CN110781616B

CN110781616B - Behavior equivalence process tree generation method based on complete finite prefix expansion

Info

Publication number: CN110781616B
Application number: CN201911016696.6A
Authority: CN
Inventors: 朱锐; 黄月; 李彤; 唐明靖; 马自飞; 郑明�; 汤雅惠
Original assignee: Yunnan University YNU
Current assignee: Yunnan University YNU
Priority date: 2019-10-24
Filing date: 2019-10-24
Publication date: 2023-04-18
Anticipated expiration: 2039-10-24
Also published as: CN110781616A

Abstract

The invention discloses a behavior equivalence process tree generation method based on complete finite prefix expansion. The method comprises the following steps: expanding the process model into a branch process containing a truncation event by using a full prefix expansion method; extracting an activity relation group meeting the reconfigurable condition from the branch process; and according to a preset conflict judgment condition, when the extracted activity relation groups are judged to have no conflict, reconstructing the extracted activity relation groups according to a preset rule to obtain a corresponding process tree. The invention can convert the non-BSPM process model into the process tree with equivalent behavior, and has high conversion efficiency, high accuracy and simple structure of conversion result.

Description

Behavior equivalence process tree generation method based on complete finite prefix expansion

Technical Field

The invention relates to the field of process models, in particular to a process model algorithm for converting a process model described by a Petri net into a process model described by a process tree.

Background

The process model is a special data with both graphic representation and behavior semantics. At present, the Petri net is the most commonly used modeling language when a process model is described, but the Petri net is used for representing the process model, and deadlock, livelock and other exceptions in the model are difficult to avoid. Secondly, when analyzing the behavior structure of the process model described by the Petri net, observing the behavior relation among a group of activities in the Petri net is relatively easy, but observing the behavior structure of the whole Petri net becomes very difficult, and particularly when the Petri net is very complex, a 'pasta-like' process model is formed.

Using the process tree to represent the process model not only avoids these unreasonable constructions, but also integrates the behavior of the process model, simplifying the complexity of the model. The process tree is used as a representation of a tree structure, a process model is described by using the process tree, and the process tree has the advantages of being clear and simple in structure, unified in structure and behavior, capable of determining behavior semantics among activities and capable of meeting basic properties of a Petri net. The use of the process tree to represent the structural information of a Petri net has many advantages, and a BSPM can perform behavior equivalent transformation with the process tree, that is, the process tree can be generated by directly analyzing the structure.

The process tree is a tree structure diagram containing process model information, leaf nodes in the process tree represent activities, branch nodes represent activity relationships, and the relationship between any two activities (leaf nodes) in the process tree is the nearest common parent node. Although the process tree structure is clear and the process model is easy to express, for the process model expressed based on the Petri net, the process model of the following two cases can not be converted into the process tree on the premise of behavior equivalence:

(1) There are a variety of relationships between a certain set of activities;

(2) Relationships between a certain set of activities do not belong to a basic relationship.

We classify the process models according to whether they can be equivalently converted into process trees (fig. 1 shows a process model classification diagram which can be equivalently converted into process trees). The slashed part is the process model equivalent to the process tree behavior, and refers to the process model without the above two special cases. Currently, the research on the Process tree is mainly based on the Process Model of Block structure (BSPM, see Wilvan der Alst et al for specific references: methods for Improving the reliability of the Process of the complex structure of the Process of Mining, J.Buijs et al for the objective algorithm for the discovery of the Process of Mining, and 8978 xzft 8978 et al for the complex structure of the mixed Process of Mining). However, in a number of studies we have found that some non-BSPMs can be equivalently transformed into process trees, and we therefore classify the process model as fig. 1. FIG. 2 is an example of converting BSPM into a process tree, and FIG. 3 is an example of converting non-BSPM into a process tree.

For the problem of how to convert a Process model described by a Petri Net into a Process model described by a Process Tree, the existing Conversion Algorithm (refer to the literature "Conversion and Conversion Algorithm of Well-Structured Workflow Net to Process Tree") can convert a BSPM into a Process Tree by identifying the basic relationship (occurrence weight relationship) between activities, and the literatures "hybrid Process mining method of Complex Structure", "Conversion and Conversion Algorithm of Well-Structured Workflow Net to Process Tree" and "converting block-Structured Process model from event volumes-a dependent Process Tree" propose that the expressive power of a Process Tree can be increased by supplementing basic blocks. However, the above document only discusses the method of converting the BSPM type into the process tree, and in fact, not only the BSPM can be equivalently converted into the process tree, but also some non-BSPM can be equivalently converted into the process tree. For example, the process model P1 shown in FIG. 3, compared to the process model P0 shown in FIG. 2, which has more libraries c7 and c8 in structure, the process model P1 is obviously a non-BSPM. However, from the perspective of the Petri net "trace", it can be found that P1 is trace equivalent to P0. The process model P1 can thus likewise be expressed in a process tree. However, the problem of how to convert the non-BSPM model equivalent to the process tree into the process tree is not solved at present.

Disclosure of Invention

The invention aims to: aiming at the existing problems, the method provides a behavior equivalence process tree generation algorithm which converts a process model equivalent to the process tree behavior into a process tree and is based on complete finite prefix expansion. The problem that non-BSPM equivalent to the process tree can not be converted into the process tree in the prior art is solved.

First, the definition and explanation about technical terms involved in the present invention are as follows:

1. a process model: a quadruple p = (C, A; F, M0) as a Petri net, where C is the set of conditions,

referred to as a condition; a is the active set>

Referred to as an activity, the occurrence of a is referred to as a being performed or a firing; f is a flow relationship on the Petri net, and>

for the initial instance of p, a d ∈ M0 is called Token.

2. And (3) process tree: for a process model p = (C, A; F, M0), A is its active set, then the activity a ∈ A { [ tