
the predicate
a related to the places 1 and 1 is "pr x i- <0, 0 0>" 15 16 2, 3, . . . ,n+l a & "pr x i- 0". When a token moves from place 1 to place 1 , the to1 15 16
a ken a receives a new characteristic: pr x decreases by 1. 1 Itie transition Z
stands for the distribution of
airplanes to
2 the dispatcher.
A token s moving through this transition, regardless J
of the place
where it is coming from, goes to place 1 (the corres4+i j ponding airplane is distributed to the i-th dispatcher), where i = c . 1 The transition condition of Z
is 2
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
11 55 11 33
rr
= = 22
11
11 66
II W W 11 33
11 77
167
8
W W 44
W W 55
W W 6
11 II W W 17 II 17 33
W W 44
W W 55
W W 66
11 II W W 18 II 33 18 II 1 1 II W W 19 I 19 I 33
W W 44
W W 55
W W
W W 44
W W 55
W W 66
1 1 II W W 220 0 II 33
W W 44
W W 55
W W 66
11 II W W 16 II 16 33
W W 44
W W 55
W W 66
66
where W
j = "c = i" for 1 <. i i 4. 2+i 1 Transition Z 3
1 . 14
contains two output (of the GN) places - 1 and 13
Tokens, moving to place 1 14
correspond
to airplanes Just flown
off, which are leaving the airport area under the dispatcher's guidan ce. A toKen s can move to place 1 , if it cones from place 1 for j 14 4+i j a 1 i i £ 4, and if c = 1 . Then pr x 2 1 Tokens
and
p pr x i
moving to the other output place 1 13
each increases by 1.
correspond to airplanes,
which are landing at the moment. A token s can move to place 1 if j 13 j « if it has c = 0 and "pr x ^ <0, 0, . . . , 0>" (at least one 3 2, 3, ...,n+l a p track is free). After this, pr x and pr x each increase by 1, where 1 i i
d e p e n d s o n t h e G N - p l a c e w h e r e t h e t o k e n comes /' j
- 4,
f or 5 i j
\ j
- 20,
f o r 21 S j
<. 6 i
24
from
168
CHAPTER 3
1 and 7
1 8
1 and 23
respectively, with the exception that in places
1 24
1 , 1 , 21 22
tokens corresponding to airplanes just taking off cannot
j appear. If the equation c = 0 (airplane waiting for landing) is valid 3 for a token s , it j (1 i x i 4), and
comes from place 1 (i £ I i 4) or place 1 4+i 20+i cc pr x - <0, 0, . . . , 0> at this moment, 2, 3, . . . ,n+l
moves to place 1 (the corresponding airplane 8+i corridor for waiting for a track to be free, der
the control of the same dispatcher).
that the transition condition r
receives
a free air-
the airplane staying un
From the above,
it follows
has the form 3
1 99
= 3
10 10
1 5
I 1
W 7 7
false
1
I false false 1
W
6
r
1
1 11 11 false
12 12
1
1 13 13
false
W
false false
W 7
1
false
false
| false I I 1 I W 21 I 7 8
I false
22 I
false false
W
false W 7
W
W 8
W 9 W
false
8
9
false
W 8
W 9
false
W
false
false
W
W
false
false
W 7
false
1 I false 24 I
false
false
W 7
1
false
false
8
9
W 8
W 9
W
W 8
false
false
false false
9 false
W 10
where a n+1
false
false
1 I false 23 I
x
false 9
false
7
I false 15 I
false
8
W
16 16
9
W
7 false
W = "pr 7 2, 3
1 14 14
8
1 I false 7 |
1
1
j = <0, 0, . . . , 0>" & "c
= 0", 2
it
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
W
a x * <0, 0
= "pr 8
169
j 0>" & "c
= O",
2, 3, . . . , n+1
2
j
W
= "c 9
=1", 2
W
a x * <0, 0
= "pr 10
a 0>" & "pr x * O".
2, 3, . . . , n+1 The transitions
1 (1 & j £ 4)
Z
are completely analogous
to
3+j one another. The
j-th of them corresponds to the functioning of the j -
th dispatcher. Because the airplanes come in from the j-th or fly
Then under his control and at some b
direction,
out in the j-th direction, the j-th dispatcher certainly gets.
= pr x
j
< 0).
moment he can be overworked
(then
That is why when a token moves from place 1
J
(the
a+j
corresponding airplane is waiting for landing under the control of the j-th
dispatcher)
through the transition Z
transition condition,
if b
£0
. After a check 3+j
of the
the token moves to place 1
j
(the
20+j
corresponding airplane continues waiting under the control of the same dispatcher),
else the token moves to place 1 and receives as new 16+j j characteristic the last one but with c = s, where b = max b (the 2 s liti* t corresponding airline will be transferred to the control of a new dis patcher
who at the moment
control
the smallest number of airlines).
(1 i j i 4) is
The transition condition of Z 3+j
1 16+j r
= 3+j
where P W
= "b
11
j
: pr x
j
< 0",
1 fl+j
I W I 11
1 20+j W 12
170
CHAPTER 3
Fig.
3. 7
SCME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
W 12
171
= nW . 11 All the transitions and places in
priorities;
the tokens in places 1 , 1 2
the described 1
3
are
GN have equal ordered
as to
24
their second current characteristics in ascending order. The GN starts functioning at fixed moment
a fixed time-moment
T
and it functions until another
T+t . It works tact by tact with an elementary time-step
o t (=15 sec. , e. g. ). All transitions are activated at every tact. All arcs have infinite capacities. All transition types are disjunctive.
m
CHAPTER 3 § 3. 6: GENERALIZED HET MODELS PQR FLEXIBLE HAHUFACTURIBG SYSTEMS Maria Stefanova-Pavlova
§ 3. 6. 1
and Krassimir Atanassov
Introduction
Flexible Manufacturing Systems (FMSs) [1-58] are an attempt to reconcile the efficiency of the production line with
the flexibility
of the job shop in order to satisfy a versatile demand at low cost. Generally two Kinds of flexibility are distinguished. The longterm flexibility corresponds to the possibility product group in the manufacturing
of introducing a new
systems during
its operation and
with little effort. The short-term flexibility corresponds to the pos sibility of handling a large variety of product groups at a given ti me in the manufacturing
system [7]. In order
to meet these require
ments, an FMS is made up of (a) a set of flexible machines, (b) an automatic transport system, (c) a sophisticated decision-making system to decide at each instant what has to be done next and on which machine. Flexible machines have
the capability of performing
operations. They have an automatic tool storage with
various
a retrieval sys
tem and the machine programs can be downloaded. A n automatic transport system is required to transport the parts for the next operation. The decision making system organizes the production schedules and synchro nizes machine
utilization. This system is very important
for solving
problems such as idle time minimization and manufacturing without pe ople, which appear in organizing the work of FMSs. A FMS is structured cell
into
is an elementary system
some local storage facilities
flexible manufacturing
cells.
Each
consisting of a flexible machine tool, for tools and handling devices
such as
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
robots in order to transfer parts and
tools
173
between the cell and the
global transport system [7]. The FMSs are a class of automated manufacturing systems. An EMS comprises processing modules or machines linked by a material handling system, all under central computer control (cf e.g. [3]). The flexibi lity of such a
system is therefore dependent on its components' capa
bilities and their
interconnections as well as on the mode, of opera
tion and control. Some of the basic properties of the EMSs, related to the models below are (a) the possibility to work autonomously without any human interventi on; (b) automatic realization of basic and supplementary operations; (c) elimination of production line idle time; (d) maximum possibility for a complete preparation of a worKpiece on a single machine; (e) high economic efficiency; and others (cf e. g. [4]). The rapidly expanding area of FMSs requires new methods for mo delling,
simulation,
control and optimization,
as classical methods
are not always adequate. From the mid 80s, ( the earliest papers are probably [3-5]) the Petri nets (PNs) which have established themselves as a convenient me ans for FMS modelling. First of all, the FAB is a discrete system. Any modelling has to be based
on the concepts
An event corresponds to a state change. are associated with transitions.
of events
and activities.
when using Petri nets, events
Activities are
firing of transitions and/or marking of places.
associated with
the
Petri nets are appro
priate for representing parallelism and synchronisation in an enginee ring environment. They are very useful at the global coordination
le
vel where real-time decision making is required to execute the planned and scheduled production in the manufacturing shop.
The real-time de-
174-
CHAPTER 3
cisions of the coordination level also have to be consistent with manufacturing shop. of
The Petri net
rule-based system
where
the
can be seen as a control structure
the rules operate
on object
attributes
(parts, machines, tools) represented by tokens. The Petri net utiliza tion at the global coordination level increases the security and reli ability of the FMS Control. The application of Petri net theory to the FMS
is a very rich
domain. Let us now briefly sum up some advantages of this approach: (a) partial order among events
can be described
to meet
flexibility
requirements, (b) states and events are represented explicitly, (c) a unique family of tools is used throughout the specification, mo delling, qualitative validation and performance evaluation, imple mentation and the operational processes, (d) the same
family of
tools is used within
various functions of an
FMS and at various levels (scheduling, global coordination and lo cal control). It represents an important
aid for
integrating the
whole system, (e) an accurate,
formal description of the synchronization
mechanism
is provided, which is something essential to achieve security [7], (f) the graphic nature of Petri nets, with state graphs,
their conciseness in comparison
the possibility
of
analysing
them , and
the
existence of techniques for a direct implementation, (g) the
Petri net permits
direct realization of the modelled
system
from the analyzed net [8], One outstanding aspect ing
analysed to see
net
must satisfy
of Petri nets is their capacity for be
if the constructed model is valid.
a set
of properties which are characteristic
"good model". Some of the properties are: tokens which any
For this the of a
boundedness - the number of
place in the net can accumulate during its evolution
is bounded; liveness - this does not only mean that the model is dead-
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
lock-free, exist
but
also that from each state reachable by the net
175
there
evolutions which can lead to the firing of any transition;
versability - if
re-
this is verified it means that it is always possible
(from any reachable state)
to return the system to its initial state.
For analyzing the properties of the Petri nets various techniques have been developed (a) Reachability
graph
analysis - this consists of analyzing the net
on the basis of the graph of the reachable marking from the initi al marking. This method allows a complete analysis of the net. (b) Reduction analysis - this analysis consists of applying to the net a set of rules which, while preserving the properties that are be analyzed,
gives a smaller model.
to
This method is complementary
to the previous method By first applying this method, later "rea chability analysis" is made easier. (c) Structural analysis - this method allows the analysis of many pro perties of Petri nets, independently of its initial marking. The main disadvantage of Petri nets arises from the size of the nets modelling very complex systems, such as FMSs. Another disadvantage is the fact functioning forces a
that a change in the system's
substantial change in the net structure.
Another suitable means for an FMS modelling are
coloured Petri
nets. They consist of places and transitions. A place can be marked by the elements
of a subset of colours
(none,
one or
various items of
each colour). The tokens of a place will be called coloured tokens
or
sinply, colours. The marking of a place can be represented by a formal sum of colours.
The resulting models can
be more concise than
obtained from Petri nets and the functions appearing
on the
those
net arcs
have a clear physical meaning. The main disadvantage of coloured Petri nets is the fact realization are
that techniques for model construction, analysis and not yet fully developed This is a result of this to-
176
CHAPTER 3
olJs relative newness. *
3. 6. 2
First
K
GN-model of fiMSs
In the FN-models of FAJSs the processes
constructed so far,
the dynamics
of
in FMSs is described and the data for main time parame
ters of the modelled processes
are investigated (cf. [5]). Analogous
results can be obtained by GNs. On the other hand, the constructed ge neralized net models (QN-models) of FMSs can be used as tools for con trol of the processes, me.
which function sufficiently slowly in real-ti
GN-models also give a possibility of optimization of the control
led process
(we shall discuss this problem in
a further communicati
on). Here we shall discuss a process, as described by PNs in [5]. The problem is to develop general understanding of which
the way
the performance of an FMS is affected by disturbances and
out how flexibility can attenuate these affects.
A GN,
in find
modelling the
same process is shown on Fig. 3. 8. It contains the transitions Z , Z , 1 2 Z , Z , having the following described components. Each of these tran3 4 sitions is activated on
every elementary
the functioning of the GN-model). arcs and of the places
The
are infinite.
time step
(this simplifies
capacities of the transition's The transition
Z
is activated 1
when there is at least one token in the places 1 , 1 , 1 and 1 and 1 3 12 13 simultaneously there is at least one token in place 1 . The transition 6 Z 2
is activated
the places 1 and 2
when there is at least one token in at least one 1 and simultaneously 14
of
there is at least one token
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
Fig. 3. fl
177
178
CHAPTER 3
in at least one of the places 1 and 1 5 9 The transitions Z
and Z 3
are activated -when there is at least 4
one token in at least one of their input places. The priorities of the places and transitions are 0 (i.e. they have no priorities). The tran sition conditions are 1 2
r
1
1
3
4
5
1 1
I W 1 1
W 2
W 3
false
1
I W 1 1
W 2
false
false
= 1
1
3
1 I 12 I
W 1
W 2
W 3
false
1 I 13 I
W
W 2
W 3
false
false
false
W 4
1 8
1
I false I
mtoere - "In place
W 1
1
there exists a token with
a characteristic which
8 corresponds to
the current identifier
of an available
machine
for the token (workpiece)"; W 2 W
= nW ; 1 = "In places 1 , 1 and 1
3
5 tic
8
there are no tokens with a cbaracteris-
9
which correspond to the current
identifier of an available
machine for the token (workpiece)"; - "There exists a workpiece which mist be machined";
W 4
SC&E APPLICATIONS OF GNs IN ECONOMICS,
1 6 6
1
1 II W W 22 || 55 rr s= 2
11 || 14 14 II
11 77
w w 7
INDUSTRY AND TRANSPORT
11 88
10 10
W W 6 6
ffalse alse
ffalse alse
false false
ffalse alse
ffalse alse
ffalse alse
false false
1 II ffalse alse 55 | |
ffalse alse
1 9
ffalse alse
W W
W W 66
W W 8 8
W W
W W 66
W W 88
55
II ffalse alse 11
11 99
55
179
where W
= "The machine which is machining the workpiece is at worK"; 5
= ~m ;
W 6
5
i. e. W
= "The machine breaks down"; 6
W
- "The machine
which is machining the workpiece
is at
work, or
7 there exists another machine
(a token in place 1 ), which can 9
machine the workpiece"; W
= "The machine is shelved"; 8 1 11 11 rr
= 3
11 II WW 66 11 99
1 12 12 WW 10
where W
= "The list of the processing machines is exhausted";
9 W 10
= nW ; 9 1 13 13 rr = 44
11 II WW 77 II 11 11
1 14 WW 12 12
where W
= "The repair duration takes a lot of time"; 11
180
W 12
CHAPTER 3
= 1W 11 In the transfer to the different places, the token
assumes the
following characteristics (the fact that the characteristic is not gi ven is reflected by the absence from the following list): $ -> "an identifier of the available machine "; 2 $ -> "the worKpiece cannot be machined"; 4 § -> "general time for processing of the token in the last token 5 transfer if the token was in place 1 ; general time for repa8 ir if the token was in place 1 "; 9 $ -> "general idle time, if the token was in place 1 "; 6 14 $
-> "the machine is not able to function"; 10
$
-> "general time for processing of the workpiece"; 11
$
-> "time for processing with the corresponding available machine". 12 The GN starts functioning at the time moment
T
according to a
* given time-scale and continues At the
t
elementary
time-steps.
start of the simulation process in place 1 there exist 8
tokens with the following characteristics; "identifier of processing machine; priority of the processing machi ne; capacity of the processing machine, etc. defined by
the GN-mo-
del' s user". At different time moments defined by a law (from the user), in the input place 1 tokens enter with the initial characteristics: 1 "identifier of the workpiece;
identifier's priority;
moment of its
arrival, a list of processing machines necessary for it; and others defined by the GN-model's user".
S(*E APPLICATIONS OF GNs IN ECONOMICS,
The token's
transfer in the GN
INDUSTRY AND TRANSPORT
181
depends on the truth-values of
the predicates of the conditions given above. As a result of the tran sfer, the tokens from both types receive, by the above functions, the following one of the GN-places,
characteristic
characteristics: idle time duration in every
general idle time
duration,
general time
for
processing in the GN etc. The tokens of the first type receive as cha racteristics the values of the general and mean times for the process ing of
the workpieces from the different types of machines
are more than 1), general and mean time for time
for the
functioning
repair,
types of machines
general and mean
of the machine between two repairs, etc.
The tokens from the second type receive values of the general
(if these
as characteristics the
and mean times for processing
in the different
(if they are more than 1), general and mean waiting
time for processing, general and mean time for repair of the process ing machine if a breakdown
occurs
during processing
user can define other characteristic function
time, etc.
values for
the
The
tokens
and as a result, corresponding new data can be received. When the real process is controlled by a GN-model in real time,
the characteristics
of the tokens are obtained from the real environment by the correspon ding interface. Ihe truth-values conditions
are calculated on the
of the predicates of basis of the ones
the transition
existing in
the
GN information, accumulated as tokens' characteristics, or as expert's (operator's) On
subjective
estimation,
the basis of a concrete
ficient following ons. cess.
information
or as
can be
accumulated
which
values
can be applied for
would show the future
ones
and thus
in real can
of
of both. time, suf
be used
the random
GN-simulation
As a result, the GN-simulation processes
approximating the real
combination
GN-model functioning
definitions for changes of
These functions
a
of
in the functi the pro
would be increasingly
the results of
the simulation
state of the process. In this case, the initial
characteristics of the tokens
ought
to be taken from the real
envi-
182
CHAPTER 3
ronment.
This information
will be necessary at different
management
levels. Further improvement of the GN-model of a FMS can be achieved by introducing the influence of the toolroom on its functioning. The con dition, flow and management of tools affect in a great extent the per formance of the system as far as a researcher wants to investigate mo re thoroughly the reasons for waiting because of a tool breakage, lack of tools, time delay for tool change, etc. In order to start machining,
a machine tool
tool set all the tools with enough tool-life.
must have
in the
Very often waiting for
machine tools is due to tool breakages resulting from uncorrectly cho sen modes of cutting or from defects in the in the machining of workpieces
are of a
blank parts. These pauses
great
significance
for the
characteristic of the system as a whole. GN-modelling the toolroom
and
allows one to introduce all these
attributes
tool-dependent conditions of the functioning
of
(i. e.
tool-life, tool breakage, availability of tools, time delay due to to ol change, etc. ) in order to achieve a more
realistic picture of the
behaviour of the FMS. M X
3. 6. 3 Second
OV-/axte7 of
K
fftSs
The constructed model is based on a real functioning the plant for metal-cutting machines in Sofia. The system (a) a store for workpieces, output operations zone
including a station for
the main
where loading/unloading is performed;
including fixing stations,
intermediate
system in
consists of
stations
input/ a fixing (ports),
and terminals; (b) machine park
including 12 processing machines of three
types depending
on the size
of the NC-palets,
washing
different machines
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
183
for clearing the workpieces, a measuring machine. To eliminate queues in front of the washing machines, are permitted.
If a measure
only two measure requests
request is not strictly
the palet with the workpiece waits on the cessing machine
determined,
rear buffer of the pro-
while at least one of the washing machines is re-
leased. (c) transport system including robocars in the machine area, in fixing zone,
robocars
robocars for transportation of tools between the
processing machines and the tool store. A model on Fig. 3. 9.
describing
the movement of
the workpieces
is
shown
It consists of four transitions with the following ele-
ments: (a) transition
Z with input places 1 , 1 and 1 and type v(l , 1 , 1 1 2 8 1 2
1 ) and output places 1 , 1 and 1 ; 8 2 3 4(b) transition Z with input places 1 and 1 and type v(l , 1 ) and 2 3 12 3 12 output places 1 , 1 , 1 , 1 . 5 6 7 8 (c) transition
Z 3
with input places 1 , 1 , 1 and 5 6 7
1,1,1 ) and output places 6 7 9
1 and type v (1 , 5 9
1 , 1 . 9 10
and output places
1 , 1 . 11 12
The capacities of the transition's arcs are 1, and those of the places are co with the exception of places 1 , 1 , 1 (which have capa5 6 7 cities equal to the numbers of the processing machines for small, middle and big workpieces)
and 1
with a capacity equal to 2 (there are
no more than two measure requests). Every transition is activated on every elementary In this model the transitions are
time-step.
activated when at least
one
184-
CHAPTER 3
Fig. 3. 9 of their input places contains at least one token. The transition conditions are 1
1
2 r i
= 1 1 1 2
IW 1 1 I W 11
1 IW A l l
1
3 W
4 false
2 W 2 W 2
W 3 W 3
where W
z "there exists a machine which can process the workpiece";
W 1
= nW ; 2
3
= "the list of the processing machines is exhausted"' '
W
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
11 55 rr = = 11 2 33 2
|| W W 11 44
11 II WW 12 I 4
12 I 4
11
11 66
11 77
8 8
W W 55
W W 6 6
false false
WW 5
WW 6
WW 7
5
6
185
7
where W = W v (-|W & W") v (nWJ & iV" 8, W ); 4 6 W = W" v (nW" & W ); 5 6 W = "the worKpiece is big"; 6 W
= "the workpiece is going for a further processing or 7 re";
and W
= "the workpiece is small";
W" = "the workpiece is a medium in size";
r
1 9
1
= 1
Il W 5 1 8
W
1
W
6
I W 1 8 I W 1 8
W
7
W
9
I W 1 8
3
1 1
10 9 9 9 9
where
W
= "the measuring machine is not free"; 8
W = iV ; 9 8 1
1 11
r
= 1 4
where
I W 10 I 10
12 W 11
to
the sto-
186
W
CHAPTER 3
= "the workpiece is shelved"; 10
w = -m . 11
10 During their movement
through the net,
the tokens
obtain new
characteristics by the characteristic function as follows: $ 3 $ 4
-> "waiting-time for place 1 ", 2 -> "general time for processing of the workpiece", "difference between the real time and the time which characteristic",
is given in
the initial
§ -> "name of the processing machine for small workpieces", 5 $ -> "name of the processing machine for medium-size workpieces", 6 $ -> "name of the processing machine for big workpieces", 7 $ -> "time for processing in 15 or 16 or 17, respectively", 9 $ 10 $
-> "waiting time in 1 ", 9 -> "value of the consuming labour",
11 $
->"time for processing separate steps". 12 Tbe places represent the following process components:
1 -> initial status of the process, 1 1 -> a workpiece is waiting in the store for processing, 2 1 -> the workpiece is going for processing, 3 1 -> the workpiece is going out of the system after processing, 4 1 -> the workpiece is processed on a machine for small workpieces, 5 1 -> the workpiece is processed on a machine for medium-size workpie6
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
187
ces, 1 -> the workpiece is processed on a machine for big workpieces, 7 1 -> the workpiece is going back to the store, 8 1 -> the workpiece is waiting for releasing of the washing machines, 9 1 -> the workpiece 10 ne,
is washed and measured from the measuring machi-
1 -> the size of the workpiece does not meet the requirements, 11 workpiece is shelved and going out of the system, 1 -> the workpiece is going for further processing, 12 is going to the store.
the
if necessary it
M *
§ 3. 6. ■*. Third
W-aode]
K
of FMSs
The system to be controlled is part of a flexible workshop the Renault company.
of
The purpose of the system is to mastic seal the
body work of automobiles. For this purpose it has six workstations (or simply stations),
each consisting of one work place P, sufficient for
one car body, and also one transfer bench with two roller tables,
one
designed to load the station (TL) and the other to unload it (TU). The bench, which is mobile,
can adopt two positions:
left,
to load, or
right to unload the car bodies. A system of transport, made up of six roller tables RT ) has 6
the task of transporting all
car bodies which
(RT 1 enter
the
system (via table RT ) to the sealing station, and of later unloading 1 and removing them to the exit via
RT6.
Each table
(TL, TU, RT ) can i
carry only one car body, and at any s t a t i o n there can be a maxirrum two (P-TU, P-TL, or TU-TL) at a given time [27],
of
188
CHAPTER 3
A coloured Petri net model delling the autonomous
coloured
of this system is Petri net
applied.
In mo
model, we can distinguish
three stages: (a) independent modelling of different subsystems, which form
part of
the system under consideration, using a coloured Petri net. In the above-mentioned system, we can distinguish two subsystems:
trans
port and stations; (b) fusion
of the modules by fusing
the transitions which
represent
synchronization between the subsystems. In the system, the loading and unloading of the stations requires synchronization between the transport and the station subsystems; (c) the eventual simplification of the coloured Petri net, obtained by eliminating implicit places. These are places whose marking can be expressed as a positive linear combination of
a set of
the net and whose elimination preserves the liveness
places of
and bounded-
ness of the net. With each place, we associate a set of colours such that we can distinguish all the tasks and resources associated with it. The urs can be simple or compound (n-tuples) depending on whether sociated
tasks or resources are defined
by one or
colo the as
several parameter
With each transition, we associate a set of colours such that its car dinal is equal
to the number of different possible
evolutions of the
state of the system which are defined by their possible firings [27]. Below we
shall construct a GN which describes the above
pro
cess modelled by a coloured Petri net. The graphical structure of this net is given on Fig 3. 10. The GN has no activated
local temporal
components.
Every transition is
when there is at least one token in its input places,
all transition types are of a disjunctive type. The capacities of the places are (cf. [27]):
i. e.
SCME APPLICATIONS OF GNs IN ECONCMICS, INDUSTRY AND TRANSPORT
189
Fig. 3. 10 c(l ) = c(l ) = oo, c(l ) = . . . = c(l ) = 6. 1 8 2 7 The priorities of the places are TI ( 1 ) = TT ( 1 ) = TI ( 1 ) >,
L
3
L
5
L
7
TI ( 1 ) = TI ( 1 ) = TI ( 1 ) = TI ( 1 > = TI ( 1 ) .
LI
L2
L4
L6
L8
The priorities of the transitions are equal. The capacities of the arcs are "a>". Every token enters the net at a certain time-moment (determi ned by a function 0 ) according to some fixed time-scale (with initial K o * time-moment T, elementary time-step t receive
an undetermined number
and time-duration
of characteristics
an initial characteristic "type of the car others)".
and
t ).
It can
enters with
body, current number
(and
190
CHAPTER 3
The GN contains 4 transitions and following
transition conditions
and to
8 places.
They contain the
the GN places, the following
characteristic functions are associated. 1 2 2 r 1
W
1 3 3
= 1 I W 1 1 1
W 2
1 I W 3| |1 3 1
w 2 2
= "there is an empty loaded table" & "there is an enpty roller tab1 le";
W 2
= iW ; 1
$ -> "
$
-> «, 3 1 4
W 3
1 5
r = 1 I W 2 1 3 2
W 4
1 I W 5 | | 3 5 3
W 4 4
a = "the workplace with a number pr x is enpty"; 1 cu
W = nw ; 4 3 $ -> "
SOME APPLICATIONS OF GNS IN ECONCMICS, INDUSTRY AND TRANSPORT
1 6 r
1 7
= l i I w 3 4 1 5
w 6
1 I W 7 1 16 7 6
W 7 7
5
a = "the unlaoded table with a number pr x is empty", 1 cu-l
e
= 1W , 5
W W
191
$ -> "«turation of the sealing of the car body, duration of waiting 6 in place 1 , duration of the transfer from the workplace to 7 the unloaded table">
$
-> »,
7 1 8 r = 1 | W 4 6 I 7 where W
; "there is an empty roller table" 7
§
-> "
of the transfer from unloaded table to
roller tab-
a le">. ■
» § 3. 6. 5 <3V-model of flexible
» manufacturing
cell
This model concerns the control system organisation
of
a fle
xible manufacturing cell. We shall show how the Petri net model of the flexible manufacturing cell from [28] can be represented by GNs. A flexible
manufacturing cell
is used
tion of metal parts with specific dimensions. an
increasing
ganization
for
automatic produc
It is
characterised by
automation of technological operations,
a better
or
and a more accurate control finalization strategy aimed at
192
CHAPTER 3
increasing production.
The configuration of
a flexible manufacturing
cell, whose functioning is controlled by this model, contains: R 1
and R , machine tools 2
robots
(NC -milling machine, NC -lathe, NC -press), 1 2 3
two auxiliary buffers (B and B ), incoming T and outgoing T trans1 2 1 2 porters, a bin for rejected parts. The cell works in the following way: Raw pieces of material co me by the incoming transporter T
up to the take up position, which is 1
defined by a step movement. Robot R
waits in front of the transporter 1
T , grips the object and goes to the working space of a milling machi1 ne
NC
and there
leaves the workpiece.
1 back
and
waits for
After
processing
takes
the processed
a milling
After that, robot R 1
machine to
complete
the object, the robot goes to object
and
the processing.
the milling machine,
carries it over
in buffer B . Robot R takes the workpiece 1 2
to a free position
from the buffer
carries it over to the lathe. Then robot R
draws
draws back.
B 1
and
When the pro-
2 cessing on lathe NC is completed, R 2 2
takes the piece from
and transfers it to the press NC . Then, 3
the robot
B
the lathe
withdraws in 2
front of the press till the workpiece is pressed. After that the ro bot R takes the workpiece and carries it over to the transporter T . 2 2 Buffer B
is used for temporary
laying down
of the workpieces which
2 the lathe has finished
and waits to be pressed.
The reject
store is
used for storing the workpieces ruined during machining. A (3<-model, describing this process is shown in Fig. 3. 11, transition conditions in the form of matrix of predicates are
The
SOME APPLICATIONS OF CTfs IN ECONOMICS, INDUSTRY AND TRANSPORT
Fig. 3. 11.
193
194
CHAPTER 3
-- 1
r 1
I II
1
1 2
1
W 1
false
3
I false
W
2 I
1
where W 1
z "3(1 , TIME) = d(l >" 2 2
(the function
3(1)
gives the capacity of the place
1 and the func
tion d(l, TIME) gives the nuntoer of tokens in the place 1); 1 4
1 5
r = 1 I W 2 2 1 2 1
F
IF
W
3 I
2
where W ; "3(1 , TIME) < d ( l )" & "the necessary time 2 2 2 p l a c e 1 i s run out". 2 1 6 1 44
r 3
I II
W 33
1
1 7
1 8
false
false
for
processing in
1 9 false
10 false
1 i false 7 1
w 3
nw & &W W 3 4
nW nW 8nw Snw & &w w 3 4 5
iw i w 8nw s n w 8nw 8nw &(W &(W vw » 3 4 5 6 7
= l1 |i f a l s e 1122 I
w W 3
iv nW & &W W 3 4
n nW W snw 8nw & &W w 3 4 5
nW -m ssnw n w snw snw &(W &(W vw ) 3 4 5 6 7
1
I false 14 14 I
W 3
nW n W& &W W 3 4
nW nW 8nW 8nW & &W W 3 4 5
nW nW 8nW 8nW 8nw 8nw &(W &(W vw ) 3 4 5 6 7
1
| false 17 I
W 3
-m & &W W 3 4
-m SnW 8nW & &W W 3 4 5
nW nW 8nW 8nW 8nw 8nw &(W &(W vW ) 3 4 5 6 7
where W = "d(l , TIME) < d ( l )", 3 6 6
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
W = "3(1 , TIME) < e(l )" & "the necessary time * 11 11
for processing
195
in
place 1 is run out", 6 W
= "d(l 5
W
, TIME) < 6(1 )", 13 13
= "the workpiece is good", 6
W
= "the workpiece is ruined". 7 1 11 11 rr = 11 || 66 | | 4
1 12 12
w w
false false 4
1 II ffalse alse 8 II
W W 4
1 13 13 r r 5
= 11 || 11 11 || 1 9
r r
W W 5
conclusion
W W 55
1 15 15
11 16 16
W W 66
W W 77
false false
false
W 7
1 I false 10 I In
false false
II ffalse alse II
:: 11 II 6 13 II 6 13
1 14 14
1 17 17
we shall note that the above constructed GN-mo-
del has important advantages in relation to the other Petri net models (a) it is considerably more universal; (b) as a result of
simulation
this model, more
of a flexible
manufacturing cell with
information on the process can be received (the
tokens
which enter the net
obtain
new
with certain initial
characteristics while moving
characteristics
through
the net by the
characterizing functions associated with the places); (c) the process (if it proceeds sufficiently slowly)
can be
control-
196
CHAPTER 3
led on the basis of the constructed model; (d) the process can be optimized on the basis of the model, K
K
3. 6. 6
Queue models Scheduling is
K
and GN-models in flexible
manufacturing
one of the important problems
in
the functio
ning of the FMSs. The classical scheduling has
formulated
tactical
literature dating back to the 1960s,
scheduling
as
the following
problem: a job shop has n jobs and m machines. composed of a sequence of m operations,
optimization
Each job has a routing
a unique machine being assig
ned to each. The goal is to find a schedule (the start time of im ope rations),
which minimizes some criteria,
e.g. makespan or tardiness.
A simple approximation of the computational complexity can be calcula ted as follows: a job has m operations, one on each machine, and since each machine can be scheduled in n! ways, m In!|
schedules.
it follows that
there are
A method for finding the optimal schedule would
be
the enumeration of all these schedules and choosing
the best one. The
search space, consisting of the possible number of
valid schedules is
very large. phase of
Traditionally,
scheduling has been
the planning process,
considered
the last
which gives a solution of the problem
as when to allocate a particular resource to an operation [59]. In practice this problem is solved in the following way: (a) the priority of the job is defined; (b) the jobs are ordered according to their priorities; (c) if the planned resource is not available,
the corresponding
jobs
with a lower priority must be transferred. In flexible manufacturing
the machines are ordered
way that they can meet the requirements
in such
a
of ful1 processing of a work-
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
197
piece in one place. The FMSs can be
considered
as one
capacity unit. Because of
centralizing according to objects, but not according to functions, the material flow can be
easily
observed.
In this way the transfer of a
job does not attach to the other capacity units. To solve the scheduling problems in the FMSs, some methods are used: (a) queue models; (b) methods of mathematical programming; (c) methods with priority rules, The GNs
are also used
for modelling
the functioning of
the
FMSs. We shall try to compare the queue models with the GN-models ac cording to the method of solving scheduling problems in FMSs. The queue model consists of
knots
workpieces in front of them The time
(machines) with a queue of
for processing
is random
The
workpieces are processed according to the order of their entrance into the system (first in, first out) or in a random way. After processing, the workpiece goes to another machine with some probability. cal
Analyti
results are as follows: middle grade of loading up the machines,
middle quantity of workpieces in the system,
middle quantity
of wai
ting workpieces [60]. One of the main disadvantages of the queue models for modelling FMSs
is the
value of
the analytical
results.
These results do not
match the real situation. It is due to several factors: (a) the often-used exponential distribution for the processing time is not close account
to the real distribution;
it is possible
deterministic processing time-moments only
cessing times are identical; (b) the processing sequence is strictly determined;
to take into when the pro
i9fl
CHAPTER 3
(c> for the queue models an unlimited
capacity
is implied; this fact
does not match with the capacities of the machine buffer and
sto
rage; (d) only one priority rule for scheduling
is used
(FIFO - first
in,
first out); (e) the workpieces
cannot be individually traced, the description be
ing for the whole system; (f) the disturbances
(machine breakdown or lack of
tools)
cannot be
taken in account. The GN-model of FMSs removes the drawbacks of queue models: (a) we can use an arbitrary distribution of
the processing
times so
that in future extensions of the GNs one distribution can be ex changed with another; (b) the processing sequence is not strictly determined;
it depends on
the optimal choice at a particular time-moment; (c) the capacities correspond to the real system (machine buffer, sto rage ); (d) different priority rules can be used - rules for the optimal choi ce of the workpieces; (e) the workpieces can be individually traced; (f) the disturbances can be taken into account in the model. Comparing the de
two approaches of modelling FMSs, we can conclu
that the GN-models
are more suitable
for deciding
the sheduling
problems in flexible manufacturing, K K
3. 6. 7:
X
Conclusion
In conclusion, note that
the above constructed GN-models have
the following important advantages in relation to other PN-models. 1) They are considerably more universal compared with
other PN-
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
199
models. A PN-model is constructed for a concrete type of FMSs. The GNmodel discussed here
describes the functioning of the results of
the
work of every element of a complete class of FMSs. 2) As a result of the simulation of more
information about the process
can
the FMS with this be extracted.
characteristic functions defined are constructed
in
GN-model
Note that the
such a way that
they are maximally similar to those for the PN-model. 3) The process
(if it proceeds sufficiently slowly) can be con
trolled on the basis of the constructed model. 4-) The process of the
of the transfer of the workpieces and the choice
processing machines can be optimized on the basis of
the GN-
model. Ibis paper is based on [61-64].
REFERENSES: [1] J. Buzacott and J. ShanthiKumar, Models for understanding flexib le manufacturing systems. AIIE Transactions, 1980, 12, 339 p. [2] V. Vassilev, An organization, control and economy of flexible ma nufacturing production in manufacturing,
Moscow,
"Machinostroe-
nie", 1986, 312 p. (in Russian). [3] D. Dubois and K. Stecke, Using Petri nets to represent production process, Proc. of IEEE Conf. on
Decision and Control, Dec. 1983,
1062. [4] P. Alanche, F. Benzakour, F. Dolle, P. Gollet, R. Vallette, PSI: A Petri net based simulator
P. Rodrigues
and
for flexible manu
facturing systems, Advances in Petri nets (ed. by G. Rozenberg et al. ) in Lectures Notes in Computer Science, Vol. 188, 1985, 1-14. [5] M
Barad and D. Sipper, Flexibility in manufacturing systems: de-
finitios and Petri net modelling, 1988, No. 2, 237-248.
Int. J. Prod.
Res. , Vol. 26,
200
CHAPTER 3
[6] H. -J. Foerster, Entwicklung von Anforderungsprofilen f lexibel automatisierter Fertigungskonzepte an die
Produktionsplanung und -
Steurung, Schlussbericht, Aachen 1987. [7] M
Silva
and R. Valette,
Petri nets and flexible manufacturing,
Advances in Petri nets 1989 (G. Rozenberg, Ed. ), Lecture Notes in Computer Science 424, 1990, 374-417. [8] J.Martinez, HAlla, M Silva, Petri nets for the FMSs,
Modelling and
(A Kusiak, Ed. ),
design of Flexible
Elsevier Science
specification of
Manufacturing
Systems,
Publishers B. V. , Amsterdam,
1986, 389-406. [9] J. Ahuja mand K. Valavanis, for
A hierarchial
flexible manufacturing
modeling
systems using extended
1988 International Conference on
methodology Petri
nets.
Computer Integrated Manufactu
ring-Washington, DC, USA: IEEE Comput. Soc. Press, 1988, 350-356. [10] A. Bobbio and A
Savant,
Modelling automated production
systems
by deterministic Petri nets. IFS Simulation in Manufacturing,
3-
rd Intern. Conf. , 1987, 127-136. [11] H, Boerner and I. Pritika, Aspekte der Modellierung von Flexiblen Fertigungssystemen
(FFS)
auf Basis
von
schaftliche Zeitschrift der Technischen
Petri-Netzen.
Oniversitaet
Wissen-
Karl-Marx-
Stad, Vol. 29. No. 6, 1987, 793-797. [12] F. Bukanov and K. Belikov, Technology of design flexible manufac turing Systems by means of coloured Petri Nets.
Systems of Robot
and Flexible Automation, Kuibishev Politechnical Institute, 1987, 62-65 (in Russian). [13] F. CapKovic,
Decision support algorithm for flexible manufactu
ring Systems Control. Computers in Industry,
Vol. 10, No3, 1988,
165-170. [14] H, Carstensen, Decidability Questions for Fairness in Petri Nets. Lecture Notes in Computer Science, Vol. 247, STACS 87, 4th Annual Symposium on
Theoretical
Aspects of
Computer Science,
Passau,
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
201
1987, Proceedings (F. J. Brandenburg, G. Vidal-Naquet and M
Wir-
sing, Eds. ), Springer Verlag 1987, 396-407. [15] B. Descotes-Gennon and F. Hemon, Mercier des Rochettes, R. ISATIS: An Editor-Simulator of Manufacturing Systems Using Timed coloured Petri Nets. Databases for Production Management. the IFIP TC5/WG
/ Working Conference, 1989,
Proceedings
Barcelona,
of
Spain -
ENSIEG, St. Martin d*Heres, F. ASTEC, Grenoble, 1989, 333-348. [16] K. Srihari, C. Bnerson and J. Cecil,
Modeling Manufacturing with
Petri Nets. CIM Review, Vol. 6, No. 3, 1990, 15-21. [17] S. Teng and J. Black,
Cellular
Manufacturing Systems Modeling:
the Petri Net Approach. Journal of Manufacturing Systems, Vol. 9, No. 1, 1990, 45-54. [18] S. Tzafestas, Petri-Net and Knowledge-Based Methodologies nufacturing Systems Modelling
Simulation
and Control.
Integrated Manufacturing. Proceedings of the
in Ma
Computer
5th CIM Europe Con
ference, 1989, Athens, Greece (G. Halatsis, et al. , Eds. ), Kempston, UK, IFS Publications, 1989, 39-50. [19] K. Jensen, Coloured Petri Nets. Lecture Notes in Computer Science Vol. 254: Petri Nets: Central Models and Their Properties, Advan ces in Petri Nets, 1986, Part I, Proceedings of an Advanced Cour se,
Bad Honnef,
1986
(W. Brauer,
W. Reisig
and G. Rozenberg,
Eds. ), Springer-Verlag, 1987, 248-299. [20] F. Itter, zen.
Integrierte Modellbildung und Simulation mit Petrinet-
Zeitschrift for wirtschaftliche Fertigung und Autcmatisie-
rung ZWF/CIM, Vol. 84, No. 2, 1989, 90-92. [21] F. Janzen, H. Kath , M
Mohrle and H. Seifert,
Petrinetze in der
Produktionstechnik. Zeitschrift for wirtschaftliche Fertigung und Autcmitisierung ZWF/CIM, Vol. 84, No. 3, 1989, 141-145. [22] S. Menon and P. Ferreira, Coloured Petri net
based
architecture
for coordination control of flexible manufacturing systems. Proc. of "Advances in Manufacturing
System
Engeneering"
presented at
202
CHAPTER 3
the Winter Annual Meeting of the ASME,
1988,
Chicago, IL, USA -
New York, NY, USA: ASME, 1988, 69-88. [23] S. Menon and P. Ferreira,
Analysis of
models for coordination control of
coloured
flexible
Petri net based
manufacturing
sys
tems. Proceedings of SME 17th NAMRC, Ohio State Univ. , May
1989,
331-338. [2*] A. MowafaK Hassan,
The modelling and analysis of
flexible manu
facturing systems using Petri nets. Ph. D. Thesis, Technical
Uni
versity of Wroclaw, Institute of Industrial Engeneering and Mana gement (Oct. , 1988). [25] Y. Nakamura, I. Hatono, Y. Kohara, K. Yamagat and H scheduling using
timed Petri net and rule base.
the USA-Japan Symposium on
Flexible
Tamura,
FMS
Proceedings
of
Automation - Crossing Brid
ges: Advances in Flexible Automation and Robotics, 1988, Minnea polis, MM, USA, 883-890. [26] U. Negretto,
A functional
process model
for flexible
assembly
cells. Proceedings of the 19th International Symposium on Automo tive Technology and Automation (ISATA), 1988, Monte Carlo,
Mona
co; with Particular Reference to Cell Control and Quality Manage ment Systems for the Manufacturing Industries, Vol. 1. - Croydon, UK. Allied Autom , 427-443. [27] J. Martinez and M
Silva,
A language for the description of con
current systems modeled by colored Petri nets: Application to the control of flexible manufacturing systems, IEEE Workshop on
Lan
guages for automation, New Orleans, Nov. 1984, 72-77. [28] M
Jocovic, An application of Petri-nets in the
controlsystem of
the FTC, Microprocessing and Microprogramming 24 (1988), 681-686. [29] J. Campos, J. Colom and M
Silva, Performance evaluation of repe
titive automated manufacturing systems,
Proc. of Second Interna
tional Conference on Computer Integrated Manufacturing, New York, May 1990, 74-81.
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
[30] P. Muro, J. Villarroel and M. Silva,
A Knowledge
representation
environment for manufacturing control systems design typing, 6th Symposium on information control facturing technology
(E. Puente, Ed. ),
203
and proto
problems
Madrid,
in manu
September 1989,
585-569. [31] S. Veli 11a and M tation of
highly
Silva, The spy: a mechanism for concurrent systems,
safe implemen
Proceedings of
the 15-th
IFAC/IFIP Workshop "Real Time Programming' 1988" (A. Crespo and J. A. de la Puente, Eds. ), Valencia, Spain, May 1988, 75-81. [32] J. Villarroel, J. Martinez and M
Silva, GRAMAN: A Graphic system
for manufacturing system design, Proceedings of the IMACS Sympo sium on System Model lingand Simulation berg and L. Carotenuto, Eds. ),
(S. Tzafestas,
Cetraro, Italy,
A, Eisen-
September, 1988,
311-316. [33] P. Muro and J. Villarroel, KRON:Redes orientadas a la representacion del concomiento, 3 Reunion Tecnica de la Asociacion Espanola para la Inteligencia Artificial,
ACTAS,
Madrid
de noviembro de
1989, 3-22. [33] U. Negretto and M tigungs-
Rillo, Erweiterte Petri-Netze in flexiben Fer-
and Montagesystemen,
Robotersysteme,
Vol. 4,
No. 1,
1968, 34-42. [34] Y. Narahari and N. Viswanadham,
A Petri net approach
to the mo
delling and analysis of flexible manufacturing systems, Annals of Operations Research 3, 1985, 449-472. [35] H. Alayan and R. -W. Newcomb,
Petri-net
robot
models
for robot
networks, IEEE, Vol. 3, 1986, 996-999. [36] F. Archetti,
Performance
evaluation
of flexible
manufacturing
systems using Petri nets. One-Day Seminar at the Bocconi Univer sity of Milan
"Applicability
of Petri nets
to Operations Rese
arch", 1986, 143. [37] C. Beck and B. Krogh,
Models for simulation and discrete control
204
CHAPTER 3
of manufacturing systems,
Proceedings
1966
IEEE
International
Conference on Robotics and Automation, San Francisko, 19*6,
IEEE
Comput. Soc. Press, Washington, 1986, 305-310. [38] J. Martinez and M
Solva,
New methods for specifying controls of
FMS, Autominstrum (Spain), Vol. 20, No. 156, 185-195 (in Spain). [39] M. Kamath and N. Viswanadham,
Applications
models in the modelling and analysis systems,
of
of Petri nets
flexible
based
manufacturing
Proceedings 1986 IEEE International Conference on Robo
tics and Automation,
San
Francisko,
1986,
IEEE Comput.
Soc.
Press, Washington, 1986, 312-317. [40] R. Ravichandran, A. Chakravarty, Design support in flexible manu facturing systems using timed Petri
nets,
Journal of Manufactu
ring Systems, Vol. 5, No. 2, 89-101. [41] M
Shaw and A
Whinston, Task bidding and distributed planning in
flexible manufacturing, The Second Conf. on AI Applications, 1985, 184-189. [42] M
Silva and J. Martinez,
A software
environment
for designing
with colored Petri nets and their implementation. Application flexible manufacturing systems. University of Milan
One-Day
"Applicability
Seminar
to
at the Bocconi
of Petri nets
to Operations
Research", 1986, 95-127. [43] T. Klotz, R. Phillips, R. Sprattling and H. Sutherland, Real-time performance evaluation of local-area networks
used
in automated
manufacturing systems, Proceedings 1986 IEEE International Confe rence on Robotics and Automation, San Francisko, 1986, IEEE
Com
put. Soc. Press, Washington, 1986, 1723-1730. [43] R. Valette,
M
Courvoisier, H. Denmou, J. Bigou and C. Desclaux,
Putting Petri nets to work for controlling flexible manufacturing systems. Proceedings of 1985 IEEE International Symposium on Cir cuits and Systems, New York, 1985, 929-932. [44] J. Bourney and J. Gentina,
Structuring of the procedural part of
SCME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
205
the control system of flexible manufacturing cells, Congres Automatique 1988, Grenoble, 233-24-2 (in French). [45] P. Stotts, R, Newcomb
and
Z. Ning Cai,
Modelling
the
logical
structure of flexible manufacturing systems with Petri nets. Com puter Communications, Vol. 12, No. 4, 1989, 193-203. [46] H
Tamura, K. Yamagata and I. Hatono, Decision making for flexib
le manufacturing-OR and/or AI approaches
in scheduling.
Systems
Analysis, Modelling, Simulation, Vol. 6, No. 5, 1989, 363-371. [47] J. Ahuja and K. Valavanis,
Modified Petri nets and comprehensive
modeling of flexible manufacturing systems.
Proc.
of the
12-th
IMACS World Conf. , Vol. 3, 1988, 532-534. [48] J. Barbez, E. Craye, J. Gentina
and J. Mayet, Hierarchical level
and implementation for analysis and synthesis of control and re liability of flexible manufacturing systems, Proc,
of the
12-th
IMACS World Conf. , Vol, 3, 1988, 552-558. [49] J. Gentina and D. Corbeel, Petri nets in manufactoring, Control Encyclopedia: Theory,
Technology,
System &
Applications,
Vol. 6
(M Singh, Ed.), Pergamon Press, Oxford,, 1987, 3670-1673. [50] R. Mercier Des Rochette,
B. Descotes-Genon and P. Ladet, Model
ling and F.MS. Control implementation, Proc. of the
12-th IMACS
World Conf. , Vol. 3, 1988, 559-661. [51] A. Pagnoni, Computer management of coordination in flexible manu facturing systems. Proc. of the 12-th
IMACS World Conf. , Vol. 3,
1988, 500-502. [52] W. Zhang, Automatic robot planning by marked Petri nets. Proc. of the Int. Workshop on
AI for
Industrial Applications,
New York,
1988, 358-361. [53] H. Kodate,
K. Fujii and
K. Yamanoi,
Representation of FMS with
Petri net graph and its applications of simulation of system ope ration. Robotics & Comput. Integrated Manufacturing, Vol. 3, No. 3, 1987, 275-283.
206
CHAPTER 3
[54] G. Balbo,
G. Chiola,
G. Franceschinis and G. Roet,
Generalized
stochastic Petri nets for the performance evaluation of FMS, Pro ceedings 1987 IEEE International Conference on Robotics and Auto mation, IEEE Comput. Soc. Press, Washington, 1987, 1013-1018. [55] A. Likic and V. ZivKovic, and analysis of
A software package
for representation
flexible manufacturing systems
nets, Proc. of the 3rd European
based
on
Petri
Simulation Congress, 1989, Edin-
burg, (D. Murray-Smith, Ed. ), 502-507. [56] M
Silva, Petri nets and flexible manufacturing, Lecture Notes in
Computer Science, Vol. 424, 1990, 374-417. [57] A. Zebiri, M
Bormey and M
Head, The use of Petri nets as
sis for simulating manufacturing systems,
a ba
Proc. of the 3rd Euro
pean Simulation Congress, 1989, Edinburg, (D. Murray-Smith, Ed. ), 825-829. [58] F. Archetti and A. Sciomachen, Representation, analysis and simu lation of manufacturing systems by Petri net based models, crete Event Systems: Models and Applications. Varaiya and A. Kurzhanski, Eds. ),
Dis
IIASA Conf. ,
Springer-Verlag,
1988,
(P. 162-
178. [59] R. Stokey,
AI Factory Scheduling:
Multiple Problem Formulation,
Sigart Newsletter, October 1989, Number 110, 27-29. [60] G. -W. Hintz,
Ein wissensbasiertes System zur Produktionsplannung
und -steurung fuer flexible Fertigungssysteme, VDI-Verlag,
Dues-
seldorf, 1987. [61) K. Atanassov, E. Dincheva, D. Matev, M ralized net-representation of tems, Proc. of the 14 Symp.
Stefanova-Pavlova, Gene
the flexible
on Operation
manufacturing
Research,
sys
Ulm, Sept.
1989, 521-528. [62] M
Stefanova-Pavlova,
facturing system, Seminar,
Sofia,
Generalized net
First Sci.
representation of manu
Session of
the Math.
Found.
Oct. 10, 1989, Preprint IM-MFAIS-7-89,
AI
Sofia,
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
207
1989, 60-62. [63] M
Stefanova-Pavlova, R. Christov, E. Dincheva,
net model of a flexible sion of the Math.
manufacturing
Found.
AI Seminar,
Generalized net
system, Second Sci. Ses Sofia,
March
30, 1990,
Preprint IM-MFAIS-1-90, Sofia, 1990, 40-42. [64] M
Stefanova-Pavlova, Queue models and Generalized net models
in flexible manufacturing. Found.
AI Seminar,
Third
Sci. Session of the Math.
Sofia, June 12, 1990, Preprint IM-MFAIS-
2-90, Sofia, 1990, 13-14.
208
CHAPTER 3
§ 3. 7: GENERALIZED HET MODELS OF THE ACTIVITY OF NEFTCCHIM PETROCHEMICAL COMBINE IN BOURGAS Stela Dimitrova, Ludmila Dimitrova, Trajana Kolarova, Plamen Fetkov, Krassimir Atanassov and Rumen Christov
Generalized Nets complex
objects
which
(GNs)
are a comfortable
means for designing
are characterized by a large scale of various
parallel processes taking place in the real time. In the general case, it is impossible to show analytical formulas formalizing these proces ses.
For their modelling,
various simulation methods are used. Petri
nets are one of these (new) basic methods. As discussed in
§1,1, the
GNs can also be used as a means for simulating such (real) processes, but (in contrast with the Petri tion condition predicates
nets and
by virtue
of their transi
and characteristic functions) they can also
include in themselves elements of analytical methods for
description.
For example, some characteristic functions can assign to a given token a value that is calculated by analytical functions. "The NEFTOCHIH of the integrated strong
Petrochemical Combine (FCC)
industrial enterprises in
in Bourgas
is one
Bulgaria which exerts
a
influence upon the dynamic development of the State Industrial
Chemical Corporation, the economy of the country,
and the internatio
nal division of labour. " [1] The processes going in PCC can be modelled by GNs and structed
models
can be used
for simulation
combine. The models are investigated following
the con
of the processes in the the ideas
for hierar
chical models and models based on the union and the composition of ot her simpler GN-models introduced in § 1. 1. The modelling starts with the construction of a simplified glo bal
model
of the combine in general.
It has the from of
Fig. 3. 12,
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
209
Fig. 3. 12 The GN used can be, e. g. , a second type intuitionistic fuzzy GN,
i. e.
a ON in which the tokens are replaced by "quantities" (see § 3 in App. 1) which
"flow in the net.
So a token ("quantity") enters place 1 1
and this
action
represents
ceived. This place symbolizes distribution
of
de: "chemical"
the quantity of raw materials (oil) re Petrochemical Sea Port (Terminal) where
raw material through the basic types of units is ma (1 ) and refinery.
The second one has the following
3 two components: a "Catalitic Reforming" (1 ) and an 4
"output of fuels"
(1 ). A part of the production from 1 is directed to 1 and 5 4 6
together
210
CHAPTER 3
with the production from 1 , is going (according to preliminary by set 3 conditions) to 1 9
(for production of plastics, fibres and rubber), or
to 1 (for production 10 A part of the 1 11
of petrochemicals - in "petrochemical units").
"quality" (token) is going from the last place to place
(for production of solvents),
and another
place 1 to the "catalytic cracKing unit" 12
part returns
through
and there takes part
with
newly entered "qualities", in the production of fuels (1 ) and 7 products (1 ). The toKens (qualities) 8 go to place
1 14
other
from places 1 , 1 , 1 , 1 , 1 9 11 7 a 5
and that symbolizes the outcome of the finished pro-
duct from the PCC. By this part of the GN the movement (in general outline) of fuels, petrochemicals and plastics, produced and processed in PCC can be observed. Besides, the GN has a second (not differentiated separately) part
that symbolizes
the entry of information on
the demand
of pe-
trochemicals (1 ), workout of this information (orders) (1 ) and re2 13 port on the orders (1 ) - the effect that they are satisfied or not. 15 The tokens ("quantities") that go into 1 have as initial cha1 racteristics the type of raw material, and during the time of movement in
the GN
they receive the following characteristics:
"the kind and
quantities of processing intermediate fractions" and "the kind and quantities of the finished product" depending on the type (Kind) of places they enter. obtain the
In the end in place
1 these tokens 14
("quantities")
finished characteristic "working expenses" and other
racteristics specified by the model's consumers.
cha-
SOME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
211
The other type of tokens are going to place 1 with the initial 2 characteristic "information about the demand of goods with sary description, quantities, with
the final
the neces
parameters, etc. and are leaving the GN
characteristic "the order is satisfied" or "the order
is not satisfied" (as well as other data). More detailed processing
models can report on
the moment when the raw material enters the PCC, cesses,
such factors as
the duration of pro
the moments when the products leave the PCC, the duration of
waiting for processing of raw materials or fractions etc. The other tokens can have their own priorities which determine the ways of
the transfer. The transition condition predicates can ha
ve explicite
forms. During simulation of the PPC-processes, different
situations can occur in relation to the initial characteristics of the tokens in places 1 and 1 , and also in relation with the characteris1 2 tics
of the tokens which are in the net at
the initial
time-moment
(they correspond to the available quantities in the PPC). The described GN reflects only the most global connections exi sting
between the different
units of PPC. Mien more details
are re
quired in the positions of the places and the transitions of the above net, new GNs will be
standing.
They will describe
the separate sub-
processes which are included in the most global processes. Another situation is the one when most global one)
corresponds
a new GN
(a subnet
of the
to a part of the places and transitions
of the above GN, but does not cover their content. A GN shown in Fig. 3. 13
is an example,
illustrating this. The
components of this net correspond to places 1 , 1 , 1 and the transi1 4 5 tion Z . 1 Briefly, the sense of the places of the new net are as follows: 1 - Petrochemical Sea Port (Terminal), 1
212
CHAPTER 3
Fig.
3. 13.
SOME APPLICATIONS OF GNs IN ECONCMICS, INDUSTRY AND TRANSPORT
213
1 - production of fuel oil 2 1 , 1 , 1 , 1 , 1 , 1 3 4 5 6 9
,1 - production of other cuts 13 30
1 - production of oil cut 7 1,1 - production of cut C 10 11 4 1 - production of high-octane gasoline 12 1 , 1 ,. . . , 1 16 17 25
- tanks
1 - production of cut C 26 5 1
- production of toluen.
27 The GN-models described here reflect the first steps modelling of
the NEFTOCHIM Petrochemical Combine
proposed text, other
of global
in Bourgas.
In the
aspects and methods for describing the petroche
mical processes are not included. REFERENCE: [1] NEFTOCHIM Economic Combine-Bourgas, 1988.
BulgarReklama Agency,
Sofia,
214
CHAPTER 3
§3.8: A SBCCHD TYPE OF IHTUITIOHISTIC FUZZY GEHERALIZED HET-HCDEL IH THE CHEMICAL INDUSTRY Rumen Christov
and
Stoian Garbov
On the basis of the second type of intuitionistic fuzzy genera lized nets (IPGH2) (see § 3 in App. 1) we shall describe a part of the technological
scheme of the functioning of
a silo-farm with pneumo-
transport in a plasticware company. The raw materials are They are
ferried with
auto-cisterns
provided from 4 places and transferred to 10
and tanks. store-bunkers
(silos) or (if necessary) directly to the production lines. rials must flow continuously to these lines. duction lines from the silos
Raw mate
The resin enters
6 pro
by a distributor. Bach line has an indi
vidual capacity and the material available for a definite time is con sumed. The IFGH2 describing this process has the form of Fig. 3.14. It has 3 transitions and 23 places with: - places 1 1
1 - symbolising the accepting places; 4
- places 1 ,..., 1 - symbolising the silos; 5 14 - place 1
- showing the transportation of the material
to
the pro-
15 duction lines; - place 1 - modeling the distributor; 17 - places 1 1 - symbolising the flow production lines. 18 23 All transitions have equal priority. The places 1 and 1 15 ve
the highest priority;
the other places have equal priority.
transition conditions have the form:
ha-
17 The
SC*E APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
Fig. 3. 14
215
216
CHAPTER 3
1
1 5
1
=
1 7
1
1
8
1 10
9
1 11
1 12
1 13
1 14
15
I 1 1
W
W 2
W 3
W 4
W 5
W 6
W 7
W 8
W 9
W
W 10
11
W 1
W
W 3
W 4
W 5
W 6
W 7
W 8
W 9
W
W 10
11
1 3
I | I I 1
W 10
11
1 4
I 1
W 10
11
W 10
11
1 r
1 6
1 2
1
1
I 16 I
2
W 1
W 2
W 3
W 4
W 5
W 6
W 7
W 8
W 9
W
W 1
W 2
W 3
W 4
W 5
W 6
W 7
W 8
W 9
W
W 2
W 3
W 4
W
W
W 1
5
W 6
W 7
W 8
W 9
where W = "the r e s i n m i s t go t o t h e i - t h s i l o " i - "the resin nust go to
W
(1 i i i
10),
the flow production lines,
because there
11 exists a flow
production line with resin less than the critical
quantity"; 1
1 5
r 2
=
I 1
W
1 I 6 |
W
1 7
I 1
W
1 8
I 1
W
1
I
W
9
16
1 17
12
W 13
12
W 13
12
W 13
12
W 13
12
W 13
12
W 13
12
W 13
12
W 13
12
W 13
12
W 13
1 1
I 10 I
W
I 11 I
W
I 12 I
W
1 I 13 I
W
1
W
1 1
I 14 |
SCME APPLICATIONS OF GNs IN ECONOMICS, INDUSTRY AND TRANSPORT
217
where W
; "the resin must go to another silo"; 12
W
= "the resin must go to
a production
line with
resin less than
13 the critical quantity or with a minimal resin quantity"; 1
r
= 3
IS
1 19
1
1 20
1 I W W W 15 I 14 15 16 I 1 I W W W 17 I 14 15 16
1 21 W
1 22 W
17 W
18
W 19
18
W 19
W 17
23
where W
= "the resin mist go to the i-th production line which has
the lo-
i west resin quantity"
and
"the resin is below the critical li
mit" (14 i i i 19). The characteristic functions $ give the infinvm and the suprei mum of the available resin quantity in the place 1 (1 i i £ 23). i The IPGN2-models functioning starts at a certain time-moment according to a certain absolute time-scale o step (t ) being 2 minutes.
and the elementary
T
time-
The features of this net are the discrete values of the predi cates of the transition the tokens are
conditions but the net
is an IFGN2,
"quantities" which flow in the net and
part
because of these
quantities is lost in the process of transportation, i. e. there exists a degree of indetermination. The following problems are solved (addaptively) by the model: a) The optimal schedule of the raw materials ferrying is determined. b) TTie optimal loading of the machines is determined. c) The optimal capacity of the connected elements of the manufacturing is checked.
218
CHAPTER 3
d) The possible waiting of the machines are determined. On the basis of the IFGN2-model, the processes in a large class of productions of the chemical industry can be simulated and led (in real-time).
control
Chapter
4:
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
The first research in the area of medical
diagnosis dates back to 1983.
structed nets are collected. scribed models
by means of
generalized net modelling of In § 4. 1 almost all
The authors plan to realize all program package for
the de
generalized nets.
more general model of diagnostics is introduced in § 4. 2.
219
the con
A
220
CHAPTER 4
H I :
APPLICATIONS OF GENERALIZED NETS IN HEFHROLOGY
Joseph G. Sorsich
and
Krassimir T. Atanassov
In this chapter, an exanple of the application of
GNs in medi
cine is given, Nephrology (see e.g. [1-16]), as a relatively new field in medicine,
has been used to make things clearer
and to demonstrate
the possibilities of GNs. Some other applications in medicine are also available (e.g. [17]). Hie mathematical
foundations of
our research
are essentially
different from other basic mathematical methods applicable in medicine (see e. g. [18-21]). A renal
damage is usually suspected in the course of a routine
examination of a patient with some clinical signs and symptoms ria, renal colic,
edema,
turbid urine),
some disturbances in blood sample or urinalyse are present themia, proteinuria, etc. ). kidney deseases,
(byperazo-
Based upon these common signs relating to
the following GN is constructed. Ibis is not a defi
nite model and it may be a subject of further modifications. which will
be
(dysu-
arterial hypertension or if
New GNs,
the subnets of the general GN described below,
added, and/or the subnets themselves may be modified. to demonstrate that such a subject in biological
may be
Our aim is also
sciences as medicine
is susceptible to comparatively simple and easy to understand mathema tical formulation by GNs. The same may be realized in almost all other areas of medicine. Initially, we shall describe one as subnets other GNs
(general)
describing the process
signs and symptoms in nephrology.
GN
which contains
of diagnosing
Let this GN be E.
different
It has one input
place L . The tokens of this GN will represent the patients of a nephrologic unit. Everywhere we describe the transfer of one token the tokens generated by it,
(patient)
or
but this is valid for every set of tokens
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
221
(patients) transferring in the GN. At the end of this chapter we shall discuss these problems more widely, as well as
the possibilities
for
application of this and similar GNs in different medical units. The token goes to place
L
with
the initial
characteristics
1 "patient with probable Kidney disease", kens which
represent the appropriate
next it splits into investigations of
four to
the patient.
Later these tokens may be united into one token whose characteristics correspond to all the applied investigations, Part of these tokens may leave the GN if some of the signs are missing. The sense of the diffe rent GN's places are as follows: L - abnormal findings (changes) in urine are present, 2 L
- elevated blood urea nitrogen (BON) is present, 3
L - clinical signs and symptoms of renal disease are present, 4L
- arterial hypertension is present, 5
L
- permanent proteinuria is present, 6
L
- hematuria is present, 7
L
- turbid urine is present,
a L
- renal colic is present, 9
L
- dysuria is present, 10
L
- edematous patient is present. (Edemas of renal origin are always 11 related to proteinuria, refer to § 9. 1),
L
- evaluation of arterial hypertension, 12
L
- a specifying the type of renal damage, 13
L
- arterial hypertension of renal (renoparenchimal) 14
origin is sus-
222
CHAPTER 4
pec ted, L
- renovascular hypertension is suspected. 15 The transition conditions are as follows: L 22 rr
= L 11
L
ll ttrue rue
4 4
ttrue rue
true true ;;
11 I I L
L 66
rr
L 33
8 8
V V
= = L
II W W 22 11 11
2
L 77 W W
22
3
where W
= "permanent proteinuria is present", 1
W
= "hematuria is present", 2
W
= "turbid urine is present"; 3 L 99 rr
L
L 10 10
= L II W W 44 11 44 3
11 11
W W 55
W W 6
where W
= "renal colic is present", 4
W
= "dysuria is present", 5
W = "edematous p a t i e n t i s present"; 6 L r
= L 4
L 12
13
l true
true
5 I L
L 12
r
= L 5
where
I 113 3 I
W 7
13
W
fl8
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
223
224
W
CHAPTER 4
= "there is suspicion of renoparenchimal arterial hypertension", 7
W
= "there is suspicion of renovascular arterial hypertension". 8 From the fact that a token
(a unique one or one of many repre
sentations of a given token (patient) that enters GN
E) enters places
L , L , L , . . , , L , L , L , i t follows that this token is directed 3 6 7 42 14 15 to the following subnets: L
- to the GN from § 4. 1. 1, 6
L
- to the GN from § 4. 1. 2, 7
L
- to the GN from § 4. 1. 3, 8
L
- to the GN from § 4. 1. 4, 3
L
- to the GN from § 4. 1. 5, 9
L
- to the GN from § 4. 1. 6, 10
L
- to the GN from § 4. 1, 1, 11
L
- to the GN from § 4. 1. 7, 12
L
- to the GN from § 4. 1. 8, 14
L
- to the GN from § 4. 1. 9. 15 On the basis of this model, we can achieve the following:
- control - optimization -simulation of real processes. Control initial
is realized
state of the patient
the token in L ). 1
as follows:
the physician describes
(this is the initial characteristic
the of
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
225
In the other offices, the token (patient) gets as current characteris tic the
results of the corresponding investigation. Thus on the basis
of the total
(at the moment)
the basis of all existing
information about the patient
(i.e. on
characteristics of the tokens, generated by
the initial token (patient)),
a decision for the subsequent direction
of the tokens is determined (i.e. the subsequent patient's
investiga
tions). When the transfer of more tokens (patients) we can optimize
in the net occurs,
their transfer with the aim of achieving minimum wai
ting-time in queues. Knowing the probable distributions
connected with
the tranfer
of the patients in different offices, on the basis of the total GN-model, the process can be simulated. pital units can be determined
Thus, different parameters of hos
(e. g. workload of offices, average time
for servicing a single patient and of stay of a patient in a hospital
a single office,
unit etc. ).
average time-
This information
can be
used for optimal organization of different medical units. Note, that direct use
of our model is possible
(in this form)
in specialized (nephrological) units because only the processes of the nephrological disease diagnosis are described in it. If we want to ap ply similar models for other medical purposes
(cardiology, neurology,
obstetrics, laboratory investigations and their interpretation, etc. >, we must include in it GN-models of the corresponding case. Some Bulga rian physicians are working on this problem There already exist more expert and/or informative medical sys tems to help physicians in diagnosis and therapeutics. Such expert and informative systems may be included in the framework of the described total GN-model or in some of its parts (subnets).
For example, expert
and/or informative systems for diagnostics can be used frames of the global GN's memory
(see App. 1)
(e. g. , in the
for calculation of the
226
CHAPTER 4
truth-values of some
transition condition predicates and for calcula-
tion of some toKen's characteristics. On the other hand, expert and informative systems for therapeutics may be included in the described GN-model, when the final token's characteristics are to be determined. Below we describe other GN-models of and symptoms
basic nephrological signs
but these models have an independent sense because their
entire inclusion in the GN
E deprives them of
the clearness which is
easyly attainable by the GN-tools. This paper is based on [22-30],
. § 4. 1. i Permanent
»
«
*
proteinwla
Any permanent proteinuria in adults is
a pathological situati-
on. Proteinuria may be detected in different conditions. Usually it is asymptomatic, but the practitioner's first step must always be a careful physical examination of the patient - not only
the urinary tract,
but the cardiovascular system (including arterial blood pressure, eyeground) and other systems as well. The tokens (patients) get into place racteristic
"permanent proteinuria".
1 l
Next they
with the initial chasplit and get in the
sections with initial places 1 , 1 and 1 2 3 4-
so that each gets
propriate
the complete
characteristic:
"the result of
the ap-
urinalysis",
"plasma creatinine's level (or gromerular filtration rate)" and "X-ray examinations of the urinary tract". In the last case, the token leaves this GN and gets in the GN and r
2
have the forms:
from § 4. 1. 2. The transition conditions r 1
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
r 1
= 1 | 1 I
1 2
1 3
true
true
1 5 rr 22
= = 11 II 22 II
The token in place
1
1 2
1 4 true
1
1 7
6
ttrue rue
ttrue rue
227
ttrue rue
8 true true
splits into four tokens which get
into
places 1 , 1 , 1 and 1 according to their characteristics: amount of 5 6 7 8 24-hours proteinuria, presence of glucosuria, microscopic investigati on of urine sediment, urocultures. The transition condition
rr 33
11 99
11
- 11 II W W 33 || 11
W W
10 10 2 2
where W
= "finds out renal insufficiency", 1
W 2
= ~\V . i
The token leaves this GN also and gets into the GN from § 4. 1. 4. 11 11 11 rr = = 11 II W W 44 55 | | 33
11 12 12 W W 4 4
where - "proteinuria more that 3 gr, per 24 hours",
W 3
w - -w . 4
3 11 113 3 rr 55
where
= = 11 II W W 66 II 55
11 14 14 W W
6
228
CHAPTER 4
W = "there is glucosuria" 5 (the token leaves this GN - the patient is directed to diabetologist>,
w =
5
6
r = 1 1 6 7I The token in place places 1 , 1 15 16 results
of
and 1 17
1 7
1 15
1 16
true
true
1 17 true
splits into three tokens which get into
with the corresponding characteristics:
the microscopic examination of
urine sediment
the
for white
blood cells, red blood cells, bacteria The transition conditions r , r 7 8 1 18 r
7
- 1 I W 8 1 7
r
have the forms 13
1 19 W 8
wher W = "urine cultures are positive" 7 (the token leaves
this GN and
the patient is
directed toward the GN
described in § 4. 1. 3), W = -W ; 8 7 1 20 r = 1 8 H
| W I 9
1 21 W
10
where W = "there is a paraproteinuria" 9 (the token leaves the GN - the patient is directed to hematologist), W
10
: nW ; 9
APPLICATIONS OF GENERALIZED NETS IN JffiDICINE
229
230
CHAPTER 4
1 22
1
r = 1 I W 9 12 I 11
W
23 12
where W
= "the proteinuria is combined with hematuria" 11
(the token goes to the GN from § 4. 1. 2); W 12
= nW ; 11 1 24 24 r 10 10
= 1 I W 15 15 I I 13 13
1 25 25 W 14 14
where W
= "leucocyturia is present"; 13
W 14
= nW ; 13 1 26 26 r 11 11
= 1 I W 16 I 15 16 I 15
1 27 27 W 16 16
where W
= "hematuria is present"; 15 = IV
W
16
15
r 12 12
1 28
1 29
= 1 | W 17 17 I I 17 17
W 18 18
where W
= "bacteriuria is present"; 17
W 18
= nW ; 17
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
231
1 30
r
= 13
If the token
1 I 21 I
true
1 I 23 I
true
1 I 14 I I| 1 I 25 I
true true
1 I 27 I
true
1 I 29 I
true
1 I 19 I
true
1 I 10 |
true
gets into place 1 or 1 , it enters the GN from 2428
§ 4. 1. 3 and from place 1 26
to the GN from § 4. I. 2.
Acute nephritic syndrome takes shape if 1 , 26
in places
1 , 1 5 9
the tokens arising from an initial one are gathering
and
(often the
hematuria is macroscopic, the renal insufficiency is transitory, there are edema and arterial hypertension).
The tokens
in places
1 , 1 , 10 14
1 ,1 ,1 ,1 ,1 and 1 transit in place 1 - the patient 19 21 23 25 27 29 30 directed to a nephrologist,
is
where complementary investigations are to
be undertaken (umunological, renal biopsy, etc. ). K K
»
§ 4. i. 2 Hematuria The
presence of blood in urine (hematuria) is a conmon symptom
in nephro-urology. The causal diagnosis sometimes is very complicated.
232
CHAPTER 4
Hematuria may
be isolated
or combined with other signs and symptoms.
Thus a complete examination
of the patient is always indispensable.
The token gets into place 1 with the initial
characteristic -
1 presence of hematuria,
which is confirmed by a microscopic investiga
tion of the urine sediment. The transition condition 1 2 r
= 1 1
1 2
I W
W
1 1 1
2
where W
= "there is a macroscopic (gross) hematuria"; 1
W 2
= nW . 1 The token in place 1 2
splits and gets into places
with the appropriate characteristics:
results of the test
glases and X-ray examinations of the urinary tract,
1 4
and
1 5
with three
because the tran
sition condition is 1 4 r = 1 | true 2 2I Place 1 6 stration
1 5
1 6
true
true .
is included in the framework of this
GN
as an illu-
of one of the methods for constructing a general GN.
presents the relation between the GN from § 4. 1. 1 and
It re
the net presen
ted here. In the framework of the general GN, tokens will transfer he re from place 1 of the GN from § 4. 1. 1 and this corresponds to trans4 ferring from one net to another. 1 9 9 r = 1 | I W 4 4 1 3 4 4 1 3
1 10 10
1
W
W w 4 4
11 11 5 5
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
233
234-
CHAPTER 4
where W
= "the presence of initial hematuria"; 3
W
= "the presence of terminal hematuria"; 4 - "the presence of total hematuria".
W 5
When the token gets into places 1 9 directed to a urologist
r
= 5
(1
and/or
1
the patient is 10
). 19 1 12 12
1
1 5
I true 1
W
1 6
I true 1
W
1
13 13 6
14 14 W 7 W
6
7
1 I true 7 |
W 6
W 7
1 l true 15 I
W 6
W 7
where W
= "there are no contraindications for intravenous urography"; 6
W 7
= ~W . 6
Thus this examination is performed, Transition conditions
r , r , r , r ,r ,r ,r 7 a 9 10 12 14 15
have the forms 1 19 r
= 7
1 | true ; 9 I 1 I true 10 I
and
r 16
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
11 21 21
235
11 20 20
11 II W W 12 II 99 12
W W
rr = - 11 II W W 88 113 3 II 99
W W
11 II W W 114 4 II 99
W W
10 10 10 10
10 10
where W
= "there are
no urographically
detectable changes or
the results
9 are uncertain"; = IV
W
10
9
(in this case the patient is directed to a urologist); 11 222 2
11 23 23
rr = = 11 II W W 99 220 0 II 11 11
W W 12 12
where - "there are no cystoscopic changes
W
or the diagnosis
is uncerta-
11 in";
. -m ;
w 12
11
rr 110 0
11 224 4
11 25 25
11 26 26
= = 11 II W W 222 2 II 113 3
W W 114 4
W W
11 II W W 13 333 3 II 13
W W 1144
W W
15 15 15 15
where W
& IV
= 1W 13
14
15
(the token gets
into place 1 - the patient 24
is left under
lance); W = " t h e r e i s a n e e d f o r complementary X-ray 14 W = "renal biopsy i s 15
needed"
examinations";
surveil-
236
CHAPTER 4
(the token in place 1 usually gets 26 agnosis); 1 29 29 r
= 1 I W 25 I 16 25 I 16
12 12
as characteristic a definite di-
1 30 30
1 31 31
W
W 18 18
17 17
where - "an extended examination of the
W
renal parenchyma is needed";
16 W
= "an examination of the bladder is needed"; 17
W
= "an extended examination of the
vessels is needed" ;
18
r 14 14
1 34 34
1 35 35
1 36 36
= 1 I W 29 19 29 I I 19
W 20 20
W 21 21
where W
= "an arteriography is required"; 19
W
= "an sonography is required"; 20
W
= "a scanning is required"; 21
r 15 15
1 37 37
1 38 38
1 39 39
= 1 I W 30 I 22 30 I 22
W 23 23
W 24 24
where W
= "a retrograde cystoscopy is necessary"; 22
W
= "a retrograde ureteropyelography is necessary"; 23
W
= "an anterograde pyelography is necessary"; 24
r 16
1 40
1 41
= 1 | W 31 I 25
W 26
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
237
where W
= "an investigation of the renal arterial system is indispensable" 25
(aortography, selective renovasography), - "an investigation of the renal venous system is indispensable"
W 26
(cavography, renal phiebography). The section of the GN described above ends by a transition with condition 1 42
1
| 11 I
W
W
1 I 34 I
W
1
I 35 I
W
I 36 I I| 1 | 37 I
W
1
1
r
= 17
1
W
( 39 I
W
1 I 40 I
W
1 41
W
1
I
27
28
27
W 28
27
W 28
27
W 28
W
I 38 |
43
W 27
28
27
W 28 W
27
28 W
27
28 W
27
28
where W
= "urologic cause of hematuria is found" 27
(the patient is directed to a urologist - 1 ); 44 W
= "a nephrologic cause of a hematuria is found". 28
(the patient is directed to nephrologist - 1 ). 45
238
CHAPTER 4
Transition conditions r and r have the forms 9 13 1 27 r = 1 I W 9 23 I 29
1 28 W 30
where W
= "bleeding from the lower urinary tract is discovered" 29
(and the token gets into place 1 ); 44 W
= "bleeding from the upper urinary tract is discovered"; 30 1 32 r
1 33
= 1 I W 13 28 I 31
W 32
where W
= "the bleeding is unilateral", 31
(and the token gets into place 1 ), 44
w
= nw
32
31
(the toKen gets into place 1 and the condition r is verified). 33 10 Transition conditions r 3
and r have the forms 6 1 7 7
r 3
1 8 8
= 1 | W 3 I 33
W 34
where W
= "presence of hematuria combined with proteinuria
that cannot be
33 explained by the presence of blood in urine only"; - "presence of isolated hematuria";
W 34
1 1 1 15 16 r = 1 | W 6 8 I 35
W
1 17 W
36
18 W
37
38
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
239
where - "X-ray examination is needed"
W 35
(and a token's transfer to a 5-th transition follows); W
= "there are definite signs of nephropathy" 36
(the patient is directed to a nephrologist - 1 ); 45
W = "investigations of the concomitant proteinuria i s needed" 37 (the token leaves this GN and gets in a GN from § 4. 1. 1), W
= "investigation of the concomitant arterial hypertension is need38 ed"
(the token leaves this GN and gets into place L
of the basic GN E and 5
from there it goes to some of the GNs from § 4. 1. 7, § 4. 1. 8 or
§ 4. 1.
9). K K
K
§ 4. I. 3 Turbid urine Here some advantages of GNs and their applications
in the ini
tial steps of identification and investigation of a turbid
urine sam
ple will be illustrated Tokens (patients) enter place 1
with the initial characteris-
1 tic
"complains of
voiding turbid urine".
The tokens'
place 1* to place 1" shall be denoted by V -> 1". The transition condition of the GN are 1 12 2 r 1 I true r - 1 I true ; ; 1 1 1 1I I
transfer from
240
CHAPTER 4
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
1
1 2
r
= 1 2
3
I W 2 1 1
W 2
where
W
= "acidification of the urine specimen is performed"; 1
W
= "investigation of urine with test strip"; 2 11 55 rr 33
11 6
= = 11 II W W 33 11 33
W W 4 4
where W
= "the turbidity of urine disapears"; 3
W 4
= IV ; 3 11 77 rr = = 11 4 4
4
11 8
II W W 1 5
W W 6
4 1 5
6
where W = "there a r e no changes i n t h e t e s t 5
strip";
w = -m ; 6
5 1
1 9
1 5
I I
W 7
1 r
= 5
26 W 8
| true 6 1 I 1 I W 7 | 9
false
1
false
| true 8 I
W 10
where W - "the t r a n s f e r 7
1
-> 1 4
i s a c c o m p l i s h e d "; 8
24-1
242
W 8 W
CHAPTER 4
= nW ; 7 = "the transfer 1
9
3
-> 1 is accomplished "; 6
- -\V ;
W
10
9 1 lO r 6
1 11
1 12
= 1 I W W W 9 I 11 12 13
where - "bacteriological investigation of urine is undertaken";
W 11 W
= "microscopic
urine sediment
examination is carried out";
12 - "macroscopic examination of urine is accomplished";
W 13
1 1 13 14 13 14 r = 1 I W W 7 10 14 15 7 10 I I 14 15 where W
= "there is bacteriuria"; 14 - iW
W
15
;
14 1 15
1 16
r = 1 I W S 8 11 I 16
W 17
where - "there is leucocyturia";
W 16 W 17
= iW ; 16 1 17
1 18
r = 1 I W W 9 12 18 19 9 12 I I 18 19 where
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
- "there are urethral filaments in the urine specimen",
W 18
- -W ;
W 19
IS 1 9
1 26
1 r
= 10
I W false 15 I 20 I| 1 I false W 16 I 21
1
I W 17 I 20 17 I 20
W 21 21
where W
= "the transfer 1 20
W
-> 1 11
= nW 21
is accomplished"; 15
; 20 1 9 r
= 11
1 I true 20 I
;
1 I true 21 I
r 12
1 25
1
= 1 I W 24 I 22
W
31 23
where W
= "there is no prostatic hypertrophy"; 22
W
= iW 23
; 22 1 19 1
| true 13 I r = I 13 1 | true 14 I 1 I true 25 I
;
243
244
CHAPTER 4
1 27 r
= 14
1 I true 26 I ; 1 | true IS I 1 22
r 15 15
1 23
= 1 I W W 19 19 I I 2424- 25 25
where W
= "there
are no
X-ray
alternations
in
the kidney
and urinary
24 tract"; W
= nW ; 25 24
r 16 16
1 ] 29
1 30
= 1 I W 23 23 I I 26 26
W 27 27
where W
= "X-ray alternations in the urinary tract are demonstrated" 26 & -W
;
27 W
= "X-ray alternations in the Kidney are present". 27 The transfer 1 -> 1 means that it is indispensable to inves1 2
tigate a fresh urine specimen.
The transfer 1 -> 1 means that the 5 26
turbidity of urine disapears and thus the investigation is discontinu ed. The transfers 1 - > 1 , 1 - > 1 and 1 -> 1 impose an extension 5 9 7 9 8 9 of investigations. The transfers 1 15
-> 1 20
and 1 17
taneously performed, which means that leucocyturia ments are found. The transfers also
-> 1
are simil-
20 and urethral fila
1 -> 1 and 1 -> 1 , which are 16 21 17 21 simultaneously performed, signify the absent leucocyturia, but
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
245
there are urethral filaments. The token in place 1 takes ascharacte24 ristic the result of the investigation of the prostatic gland, place
1 , the characteristics 19
of the results from
and in
the intravenouse
urography. The transfers 1 -> 1 and 1 -> 1 mean the end of the 26 27 18 27 investigation; 1 -> 1 means that the patient is directed to a spe23 29 cialised nepbrologic unit; 1 -> 1 30 32 patient is
and
1 -> 1 means that the 31 32
directed to urological unit; 1 -> 1 means that the pa22 28
tient remains under observation after appropriate therapy.
• § 4. 1.4 Elevated blood urea Estimation of importance
»
nitrogen
the Glomerular Filtration Rate (GFR) is of great
both in the initial evaluation and in subsequent
low-up of patients with renal disease.
the fol
Measurement of blood urea nit
rogen (BUN} is probably the most commonly used, albeit the least accu rate, clinical method for estimating GFR. The serum creatinine concen tration provides a more reliable estimate of GFR. In place 1 we have a token (patient) with the initial charac1 teristic
"elevated BON". Next it splits and gets in places 1 , 1 and 2 3
1 , as the transition condition r has the form 4 1 1 2 - 1
r 1 Each
place
namics
I true 1 I
1 3 true
1 4 true
gets an appropriate characteristics
investigation"
(1 ); "state 2
"results of barody
of bydratation - evaluation
of
24-6
CHAPTER 4
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
247
extracellular volume" (1 ), and "urinalysis" (1 ). 3 4 The transition condition 1 5 5 r 2
1 6 6
= 1 I W 2 1 1
W 2
where W
= "there is a reduction in effective arterial blood volume"; 1
W 2
= IV . 1 The t r a n s i t i o n c o n d i t i o n
r r
1 7
11 88
= 11 II W W 3 33 11 33
W W 44
1 9 W W , 5
where W
= "there are symptoms of hyperhydratation"; 3
W = "normal findings"; 4 W
= "dehydration is present". 5 The transition condition
r 4 4
1 10
1 11
1
= 1 I true 4 4 I I
true
true
1 I true 18 |
true
true
12
From place 1 the token splits and gets into places 1 (deter4 10 mination of urea in the urine), protein), 1 (examination 12
1 11
(evaluation for
the presence of
of urine sediment for broad waxy casts).
The transition conditions
r , 5
r
and 6
r
have the following 7
248
CHAPTER 4
forms: 1 13 13 r
= 1 I W 5 10 I 6
1 14 14 W 7
where W
- "there is an increase of urea excretion in urine"; 6
W
= "there is a decrease of urine urea" ; 7 1 15 15 r = 1 I W 11 I 8 6
1 16 16 W 9
where - "proteinuria is present" (see § 4. 1. 3);
W
a w = nw . 9
8 1 17 17
1 ia ia
r = 1 I W 7 12 10 7 12 I I 10
W 11 11
where W
= "broad casts are present"; lO = lW
W
11
10
and this investigation may be repeated (see § 4. 1. 3). Ihe tokens from
places 1 , 1 and 1 5
9
get into place 1 13
(the-
19
re are signs of functional, reversible renal failure). The transition conditions
r
and r 8
have the following forms; 9
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
249
1 19 1 6 r
I I I | I
= 8
1 9 1
I 13 I
true true true
1 20 r
= 1 9
I W 19 I 12
1 21 W 13
where W
= "BUN is normalized after an appropriate therapy"; 12
.
= IV
W 13
12 The tokens from places 1 , 1 , 1 , 1 ,1 ,1 6 7 8 1415 17
and
1
get 21
into place 1 because organic renal failure is suspected. 22 The transition conditions r
and r lO
have the following forms: 11
1 22 1 5
r
I | I
true
1 I 21 I
true
1 7
true
= 10
1 8 1
l I I I I
true
| 14 I
true
1 | 15 I
true
1
true
| 17 I
250
CHAPTER 4
1 23 r
1 1 2* 24 25
= 1 I true 22 I
11
true
true
Tokens from place 1 mist be submitted to further investigati22 ons: blood (1 ), ultrasonographic (1 ), and radiologic (1 ) evalua23 24 25 tions of kidneys. The transition conditions r
, r , r and r have the follo12 13 14 15
wing forms'.
r 12
1 26 26
1 27 27
= 1 I true 23 I
true
1 1 1 32 33 32 33 r 13 13
= 1 I W 24 I 14 24 I 14
W
1 I W 25 I 14 25 I 14
W
1 34 34
35 35
16 16
W 17 17
16 16
W 17 17
W 15 15 W 15 15
where W
= "the size of both kidneys is normal"; 14 - "presence of bilateral small kidneys";
W 15 W
= "the kidneys are enlarged"; 16
W
= "there is enlargement of the pyelocalicial system"; 17
and
urologist must be consulted because obstructive
pected;
r 14 where W
= "anaemia is present"; 18
1 28
1 29
= 1 I W 26 I 18 Id
W 19
uropathy is sus
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
w
- -m
19
251
;
18 1 30 r 15
1 31
= 1 I W 20 27 I
W 21
where W
= "there is hypocalcemia"; 20
w
- -m ;
21
20 1 36 1 I 28 I r
=
1
true
1 30 I
true
1 I 33 I
true
1 I 34- II 34
true
16
1 37 1 I 29 I r
=
1
1 31 I
true
1 | to I i 32
true
17
Tokens from places 1 , 1 , 26 30 with the
true
1
, and 1 33
get into place 34-
characteristics of chronic Kidney failure.
1 and 1 31 32
the tokens get into place 1 37
1 36
From places 1 , 29
with the characteristic of
acute kidney failure. The above constructed GN describes the a patient presents high level of BUN. K K
It
diagnostic process when
252
CHAPTER 4
§ 4. i. 5 Sena]
colic
The diagnosis of renal
colic is modelled in
the framework of
this GN. The tokens (patients) get into place 1 with the initial chai racteristic "symptoms of renal colic". Then they split up and get into the sections with initial places 1 and 1 and the appropriate charac2 3 teristics - investigation symptomatic
treatment
of
urine specimen
(the subside
and the
of renal colic),
result of
the
because of the
transition condition 1 2 r =1 1 1 1 1 Depending on the values of
1
true
3
true
these characteristics they
continue their
way in the GN. The transition :ondition
r 2
1 4
1 5
= 1 1 2 I
true
true
true .
1 I 31 1
true
true
true
The tokens in places 1 and 1 split 2 31
1
6
and
subsequently
places 1 , 1 and 1 , with the corresponding 4 5 6 the
their
characteristics:
They advance further depending on the values of the transition
conditions r , r , r and r : 3 4 5 6
r -- 1 1 3 41
1 7
1
W
W 2
1
8
1
1 9 W
3
lO W
4
where 1
into
results of macroscopic, test-strip and cytobacteriologic examina-
tions.
W
get
= "the urine is without any microscopic change";
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
253
254-
W
CHAPTER 4
= "there are heiraturia and/or blood coagulums" 2
(the token enters the GN from § 4. 1. 2); W
= "the urine sample is turbid" 3
(the token enters the GN from § 4. 1. 3); W
: "a calculus is emitted"; 4
1 1 11 r = 1 I true 4 10 10 I 1 1 1 12 13
r
= 1 I W 5 5 1 5
W
1 14
W 6
15 W
7
8
where z "there are no changes";
W 5 W
= "there is proteinuria" 6
(the token enters the GN from § 4. 1. 1); - "there is blood in urine"
W 7
(the token enters the GN from § 4. 1. 2); W
= "nitrites are present";
a
1 116 16 1 I true 7 I r = 1 I true 6 12 I 1 I true 5 I The transfer of tokens into place 1
means that a cytobacteri-
16 ological
examination is performed,
of the urine's sediment (1
which is a parallel investigation
) and microbiology (1 ) because 17 18
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
1 17 17 r 7 7
= 1 | true 16 16 I I
255
1 18 18 true
The transition conditions r , r and r have the forms 8 9 10 1
1 19
r = 1 | W 8 17 I 9
1
1 21
20 W
W lO
1 22 W
11
23 W
12
13
where W
= "the urine sediment is normal"; 9
W
= "there is heroaturia" 10
(the token enters the GN from § 4. 1. 2); W
= "there is leucocyturia" 11
(the token enters the GN from § 4. 1. 3); W
= "there is crystaluria"; 12
W
= "there is bacteriuria"; 13 1 24 24r = 1 r I t true rue 9 9 2 23 3I
r r
1 1 25 2 5
1 26
I W 24 I 14
W 15
= 1
10
;
where W
= "there is no bacteriuria"; 14
- -m .
w 15
14 The token's characteristic in place
infection is present" and in place 1 , 26
1 25
is
"appropriate
"no urinary tract therapy must be
256
CHAPTER 4
undertaken". The tokens in
places
1 and 3
1 split and get 22
on parallel
ways. The transition conditions r
, r 11
, r and r have the forms 12 13 14
1 1 1 27 28 r
1 29
30
= 1 | true 11 3 I
true
true
true
1 I true 19 I
true
true
true
r 12 12
1 31
1 32
= 1 I W 27 I 16 27 I 16
W 17 17
where W
= "the transfer 16
W 17
1 -> 1 is previously acconplished"; 3 27
= iW ; 16
r 13
1 33
1 34
= 1 I W 28 I 18
W 19
where W
= "there is no decrease in diuresis"; 18
W 19
= iW 18
(the token enters the GN from § 4-. 1.4);
r 14
1 35
1 36
= 1 I W 29 I 20
W 21
where W
= "there is no abnormal increase in body temperature"; 18
w
= nw . 21
20
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
257
The transition condition
r 15
1 37
1 38
1 39
1 I true 22 I
true
true
= 1 I true 11 I
true
true
1 | true 32 32 I I
true
true
1 I true 30 I
true
true
Therefore, when a token gets into place 1 , 1 , 1 or 1 it 11 22 30 32 splits into three. The transition conditions
r
, r 16 17
the forms:
r 16 16
1 40
1 41
= 1 I W 37 37 I I 22 22
W 23 23
where W
- "the renal function is normal"; 22
w
- -m
23 22 (the token enters the GN from § 4. 1.4);
r 17 17
1 42
1 43
= 1 I W 38 38 I I 24 24
W 25 25
where W
= "the X-ray of the urinary tract is normal"; 24
W = nW ; 25 24
r 18 18 where
1 45
1 46
= 1 I W 39 39 I I 26 26
W 27 27
r 20
have
258
W
CHAPTER 4
= "the metabolic functional tests are normal"; 26 - iV
W
27
;
26 1 44 r 19
- 1 ; I true 43 I 1 47
r 20 When
tokens
= 1 I true 46 I
get into places 1 , 1 , 1 ,1 ,1 ,1 ,1 , 8 9 13 14 20 21 34
1 and 1 , they leave this GN and get to the GN from § 4. 1.4. 36 41 In places 1 , 1 , 1 , 1 and 1 they end their movement in 25 33 35 42 45 the net with the characteristic "normal findings", ces 1 , 1 and 1 end with the characteristic 26 44 47
and tokens in pla "directing towards
appropriate treatment". x it
§ 4. 1. 6
*
Dysui-ia In normal
the ability
micturition cycle,
the continence
phase depends on
of the bladder adapt itself to urine volume
and maintain
pressure in urethra. During micturition, the bladder contraction asso ciated with sphincter relaxation empties the bladder. difficulty or pain associated with voiding.
Dysuria denotes
It may result from a wide
variety of pathological conditions and it is generally due to a cervico-urethro-prostatic obstacle. When this it is not the case one should investigate for a functional obstruction of the striated sphincter and the cervix, or a bladder hypocontractility induced by denervation, in hibition or muscular insufficiency.
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
The tokens (patients) get into place 1 with 1 racteristic "dysuria". Next they split up and get and
1 4
259
the initial cha-
into places
so that in each place they get the characteristics:
1 ,1 2 3 "results
from the case history" (1 ), "general examination" (1 ) or "complemen2 3 tary laboratory investigations" (1 ), because 4 1 2 2 r 11
= 1 I true 11 I I
1 3 3
1 4. 4
true
true
The transition conditions r , r and r have the forms 2 5 4 11 55
rr
= = 22
11 66
i1 iI w W 2 | | 11 2
w W 22
11 33
W W
II W W 11 11 II 11 II ffalse alse 17 II 17
11 II ffalse alse 55 55 II
2 true true true true
where W
= "there is a certain disease that may cause dysuria"; 1
w = -m ; 2
1 1 1 9 9 rr 55
= 11 II W W = 66 11 33
11 11 11 110 0
W w 44
W W 5
where W
= "consultation with a specialist is needed"; 3
W
= "consultation with a neurologist is needed"; 4
260
CHAPTER 4
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
W
261
= "urodynamic investigation is needed"', 5 1 12 r = 1 | true 4 5 I 1 I true 9 I From places 1 and 1 5 9
following characteristic:
the tokens get
into place 1 with the 12
"a specialist's opinion is required"
urologist, gynecologist, etc. ). From there the
(e.g.
token gets back to 1 , 6
passing through place 1 or to 1 according to the condition 17 16 1 16
1 17
r = 1 I W 8 12 I 6
W 7
where W
= "a definite cause of dysuria is found", 6
w = nw . 7
6 The transition condition 1 13 r 9 9
= 1 I W 11 11 I I 8 8
1 1 14 15 W
W 9 9
10 10
where W
= "the urodynamic tests are normal"; 8
W
= "the urodynamic tests are pathologic"; 9
W
= "the urodynamic tests are inconclusive". lO When tokens reach place 1 , 15
they leave
the net with the cha-
racteristic "further medical observation is needed".
262
CHAPTER 4
The transition condition 1 IS 18 r 10
= 1 I lO I
1 19
W
W 12
11
1 I false 13 I
true
where - "a neurological condition
W
that explains
the micturition disor-
11 ders is present";
- -m .
w 12
11 From place 1 (through place 1«
1 ), 35
the token
leaves the net
with the final characteristic "the micturition disorder is symptomatic of
a neurological condition".
Depending on the need
consultation (the transition r
of
psychiatric
has a form: 15
r 15 15
1 20 20
1 21 21
= 1 I W 19 I 13 19 I 13
W 14 14
where W
= "the patient must be consulted by a psychiatrist"; 13
W
= nW 14
13
). The tokens get into place 1 or 1 . From place 1 , the 20 21 20
token gets into place 1 or 1 (the transition r has a form 22 23 10
r lfl 18
1 22
1 23
= 1 | W 20 | 15
W 16
where W
= "the micturition disorder is of functional or anorganic origin"; 15
w
= nw ). 16
15
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
263
The transition condition 1
1 24
r
=
l | 14 I
16
W 17
1 25
W
1
1 27
1 28
29
W
W 20
W 21
22
26 W
18
19
where
W
= "vesical hypercontractility is present"; 17
W
= "a clearcut
pathologic
vesico-urethral dysfunction is manifes-
18 ted"; W
= "vesico-sphincteric dyssynergism is present"; 19
W
r "cervical obstruction is detected"; 20
W
= "there is vesical hypocontractility"; 21
W
= " other findings are present". 22 The transition conition 11 330 0 rr 114 4
= = 11 II W W 16 II 223 16 3
11 31 31
11 32 32
W W W W 25 224 4 25
where W
= "there is a definite diagnosis"; 23
W
= "there are uro-gynecological disorders"; 24
w 25
= nw . 24 Thus, from place 1 the token gets into places 1 , or 16 31
or 1 depending on 30
the characteristics it has acquired
(there is or is not uro-gynecological disorder, or another
1 , 32
in place 1 »2 disease is
present). When a token splits and engenders tokens in places
1 and 2
1, 3
264
CHAPTER 4
and when at a given moment these newly engendered tokens get into pla ces 1
, 1
31
and 1
21
(i. e. there is an accumulation of data for a gi24
ven patient from different channels), these three tokens fuse into one in place
1
with the characteristic
"micturition disorders of uro-
33 gynecological origin", because of the transition condition 1 33 1 I 21 I r = I 19 1 I 31 I 1 I 24 I
true true true
The transition condition 1 34
-=
r 21
1 I 32 I
true
1 I 23 | I 1 I 24 I
true
1 I 25 I
true
true
Therefore, when tokens engendered by an initial one simultaneo usly get into places 1 , 1 , 1 32 in place 1 34
with
23
and 1 25
,
they fuse into one
token
26
the characteristic "micturition disorders of
the
isolated neurogenic bladder". From 1 the token leaves the net with the characteristic "non29 dysuric disease". The transition condition
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
265
1 35 1 I 30 I
true
1 I 10 18 I|
true
1 I 22 I r = | 25 1 I 33 I
true
1
Therefore,
true
I 34 I
true
1 I 27 I
true
1 I 28 I
true
1 I 54 I
true
from each of the places
and 1 , the token gets into 1 54 35
1 ,1 ,1 ,1 ,1 ,1 ,1 30 18 22 33 34 27 28
which symbolizes
the performance of
an appropriate treatment. The transition condition 1 7 r 3
- 1 l 4 I
true
1 8 true
The token from 1 splits into two new tokens which get into 1 (inves4 7 tigation for hyperazotemia) and 1
(urinalysis).
8 The transition condition
r 6
= 1 I 7 I
1 136 36 W 26
1 1 37 37 W 27
where W
= "elevated blood urea nitrogen is present"; 26
266
w
CHAPTER 4
= nw 27
26 From 1 the token gets into place 1 or 1 according 7 36 37
dition r 7
to con-
(hyperazotemia is present or is not). In the first case the
token leaves the net and gets into the GN from § 4. 1. 4. From 1 , the token splits into three which 0 to places: 1 ("two glasses" test), 38 ons, including for b. Koch) and 1 40
1 39
(bacteriologic investigati-
(micturition flow), because
1 38
1 39
1 40
r = - 1 I true 7 7 I
true
true
The transition conditions r
, r 11
r 11 11
get correspondingly
and r 12
= 1 | 38 38 I I
have the form 13
1 41
1 42
W 28 28
W 29 29
where W
= "turbid urine is found"; 28
= -m
w 29
28
1 1 43 r 12
= 1 I 39 I
W 30
1 1 44 W 31
where W
= "a positive bacteriologic finding is present"; 30
w 31
= nw 30
APPLICATIONS OF GENERALIZED NETS IN HEDICINE
1 45 45 r 13 13
s 1 | 40 I 40 I
W 32 32
267
1 46 46 W 33 33
where W
= "there is an increased urine flow"; 32 = iV
W 33
32 The increase of urine flow means polyuria, so the token in pla
ce 1 leaves the net with this characteristic. 45 From places 1 , 1 , 1 and 1 the tokens get into place 1 37 42 44 46 48 (combined excretory and mictional urography) because 1 48 1 | true 37 I r = 1 I true 20 42 I 1 I true 44 I 1 I true 46 I The transition conditions r , r , r and r have the form 23 24 25 22 1 49 49 r = 1 I W 22 48 22 48 I I 34 34
1 50 50
1 51 51
W
W 36 36
35 35
where W
= "the X-ray picture reveals distinct pathological finding", 34 - IV
W
35 W
,
34 = "the X-ray report is in conclusive".
36
268
CHAFFER 4
r
= 1 I 23 51 I
1 52
1
W
W 37
53 38
where W
= "suprapubic cysto-urethrography is performed"; 3?
W
= "retrograde urethrography is performed" ; 38
r 24
1 54
1
1 I 49 I
W
W 40
= - 1 I 50 I
W
1 I 52 I
W
1 I 53 I
W
39
W 39
40
39
W 40
39
W 40
where W
= "there is a definitive diagnosis"; 39 - iW
W 40
. 39
Finally, 1 47 r
= 17
1 I 41 I
true
1 I 43 I
true
K
It
It *
55
APPLICATIONS OF GENERALIZED NETS IN IVEDICINE
§ 4. 1. 7 Arterial
269
hypertension
The purpose of the diagnostic examination of a patient
is with
hypertension to determine both the severity of the disease and whether the hypertension is primary (essential) or secondary to another disor der. The results of this procedure
will determine the need for treat
ment as well as the type of treatment. The systemic arterial hypertension is usually astolic pressure
defined as a di-
(DBP) of 90 nnHg or higher recorded on
minations of a subject with normal salt intake
several exa
( 1 0 + 3 g/day) regard
less of age and sex. A token in place 1 means a patient 1
with arterial hypertension
(as initial characteristic). The transition conditions of the GN have the form 1 2
r 1
- 1
I W 1 1 1
1
1
1
3
4
W 2
W 3
1 5
W 4
6
W 5
where
W
= "recording the DBP under appropriate conditions"; 1
W
= "fundoscopic grading of retinal changes
(according Keith-Wagner-
2 Barker classification)"; W
= "each of left ventricular hypertrophy"; 3
W = "investigation for renal damage"; 4 r
- "evidence
of target organ damage or
presence
5 risk factor other than hypertension";
r 2 where
1 7
1 8
1
= 1 I W 2 1 6
W 7
W 8
9
of cardiovascular
2T0 270
CHAPTER 4
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
W
= "the DBP is within the limits 90-1O4 imHg"; 6
w
= "the DBP is within the limits 105-114 mnHg"; 7 = "the DBP is i 115 imHg";
W ft
1 10 10 r 3
= 1 I W 3 1 9
1 11 11
1
W
W w 11
12 12
10
where W
= "no retinal changes or such of KWB group 1"; 9
W
= "retinal changes KWB group 2"; lO
W
= "retinal changes KWB groups 3 and 4"; 11 1 13
1 1 14 15
r = 1 I W W 4| 12 13 4
W 14
where
- -m
W 12 W
& nW ;
13
14
= "paterns of left ventricular 'strain'"; 13
W
= "left ventricular hypertrophy is present"; 14 1 16 16 r 5
1 17 17
= 1 I W W 15 16 5 I
where W
= "no renal damage"; 15
W 16
= nW ; 15 1 1 1ft 18 1 19 9 r = 1 I W r 6 6 7 6 6 II 117
W 1ft 1ft
271
272
CHAPTER 4
where W
= "no evidence of other target organ damage
or other cardiovascu-
17 lar risk factors"; W
= nW 18
; 17
- 1
r 7
1 20
1
I W 10 19 10 I
W
21
20
where W
= "there is a token, related to the present token, at least in one 20 of the following places: 1 , 1 , 1 , 1 ,..., 1 , 1 "j 8 9 13 14 17 19
= -m ■,
w 19
20
r
= 1 8
1 22
1
I W 13 I 21
W
23 22
where W
= "there is a token, related to the present token, at least in one 22 of the following places: 1 , 1 , 1 8 9 = iV
W 21
,1 11
,1 ,1 "; 12 17 19
; 22
r
= 1 9
1 22
1
I W 16 23 16 I 23
W
25 24
where W
= "there is a token, related to the present token, at least in one 24 of the following places: 1 , 1 , 1 , 1 , 1 , 1 ■'; 8 9 13 14 15 19 = iV
W 23
; 24 1 26 r
= 1 I W 17 I 10 25
1 27 W 26
APPLICATIONS OF GENERALIZED NETS IN MDICTNE
273
where W
26
= "there is a token, related to the present token, at least in one
of the following places: 1 , i , i , i "; 9 12 16 19 W = iH ; 25 26 1 1 28 29 r
11
= 1 | W 27 19 1
W 28
where W
28
= "there is a token, related to the present token, at least in one
of the following places: 1 , l , 1 , 1 "; 9 12 15 17 W = nW . 27 28 The remaining transition conditions' predicates
(of the 12-th,
13-th and 14-th transitions) are only "true". K K
§ 4. i. 8 Arterial
hypertension
«
of renal
The tokens get into place 1 1
origin
with initial characteristics for-
med by three groups of data: - group X: "signs and symptoms
directing to renal origin of the arte-
rial hypertension" - group Y: "signs and symptoms directing to renovascular origin of the arterial hypertension" - group Z: primary
"there are no signs, symptoms or conditions, excluding the participation of kidneys in the
origin of the hypertensive
state". Hie transition condition
r
1 2
= 1 1 true 1 11
1
3
true
274-
(
CHAPTER 4
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
275
Thus the token splits into two, which get into place 1 (urine is sent 2 for analysis)
and place 1 3
(blood sanple is taken
for examination).
The token in place 1 splits into three new tokens that simultaneously 2 get into places 1 (biochemical examination of urine), 1 (urinary se4 5 diment examination) and sition condition r
1 6
(bacteriologic examination) because tran-
has the form 2 1 4
1 5
r = 1 I true 2 2 I i 2 2
1 6
true
true
The transition conditions r and r have similar form 4 5 1 10 r = 1 I true 4 4 I 1 1
12 12
r
= 1 I true 5 5I
when the token gets into places 1
true ;
1 1 1
1
13 13
true
14 14 true
, . . . , 1
10 taneously for proteinuria,
1 11
,
if means testing simul-
14
glucosuria,
hematuria,
leucocyturia
and
bacteriuria. The transition conditions r , r , r 6 lO 11
r = 1 I 6 6 1
1 15
1 16
W
W 1
where W
= "there is positive urine culture", 1
2
r
have the form 14
276
W
CHAPTER 4
= nW ; i
2
1 ) 24 24 r 10
= 1 I W 3 10 I
1 25 25 W 4
where W
= "there is protenuria"; 3
W
-- nW ; 3
4
1 26 r
= 1 I W 11 10 I 5 11 10 I 5
1 27 W 6 6
where - "diabetes mellitus is suspected";
W 5 W 6
= lW ; 5 1 28 28 r 12 12
= 1 I W 12 I 7 12 I 7
1 29 29 W 8 S
where W
= "there is hematuria"; 7
w = nw ; a
7 1 l 30 30 r 13
= 1 | W 13 I 9
l 1 31 31 W 10
where W
= "there are bacteria in urine sediment"; 9
W 10
= nW ; 9
APPLICATIONS OF GENERALIZED NETS IN MBICTNE
1 32 r
= 1 | W 14 1414 I 11
277
1 33 W 12
where W
= "there is bacteriuria"; 11
W 12
= iW ; 11 The transition condition
r
= 17
1 36
1
1 I 25 I 25 I
W 13 13
W 14 14
1 I 27 I | 1 | 29 I
W 13
W 14
W
1 I 31 31 I I
W
1 I 33 I 33 I
W
1 16 16 | |
W
37
W 13
14
13 13
W 14 14
13 13
W 14 14
13 13
W 14 14
where W
= "glomerular nephropathy is suspected"; 13
w
= nw . 14
13 The token in place 1 splits into three tokens which get into 3
places 1 (serum creatinine by urea determination), 1 (serum potassi7 8 urn), and 1 (blood sugar), because the transition condition r has the 9 3 form: 1 7 r 3
= 1 I true 3I
1 8 true
1 9 true
278
CHAPTER 4
The transition conditions r , r , r , r and r have the form 7 ft 9 10 17
r 7
1 17
1
= 1 I W 7 I 15
W
18 16
where W
= "there
is increase in serum creatinine level";
15
w
= nw 16
15 1 19 19 r = 1 I W ft 17 ft ft ft I I 17
1 20 20
1 21 21
W
W 19 19
18 18
where W
= "hyperpotassemia is present"; 17
W
= "nomaopotassemia is present"; 18 - "hypopotassemia is present";
W 19
r 9 9
1 22
1
= 1 | W 9 20 9 I I 20
W
23 21 21
where W
= "there is hyperglycemia"; 20
W 21
= iW ; 20 1 34r
= 1 | W 18 I 22 10
1 35 W 23
where W
= "there is a low glomerular filtration rate"; 22
w 23
= iw 22
;
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
279
1 39
Al1 tokens place
1 39
1 I 37 I
true
r - 1 | 17 35 I
true
1 I 21 I
true
1 I 23 I
true
pertaining to a specific patient 'when getting
into
fuse into one token with characteristics incorporating all
the previously
acquired characteristics by the tokens
the investigations carried out up to now. Place 1
resulting from
symbolizes "a test
39 for iodine sensitivity is performed". The transition conditions r 19
r 19
and r have the form 20 1 1 1 1 40 41
= 1 I W 2439 I 24
W 25
where W
= "the test for sensitivity is negative"; 24-
W 25
= nW ; 24
r 20 20
1 42
1 43
= 1 I W 40 40 I I 26 26
W 27 27
1 | false 41 I
true
where W
= "intravenous urography is performed"; 26
W
= "other 27
instrumental
investigations
of kidneys
are
necessary
280
CHAPTER 4
(isotopic, sonography, etc. )". The place
1
corresponds
to the following actions: isotopic
43 investigations (renography, scintigraphy,
gamma-camera >,
sonography,
computer tomography, and the place 1 to excretory urography. 42 The transition conditions r , r and r have the form 21 22 23 1 1 44 45 44 45 r = 1 I W 21 42 21 42 I I 28 28
W 29 29
where W 28
= nW ; 29 - "there are urographic data for renovascular hypertension";
W 29
1 46 46
1 47 47
r = 1 | W 22 43 22 43 I I 30 30
W 31 31
where W 30 W
= nW ; 31 = "there are signs of renovascular hypertension";
32 1 48 48
1 49 49
r = 1 I W 44 I 32 23
W 33
1 I W 46 I 32 46 I 32
W 33 33
where W 32 W
= nW ; 33 = "the initial
characteristics
of the token
has positive
siqns
33 from group Y". When tokens get into place
1 24
there is
a call
to
GN
from
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
§ 4. 1. 3, in place 1 28
281
- to GNs from § 4. 1. 2 and § 4. 1. 5, in places 1 15
and 1 to GNs from § 4. 1. 3 and § 4. 1. 5; in places 1 , 1 30 17 19 GN from § 4. 1.4; in place 1 - to GN from § 4. 1. 3; 32
and 1 34
in places 1 47
to and
1 - to GN for § 4. 1. 9. 49 The tokens in places 1 , 1 and 20 22
1 26
leave the GN because of
the indications for other diseases, and in places 1 and 1 they re38 48 ceive the characteristic: appropriate treatent and medical observation of the patient.
* X
§ 4. i. 9 Renovascular The tokens
*
hypertension (patients) get into place 1 from a GN from § 4. 1. 8 1
with the initial characteristics
"signs and symptoms directing to re-
novascular origin of the arterial hypertension", The transition condition r
has the form: 1
1 2 r
- 1 1 true 1 11
A token from place places
1
1 1
1
3
true
1 4 true
5 true
splits into four tokens
1 ,..., 1 . These places signify: 2 5
that get
into
1 - check-up for endocrine 2
ne disorder (primary hyperaldosteronism, pheochromocytama, hypercorticism etc. ), 1 - performing an iodine sensitivity test, 3 gation
of
plasma renin activity, 1
tests. The transition condition
1
4
- investi-
- performance of pharmacological
282
CHAPTER 4
1 6 r 2
1 7
= 1 I W 2 | 1
W 2
where W
= "an endocrine
disease as
a probable cause
of the
hypertensive
1 state is found", W 2
-- -W . 1 Place 1 is the final one for the net. 6 The transition conditions r , r , r , r , r 3 4 5 8 9
and
r 6
have the
forms
1 3
1 8«
1 9
II 11 rr = II 3 11 II 10 II 10
W W 33
W W 4
W W 3
W W 4
1 II 17 I 17 I
W W 3
W W 4
1 II 19 I 19 I
W W 3
W W 4
where W
= "pharmacological test gives negative result and it is possible to 3 perform a vasography";
w
= w 4
; 3
- "high
W
level of plasma renin activity is found";
5 W 6
= nW
; 5
1
1 12
r
= 1 5
where
I W 5 | 7
13 W 8
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
W
283
= "angiotensin test is performed"; 7
W
= "test with saralasin is performed";
a
1 1 17 17 r d 8
= 1 | W w 12 12 I I 9 9
1 1 18 18 W w 10 10
where - "the angiotensin test is positive";
W 9 W
10
= nW ; 9 1 19 19 r - 1 I W 13 I 11 9
1 20 20 W 12
where W
= "the saralasin test is positive"; 11
W 12
= nw ; 11 1 14 r = 1 I 6 7I
true
1 | 8I
true
The tokens from places 1 and 1 reach directly place 1 (per7 8 14 forroance of
vasography - abdominal aortography and/or selective renal
arteriography). The transition conditions r , r , r and r have the form 71 10 11 12 1U 11 1-
r 10 where
1 21 21
1 22 22
= 1 I W w 14 I 15
w 16
284
W
CHAPTER 4
= "there are signs of renal arterial stenosis"; 15
W 16
= nW ; 15
r 11 11
1 23
1 24
= 1 I W 21 I 17 21 I 17
W 18 18
1 | true 15 I 15 I
false
where W
= "there is
evidence that
the stenosis is
functionally signifi-
17 cant", = iW ;
W
18
17 1 25 r
= 1 | true 12 23 I
Places 1 , 1 25 26 tests to
26
1
1 27
1 28
29
true
true
true
true
1 indicate the performance of appropriate 29
determine whether
1 (Howard's test), 25
1
the stenosis has functional significance:
1 (Rapoport or Stanley test), 1 (separate he26 27
modynamic investigation of Kidneys), 1 (determination of 28 nin activity in renal
plasma re
veins separately), 1 (other investigations of 29 separate Kidney functions).
r 13
1 30
1 31
= 1 | W I 25 | 19
W 20
where W
= "the test of Howard is positive", 19
w 20
= -m ; 19
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
r 14 14
1 32
1 33
= 1 | W 26 26 I I 21 21
W 22 22
285
where W
= "the test of Rapoport or Stamey is positive"; 21
W 22
= nW ; 21 1 34 r
= 1 I W 15 27 I 23
1 35 W 24
where W
= "unilateral decrease of ERBF is found during
separate hemodyna-
23 mic isotopic investigation of kidneys"; w
- -m ■, 24 23 1 36 r 16 16
= - 1 I W 28 2ft I I 25 25
1 37 W 26 26
where w
= "augmented plasma renin activity is determined unilaterally"; 25
W 26
= nW ; 25
r 17
1 38
1 39
= 1 I W 29 I 27
W 28
where W
= "other functional 27 on"; 28
27
tests for
determining unilateral renal lesi-
286
CHAPTER 4
1 40
1 41
1 I 30 I 30 I
W 29 29
W 30 30
1 I 32 I r = I 18 1 I 34 I
W 29
W 30
W
W 29
30
1 I 36 36 I I
W 29 29
W 30 30
1 I 38 I 38 I
W 29 29
W 30 30
where - "a biopsy of the contralateral kidney is necessary";
W 29 W 30
= nW . 29 Place 1 signifies the performance of renal biopsy. 40
r 19 19
1 42
1 43
= 1 I W 40 40 I ! 31 31
W 32 32
where W
= "evidence for lesion of the contralateral kidney is present"; 31
W 32
= nW . 31 1 44 r 20
= 1 | true 43 I 1 I true 41 I
Place 1 signifies aa check-up for contraindications to surgi44 cal intervention. The transition condition
r 21
1 45
1 46
= 1 I W 44 I 33
W 34
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
287
288
CHAPTER 4
where W
= "there are contraindications"; 33
W 34-
= nW 33
(a surgical investigation (1 ) is possible). 45 From
places
1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 11 16 18 20 22 24 31 33 35 37
1 , 1 and 1 the tokens get into place 1 . Here, they acquire the 39 42 46 47 characteristic
"appropriate treatment and medical observation
of the
patient".
REFERENCES'. [1] F. Balag, G. Petranyi and F. Renyi-Vamos,
Nephrologia,
Medicina
Acute renal failure,
Saunders
Konyvkiado, Budapest, 1980 (in Hung. ). [2] B. Brenner and J. Lazarus (Eds. ), Co. , Philadelphia, 1983. [3] B. Brenner and F. Rector (Eds. ), The kidney, Vol. 1 and 2,
Saun
ders Co. , Philadelphia, 1986. [4] N. Bricker and M. Kirschenbaum (Eds. ), The kidney - diagnosis and menagement, Wiley Medical Publ. , New York, 1984. [5] J. A. Halsted and C. H
Halsted (Eds. ) The laboratory in clinical
medicine - interpretation and application,
Saunders Co. , Phila
delphia, 1981. [6] J. Hamburger, J. Crosnier and J. -P. Grunfeld (Eds. >, Nephrologie, Vol. 1 and 2, Flanmarion, Paris, 1979 (in French). [7] J. Jones, J. Briggs and T
Hargreare, Diagnosis and menagement of
renal and urinary diseases, Blackwell, Oxford, 1982. [8] S. Klahr and S. Massry, Contemporary Nephrology, Plenum Publ. Co. New York, 1981. [9] H
Kremling, W. Lutzeyer and R. Heintz,
Gynakologische
Urologie
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
289
und Nephrologie, Urban Schwarzenberg, Munchen, 1982 (in German). [10] M
Legran and J. -M Sue,
Abrege de nephrologie,
Masson,
Paris,
1981 (in French). [11] G. Majdrakov and . Popov (Eds. ) Gidney diseases, Sofia,
Medicina
i Fizkultura, 1980 (in Bulg. ). [12] F. Marsh (Ed. ) Postgraduate nephrology,
W. Heinemann Medical Bo
oks Ltd. , London, 1985. [13] S. Massry and R. GlassocK, Textbook of nephrology,
Vol. 1 and 2,
Williams and Wilkins, Baltimore, 1983. [14] E. Patev, Acute renal failure, Sofia, Medicina i Fizkultura, 1981 (in Bulg. ). [15] D. Vasilev, Vasorenal hypertension, Sofia, Medicina i Fizkultura, 1989 (in Bulg. ). [16] D. Vasilev and G. Stefanov
(Eds. ),
Clinical nephrology,
Sofia,
Medicina i Fizkultura, 1990 (in Bulg. ). [17] J. Sorsich and K. Atanassov.
Application of generalized
medicine (Acute Attack of Gouty Arthritis),
nets in
Proc. of Third Symp.
Int. Ing. Biomed. , Madrid, Oct. 1987, 643-64-5. [18] R. Bellman, Mathematical methods in medicine.
World Sci. Publ. ,
1983. [19] S. Kiforenko, Methods of mathematical biology, Kiev,
Vischa sko-
la, 1984 (in Russian). [20] J. Piemann, Modeling hospital information systems with Petri nets Methods Inf. Medicine, Vol. 27, No. 1, 1988, 17-22. [21] Mathematical methods
in medicine
and biology,
Proc. of
UNC of
Academy of Sciences of USSR, Sverdlovsk, 1986 (in Russian). [22] J. Sorsich,
An example of application of generalized nets in me
dicine. Proc of II Int Symp. "Automation
and Scientific
Instru
mentation", Varna, May 1983, 387-389. [23] J. Sorsich and K. Atanassov, medicine
Application of generalized nets
in
(Diagnostics of arterial hypertension of renal origin).
290
CHAPTER 4
Proc. of III Int. School
"Automation and Scientific Instrumenta
tion", Varna, 1984, 233-236. [24] J, Sorsich and K. Atanassov,
Application of generalized
nets in
medicine (Renal colic), Lect. Notes in Medical Inform 24, 1984, 352-355. [25] J. Sorsich and K. Atanassov, medicine
(Renovascular
Application of generalized nets
in
hypertension). Proc. of Third Int. Symp.
"Automation and Sci. Instrumentation", Varna, 1985, 167-169[26] J. Sorsich and K. Atanassov,
Application of generalized nets
in
medicine (Permanent Proteinuria), Proc. of National Sci. Session "Automation of Biotechnology Processes",
Sofia, Oct. 1985,
138-
141 (in Bulgarian). [27] J. Sorsich and K. Atanassov,
Application of generalized nets
medicine (Haematuria), Proc. of III International Symp.
in
"Automa
tion and Scientific Instrumentation", Varna, Oct. 1985, 163-166. [28] J. Sorsich and K. Atanassov,
Application of generalized nets
in
medicine (Dysuria), Proc. of III Symp. Int. Ing. Biomed. , Madrid, Oct. 1987, 639-642. [29] J. Sorsich, Application of the generalized nets in medicine (Ele vated blood used nitrogen),
First Sci. Session of
the Math. Fo
und. of AI Seminar, Sofia, Oct. 10, 1989, Preprint IM-MFAIS-7-89, Sofia, 1989, 57-59. [30] J. Sorsich,
Application of the generalized nets in medicine (Ar
terial hypertension), Ninetieth Session of
the Nat. Seminar
Informatics of the Union of Bulgarian Mathematicians
and
of
Fourth
Sci. Session of the Math. Found. AI Seminar, Sofia, Nov. 5, 1990, Preprint IM-MFAIS-5-90, Sofia, 1990, 37-39.
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
291
§4.2: MODELLING OF DIAGNOSTIC AND THERAPEUTIC PROCESSES IN MEDICINE BY GENERALIZED NETS Borjana Jordanova
and
In this article we describe
Joseph Sorsich
a general outline of
the approach
to a patient who needs medical advice. This net is more general than the ones lacks the details of the others, that most patients pass through
described in § 4. 1.
It
but it gives some idea of the stages in a medical department.
So the nets
described in § 4.1 are its concrete applications. In the
net presented it is possible to
see more
clearly the
ways of introducing waiting and performing times of medical examinati ons
and carrying out the
time may
appropriate investigations,
duration of the whole cycle of treatment.
as well as the
These time-parameters
be obtained by a characteristic function of
the net as presented
below. The GN models
of
the diagnostic process
and
the appropriate
treatment of other (non-nephrological) diseases may also be settled in this scheme. The token
(patient)
enters place
1
with the characteristic
1 "need to consult a physician"
(with some complaints or for prophylac
tic needs). The GN below described is a reduced GN from the class A ,A ,A ,A ,n ,n ,c,6 ,6 , n ,G ,b 3 4 6 7 A L 1 2 K K
I
(everywhere the transition type is "v", ties are
"
the places' and arcs' capaci
the places' and transitions' priorities are "0",
every token a: b(a) =
is 1
for
292
CHAPTER 4
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
Z
=
,
1
1
),
(1 , 2
14
1 ), 3
293
r >, 1
where 1
1 2
r
=
1
1
I i
1
l
I 14 I 14 I
W
3 W
l
2 W 1 1
W 2 2
where
W
= "the result of the medical examination is that the patient is hei althy";
w = nw . 2
1 The ransition Z
2 2
is
Z == <[1 <[1 ,, 11 ,, 11 ,,11 ,,11 )), , {1 {1 ,, 11 ,, 11 1, 1, rr > >,, 2 2 33 66 lO lO 12 12 220 0 44 55 66 2 where 1 1 4 4
rr
= = 22
11 55
11 6
11 II W W 33 | | 3 3
W W 4 4
W W
11 II W W 66 | | 3
W W 4 4
W W 55
11 II W W 110 0 II 33
W W 44
W W
11 II W W 112 2 I I 33
W W 4 4
W W 55
11 II W W 220 0 II 33
W W 44
W W 55
55
5 5
where W
= nW 3
W
4
& nW , 5
= "there is need for appropriate investigations"; 4
W
= "there is need to consult a specialist". 5
294
CHAPTER 4
The transition Z
is 3 3
Z = <(1 >, <(1 i, i, (1 (1 ,, 11 1, 1, rr >, 33 55 77 88 3 3 where 1 77 rr
= 3
11 55
1 8 8
Il W W 11 66
W W 77
where W
= "laboratory investigations are necessary"; 6
W
= "instrumental investigations
(e. g. X-ray, sonography,
7 are necessary". The transition Z is 4 Z = <[1 <[1 ,, 11 1, 1, II II ,, 11 ), ), rr >, >, 44 77 88 99 10 4 10 4 where 1 99 rr = 4
1 10 10
11 77
II W W 11 88
W W 9
1 88
II W W 11 88
W W 9 9
where W
= "abnormal findings are present"; 8
w = nw . 9
a The transition Z
is 5
Z : <(1 , 1 I, 5 4 9 where
[1 11
,1 12
,1 13
,1 14
), r >, 5
ECG etc. )
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
1 11 r
= 5 5
1
1
1 13
12
1 4 4
I W I 10 I 10
W
1 9
I W I 10
W
295
W 11 11
14 W
12 12 W
11
13 13 W
12
13
where W
= "patient needs appropriate treatment"; 10
W
= "patient needs some complementary investigations"; 11 - "patient is estimated to be healthy";
W 12
- nw
w 13
& -m
10
& iv
11
. 12
The transition Z
is 6
Z = <(1 }, 11 , 1 , 1 ), r >, 6 11 15 16 17 6 6 11 15 16 17 6 where 1 15 r = 6 6
1 I W 11 14 11 I I 14
1 16
1
W
W 15 15
17 16 16
where W
= "ambulatory treatment is suitable"; 14
W
= "treatment at home is reccnmeanded"; 15
W
= "hospitalization is necessary". 16 The transition Z
is 7
-.
Z 7 where
,1 15
,1 i, [1 , 1 , 1 !, r >, 16 17 18 19 20 7
296
CHAPTER 4-
1 18 1 I W 15 I 17 r = I 7 1 I W 16 | 17 1 I W 17 17 17 I I 17
1 19
1 20
W 18
W 19
18
W 19
18 18
W 19 19
W W
where W
= "patient is restored to health"; 17
W
= "patient is dead"; 18
. iw
w 19
17
& -m . 18
The tokens receives in the different places racteristics by the characteristic function $ = §
the following cha-
20 U , defined by: i=2
-> "healthy person"; 2
$
-> "duration of medical examination"; 3
$ -> "K"; 4 §
-> "Kinds of appropriate laboratory examinations, the duration to 5 carry out these examinations";
$
-> "Kind(s) of appropriate consultation)s), duration of these con6 sultation(s)";
$
-> "result(s) of the laboratory investigation)s)"; 7
$ -> "result(s) of the instrumental investigation)s)"; 8 §
-> "presence of abnormal findings"; 9
$
-> "*"; 10
$
-> "patient needs appropriate treatment"; 11
APPLICATIONS OF GENERALIZED NETS IN MEDICINE
$
-> "patient
needs
complementary investigation(s)
297
or consultati-
12 tion(s)"; $
-> "patient is estimated healthy after the consultation(s)"; 13
$ -> "further surveillance or control is needed"; 14 $
-> "ambulatory treatment"; 15
$
-> "treatment at home"; 16
$
-> "treatment in hospital"; 17
$
-> "the patient is cured with
end of treatment,
duration
of the
18 treatment"; $
-> "exitus letalis"; 19
20 The characteristic function determines some
time-parameters of
the model. For their calculation statistical data,
as well as results
from concrete observations, can be used.
Chapter
5:
GENERALIZED NETS AND COMPUTER S C I E N C E
Some applications of the generalized nets as a means for model ling of computers are introduced in § 5, 1, § 5. 2 models are related to
the various
and § 5. 5.
The
processes in the computer and they
are written by the postgratuate students of
the Institute for Micro
systems, Sofia. § 5.6 the
deserves special attention where a short
Program Package for Generalized Nets
scription corresponds
(PPGH)
to the current state of
description of
is given.
the work on
contains the basic directions of its further development.
298
This de PPGH
and
GENERALIZED SETS AND COMPUTER SCIENCE
299
§5.1: IMPLEMENTATION OF GENERALIZED NETS IN DISTRIBUTED OPERATING SYSTEH DEVELOPMENT NIKo Fetalivanov
§ 5. i. 1
Intro&ictian
A general view upon
the latest developments
in operating sys
tems snows that distributed processing is a subject of tion.
This interest may be explained,
growing atten
on the one band by the increa
sing demands for more computing power and, on the other by cements to hardware technology.
The advantage of distributed
ting systems has now become a fact, in computer science,
its enhan
after a long
particularly in the field
opera
period of evolution
of operating systems.
In the last several years that local area networks (LANs) and personal workstations have become so popular and widely problems associated with these arise
and
used and that serious
required solution.
Present
operating systems provide time-sharing, multiprocessor, graphics, net working, etc. shared
However,
computing and
the answer
to
demands for efficient
storage resources is
in distributed
use of
operating
systems. The design of such distributed systems has proved to be and difficult process, which sometimes takes years is caused
by
ning on them
a long
of hard work. This
the presence of multiple nodes with separate tasks run that share system resources and require synchronization,
control and load balancing.
A major consequence of the latter
is the
numerous set of states in which the system may appear. Its performance is therefore determined by the ability of the software to act tely in each
adequa
particular state, which itself is defined by the respec
tive algorithms. A general way to facilitate software
design is
to present the
algorithms, the corresponding system states and the transitions betwe en
them in
a graph scheme.
The implementation
of
Generalized Nets
300
CHAPTER 5
(GNs)
is particularly suitable for this purpose,
as they
provide
a
universal method for simulation of system behaviour. The purpose of implementation in
this article is
one specific area
to outline some aspects of
of distributed
GN
operating system
development - the design of distributed file system (DFS).
§ 5. 1, 2 Generalized The DFS is
Net model for a Distributed an element of
File
every distributed
System operating system,
which carries out functions of vital importance. It has to provide the availability and consistency of all files, to access,
giving distinct
control concurrent file
response times to client requests.
To meet
these requirements, a DFS manages all shared storage resources, placed at the nodes, maintaining all data integrity and security. One generalized presentation of shows several nodes in rage resources
a DFS is given in Fig. 5. 1, It
a distributed system,
attached to them,
seme of which have sto
while the others
do not have such.
All system services and hardware components that accomplish between
the nodes are presented as
the links
a cowminication medium,
carrying
out data transfer. This picture will make it easy to follow the client requests and server responses when the discussion
is
focused at the
GN model of the DFS. A common approach
to maintaining high availability
response time is the use of 'replicated' files. have
copies at multiple nodes across
their modification quite efficient.
does not
take
and
short
These are files that
the distributed
system
When
place frequently, this method is
However, attempts to modify a replicated file invoke
special synchronization mechanism, traffic
and
which
affects system performance.
causes
heavy communication
Ordinary files do
multiple copies and are processed in a simpler way,
not have
but their average
access time is usually higher. Normally
a
DFS allows
several operations
to be
carried out
GENERALIZED NETS AND COMPUTER SCIENCE
301
Fig. 5. 1 upon files, such as read, write, create, etc.
There is
a special way
in which all file modifications take place: they are either successfully, or the system is left unchanged. sociated with
this type
of action,
completed
A specific term is as
reflecting
its nature - 'atomic
transaction'. Taking into consideration the above features of been possible to form
a
Fig. 5. 2.
DFS, it has
GN model of the distributed system currently
designed at the Institute of Microsystems, Sofia. in
a
The places in the
net represent
This model is shown the states,
in which
client requests appear, while they are being serviced by the DFS.
The
302
CHAPTER 5
toKen stands for
a
particular client request at
transitions of these tokens between the places
a
are
given time.
The
determined by the
respective predicates. An important advantage of GNs
is
their ability to assign cer
tain characteristics to the tokens. They may be further used to analy se the events that have taken place in the DFS, their causes
and re
sults. Some definitions have to be made prior to their use in predica tes and token characteristics. Let every file 'f
in the system have a
function 'r(f)' defined for it, such that r(f) = 1, if 'f is an ordinary file, and r(f) > 1, if 'f
is a replicated one,
On this basis, the sets 'Fo' and 'Fr' of ordinary and replicated files in the DFS are defined as follows: Fo = ( f / r(f) -- 1 !, Fr = ( f / r(f) > 1 ). Every token 'a' tic
entering the net is assigned
a x - "
the initial characteris-
where 'clientnode'
which the respective request is serviced,
is the node
in
'f is the file identifier,
and 'o' is the operation to be applied on ' f . In places
1 and 1 the tokens are assigned new characteristic 3 5
expressed by the respective characteristic functions $ 3 $
and $ : 5
-> "<server_node>", 3 where 'servernode'
is a node in the system where the request
is to be serviced. $ -> "
'TIME' is the current time,
assigned by the
'6' is the period during
which the transaction is active, 'result' signifies whether the
GENERALIZED NETS AND COMPUTER SCIENCE
operation has or has not been successful: result € ( true, false ).
Fig. 5. 2. The t r a n s i t i o n c o n d i t i o n s are defined as f o l l o w s : 1 2 r
w*iere W ='f 1
E Fr";
W = "f € Fo"; 2
1
=
1
I W 1 1 1 I 1 I W 2 I 3
1 3 W 2 true
303
30*
W
CHAPTER 5
= oc a = "c(l , X ) < r(pr2 x )" 3 3 0 0
where
c(l, X)
being a function giving the number of tokens
in place
'1' with characteristic 'X';
rr -2
1 i 4 4 W W 44
1i
5 5 W W 5
1 II 33 || II 1 II ffalse alse 4 II
true true
where
4
a a = "x - pr x "; cu 10
5
= IW ; 4
W W
rr 3
::
1 6 6
11 77
1
1 II W W 55 || 66 II 11 II W W 6 1 9 6 1 9
W W 77
W W 8
W W 10 10
W W 11 11
88
where W
= "f e Fr"; 6
W 7
a a = "f e Fo" St "pr x i Tr" & "pr x = true" 3 cu 4 cu where 'Tr' being the maximum time allowed by
the system
which a transaction may be active, a a W > = W = = "f "f ee Fo" Fo" & & "pr "pr xx > Tr" Tr" vv "pr "pr xx = ff aa ll ss ee " " ;; 8 3 cu 4 cu 8 3 cu 4 cu = a a w = "c(l , x ) < r ( p r x )"; 9 6 0 2 0 = a a a a W = " c ( l , x ) = r ( p r X )" & "pr x <. Tr" & "pr x = true"; 10 6 0 20 3 cu 4cu W
11
= a a a = "c(l , x ) ; r ( p r X >" & "pr x
6 0
20
3cu
a > Tr" & "pr x
4cu
= false";
during
GENERALIZED NETS AND CCWUTER SCIENCE
r
l1 I 7| I 1 I fl II
= 4
1 99 wW 12
1
W
W
305
10 10 Ww 13
12
13
1 I false 9 I
true
where - 1W
W
;
12
13 a a W = "x = pr x ". 13 1 10 The presented GN model of a DFS does not describe all sible situations in ally during are made and
the pos
a full-scale system, which is rarely needed. Usu
the time when
a DFS is designed,
some restrictions are imposed.
particular assumptions This is done to prevent
the possibility of eventual indefinite states in the system, to avoid deadlocks and other undesired ten achieved after
effects.
a reasonable
A predictful behaviour is of
compromise between the complexity of
a model and the clarity of its operation. In this particular case some potential events are ignored, such as conmunication failures,
invalid client requests or
These situations are easily solvable,
access rights.
but the model would be unreaso
nably complicated and will not be adequate for its present purpose. Further research work in the field of DFS design will probably bring forth
other problems, and methods for
thor's belief is that a precise GN model of
their solution.
The au
any distributed file sys
tem could suggest helpful ideas about its design and development.
306
CHAPTER 5
§ 5. 2: GENERALIZED NET MODELS OF DATA TRANSFER IN INDUSTRIAL LOCAL AREA NETWORKS Alexander Savov
§ 5.2.1
Introduction This
paper by means of the Generalized Nets (GNs) presents the
results of a simulation study carried out by an Network (ILAN).
Industrial Local Area
The results are very close to the designed specifica
tions, The ILAN is composed of three layers: physical, data
link, and
application [1]. The data link layer
is built on a polling mechanism
because of its deterministic behavior.
All stations are polled either
centrally
by a supervisor or decentrally by a token
[£]. There are two types of stations:
passing protocol
initiators and responders. Ini
tiators provide the following data transfer services [3]: (a) SDA - send data with acknowledgement, (b) SDH - send data with no acknowledgement, (c) RDR - request data with reply. The initiators connected to the bus cooperate by using the sha red channel. The right to use the channel is named a token. A token is passed
from a station to a station in a cyclic way,
thus defining a
logical ring. The services SDA and RDR
are a couple of frames
sent via the
medium: the frame - request - passes from the initiator to the responder, the frame - answer - passes from the responder to the initiator. The service SDN has only one frame request passed from the ini tiator to the responder. The stations
included in the ILAN
are Programmable
Devices
(PDs) - components of manufacturing systems such as programmable logic controllers, robots, process controllers, personal hers.
computers, and ot
Their base functions are calculating the control
programme and
updating inputs and outputs. The control programme works cyclicly. The
GENERALIZED NETS AND COMPUTER SCIENCE
cycle is called
307
a Treatment Cycle (TC). Tie PD takes inputs -when the
TC starts. This is done in order to avoid hazards due to value changes during a cycle.
For the same reason data exchange is impossible
when
the TC runs. Two models of ILAN exist: (a) model 1 - only initiators are included in a logical ring, (b) model 2 - all stations are included in a logical ring. Model 1 behaviour 1. The initiator receives
the toKen and sends
the frame - request to
the responder, 2. The receipt of this frame causes
the responder to wait
the
TC to
end, to compose and transmit the frame- answer. 3. the receipt of this frame causes the initiator to pass the token to another initiator. Model 2 behaviour 1. The initiator receives the token,
sends the frame - request to the
responder, and passes the token to the next initiator. 2. The receipt
of this frame causes the responder to wait
for the TC
to end, to compose the frame-answer, and to wait for the token. 3. Tne responder receives the token and transmits the frame - answer.
§ 5. 2. 2 Generalized
net model
By means of GNs we shall describe
a GN-model
(see Fig. 5. 3)
of data transfer in ILAN. N is the number of stations included in the ILAN. I is the number of initiators included in the ILAN. Lmin is the minimum length of the data field
that can be tras-
ferred by one service. Lmax
is the maximm length
trasferred by one service.
of the field
of data
that can be
308
CHAPTER 5
Fig. 5. 3 CI1 is a TC in the initiator when it composes a frame-request. Cli is a TC in the initiator when it processes a frame-answer. CR
is a TC in the responder when it processes a
frame-request
and proposes a frame-answer. The GN-model consists of two parts. tions of
the ILAN and the second
The first models the func
(transition r , place 1 and token 5 9
GET'ERALIZED NETS AND COMTJTER SCIENCE
309
a) - the determined protocol on the data link level. The tokens p , . . . ,p 1
in the GN describe
the transfer and pro-
N
cessing of the data in the media and the station. The tokens in the first part of the GN have the following tial characteristics:
ini
<SA, DA, L, typo, where
(a) SA is the address of the station initiator, the source of the ser vice; (b) DA is the address of destination station - responder, the receiver of the service; (c) L is the length of the
data field
that is transferred during the
service; (d) type is the type of service (SDA, RDR, SON [2]). These tokens pass consequently through the places 1 , 1 , 1 , 1 , 1 , 1 , 1 6 3 6 4 6 5 7
1 , 1 , 1, 8 1 2
and 1 . 8
One service is modelled by passing a token through the net. The places model the following: 1 - setting up of the frame-request in the initiator; 1 1 - frame-request's buffer in the initiator; 2 1 - transmission of the frame-request and frame-answer in the media; 6 1 - processing 3
of the
frame-request and setup of the frame-answer's
buffer in the responder; 1 - frame-answer's buffer in the responder; 4 1 - no physical meaning; 5 17 - processing of the frame-answer in the initiator. The time intervals,
when the tokens are in
the different pla
ces, depict the processing of the data in the stations and the time of transmission in the media
These time intervals are a function of the
310
CHAPTER 5
length of the data-field and depend mostly on U ) the speed
of the microprocessors
and
the hardware in
the data-
transfer section (DMA or other), (b) the data transmiting speed in the media, (c) the frame organization and the data representation. The following equations are used for estimation of the duration Pf a tokens' stay a certain place: (a) for 1 : Tl - the time for setting up the frame-request 1 Tl = Kl+L. tl+RCIl (for the services SDA, SDN), Tl = Kl+R. CI1 (for the service RDR), where Kl - a constant, tl - a time constant depending on the speed of the hardware, R - a random number in the interval [0, 1],
depending on the speed and the method of
the media transmission, S - the number of the service characters in the frame, (c) for 1 : T3 - process time of 3
the frame-request and setting up of
the frame-answer in the responder T3 : K3+L. t3+R. CR;
GENERALIZED NETS AND COMPUTER SCIENCE
311
(d) for 1 : T7 - process time of the frame-answer in the initiator 7 T7 - Kl+L. ti+R. CI2 (for the service RDR) TY = Kl+R. CI2 (for the services SDA, SDN) (e) for 1 : Tfl - time for generation of a request, 6 T8 = R. t4, where t4 is a time constant. The token a in the second part of the net has
a characteristic
"a" which determines the number of the stations that have an access to the media, "a" is changed on every firing of the transition.
= 1 1 11
W 1
1 1 false 6 1
1
1
1 4
3 false
2
false W
W 3
5 false W
4
where
w
: "the time parameter of the token is <. the current time",
w
: "the token has been in place 1 before going to 2
1
2
16" &
"the time parameter of the token is <, the current time",
w = "the token has been in the place 1 before going to 3 3
16" &
312
CHAPTER 5
"the time parameter of the toKen is £ the current time", W
= "the token has been in the place 1 before going to 4 4
16" &
"the time parameter of the token is £ the current time"; 1 6 rr
= 11
1 77
II W W 2 || 5
false false
1 II Wl 3 II
false false
1 II ttrue rue 4 II
false false
1 || ffalse alse 5 II 5
true true
2
where W
= "a = SA for the token", 5 1 1 88 rr
= = 11 33
II W W 77 || 11 11 11
rr = = 11 II W W 44 88 || 11 11 99 rr
= = 11
5 5
9 9
II W W
1 1 1 1
The predicates in the transition conditions for model 2 are si milar to the above-mentioned ones with the following difference: 1 1 16 17 6 7 rr = ff aa ll ss ee = 11 II W W 22 22 11 55 I 1 I Wl false 3 I I 1 I
W
4 I 6 I 1 I false 55 I
false
true
GENERALIZED NETS AND COMPUTER SCIENCE
313
where W
= "a = DA for the token". 6 The characteristic
places
functions give
to the toKens
characteristics which are related to the
at different
time-stay of the to
Kens at these places. The models so sults
described
for the mean time of
are simulated.
a service,
There are certain re
the optimum traffic
and
the
length of the buffers in the different ILANs. REFERENCES: [1] ISO - Open System Interconnection (DIS-ISO 1*96) [2] ANSI/IEEE
Std.
802.4-1985.
Token passing bus access method and
physical layer specifications. [3] IEC - Process data highway - PROWAY - A, B, C.
314
CHAPTER 5
§ 5. 3: GENERALIZED HET - MODELS OF FFT AND TRANSFUIER SYSTEMS Antonia Dimitrova
Many signals cribed
and
Rossen Fetrov
and phenomena exist in nature which could he des-
and explained by the power analytical apparatus of
formation
the trans-
named after the notorious French mathematician Jean Batiste
Joseph Fourierr. By the Fourier analysis
an arbitrary spatial or time
function can he decomposed into sinusoidal constituents, each of which possesses its own amplitude, represents
frequency, and phase. The transformation
a transition from the time area to the frequency area;
it
is a basis for solving many research problems. In essence, the Fourier transformation performs a spectral analysis of the examined
values. A
basic goal of the implementation of the Fourier transformation is minimizing
the time for its computation. For this purpose a variant for
accelerated
transformation computation
was developed
by Cooley
and
Tukey in 1965, as Fast Fourier Transformation (FFT). By computation of the FFT upon
K discrete readings (reports, samples), the acceleration
which is obtained by the FFT implementation compared with the standard Fourier transformation is K/Log K, r
where
r is the radix of the FFT.
The radix of the FFT could be 2, 4, fl, 16, etc. variant
The most
widely used
is the one with radix 2. In this case the number of
cessed readings algorithm is
is an exact power of 2.
its implementation in m
the pro-
A typical feature of the FFT consequent phases,
Log K. The Fourier transformation is applied to 2
where
m =
both real and complex
input signals. The output signal is always complex. All the intermediate
results
set
up grave requirements on computational environments in which
is implement. of
the FFT
cells is
are complex values too.
That is why with
These peculiarities
the problem for
a maximum load
of parallel
topical and very important.
parallel working
A detailed
(features) FFT
implementation computational
analysis of the FFT
GENERALIZED NETS AND COMPUTER SCIENCE
315
algorithm shows nonlinear dependency of the algorithm
acceleration as
a function of
the number of parallel processors. The aim of the built
up
to
model
is
set up
the number of
parallel
working
which will accelerate the FFT algorithm in a maximum way. tational environment
an INKS IMS T80O
processes
As a compu
transputer network is chosen.
The number of transputers building up the 16. The basic features of these processors,
transputer network is 2 concerning
to
the built mo
del, are as follows: (a) four
serial communication channels for connection with processors
of the same type, (b) independency of the data transfer on the computational process; (c) built in device for floating point arithmetic with
10 1WLOPS per
formance; (d) tact frequency 20 Mfe. Let us have possibility to include tational network.
p transputers in the compu
In the common case, it is desirable for
power of 2 due to the
p
to be a
fact that the algorithm is implemented by using
radix 2 in order that a maximum and uniform load of
the computational
network is obtained. In this presentation we have a maximum of 16 com putational cells which
are connected under a rule each by each. Beca
use of the large period for the data transfer, conmensurable with computational
time,
a large number
of computational cells
the
does not
correspond to the great speed. This sets up the considered problem It has been examined that for the considered problem bet of cells to over
16 is not effective and this
increasing the numis
the reason why
such cases not to be considered here. Fig. 5.4 ve
represents a generalized net. Using it we shall
the problem about the number of the necessary computational
to achieve a maximum performance. The data volume is significant
sol cells when
this problem is discussed. That is why it is set up as an initial fea-
316
CHAPTER 5
ture of the token in place 1 . By changing this feature, a new initia1 lization of the network is necessary. In addition to the token in pla ce 1 , another token in place 1 exists and it has a "maximum time for 1 S problem execution" feature. This "maximum time" depends on the will of the operator. For instance, it can be taken from the given task,
e. g.
some boundary time. The first
token
represents a problem in its initial look: the
i FFT which is to be implemented on K points, over p transputers (p = 2 for 0 s i <, 4) maximum speed.
and it is required to determine which variant After the transition
Z
has the
this problem is split into
N
1 2 subproblems, where N - (p - 1), as on each of the subproblems a serial number
is set
in order that they
be distinguished from each
other.
Each subproblem is given a priority in so that the next token does not pass
the transition which has as the input both the next and previous
tokens.
Fig. 5. 4 The first token possesses the highest priority and the priority is
reduced consequently
to the last token.
The transition
by which
GENERALIZED NETS AND COMPUTER SCIENCE
317
such a situation can arise is Z . 4 In the
transition Z
the features of the tokens
in places 1
5
7
and 1 containing the total subproblem execution time are compared. As 8 a result, in the place 1 the token which has the least execution time 9 will be obtained and the problem will be solved at this moment.
where
Z = <[1 , 1 ), 1 1 3
(1 , 1 1, r , M , v ( l , 1 )>, 2 3 1 1 1 3 1 2
-
r 1
1 I false 1 I I 1 I true 2I
1 3 W l false
where W
= "The number of the firings up to the moment, n < N"; 1
Z =
= <{1 !, [1 ), r , M , v(l )>; 4 5 3 3 4
where the transition condition predicates of r
and r 1
Z = <(1 , 1 J, (1 , 1 ), r , M , v(l , 1 )> 4 5 6 6 7 4 4 5 6 where 1 6 6 1 I W r - 5 I 2 I 4 1 I false 6 I
1 7 7 W 3 true
where W
= the number of transputers is > 1", 2
= -m ■,
w 3
Z
2 : < i l , l ], [ 1 , 1 , 1 ], r , M , A(1 , 1 )> 5 7 8 8 9 10 5 5 7 8
are "true"; 2
318
CHAPTER 5
where 1 8 r
5
=
1 1 7 i I 1 I 80 I!
w
v W" false
1 9
1 10
w v V"
nw & ivi"
W" & -m'
W
where W -
"The l a s t c h a r a c t e r i s t i c of the token
in place 1 7
i s l e s s than
the last characteristic of the token in place 1 " 8 ' W" = "The transition firings are equal to 0", W " = "The transition firing number is N". All elements of M (1 i i i 5) are 1. i
For every token
a
the
following (consecutively) is valid: Ti (a) = (Log p+l, Log p K
2
Tokens enter with the
1).
2
initial characteristic
"Data volune, k"
and they receive the following ones: $ -> "current number of transition firings"; 3 $ -> "k. T ", where T is the one data item transfer time", 4 t t $
2 -> "k. T . ((Log k) - N + l)/(2. (N-l) )", where T 5 c 2 c r a t i o n computation time; 2
$
6
$
-> "k. (T
+ T )/(2x(N-l) t C
-> "k. T ";
7
t 2 -> " ( n - l ) ".
$ 9
);
is thetoasicope-
GENERALIZED BETS AND COMPUTER SCIENCE
319
§ 5. 4: IMPLEMENTATION OF GENERALIZED NETS FOR SIMULATION CF HQDE-TO-NQDE CQMMUNICATIOU IN A REGULAR STRUCTURE Stefan Stefanov
The case
of node-to-node
commmcation in
a massively linked
network of nodes is discussed. A structure
which
we shall
call
a massively linked
regular
graph (MLRG) is shown in Fig. 5. 5. Each node must have 8 bidirectional links. In case the hardware has only
6 or 1 bidirectional
cluster
like the ones shown in
links each node may Fig. 5.6.
be replaced
In that case
by a
we call
the
structure a massively linked regular cluster graph (MLRCG). Bearing in mind that a) the nodes are very close to one another in space, and, b) the main difficulties
in conmini cation
are
the conflicts between
the messages trying to pass through the same link at the same time, the distance (or the message price) between two nodes N
and N i
be defined as the number of links which the message in order to reach S
may j
passes through
starting from N . j i
The distance (message price) between N i
and N is J
P(N , N ) = max (lv(H ) - v(N ) I, lh(N ) - h(H ) I) , i j i j i j where h(N ) i
v(H ) i
is the relative vertical position of H
is the relative horizontal position.
in the graph and i
Note that this result
is
optimal (even better than could be expected). Let D = IE, W, H, S, NE, HW, SE, SW] be the set of the possible "world" directions. Let us substitute v(N ) - v(N ) = v i j
320
CHAPTER 5
h(N ) - h(N ) = h. i j Then P(v, h) = max(|v|,
|h| ).
The primary direction calculation mechanism (PDCM)
is given in
the table below, where /'"-", if x < 0 sgn(x) -)
"0", if x : 0 , I " + ", if x > 0
d(v, h) € D, "NONE" means that no direction is calculable. ( v , hh) ) I 1 1I sgn(v) sgn(v) I ssgn(h) g n ( h ) I dd(v, 11 0 0
1 00
1 None
II
I 1
00
I
--
IW W
||
1I
00
I
++
I EE
I1
1I
-
I
00
I NN
II
1I
-
I
--
I NW
II
1I
-
I
++
E I NNE
||
1I
++
I
00
I SS
II
1I
++
I
-
I SW SW
I|
1I
++
I
++
E I SSE
II
The secondary direction calculation mechanism type of A (SDCMAI is defined recursively: applicable if h / 0 and v ^ 0 and v / h (if (|v| > Ihl) then d(v, 0) d'A(v, h) -1 \ if (Iv| < |h|( then d(0, h) The secondary direction calculation mechanism type of B (SDGMB) is given by the table below which is based on the PDCM The "nearest possible direction" to d(v, h), where D, is defined as follows:
d'B(v, h) €
GENERALIZED NETS AND COMPUTER SCIENCE
321
I1 d(v, d ( v , h) h ) I d'B(v, d'B(V, h) h ) 1I
1 I
NE
I NNor orEE
I 1
E
I NE NEor orSE SE 1I
1 I
SE
I E or S
II
1 I
I 1
S
I SE or SW 1 I
1 I
SW
I S or W
I 1
W
I SW or NW I
1 I
NW
I WWor orNN
I 1
N
I NW NWor orNE NE 1I
I
1I
The tertiary direction calculation mechanism (TDCM) determines d"(v, h) € D - {d(v, h)I, i.e. , d"(v, h) is any member of D different from d(v, h). In the case of a MLRCG
each node is a cluster and the in-clus
ter directions are calculated similarly (using tables). Note that if the SDCMA is chosen, instead of the FDCM, the mes sage' s price
P does not increase, i. e. an optimal routing is achieved
if the FDCM or the SDCMA, is used. If the SDCMB is chosen instead of the FDCM then P
is increased
If the TDCM is chosen instead of the PDCM then
P
is increased
Tertiary directions are chosen in emergency cases
(hardware or
by 1.
by more than 1.
software malfunctions). A problem that Let
is solved by the simulation is the following.
C be the set of the vertices of the graph,
Cs c C be
the
set of message senders, Cd c C be the set of message destinations, nl - |Cs|, n2 = |Cd|. The time W ; W(C, nl, n2) during which all messages wait to re ach their destinations is a topic of importance.
322
CHAPTER 5
The generalized net (GN) representing the case is shown in Fig. 5. 7. The model
consists of 7 transitions
further minimized,
but this is not done to
and 11 places.
It can be
preserve clarity
and the
closeness to physics of the process. Tokens from the initial place L
pass through the trivial trani
sition
T
where they receive their initial location coordinates
cc €
1 Cs, destination coordinates
cd € Cd,
with priority equal to the time
when they enter the transition, Wc and Wg being zeroes and all the ot her elements of the characteristic vector blanks. Transition T
stays inactive until all nl tokens gather in 2
Passing through the trivial
T
a token receives
L, 1
its next partial de-
2 stination en e c,
calculated
using the PDCM Then it continues thro
ugh T and goes either to L (1 £ i £ K, and K is the number of no4 7, i des in the graph) if it is empty, or to L if otherwise. From L 8 7, i tokens go to
L
and from there either to L 9
if cc = cd,
or to
o
L
to 2
continue moving in the net. A token from
L
goes either to 8
tial destination, or, Passing
through
T
L
to retry to reach
its par-
5
if it has tried for more than a new partial destination
X
times, to L , 6
en € C
is calculated
3 using the SDCM. The simulation stops when all tokens gather
in L . o
is the maximum retry count. o * The time-parameters of the net are: T = o , t = 1 , t =oo. The transition conditions are
x
GENERALIZED NETS AND COMPUTER SCIENCE
323
L 1 r
= L
1
I
W
I I I
where W = "TIME > nl"; 1 L 3 r 2
= L I true 1 I L l true 2 I L 4
r 3
- L | true 6 I
L 7,1 r 4
L 7,2
...
L 7,N
L fl
- L I W 3 1 2
W 2
...
W 2
W
L I W 4 -| 1 22
W 2
...
W W 2 2 3
L I W 5 1 2
W 2
...
W 2
3
W
where
W = "C(L "c(L 2
, TIME) = 0" & "Cn = : cC(l (l 7, i
)", 7, i
where "i" i s the ni«l>er m«i4>er of the current place; W
= nW ;
3
2
L 5 r : L I W 5 8 I| 4 4 where W
4
= "W
C
W - iv ; 5 4
< X",
L 6 W 5
3
324
CHAPTER 5
Fig.
5. 7.
GENERALIZED NETS AND COMPUTER SCIENCE
325
L 7, 1 L r r = = 6 6
| true 7, 7, 1 1I I l I L | L | true true 7, 2 7, 2 I I
L
| true 7, N I L O
r =L I W 7 9 1 6
L 2 W 7
where W 6
; "C = C "; c d - nW .
W 7
6 Every toKen has only one (its current) characteristic "
en, Wc, Wg>", where (a) cc are the current coordinates; (b) cd are the destination coordinates; (c) en are the coordinates of the next partial destination; (d) Wg is the global wait time (retry count); (e) Wc is a scratch (current retry count). The characteristic functions related to the places are $
-> "
where (a) cc is the initial location (the message source); (b) cd is the destination; (c) en is undefined; (d) Wc is 0; (e) Dg is 0;
326
$
CHAPTER 5
-> "
where
"en" is the next partial destination calculated using the PDCM,
all the rest remaining unchanged; $ = "
"en"
is the next partial destination calculated using the SDCM
(A or B), all the rest remaining unchanged; $ = "
$
-> "*";
2 $ -> "*"; 7, i $ -> "»"; 8 $ -> "<Wg, TIME - nl>", O (this is the final token's characteristic;
TIME
is the current model
time). All the places and arcs have infinite capacities. All the tran sitions and places have equal priorities. All transition types are "v".
GENERALIZED HETS AND COMPUTER SCIENCE
327
§ 5. 5: GENERALIZED HET MODELLING OF DISK SUBSYSTEMS Bistra B. NiKolova
Disk Drives (DDs) and the subsystems one of
the most important components
built on their bases, are
of the computer.
They are used
for expansion of the RAH of the computer. The productivity of the com puter depends on them An example of a structure scheme shown
in
Fig. 5.8.
It consists of
and H numbers of DDs. The whole
of a disk subsystem
an input/output path
(DS) is (i/o path)
software and hardware, necessary for
the data exchange between the RAH and DDs is represented as a separate functional unit, placed between the RAH and DDs [1]. This unit is mar ked as an i/o path. part in
If a certain
the i/o path of
part of
the hardware,
which takes
the concrete subsystem, is occupied by a DD,
the other DD cannot exchange data. Data exchange is realized by direct memory access.
Fig. 5. 8. There are some kinds of architectures
in which in order to in
crease the parallelity in the functioning of the DS, new i/o paths are added. They are
called "subsystems with a dynamic distribution of the
control between a certain number of i/o paths". They consist of H num bers of DDs and connected DDs and
H
numbers of i/o paths,
to every i/o path.
2 i H i H [4]. Every DD is
The control of the i/o operations
with
their parts can be done with the help of as i/o path. The re
quests to an i/o path form a queue [2,5].
328
CHAPTER 5
Usually, the i/o path is used for but also added, for
of control information.
two buses
are formed
of data,
If a separate path for commands
for data and commands.
positioning at every moment of
the data bus
exchange, not only
the DD
is
The opportunity
is created,
nevertheless
and the i/o path are busy with the data exchange from/to
a DD. The result of an i/o operation is given in Fig. 5. 9, where A - the request goes into the DS (in the queue to the DDs) B - the i/o operation is started C - the positioning of the DDs is completed and the request goes
into
the queue of the i/o path D - the data exchange begins E - the request service ends and the request leaves the DS T - the time during which the request is in the queue to the DDs DQ T - position time DS T - the time, during which the request is in the queue to i/o path CQ T - data exchange time CS
Fig. 5. 9. On the other hand, an i/o operation in the DS consists of positioning, searching and exchanging data. The purpose of the article is to make a model of some DS archi-
GEI>JERALI2ED NETS AND COMPUTER SCIENCE
329
tecture in terms of Generalized Nets (GNs). Some specifications are made: 1) all DDs are equal and
the input requests
are equally
distributed
between them [2, 5]; 2) the fulfilment of the i/o
operations includes the steps from Fig.
5.9 [2,5]; 3) the queues to the DDs and the i/o path are realized by means of the FIFO [2, 3]; 4) every D has its own queue of requests,
while the queue to
the i/o
path is conmon; 5) DD positions the head over the wanted cylinder by itself; 6) the i/o operation starts from a free DD and i/o path; otherwise the request
moves either to the queue of the DD or to the queue of the
i/o path, according to the architecture of the DS; 7) the i/o path
takes
part in
the search for the block on the track
and in the exchange of information between the DDs and the i/o path [33. The requests in the DS are the tokens in the given model. token a goes in the net with an initial characteristic:
Each
"
a NB, LDBL, IOO>" (= x ), where NDS, NCYLN, NB, LDBL are natural numbers 0 and NDS is the number of the DD, 1 i NDS i MaxNDS, where MaxNDS is the ma ximum number of the DD, NCYLN is
the
number
of
the cylinder,
MaxNCYLN is the maximum NB is
the number
0 £ NCYLN <, MaxNCYLN,
where
number of the cylinder,
of the block
on the track,
1 4 NB i MaxNB,
where
MaxNB is the maximum number of blocks on the track, LDBL is
the length of the exchange
data
in
the blocks,
1 <, LDBL i
MaxLDBL, where MaxLDBL is the maximum length of the exchange data
330
CHAPTER 5
in the blocks, equal to
the information capacity of the DD in the
blocks, IOO is the code of the i/o operation, 100 =
the code of
where READ
the read-operation, WRITE is the code of the wri-
te-operation. A request
service has
been modelled
as a transfer of a token
through the GN described in Fig. 5. 10. The places in the net used for modelling of different architec tures are denoted as follows: b - models the preparation of requests, 0 bJ - distributes the requests to the corresponding one of the m-th DDs, 0 b
- a queue to the DD, 1
b
- the DD is free, 2
b' - the DD is not free, 2 b
- the i/o path is free, 3
b' - the i/o path is not free, 3
Fig. 5. 10.
GENERALIZED NETS AND COMPUTER SCIENCE
331
b" - the independent path of commands is free, 3 b"' - at least one of the N numbers of the i/o path is free, 3 b"" - none of the N numbers of the i/o path is free, 3 b
- start of an i/o operation, 4
b
- independent positioning of the DD over the desired cylinder, 5
b
- a queue of the i/o path, 6
b
- a free i/o path and search for a blocK, 7
b' - the i/o path is busy and search is not possible, 7 b" - all the N numbers of the i/o path are busy, 7 b
- exchange between DDs and RAM. 8 First, we shall discuss a DS which consists of a DD and
path,
as is shown in Fig. 5. 10. Each token
characteristic enters
the net.
an i/o
cc with the above initial
The request goes
in the queue of the
DDs. This queue is processed by means of a FIFO-discipline. The GN following
modelling
this architecture
components and
parameters.
is characterized
b 1 b I true 0 I I b I true 2I b
I true 3I b 2
r 1
= b I W 1 1 1
the
It contains 6 transitions with
conditions:
r = 0
by
b' 2 w 2
332
CHAPTER 5
where W 1
= "c(b , TIME) = 0"; 2
2
= 1W ; 1
W
(the function
c(l, TIME)
determines the number of tokens in place
at the time-moment TIME); b 3 3 r
=b 2
I W | 2 | 3
b' 3 3 W 4
where W 3
= "c(b , TIME) = 0"; 3
4
= nw ; 3
W
b 4 r
= b 3
I true 3I b 5
r = b | true 4 4I b 7 r
= b 5
| true 5I b 8
r
= b 6
I true 7I
The characteristic function is defined as follows: ^"<TIME, TIME - 6 (a)>", if a goes into b for the first time K 2 $
-> ■
2
a "<TIME, TIME - pr x >", otherwise V. 1 cu
1
GENERALIZED NETS AND COMPUTER SCIENCE
333
a where x is the current characteristic of token a; cu a $ -> "<TIME, TIME - pr x >"; 3 1 cu a a $ -> "h (pr x , pr x )", where the function h determines the ti4 1 2 cu 2 cu-1 1 me for waiting in place b ; 4 a a $ -> "h (pr x , pr x )", where the function h determines the time 5 2 20 4 0 2 for positioning; a a $ -> "h (pr x , pr x )", 7 3 20 4 0
where the function h determines the time 3
for searching for the blocK; a a $ -> "h (pr x , pr x )", 8 4 30 4 0
where the function h determines the time 4
for exchange between the DD and the RAM. Second, we shal1 discuss a DS which consists of an independent command path, an i/o path, and M DDs, as it is shown in Fig. 5. 11. The independent command path controls the search and exchange of in formation. Each toKen a with the above initial characteristic enters the net. We accept that in practice the independent command path is always free. The queue to the DDs and the i/o path, as above, are pro cessed by means of a FIFO-discipline. The GN modelling this architecture is characterized by compo nents, which are similar to the above ones. The transition condition r is 0 b b ... b 0, 1 0,2 0, M r =b I W 0 0 | 1 where
W
... 1
W 1
334
CHAPTER 5
GENERALIZED NETS AND COMPUTER SCIENCE
W
335
a = "the number of place is equal to pr x ". i 1 0 The transition conditions r
are i, i
r
= b 1, i b
b b
1, 1, i i | true 0, i |
| true 2, i I
for 1 i i i VL The transition conditions r
r , r ,. . . , r 2, 1 2, 2 2, M
are similar to
with exactly the same notation. 1 The transition conditions r
and r 3
have the forms: 4
b" 3 b r
= 3
| true 2, 1 I I b | true 2,2 I
b
I true 2, M I
b 4, 1 r = b" I W 4 3 1 1
b
. . . 4, 2
W
b 4, M
. . . 1
W 1
Noting that the capacity of place b" is 1. 3 The transition conditions r
r , r ,..., r are similar to 5,1 5,2 5, M
with exactly the same notation. 5 The transition conditions r
and r 6
have the forms: 7
336
CHAPTER 5
b 6 b r
= 6
I I true true 5, 11 II 5, I I bb II true true 5, 22 II 5,
b
I I true true 5,M 5,M II
b' b' 7
II true true II
b b 77 rr ss bb II W W 77 66 11 55
b' b' 77 W W 6 6
where W
= "c(b , TIME) = 0"; 5
W
7
= nW ; 6 5 The transition condition r
is the same as the one in the first
8 GN. The characteristic functions in both GNs coincide exactly in nota tion. Finally,
we shall discuss a DS which consists of DDs in number
M and i/o paths in number N, as it is shown on Fig. 5. 12.
The discip
line to the DD and the i/o path, as above, is FIFO. The differences between the second and the third net are in the forms of transitions conditions r , r and r . They have the forms: 0 3 7 b b .. .. .. b b b b 0, 0, 0, 0, 11 0, 22 0, M M r = b I W WJ ... W 00 0 1 11 1 1 bb"" "" II 33 1
W 11
1
W 1
...
W 1
GENERALIZED NETS AND COMPUTER SCIENCE
337
338
CHAPTER 5
b"' 3
bb"" "" 3
I 2, 1 I I b I 2, 22 II 2,
W
W 8
b
W
b r
= 3
7 W
W 77
I 2,M I
7
88
W 8
where W 7
= "c(b"', TIME) : 0"; 3
w = -m ; 8
7 b
b 7, 1
r
= b 7
I W 6 1 9
. . . 7, 2
W 9
b 7, N
b" 7
. . . W 9
W 10
where W
= "This i/o path is free"; 9
W
= "There is no free i/o path". 10 The transition conditions
r , r 8, 1 8, 2
r
are similar to 8, N
r with exactly the same notation. 8 The function $ is similar to the one from the second GN. The theory of the GNs makes it possible to create a cannon for mal description of DS. The concept GN allows
all phases of an
dependency between them when i/o requests be shown.
i/o operation
in the DS
and the
are processed to
It also allows all specialities of the separate architectu
res to be shown.
GENERALIZED NETS AND COMPUTER SCIENCE
339
REFERENCES: [1] J. Major, Processor, i/o path and DASD configuration capacity, IBM System Journal, Vol. 20, No.
1981. 63-85.
[2] I. Donev, A research over the productivity of disk subsystems, [3] D. Brown, R. EiQsen and C. Thorn, Channel and direct access device architecture, IBM System Journal, Vol. 1 No. 3, 1972. 186-199. [4-] R. Wilson,
Designers rescue
superminicomputers from
i/o bottle
neck, Computer Design, Vol. 26, No. 18, 1987, 61-71. [5] I. Donev,
A research over
the productivity
of disk
subsystems.
Proc. of Symp. "Electronic and computing technics and technology", May 1987, Sofia, (in Bulgarian). [6] I. Donev, L. Michov and N. Sinyagina,
A research over the organi
zation of highly-productive disk subsystems with stratification of information. Automation Control, Systems, Sofia (in Bulgarian).
Computing Technic
and Automated
340
CHAPTER 5
§ 5. 6: SOFTWARE TOOLS FOR GENERALIZED HETS Rumen Christov
§ 5.6.1
and
Stojan Hibov
Introduction During the last ten years concurrent processes
ve been developed.
Too many big
researches ha
and complex systems do not allow
an
analytic study. Petri Hets (PNs) and their extensions and modificati ons are a powerful theoretical tool for the study of the properties of complex concurrent systems. Too many software tools for implementation of the PNs and some of
their modifications
work of scientists easier.
are produced to make
the
These tools contribute to further develop
ment and extension of the practical applications
of PNs and their mo
difications [1], Each extention of
the FHs is
designed to solve problems in a
determined area. Howadays a trend to generalization and comparison mathematical
models
But there is no
of processes
from different areas
of
is observed.
compatibility between the FHs' extentions.
It
makes
the comparison of the processes' models very difficult. Generalized Hets
(GHs) were defined in 1982.
They are a po
werful mathematical tool for describing and studing the properties
of
complex concurrent systems. These nets are a continuation of the trend to make tools for modelling of parallel processes. GHs are an extension of the PNs. It is proved, that all exten sions
and modifications of -the PNs are particular cases of
fact proves
that each process
GHs. This
which may be described and studied
by
FHs or with their extentions may be described and studied by GHs. This means
that GNs
are a universal
language for describing and studying
the processes from different scientific and practical areas. This fact proves the neccesity
of software tools for
the im
plementation of GHs. They are developed by the group, working over GNs and the tools are named STGNs. The STGHs include:
GENERALIZED NETS AND CCMTJTER SCIENCE
34-1
(a) presentation of GN-based hierarchical models of processes; (b) computer tool support, with which it is possible to create and si mulate hierarchical GN; (c) tools for analysing results of a GN-based model simulation.
§ 5. 6. 2 GTfs - a short
description
GNs are a complex
mathematical object.
Many of
their compo
nents are presented in the other FNs extentions. However, the GNs are not an aggregate
of the components of other types of PNs.
All compo
nents of GNs have a specific content. GNs have a static and a dynamic structure, temporal components and a global memory. Static structure The
static structure includes the net transitions, net places
and arcs which connect the places with the transitions. The information necessary for each transition description is: (a) a transition
identifier,
which is
a unique name in
the net for
them; (b) a transition name (only for user applications); (c) priority towards other transitions in the GN; (d) temporal components of the transition: time of the next activation and duration of this active status; (e) type of the transition; (f) indexed matrix expressing the transition conditions; (g) a list of the input places for the transitions; (h) a list of the output places for the transitions. The transition type, condition matrix and the functions which assign temporal transition
components are included in the dynamic net
structure. The transition type contains the necessary conditions
for the
342
CHAPTER 5
transition activation.
In the GNs' theory
the transition type
has a
specific structure, but in software tools it may be an arbitrary logi cal expression, which may depend on arbitrary events in the net at the same or a certain previous step of the model simulation. The information necessary for each place description is (a) place identifier, which is a unique name in the net for than; (b) place name (only for user applications); (c) priority towards other places in the GN; (d) place capacity - maximum number of tokens which may be in the pla ce in one step from the net simulation; (e) tokens which
are in
the place
before
the GN-roodel
simulation
starts. A place may be (a) a GN input, if it is
a member only of
the input places' list for
any net transition; (b) a GN output, if it is a member only of the output places' list for any net transition; (c) internal to the net, if it is a member of for any net transition and is a member of
the input places'
list
the output places' list
for another (eventually the same) net transition. From
each input place for each transition in the GN,
each of its outputs are defined. For each arc a predicate from
A capacity
arcs
to
is assigned to each arc.
the transition condition matrix corres
ponds. Dynamic structure Components of the dynamic structure are tokens, types of tran sitions,
conditions of
the transitions and temporal functions of the
transitions. The information necessary for each token description is (a) a token identifier, which is a unique name in the net for them;
GENERALIZED NETS AND COffUTER SCIENCE
34-3
(b) a token name (only for user applications); (c) priority towards other tokens in the GN; (d) the time when the token enters the GN; (e) a list of the net input
places,
through which
a token may enter
the GN. At the moment
when the net starts to simulate, the tokens may
be in any place in the net or may be waiting to enter the net.
In the
beginning, the tokens may have a list of any initial characteristics. The token's way
through
the net depends
When the tokens are going in the net,
on
these characteristics.
to the lists of characteristics
new characteristics are appended The characteristics are the greatest difference between GNs and other kinds of PNs. dividuality
The tokens have an in
during the whole time of net simulation.
This fact means
that many of the properties of the concurrent processes may be descri bed
by characteristics.
helps models
the user
It simplifies
the static net structure and
to understand the model of the process.
are more compact
The GN-based
than models based on other kinds of PNs.
It
allows more detailed models of the studied processes to be made and to get more complex results. Token characteristics are useful but their software implementa tion is very difficult.
They may not be formalized in their full the
oretical definition. Software implementation is the implementation
of
a class of GNs. However, this class is enough to describe and to study real
and theoretical processes
and to use these software
tools for
practical applications. The characteristics nents of
have two aspects.
First,
they are compo
the net PSI-function. It is a definition of the way for cal
culating token characteristic values. Second, they are the list of va lues, which a certain token gains when it comes into
a certain place.
The following types of characteristic values are supported: - a whole number;
344
CHAPTER 5
- a real number; - logical values TRUE and FALSE; - a string; - a list, containing an arbitrary combination
of
values of
the above types; - a matrix of whole or real numbers. Conditions of transitions The conditions of
net transitions
are
an indexed matrix
of
predicates (see App. 2). The matrix elements are indexed by transition input and output place
identifiers.
Each predicate of these matrices
corresponds to one of the transition arcs. The values of
the predica
tes are calculated at each step of the model simulation for each token in the corresponding input place. Predicates may sions which return the values TRUE or FALSE. ons between all net attributes.
be arbitrary expres
They may include relati
A very important limitation
is
that
the predicates cannot depend on further events in the net. Type of transition The type of transition is a logical expression. It assigns the necessary set
of
tokens at the transition input
places for
regular
work of the GN. If the type of transition returns TRUE, then the tran sition has been activated. Global memory The global memory includes; (a) initial token characteristic values; (b) characteristic functions,
which act on
each token at
the places
corresponding to them; (c) the number of
the characteristic
token is in the net.
values which
are kept when the
GENERALIZED NETS AND COMPUTER SCIENCE
Wien the token is at a net's characteristic
values.
net. It allows the
output place,
345
it owns a list
of
The list corresponds to the token path in the
process to be studied by the history of its compo-
nents represented by tokens in the GN model. Hierarchical models based on GNs GNs allow one to create hierarchical models of the studied pro cesses. It is useful for their complex study. of subnets.
A GN may contain
a set
In the primary GN these subnets are represented by places
or transitions. If the subnet has only one input and one output, it is represented by a place. outputs,
If the subnet has
several inputs and several
it is represented by a transition.
If a token comes
into a
certain subnet input, then the subnet simulation starts. The simulati on of the primary net and its
subnets is synchronized by time towards
an absolute time scale. The subnets can use information for the events in the primary net and other subnets tion.
As a result the tokens gain
during the time of their simula
the result of the subnet action as
a characteristic. As a further development be
other PNs extentions,
it is provided
that
the subnets may
as Predicative/Transition Nets
or Coloured
Petri Nets for example or Neural Networks. This fact means: - models, based on different PNs'
extentions may be compa
red by their actions and results during the time of their simulation; - models, based model;
already built, may be included as parts of
- if a Neural Networks
based model
based model, then each Neural Network
is a subnet of the GNs
may be taught and it may define
the action of the primary net,
§ 5. 6. 3. Tools for <3V
a GN
representation
The software tools for implementation of GNs consist of:
346
CHATTER 5
- a language for description of GNs based models; - a graphic editor which allows to construct, edit and update GNs based models; - a tool for simulation of GNs based models; - a tools for analysis and representation of the results of the simulation of the model as histogranms, piecharts, etc. Language for GN description The models
based on GNs
are represented by language for
a GN
description (LGND). The LGND is designed and implemented specially for the description of the Its
reserved words
describing ment and
GNs. It
is a high-level programming language.
are notions from the GNs' theory.
the GN structure
Besides tools,
the LGND contains tools for
evaluation of predicates
the assign
and standard functions for all at
tributes of the net; current time, time, when a selected transition is active, number of tokens in the selected place and many others. Construction and editing of GNs The
editor in the STGNs allows the user to construct
and edit
complex hierarchical GNs. The user works with a GN on a computer, sup porting high resolution graphic screen and a mouse. The graphic repre sentation of the GN uses theoretically nents. of
Each editing operation
the computer.
defined symbols for its compo
is immediately reflected on the screen
The user can increase or decrease some parts of the
GN image, so that he can have a global view over the model, created by him The user has
an access to each net element.
attributes by a menu system He may move or delete a
He may change its selected element
or may observe the subnet, which this element represents, ment is a place or a transition.
if the ele
GENERALIZED NETS AND COMPUTER SCIENCE
The STGNs
recognizes the structure of GNs.
347
It makes
a syntax
(by the GN theory) analysis to the GN, created by the user and reports him about errors. GN simulator The STGNs includes a GN simulator. This is a tool for simulati on of the GN, described by the LGND. The simulation corresponds to the conmon algorithm Seme variances are allowed Certain tokens can split and join in certain net places.
If the splitting
is allowed,
token may go from one input to several output places, in the net as many new tokens are as
their
children place,
born with
when a
it splits and
the same characteristics
parent as the number of the output places is. When all the of one parent (tokens from one split level) are in a certain
they can join to become one token.
tics will not be a list.
The new token characteris
They are a net with nodes
characteristic of
different tokens at the same time moments. In a certain practical app lication, a c annum cat ion of different tokens is allowed. The user
follows the simulation on the screen step by step. He
can see the token's movement through the GN. The simulation can be manual or automatic. In manual simulation each GN step may be defined by the user. In automatic simulation, only the first step is defined by the user. the system
Each next step is generated by
(GN model). In both cases, results of the traversed
steps
are generated. The user can define the appearance
of
the results of
the GN
model simulation. He can use histogramms and tables for representation of the change in the characteristic selected by him for kens.
He can
for one
follow the relations between different
different to characteristics
or several tokens. He can follow which transition was not ac
tive in a selected time interval, kens and other features.
which place was not reached by
to
3*8
CHAPTER 5
§ 5. 6. 4.
Conclusions
The STGNs are a powerful tool the properties of
using
the GNs for
the study of
the practical and theoretical concurrent processes.
Ttieir software support and development has to be continued. Currently, the STGNs is used for description and simulation of
the GNs based mo
dels, described in the this book. The next step is to make these tools more user-friendly and to extend
the class of GNs,
which can be
described and simulated
with
them REFERENCE: [1 ] K. Jensen
and
R. Shapiro,
A tool package,
supporting to
use of
Colored Petri Nets, Petri Net Newsletter, Vol. 35 (1988), 22-36.
Appendix 1: REMARKS ON GENERALIZED NETS
In this
Appendix
we shall give
the basic definitions related
with the concept "generalized net" following the contents of
the bock
"Generalized Hets".
* a
§ 1. On the concept "generalized
«
net"
We shall give a formal definition of
the concepts
"transition
of a GET and "a GH". Every transition is described by a seven-tuple: Z =
and
L"
are finite, non-empty sets of places (the transition's
input and output places respectively); J
1.1 they are L'= (1 , 1'
1 b) t
2
for the transition in Fig.
1'J and L" = II", ]",...,]"].
1
m
2
n
is the current time-moment of the processes's firing; 1
c) t is the current value of the duration of its activity; 2 d) r is the transition's condition determining the
349
tokens which shall
350
APPENDIX 1
Fig. 1. 1. transfer from the
transition's inputs to
its outputs;
it has the
form of an index matrix (see App. 2): 1" . . . 1 1'
r
-
1
V i : 1' m
(i, j)
1" . . . 1" j n
I | I I I I I I I
denotes the element
r i,j (r
- predicates) i. j i,
( l i i i m l i j i n ) which corresponds to the
i-th input and
j-th output places; these elements are predicates and when the truth value of the
(i, j)-th
element is true,
the token from
place can be transferred to j-th output place; otherwise, possible;
i-th input it is not
REMARKS ON THE GENERALIZED NETS
351
e) M is an index matrix of the capacities of transition's arcs: I"
1
1' M
I' m f) D
.
.
1" . . .
j
1"
n
1 1 1
: 1' i
=
.
is an object having
1 m 1 i,j ; I 1 (m i 0 - natural lumbers) 1 i, j 1 | |1 U U 1 i j ( n| a form similar to a Boolean expression.
In
it the variables are all symbols which mark the transition's inputs nemes, and the Bollean operations "A" and
"v" determine the
fol-
lowing conditions: Ml
i
.1 1
,...,1 ) - every place 1 , 1 i i i i 2 u i 2
1 i
must
contain
u
at least one token, v(l
i
, 1 , . .., 1 i i 2 u
1
) - in all places 1 , 1 i i 1 2
1 i u
tain at least one token, where (1 , 1 , . . . , 1 i i i 1 2 u
must contain
) c L'.
The ordered four-tuple O
K
E = <,
is a function giving the
priorities of the transitions,
i. e. ,
TI : A -> N, where N = 10, 1, 2, . . . ) U too); A c) TI L
is a function
L -> N,
giving the
priorities of the places,
where L = pr A U pr A, and 1 2
i. e. , TT : L
pr X is the i-th projection of i
the n-diroensional set, where n € N, n i i and 1 <, k i n (obviously,
352
APPENDIX 1
L is the set of all GN-places); d) c is a function giving the capacities of the places, i. e. , c:L-> N; e) f is a function which calculates the truth values of the predicates of the transition's conditions
(for the GN described here, let the
function f have the value "false" or "true", i. e. , a value from the set
(0, 1).
In § 3. 1,
we shall
describe other
types of nets in
which this function will have a value in the interval [0, 1] or in the set [0, i ] x [0, 1 ] ); f) © is a function giving the 1
next time-moment
when a given transi-
* © (t) = t', where t, t' € [T, T + t ] 1
tion can be activated, i. e. ,
and t £ t'. The value of this function is
calculated at the moment
when the transition ends its functioning.
Below we shall discuss a
special representation of this function; g) © 2
is a function giving the
duration of the activity
of a
given
K
transition, i.e., © (t> = V , where t € [T, T + t ] and t' £ 0. The 2 value of this function is calculated at the moment when the transi tion starts its functioning.
It has
another representation as the
above one; h) K is the set of the GN's tokens. In some cases, it is convenient to consider this set in the form K = U I 1€Q where K
K , 1
is the set of tokens which enter the net from place 1, and 1
I Q i) TT K
is the set of all net's input places; is a function giving the
priorities of the tokens,
i. e. ,
TI : K
K -> N; j) 9 K
is a function giving the time-moment when a given token can en-
REMARKS CM THE GENERALIZED NETS
353
ter in the net, i. e. , 6 (a) = t, where K a 6 K,
*
t € [T, T + t ] ; K) T
is the time-moment when the GN starts functioning.
This moment
is determined about a fixed (global) time-scale; o 1) t is an elementary time-step, related to the fixed scale;
(global) time-
M
m) t
is a duration of the functioning of the net;
n) X
is the set of all initial characteristics
which can receive the
tokens when they enter the net; o) $
is a
characteristic function which gives new characteristic to
every token when it makes a given transition.
a transfer from input to output place of
As some of
the above functions
this function
can be represented in another form, shown below. p) b
is a function giving the maximum number of characteristics which
can receive a given token, i. e. , b: K -> N.
If for a certain token
a, b(a) = 1, the token will enter the net with an initial characte ristic (as a zero-characteristic). After this, it will receive only the characteristic of the previous. When b(a) = a>, the token a will receive all possible characteristics. When b(cc) = k < co, except its zero-characteristic, the token a will keep the last
k
as its cha
racteristics (previous characteristics will be "forgotten"). Hence, in general, every token a has b(a) + 1 characteristics. We must note that this definition similarly to
the ordinary ON
definition is not fully formalized, because if we fully formalize transition
conditions and the characteristic functions
the
of the GNs, a
smaller class of GNs will be obtained. It is convenient
to assume
that the functions
f, 6 , © i 2
and
354
APPENDIX 1
$ have the following forms, also: IAI f = U f , where f calculates the truth-values of i-1 i i the i-th
GN transition
the predicates of
(as we shall see in
§3.1,
they
cannot have values "true" and "false"; they can have fuzzy or intuitionistic fuzzy transition
conditions
(see can
App. 2)
be
values;
calculated
different
by
different
IAI i i 6 = U 9 , where 9 calculates the next time-moment of the 1 i= l 1 1
activati-
ways);
on of the i-th GN transition; IAI i i Q - U 0 , where 9 2 i=l 2 2
calculates the
duration of the active state
of
the i-th GN transition; ILI $ = U $ , where $ calculates the i= l i i
tokens'
characteristics,
which
they will receive in the i-th GN place. A given GN may not have some of the components. Then the places of such components will be marked by "*". The GNs with such form gene rate a special class of GNs, which we shall describe in § 2.1, The static part of a given GN is determined by the set pr
{i
i i
A, where for a given n-dimensional set X (n £ 2) i, 2, 6, 7 K pr X - n pr X i ,i i j=l i 1 2 k j
in, l i j i k , j
of
the elements of
i i- i JJ J"
for j' t j"),
a GN is determined by input and output
i.e. the static part
transition places,
dex matrix of the arcs and by the type of the transition. cal character of the net comes from the GN's tokens and ons conditions
(pr A), 5
by in
The dynami the transiti
the temporal character comes from
the compo-
REMARKS ON THE GENERALIZED NETS o * nents T, t , t and from the elements of the set pr components $, X and
b
355
A. 3,4
Finally, the
play the part of the GN's memory.
The different functions are also related to these four GN com ponents: functions TT , n , c to the static structure; f, IT to the dyA L K namical elements; 9 , Q and 6 to the temporal components. 1 2 K One very important restriction for
the GNs
is the
following.
The transition condition predicates cannot be related to future events for the GN. X K
K
§ 2. Reduced GNs For two subsets £' and E"
of E - the class
of all GNs
define (see also definitions of the relations n, n , s, * E' t- £" iff the functioning and
x
in
let us § 5:
*
the results of the work of every ele
ment of E" can be described by some element of E'; E' i- E" iff "HE* H E"), E' H E" iff <E' i- E") & <£" i- E'), E' H E" MI
T(S' H
E")-
A given GN may not have some of the components. Then the places of such components will be marKed by "*". The GNs with such form gene rate a special class of GNs. Let us define the set of the GN's components o * (A, TI , IT , c, f, e , e , K, TT , e , T, t , t , x, $, b) u A L 1 2 K K (A / 1 £ i £ 7}, i where A = pr A (1 i i <■ 7), i. e. , A € I L \ L", t , t , r, M, ol i i i 1 2 Q
:
and l e t Y € Q.
and
356
APPENDIX 1
Y By E we shall denote the class of those GNs
which do not have
the Y-component. Y Y Obviously, £ i- £ for every Y €ft.Elements of the class £
we
shall call "reduced GNs". Evidently A A
£ because
A 1
=E
2
=E
there does not exist
component A)
K
=2=1*,
a GN without
graphical structure
and without tokens (the component K);
(the
without input and
output places of its transitions (components A
and A respective). 1 2 Y Hie class £ will be called "a 1-class of reduced GNs" for eve-
ry Y € ft. If Y , Y 1 2
e 2 for s i i,
Y
then
Y , Y 1 2 E
Y s will
s
be called an "s-class of reduced GNs", Y , Y ,..., Y 1 2 s We shall note also that, if £ * 0,
A , A , Kj, then I
and
Y € (A,
Y , Y ,. . . , Y , Y 1 2 s £ i- ft.
2 O A , A , A , A ,TT ,11 , c, 0 , 0 , n ,© , T, t 3 4 6 7 A L 1 2 K K
* ,t
, X, b
THEOREM 2. 2. 1: £
H £. x *
§ 3, Conservative
extensions
K
of the OVs
Eight different types of GN-extensions are defined and for each of them it is proved that it is a conservative extension. First type intuitionistic fuzzy GNs In the definition of component f
the concept
GN
it was noticed
that the
(the function which checks the truth values of the predi-
REMARKS ON THE GENERALIZED NETS
357
cates) can give different type values. In the first case they are ele ments of the set ("false", "true") (or (O, 1J); in the second - it has values
in the interval [O, 1] and then the GN which contains such
component is called a Fuzzy GN; in the third case - it has
f-
values
in the set [O, 1] x [O, 1] and the corresponding GN is called Intuitionistic fuzzy GN, The above two nets are denoted in brief by FGN1 IFGN1 respectively. FGNs" and
We shall call
"first type IFGNs"
these first types GNs
(for the concept
and
"first type
"intuitionistic fuzzy
set" see App. 3). It is obvious
that every FGN1 is a IFGNi. Every GN of this ty
pe has the form E = <i
IT
K
o * , e >,
where the elements of the set A (the IPGN's transitions) as the
are the same
GN's transitions. All other components, without the components
f and § are also the same. The function
$ now gives to every token as
current characteristic two values: the first coincides with characteristic in the sense of the GN's;
the token
the second value is an orde
red tuple with real numbers, every one of which is an
element of
the
set [0, 1 ]. They are equal to the truth values of the predicate of the transition condition between the place, from where starts the transfer of
the corresponding
token and the place,
where this transfer ends.
The function f calculates these two values of the corresponding predi cate r i, J ff(r (r where p(r
)
> = <jj(r <jj(r ), ), T(r T(r )>, )>, i,j j i, i iiJ iJ i,i,J J
i s t h e degree o f t r u t h of the p r e d i c a t e
i,j i s i t s degree of f a l s e ,
r
, i, j
and (cf.
App. 3)
p(r ) + T(r ) <, 1. i, j i, j
T(r
) i, j
358
APFEHDIX 1
Second type intuitionistic fuzzy GHs Here we sbal1 give a complete formal definition of
the concept
"second type intuitionistic fuzzy GHs" (IPGH2), In this net the tokens are some "quantities" which move inside the net. The values of the transition condition's predicates can
be in
tuitionistic fuzzing s, i.e. they can have degrees of truth and of fal seness (see App. 3). The IPGN2 has the form o « E = <,
(cf. § i)
matrix
H
the net's transitions
which have the
ordinary
and there exists only one difference: here the in
contains as elements real numbers - capacities
of the
transition's arcs. The functions in the
first component are similar to
those for
IFGN1 and they satisfy the same conditions. The essential
difference between IFGH2S and the other
the set K and the functions related to it. some "quantities", which have as (elements of the set tics. The function 8
X)
How the
initial characteristics some
and which do not receive
gives the time-moment when a
GHs
is
elements of K are "type"
other characteris given token
will
X
enter the net as in the ordinary GHs. The temporal components are also as in the other GH-types. The function $ has new meaning. How it is related to the places and they receive
by this
function characteristics (the quantities of
the tokens from the different times in the
corresponding places).
As
in GHs, it can be extended: it can also give other data for the model led process (e. g. the time-moments for
the entry of the
"quantities"
in the places. The definition of the second type FGH (FGH2) is analogous. For IFGH2 the following are valid:
REMARKS ON THE GENERALIZED NETS
THEOREM 3. 2. 1: For every
359
IFGN2 there exists a GN which represents it.
THEOREM 3, 3. 2: For every GN there exists a IFGN2 which represents it. Colour GNs They have: (a) toRens coloured in n different ways (1 i n <. <x>), (b) arcs coloured in R different ways (1 i R i co). Arcs of
various colours will spring off from every place
not more than R in nuntoer), going into every place GN's
transition will
and also arcs of various colours will
(but not nore than have the form of
(1 i i < p) and 1 < R " <. R (1 i j i s). j
Fig. 3. 1.
R
in nuntoer).
Fig. 3. 1,
where
(but be
Thus every i i R '£ R i
360
APPENDIX 1
(1 i i i p) and i i K " i K (1 £ j <. 5), j The token
tinted in colour which is narked with number
brief "colour i"> should only pass through a coloured arc, of which is an element of the set I .
A n arc numbered
j
i
(in
the number should host
i only tokens tinted in the colour with a number which is an element the set J . Obviously: j
of
n In U I I = n , IiU= l I iI = n , i= l i k k I . I u u J J i i = = k k . J=i J=i J J
The condition of the above CGN transition (which is an arbitra ry one) has the form (1", q" ) . . . (1", q» ) . . . (1", q" ) . . . (1", q" > 1 1,1 1 1, k" s s, 1 s s, k" 1 s
(i','
> i
i 1
1, K" lI l,k' r = . 1 I : I ( J ' I q"
rr qq" , q» ii P Ji (i i
P
<. k', l i i
T
T
<, k", i <. i <. p, l <, j £ s) j
where r is a predicate related to the q' -th output arc of q' , q" i,P i, P j. T the input (for the given transition) place \J and to the q" - th i j, T put arc of the output place 1". j transition has the form
Analogously,
the
in-
M-component of the
REMARKS ON THE GENERALIZED NETS
361
d", q" ) . . • d", q" I . . • U", d", q" ) .■ .■ .■ (1", U". q" ) 11,1 1 1, K" k" s s, 1 s s, k" 1 s d', U \ cf ) I 1 1, 1 I IV, IV,'f "q"
l M
-IV,
P P
cf
)
p, i 1 p,
(1J, V P
)I
l,k' I i i I
m qJ , q" i, p P j, J, TT
|
(l s sp p ii k\ l <, <, TT i. k", 1 1 <, <, i is s p, p, l l ii jj ii s) s) (l k\ l i. k", i j i j
I I I I
)> II
P,K' | P I
where m i O sire natural numbers. q' , q" i.P J, T The net so constructed has a set of tokens
K
of the form
K =
n U K , where K is the set of tokens which are coloured in i-th coloi=l i i ur. viously, every ordinary GN is a CGN for which n = k =1. GNs with interval time for activation The transition firing time-moment of this extension of is substituted by an interval.
the GNs
In that way the transitions in the new
net acquire the form: Z -
where d - [ f , t"] and T s t' i t" < T + t . When
the tokens
in the
input places of Z are already enough to satisfy the transition type
D
and TIME € a, the transition Z will be fired. Let the GN E be a net which has at least one transition of the form described above. For this the function © is extended to function 1
362
APPENDIX 1
9 , which to every time-interval d = [V, t"] 1 1 1
juxtaposes a new timeK
interval [t\ t"j 2 2
for which (T i)
t" i t' i t" i T + t , 1 2 2
i.e. ,
the
next time-interval for firing of the transition Z will come not earli er than the moment when this transition ends its former firing. GNs with a complex structure The transition forms of GNs are graphically similar to those of the E-nets. For the Petri nets the graphical structure is not the same as for GNs.
It is proved
that if a constructed GN which
than one arc entering each of its places
allows more
and more than one arc coming
out of each of its places, then this extension of the GNs will also be conservative and there it is called "a GN with a complex structure". GNs with global memory In the modelling
of real processes it
is appropriate
to keep
data (while the GN is functioning) or to determine the values of ferent parameters related to these processes.
dif
Thus, the definition of
the concept GN can be extended with the addition of a new component B, which we shall call "global memory". The component
B can be seen as a
list of such functions that will change their values in the process of GN functioning.
Let a GN with component B be called "a GN with global
memory" (GNGM). The formal definition of this object is: *
E
O K
= <,
L
1 2
K
K
>, where all components are as in § 1,1 and the component B is as descri bed above.
REMARKS O N THE GENERALIZED NETS
363
GNs with optimization components Some GN's models correspond to real processes
whose functions
do not stop and which determine standard (given before) the process's continuation. components
solutions for
F o r these G N s it is necessary to add such
that secure their non-stop functioning.
For this aim,
we
shall construct a new extension of the ordinary GNs - GNs with additi onal clocks (GNACs). Itiey have transitions with the form: Z - < L \ L", t , t , r, t , r', M, Q>, 1 2 3 where the components conmon with the ordinary transitions are the same and a) r' is a (0, 1)-index matrix with the form 1" . . . 1
1" . . . j
1" n
1' I 1 1 I I : 1' i
r' =
I r' I i, j I I (r € (0, 11) I i. j I I (1 S I $ m, 1 < J ,j i n)
: 1' m where L' = IV,
1 b) t
1 J ) and L" = 11", 1"
1'
2
is the maximal
1
m
,
1"),
2
n
time-duration for a check
of the truth
of the
3 predicates of the transition condition
r. In the case that time is
increased the checking of the predicates of
r shall be reduced and
their respective values from the matrix r' will be assigned. Obviously, t 3 the predicate's
s t , but if 2
truth can
t 3
> t 2
the process of checking of
be stopped before
the time necessary
for
364
APPENDIX 1
this tume-moment. If A is the set of the GNAC's transitions the GNAC has the form o E
=
<
n , n , c, f, e , 0 , e , a>, A
L
1 2
3
«
n , e >,
K
<X, $, §', b>>, where a) 6
is a function
giving the duration of the transition's condition
3 K
checking, i.e., 0 (t) = t', 3
with t € [T, T + t ]
and t' i O.
The
value of this function is calculated at the moment when the transi tion starts its functioning and,
after this, at every time-moment,
when the condition's predicates are calculated and the current step of the token's
transfers are finished
and there exists
more such
steps. This function determines the values of the transition compo nent t , and if © (TIME) = t' and 3 nent t
TIME + t' > t
3 does not receive new value.
+ t , 1
the compo-
2
3 b) G
is a function which to the predicates of r Juxtaposes (0, 1)-va
lues by an advance determined way.
Its calculations are made inxne-
diatelly after each checking of the values of the function © . 3 c) #'
is a function which
tokens which
determines the new characteristics
enter the transition's
output places
of the
after applying
the specifics of the new nets mechamism,
» x
§ 4-. GNs and other
it
objects
In this chapter,
it is proved that the functioning of
the re
sults of the work of the different types of modifications of the Petri nets and of the finite automata and the Turing machines can be sented by GNs.
A GN which makes the same for each
repre
ordinary Petri net
REMARKS ON THE GENERALIZED NETS
365
is constructed. •
§ 5. Algebraic
aspect
of the theory
it
of OVs
For the two t r a n s i t i o n s Z and Z we s h a l l 1 2 Z = Z i f f ( v i : 1 i i £ 7 ) ( p r Z = pr Z ); 1 2 i 1 i 2 Z c Z 1 2
iff
define
(Vi: 1 £ i £ 2)<pr Z c pr Z ) & i 1 i 2 (Vi: 3 £ i £ 4 ) ( p r Z = pr Z ) & i 1 i 2 (Vi: 5 £ i £ 6>(pr Z c i l l
pr Z ) & (pr Z c l 2 i l 2
pr Z ), 12
where c
is a relation of inclusion over index matrices and if 1 A = (X , L , la I], B = [ K , L , 1 1 i, j 2 2 A C B iff 1
c
(K 1
lb I], then i,j
c K ) & (L C L ) & (Vi€K >(Vj€L ) ( a 2 1 2 1 l i ,
= b ) j i , j
i s a r e l a t i o n of i n c l u s i o n over Boolean e x p r e s s i o n s and, f o r two of 2 these expressions a and b a c
b iff it is getting the expression
a after removing
the argu-
2 merits of b, which are not arguments of a and the logical operations, associated to them Over the transitions Z 1),
i i i i i i i = < L , L , t , t , r , M , D > i 1 2 1 2
we shall define three operations.
It is necessary for
(i=l,
the follo
wing to be valid: if place 1 € pr Z n pr Z and 1 € pr Z , then 1 € 1 i 2 i s 3-i pr Z f o r l £ i £ 2 , l £ s £ 2 . These o p e r a t i o n s are 3-s 3-i
366
APPENDIX 1
a) a union
(the necessary conditions
for this operation are
(j = 1,2), and (as if 1 e pr Z , this is s i pr
not possible),
1 2 t - t j J that
1€
Z for 1 i x i 2, 1 <. s <, 2): 3-s 3-i 1
Z V Z 1 2
=
2 1 U L , L 1 1 2
2 1 1 1 U L , t , t , r 2 1 2
2 1 + r , M
2 1 2 + M , V(D , D >>,
b> an i n t e r s e c t i o n (with the above c o n d i t i o n s ) : 1
Z HZ 1 2
=
2 1 D L , L 1 1 2
2 1 1 1 n L , t , t , r 2 1 2
2 1 x r , M
2 xM,
A(Q
1 2 , Q )>.
c) a composition (with the above c o n d i t i o n and w i t h t h e c o n d i t i o n 1 2 L nL 1 1
= L 2
Z
=
1 2 f l L = (*): 2 1
oZ 1 2
2 1 2 1 2 1 1 U | l - L |, L U ( L - L ), t , t , r 1 1 2 2 2 1 1 2
1 2 1 2 or,M oM,
1 -2 v(D , D )>, where D results from D after removing all its arguments, whose iden1
tifications are elements of the set L 1 It is possible that L Z
o o n Z = Z , where Z 1 2
2 n L 1 1
2 n L. 2 1
- f> and
L
1 2 n L =
is the empty transition
(for
some other coro-
ponents as M, r, Q will be degenerated). "Hie operations described below are unique in the Petri net the ory.
Tliey can be transfered on practically
nets
(obviously with some modifications
the corresponding
nets).
all other types
related to
These operations are
of Petri
the structure of
very useful
for con
structing GN models of real processes. Before introducing the different GN's operations, we shall for-
REMARKS ON THE GENERALIZED NETS
367
mil ate same conditions which the arguments of these operations (diffe rent GNs) must satisfy. We shall assume that
if a place
takes part
simultaneously in
two nets, then in both locations the place has (a) equal capacities, (b) the same characteristic functions (c) equal possibility to accept tokens net
from K,
(which is a model of some process)
ments of the set
K' c K
could pass
i.e. if
in the first
the tokens which are ele
through the examined place,
while in the second net tokens which are elements of the set
K" c
K could pass through this place, then K' = K", (d) equal values of the priority functions, (e> one and the same numeration/notation. Similarly,
we shall expect that if two places are respectively
an input one and an output one for two transitions in both nets, then (a) the capacities of the connecting arcs are equal, (b) the time for passing from the input to the output place
should be
one and the same, (c) they should be in the same consequence, i. e. the place which is an input
(output) one in the first of the transitions
should be the
same in the other transition too. Let E
and E 1
be two GNs and let for i = 1, 2: 2
i i i i E = <
i i , 6 , 9 >, 1 2
i i
o *
.
A union will be called the object: 1 E
U E
1
=
<
2 2 6 U6>, 2 2
U A ,
1
n
2
1
2
1
U T I . T I
A
A
2
1
U T I . C
L
2 U c , f
1
2 Uf,
L
1 2 1 2 TI U T T , 6 U 6 > , K K K K
1
2
6
UO,
1 <min(T , T ), 1 2
1
O o GCD(t , t ), 1 2
368
APPENDIX 1
» o o o + t . t /GCD(t / G C D ( t , t ) - m i n ( T , T )>>, ))>, i i 1 2 1 2
max (T iixi2 i <X 1
U X , $ U $ , b U b >>, 2 1 2 1 2
where A
U A 1
= 2
2 U ( Z / ( Z € A ) & (V Z' i=l i 2 U fZ/(3Z' i=l
€ A
) (Z D Z'
o = Z )) U
3-i
o € A )(3Z" € A ) (Z' D Z" ?! Z > & (Z = Z' U Z " > ) . i 3-i
A composition of the above nets will be called the object: (E , if T + t < T 1 1 2 2 1 E o E = 1 1 2 \ x E , i f T <, T + t ^ 3 1 2 2 where 1
E
=<
1
2
UK, 1 2
A
2
1
un
, i
A
L
1 2 TI V T HI , K K
2
1
un,c
2
uc,f
1
2 1
uf,
e
2 1
1 2 © U © e >, K K
* o o o t . t //GCD(t G C D ( t , t ) - T )>, i i 1 2 1
o o
<X 1
U X , $ U $ , b 2 1 2
1
2
ue,e
L
ue>, 1
2
2
max (T + liiS2 i U b ». 1 2
Before introducing the next operation over GNs, we shall defiene a local operator O - an
operators for place identification, which juxI
taposes to every given GN E and to every given places 1 € Q I E O d a new GN E 0(E,
1 , 1 ) = <<(A 1 2 [
L",
(
t , t , r, 1 2
L",
t
, t , r, 1 2
M, D > / ( 1 " € L")
M, D > / 1 "
and 1 € 2
E l " l | D
&
U"!)
U
(l')>
REMARKS ON THE GENERALIZED NETS
& (
L",
TT , e >,
K
K
t , t , r, 1 2
o *
M, D> € A ) ] > ,
369
71 , TI , c, A L
f,
9 , 9 >, 1 2
<x, $, b>>.
I t can be e a s i l y seen, t h a t the s e t s f
and
[ < L \ L", t , t , r, M, D>/(1" € L") & (L" = (L" - [1")) U ( l ' l & 1 2 (
extend the operator O over two s u b s e t s of L, as f o l -
If L' = ( 1 ' , 1 and 1J * 1', 1' * i J i 0(E, L \
I 1' 1') C Q and L" = fl", 1" 2 n E 1 2 1", 1" n ' (1 H , j S n ) , then J i J
L") = 0 ( . . . 0 ( 0 ( E ,
1', 1
1"), 1', 1 2
I")--, 2
O 1"! c Q n E
1', n
1">n
Obviously, the new object will be a (34 too. The object E» for which there exist sets h' - fl*,. 1' 1 2 I O c Q and L" = (1", 1" 1") c Q and Ex = O'E, L \ L") E 1 2 n E fixed GN E for which L = L , and for every i E an oriented
1'] n for some
(1 s i <. n) there exists
path from a place 1' to a place 1" i i
in
E,
we shall call
iteration of the (24 E For every two GNs
E
and
E
1
different types of relations are 2
defined in [«], but we shall give here only these which
are used
in
the book. Let us define for a given GN
E
two sets, one of tokens K and
one of the initial token's characteristics X:
370
APPENDIX 1 «
K(E) = K(E) = la/(ce ( a / ( a €€ K> K) & & (6 (6 (a) (a) <
with which
the token a can start its transfer in the GN E. Obviously, X(E) c U X(a>. a€K For a given token a € K and a given initial characteristic x 6 X(E) let
aa x >, if if a a € € K(E) K(E) and and x x € € X(cc| X(a| ( , )) fin fin E(a, < E(a, x) x) -. =<
rr (
II x>
f
a where x fin
,, otherwise otherwise
a a aa a a
>, if a e K(E) and x € X(a)
,
, otherwise
, i f a e K(E) and x € X(a) 1 2 fin
is the final characteristic of the token a in the GN E. ,
, otherwise
Let us define E (a, x) = E fa, xi = . Below we shall introduce four relations between the GNs: E 1
c E 2
iff
(K (E ) c K (E )) & (X (E ) c X (E )> & (Va € K (E )> 1 1 2 2 1 1 2 2 1 1 (VX € X (E ) ) ( E (a, X) = E (a, 1 1 1 2
E c E 1 x 2
iff
x||
(K (E ) C c KK (E )) & (X (E > ) c X (E )) & (Va € K (E )) 1 1 2 2 1 1 2 2 1 1 (VX € X (E ) ) ( 3 i , i 1 1 1 2
i : 1 <, i < . . . < i <, f i n ) s 1 s
(E (a, xl = pr E (a, x) ), 1 i , i ,. . . , i 2 1 2 s E s E 1 2
iff
(E 1
c E ) & (E c E ), 2 2 1
REMARKS ON 1HE GENERALIZED NETS
E
s E 1 * 2
iff
(E
c 1
371
E | & (E c E ). 2 2 x 1
x
Obviously, the relations c
and K
are stronger than the rela-
The first GN-definition (in § 1. )
in some sense is a deductive
tions c and z respectively.
one.
Below
we shall introduce
a second GN-definition.
In the above
sense, the new definition will be of an inductive type. Let
Z
be a
given object,
having the
grapical structure
in
Fig. 1. 1 and the following components: a) L' - a set of places which we shall call input places, b) L" - a set of places which we shall call output places, c) t
- a time -moment about some fixed
time-scale (with an elementary
1 o time-step t >, d) t
- a real number which corresponds to the length of a time-inter2
val in the above mentioned time-scale, e) r
- an index matrix (see App. 1) having the form 1" . . . 1" . . . 1" 1 j n 1' 1 r
=
1' i : 1J m
(i, j)-th
element of
I I I I l I I I l
r i,j (r
- predirates) i, J
(i i i i m, 1 S. j S n)
which correspond to the
output places, these elements being predicates, f) M - an index matrix having the form
i-th
injwt and j-th
372
APPENDIX 1
1" .. .. .. 1 1' 1'
M = M
1" .. .. .. 1" jj n
II 1 II
1' 1' i i : : 1' rm m
I I ID m l l i,j i,j II I (m i O I 0 -- natural natural lumbers) numbers) II i,i, jj I iI (l£i£m,l£j*n) I ( l £ i £ m , l £ j * n )
g) Q is an object having a form similar to an Boolean expression. variables are exactly the symbols which marK the names; the Boolean operations
"A" and
"v"
Its
Z's input places'
determine the analogo
us conditions from § 1 and 1)A(1
, ] ,...,1 ) - every one of the places 1 , 1 i i i i i 1 2 u 1 2 contain at least one token,
2) v(l
,1 i 1
,..., 1 ) - all the places 1 , 1 ,. . . ,1 i i i i i 2 u 1 2 u
at least one token, where
1 i
must u
must contain
(1 , 1 ,...,1 ) c L'. i i i 1 2 u
The object described above will be called a transition. The inductive definition of the concept GN is as follows: (1) The object which has the form of a transition
is called a GN,
if
to it are added a) TT - a function giving a natural number, being the priority of the A transition, b) n - a function giving the priorities of the transition's places, L c) c - a function giving the capacities of the transition's places, d) f - a function
which calculates the truth values of the predicates
of the index matrix r; e) ©
- a function giving 1
the next tune-moment
when the
given tran-
REMARKS ON THE GENERALIZED NETS
373
* sition can be activated, i.e. © (t) - t', where t, t' e [T, T+t ] 1 and t i t ' . The value of this function is calculated
at
the mo
of
a given
ment when the transition ends its functioning; f) © 2
a function
giving the duration of
the activity
transition, i. e. , 6 (t) = t\ where t e [T, T + t ) and 2 The value of this function is calculated
at the moment
t' z 0.
when the
transition starts its functioning; g) K - a set of tokens; h) ii - a function giving the priorities of the tokens (IT : K -> N); K K i) © - a function giving the time-moment when a given token can enter K the net, i. e. © (a) = t, where K a e K, K
t 6 [T, T + t ]; j) T - a time-moment
when the GN starts functioning.
This moment is
determined on a fixed time-scale and the first value of t = T; 1 o k) t - an elementary time-step related to the fixed time-scale; K
1) t
- a duration of the functioning of the net;
m) X - a set of all initial characteristics which can
receive the to
kens when they enter the net; n) § - a characteristic
function
which gives
new characteristic
to
every token when it makes a transfer from input to output place
of
a given transition; o> b - a function
giving the maximal number of characteristics
which
can receive a given token (i. e. b: K -> N) transfering in the net. (2) If E, E
and E 1
a) E
U E 1
are GNs, then 2
is a GN, 2
APPENDIX I
374 b) E o E is a GN, 1
2
c) Ex is a GN. THEOREM 5, 3, t: The two definitions of the concept "GN" are equivalent. K
» J 6. Topological
»
aspect of the theory of GNs
The graphical structure of a given GN is a two-coloured graph. Thus the study of such a structure and checking of the properties of a GN following from these structures are important for the theory of the GNs.
Some operators which assign to every GN one- or two-coloured
graph and some other operators, which juxtapose to every GN one natu ral
number corresponding to the complexity of the net with respect to
a certain criterion, are defined, The class £ is ordered by relations, generated from the above two types of operators, and some topological properties of £ are studied. x K
j 7. Logical
*
aspect of the theory of GNs
Logical operators which are similar to the modal operators "ne cessity"
and "possibility"
are defined and some of their basic pro
perties are given. * *
x
§ 8. Operator aspect of the theory of GNs The operator aspect has an important place the GNs.
in the theory of
Six types of operators are defined in its framework.
Every
operator assigns to a given GN a new GN, which has some preliminaryly, introduced properties. Particular types of operators are: global (one
REMARKS ON THE GENERALIZED NETS
which
transforms globally
transforms
some
some transition
375
GN's components),
components
local
of a given GN),
(one which
hierarchical
(one which substitutes a place or a transition for some part of
a gi
ven GN, or makes reversal), reducing (one which removes some of the GN components), extending (one which transforms a given GN to the corres ponding extending GN)
and dynamical (one which determines
rent strategies for the tokens' transfer
in the net).
the diffe
The basic task
of this aspect of the theory of the GNs is a research on the connecti ons and influence between the separate operators.
* § 9. Other extensions of the 0Vs By analogy with the universal Turing machine, versal GN (UGN) as a GN E
the object Uni-
for which the following is valid: K
(V E € £)(E c is intriduced.
E ),
It is noted that at the moment of writing of [K] (NOV.
1990) , the question for the existence of an UGN is open. Let Q be a some fixed time-scale. For T' e Si we define 2 = I E / E E J SprprE tprprEST'!. T' 1 3 3 3 K
We shall call the GN E
for which the following is valid, K
(V E € £
)(E c
T'
E ),
it
a T'-UGN about the time-scale Q. Obviously, the global time-components of O
E
=
<
A
, n , c, f, e , e >, L
K
1 2
, e >,
K
, t >, <x, §, b » ,
K
it
satisfy the inequality T + t
ST'.
THEOREM 9. 1. 1: For every time-moment T' € Q about the fixed time-scale
376
APPENDIX 1
it
Q, there exists a T' -UGN E
for the class £ T'
Let £ *
= IE / E e £ & pr pr E = x & pr pr E < 00), 1 3 3 3
i. e. , every GN of J x
has finite time for its functioning.
Let E x
be a
X
GN similar to E
but without the first global temporal component. Then
we shall define GN E x
as a universal GN (UGN) for the class £ , i. e. , x x
IVK6I |(Ec E >. X
K
THEOREM 9. 1. 2: There exists a UGN for the class £ . M
Let E be some GN. Let K be a set of tokens of E, and X be a set of initial
characteristics for these tokens of K.
The object
"Self-
Modifying GN" (SMSN) is constructed. x Let K
be a set of tokens which will be called
The defining of this set is possible in the
"control
framework
tokens".
of the GN the-
x ory.
The tokens of K
can be interpreted as GN-tokens
with initial
K
characteristics
(elements of the set X ) which contain the identifier
"control token"
(this is also possible).
frame of a given GN places
(in which
characteristic
E
is that
there will be
The second addition in the
the characteristic functions of transferred control tokens)
some
give as
identifiers and parameters of some types of operators,
defined over the GN.
This is also not in contradiction with the defi
nition of GNs. Finally (really, the new): when a certain control token gets as characteristic the identifier of some operator with parameters former
or current characteristics of
this or
another control token,
then at this time-moment over the GN E the corresponding operator will be realized. Obviously, not-every operator can be applied to the net
in the
REMARKS ON THE GENERALIZED NETS
377
time-interval when it functions. Thus, we can define the SMGN as a GN, for which there exists a mechanism p for realization of the over it.
This mechanism functions in the following way:
moment when a certain control token of a given ristic the identifier
of a certain
operators
at the tiroe-
GN gets as a characte
(added) operator and the
values
of its parameters, the mechanism p realizes this operator. The basic aims in [») are: 1. to show the operators which can be realized by p; 2. to show the relations between GNs and SMGNs.
and £
If £ is the class of all SMGNs, then obviously J c J SMQN SMQN H £, because every GN is a SMGN with an empty set of con-
SMGN trol tokens. THEOREM 9. 2. 1: For every SMGN, there exists a GN which represents it. The SMGNs are the basic aim of the operator aspect of
the the
ory of GNs. x
* § 10. Methodological
aspect
of
»
the theory
of GNs
The b a s i c ways for c o n s t r u c t i n g o f a GN of a given r e a l process and same ways for u s i n g and modifying of already c o n s t r u c t e d GN-models are d i s c u s s e d . X *
§ 11. Open
X
problems
There are many unsolved
or unforroulated problems in the theory
of GNs. Here we shall introduce some of these problems which are rela ted to the text of the book. 5. To research the different classes of reduced GNs without two or mo re components, and to show the relations between these classes.
378
APPENDIX 1
7. To prove that the constructed
reduced GN in § 2. 2 from the book on
GNs (from § 1. 1.2) has the least possible number
of different GN's
components, or to give a counter example of this. 8. What other conservative extensions of the GNs can be defined? 9. To construct GNs which are universal for the classes (sets) of the different
Petri net modifications,
by analogy
with the
GNs from
§ 4. 3. 12. To construct an algorithm which for every set of transition phical structures
(obviously they must satisfy
some
gra
conditions)
and for every given GN, construct a GN for which: - all its transitions have
the forms, which are elements
of the
given set, - both nets have equal tokens and as a result of their
with equal
initial characteristics
functioning equal tokens receive equal
final characteristics. 19. To prove or disprove the existence of a universal GN for £. 20. To construct other extensions of the GNs and to prove
or disprove
their conservativness. 21. Chapter 10 is a superficial attempt to give the basics of
the m e
thodology of work with GNs. To continue and enrich the research in this area. 22. A list of the basic applications to this moment (Nov. 1990) is gi ven in Chapter 0. To extend the area of the GN's applications.
Appendix 2: GENERALIZED INDEX MATRICES
A definition of "generalized index matrix" (GIB) or "index mat rix" is given in [1-3]. Here we snail introduce only the results which are related to this basic text. Let bers.
I be a fixed index set and H be the set of all real num
A GIH with generalized index sets K and L
(K, L c I) will be
called the object: 1 1
k
1
k 2
k m
1
1 2
|a i kK , 1l
1
1
a k ,l
1
. . .
1 n
. . .
2
a k , 1
I n
la a . . . I k ,1 k ,1 I 2 1 2 2 I I . .
a k ,1 2 n
I a a I k ,1 k ,1 | m 1 m 2
a k ,1 m n
< =
[X, L,
.
fa ]]) k ,l i j
),
where X = Ik , k , . . . , k 1, L = 11 , 1 , . . . , 1 !, for l ! i ( n an
1
2
1 2
m
for 1 i j i n: a
n
€ K. Let H be the set of all GIHs with index k , 1 X, L i j sets X. and L, and let H = U H . K, Lcl
K, L
379
APPENDIX 2:
380
In [31, for the GIMs A = [K, L,
(a
k ,1 i j
1], B = [P, Q, (b
a r e defined ordinary matrix o p e r a t i o n s "addition" on"
t ,v u w
/a / k ,1
"miltiplicati-
! ] . where
, if t
- k
u
k
1
b
v "i
t , u w
a
, if t
p ,q r s
t + b
K ,1 i j
p ,q r s
v0
t ,v u w
-
a
k,l i j
, if t
. b
P,q r s
t ,v u w
u
= k
€L-Qor
j
€ K - P and v = 1 e L w j
i
€ P - K and v
r
= p
i
r
6 K DP
w
€Q-Lor = q
and
e Q
s v
= 1 w
= q j s
na
= k i
u
The b a s i c p r o p e r t i e s of
= p r
e K D P and v
=1 w
= q € j s
L n Q; fc
t
u
v
!],
where
w
if t
1 b 1 p ,q r s E a .b 1 =p 6LTP k , 1 p ,q j r i j r s
,°
=1
1], where
, for t
(v. t ,V u w
= p
u
w
= p € P and v = q r w s
e L
A . B = [K U (P-L), Q U (L-P),
c
u
€ K and v
i
, otherwise
A x B = [K n P, L n Q, fc
c
!)
and a l s o the f o l l o w i n g o p e r a t i o n s
A + B = [K U P, L U Q, fc
c
and
p ,q r s
if t
if t
u
u
u
= k
= p
i
r
€ K and v
w
=1
e P - L and v
w
j
€ L - P
= q
s
e Q
= K € K and v = q € Q i w s
otherwise the s e t M w i t h t h e above o p e r a t i o n s are
GENERALIZED INDEX MATRICES
381
studied in [3]. There some GIMs measures are introduced (for GIMs de terminants do not exist). Ihe given mathematical
apparatus may be
with elements from the sets (0, li,
applied to
the
GIMs
[0, 1), or from the class of all
predicates, etc. In the first two cases, the operations " + " and ". " in R will be substituted by
"max"
and
"min" respectively,
and in the
third case - by the operations "v" and "*". REFERENCES: [1] Atanassov K.
Conditions in
Generalized
nets,
Proc. of the XIII
Spring Conf. of the Union of Bulg. Math. , Sunny Beach, April 1964, 219-226. [2] Atanassov K. Dynamical elements in the Generalized nets,
AMSE Re
view Vol. 1 (1985), No. 4, 1-9. [3] Atanassov K. Generalized index matrices, Comptes rendus de demie Bulgare des Sciences, Vol.+0, 1987, No. 11, 15-18.
l'Aca-
Appendix 3 : INTUITIONISTIC FUZZY SETS
Following [1,2] we shall give short remarks on
the intuitioni-
stic fuzzy sets (IFSs) and other intuitionistic fuzzy objects
related
to them. Let a set E be fixed. An IFS A«
in E is an object having
the
form a A
: |<1, |l |I), 1 <X)>/X€E),
A where
the functions
A
y (x) : E -> [0, 1] A
fine the degree of membership
and
t (x) : E -> [0, 1] deA
and the degree of non-membership of the
element x€E to the set A, which is a subset of
E,
respectively,
and
for every x€E 0 i \i <x] + T (x) i 1. A A Obviously, every ordinary fuzzy set has the form |<x, v (x), 1-p <x)>/x€E]. A A 11
For simplicity, below we shall write A instead of A . For every two IFSs A and B, the following relations tions are valid: A c B iff (VX€E) <JJ (x) <, \i (x) 8 T <x) i T (x)); A B A B A D B iff B c A;
382
and opera
INIUITIONISTIC FUZZY SETS
383
A = B iff (Vx€E)(p (x| = p (x) & T (x) = Y (x)) A B A B A = (<x, T (x), p (x)>/xeE); A A A D B = (<x, minlp (X), p (x>), max(T (x), A B A
t
(x))>/xeE);
A U B = (<x, max(p (x), p ( x ) ) , min(T (x), A B A
t (x))>/x€E); B
B
A + B = (<x, p (x)+p ( x ) - p (x). p (x), T (X). T (x)>/x€E); A B A B A B A . B = [<x, p (x). p (x), A B
T (x)+T ( X ) - T (x). T (x(>/xeE); A B A B
A - B = A fl B. The concept IFS is the basis for the definition of the concepts Intuitionistic Fuzzy Propositional Calculus
(IFPC)
and such extensi
ons as IF predicative calculus, IF modal logic, etc. (see [3-5]). To each proposition (in the classical meaning) its truth value: truth
denoted by 1, or falsum by 0.
one can
assign
In the
case of
fuzzy logics, this truth value is a real number in the interval and can
be called "truth degree" of a particular proposition.
case of
IFPC is
added one more value - "falsum
[0, 1] In the
degree" - which will
be in the interval [0, 1] as well. Thus one assigns to
the proposition
p two real numbers p(p) and T(p); moreover the following constraint is valid: p(p> + T(P) i
i.
Let this be done by an evaluation function V defined so that V(P) =
[0, 1] x [0, 1] gives the truth and
falsuni degrees from the class of all propositions and it
can be defi
ned so that it assigns to the logical truth T: V(T> = <1, 0>, and to the logical falsum F: V(F) = <0, 1>. Ibe negation np of the proposition p will be defined through
384-
APPENDIX 3:
V H P ) =
P(p)>.
When T(P) = 1 - p
"3"
can be
obtained by different variants of IFPC. One of them is as follows: V(p & q) = <min(p(p), p(q)>, max
T(q))>,
V(p * q> = <max(p(p), p(q)), min(T(p),
T,
V(p 3 q) = <max(T(p), p(q)), min(p(p>,
r(q))>.
A given propositional form A propositional form; if
(c. f.
[7):
each proposition is a
A is a propositional form, then iA is a propo
sitional form; if A and B are propositional forms, A D B are propositional forms)
then A & B, A * B,
will be called a tautology,
iff
V(A) = <1, 0>, for all valuation functions V,
and an intuitionistic fuzzy tautology,
iff "if V(A) = , then a i b". There exist some forms of an intuitionistic fuzzy
Modus Ponens
(see [3,5]). The definition for the quantifiers is as follows: V(VxA) = <min p(A>, max T ( A ) > x x and V(3xA) = <max p(A), min T(A)>. x x For the proposition p for which V(p> = , in [4] are defined the following operators: V(Qp) = ,
INTUITIONISTIC FUZZY SETS
385
V(Op) =
Intuitionistic fuzzy sets,
Fuzzy Sets and Systems,
Vol. 20 (1986), No. 1, 87-96. [2] K. Atanassov,
More on intuitionistic fuzzy sets.
Fuzzy
Sets and
Systems, 33, 1989, No. 1, 37-46. [3] K. Atanassov,
Two variants
of intuitonistic fuzzy
propositional
calculus. Preprint IM-JFAIS-5-88, Sofia, 1988. [4] K. Atanassov,
Two variants of
intuitionistic fuzzy
modal logic.
Preprint IM-MFAIS-3-89, Sofia, 1989. [5] K. Atanassov, G. Gargov, Intuitionistic fuzzy logic.
Compt. rend.
Acad. bulg. Sci. , Tome 43, N. 3, 1990, 9-12. [6] A
Kaufmann,
Introduction a la theorie des sous-ensembles
flous,
Paris, Mas son, 1977. [7] Mendelson E. , Introduction to mathematical logic, D. Van Most rand, 1964.
Princeton,
NJ:
LIST QF AUTHORS
Alexander Georgiev - Institute for Microsystems, Sofia Alexander Savov -
Institute for Microsystems, Sofia
Antoaneta Eirova - High Economical Institute, Sofia Antonia Dimitrova - Institute for Microsystems, Sofia Bistra Nikolova - Institute for Microsystems Borjana Jordanova - Faculty of Mathematics, El. Ochridsky Univ. , Sofia Evgeni Tzolov - Institute for Microsystems, Sofia Ewgeni Dimitrov - High Transport School, Sofia Ilya Eazalarsky - Institute for Microsystems, Sofia Ivan Hristozov - Institute for Microsystems, Sofia Joseph Sorsich - 2-nd City Hospital, Sofia Katja Stefanova - Institute for Standardization and Sertification, Sofia Erassimir Atanassov - Institute for Microsystems, Sofia Lilija Atanassova - 105 Al. Dimitrov School, Sofia Ljubomir HadJyisKy - Technical University, Sofia Luomila Dimitrova - Institute of Chemical Engineering, Burgas Maria Stefanova-Pavlova - Institute for Microsystems, Sofia Martin Tetev - Faculty of Mathematics, El. Ochridsky Univ. , Sofia HiKo Pehlivanov - Institute for Microsystems, Sofia Favlin Gyurov - Faculty of Mathematics, El. Ochridsky Univ. , Sofia Peter Georgiev - Faculty of Mathematics, El. Ochridsky Univ. , Sofia Flamen Fetkov - HEFTCCHIM Petrochemical Combine, Bourgas Radosvet Todotov - Centre for Scientific Information of Bulg. Acad. of Sciences, Sofia Kossen Petrov - Institute for Microsystems, Sofia Rumen Christov - Institute for Microsystems, Sofia Sergej Bedev - Technical University, Sofia
386
367
Stanislav Pishmanov - Faculty of Mathematics, Kl. Ochridsky Univ. , Sofia Stefan Stefanov - Institute for Microsystems, Sofia Stela Dimitrova - NEFTOCHIM Petrochemical Combine, Bourgas Stojan Mihov - Faculty of Mathematics, Kl. Ochridsky Uhiv. , Sofia Stoian Garbov - Central Laboratory of Automation, Sofia Trajana Kolarova - NEFTOCHIM Petrochemical Combine, Bourgas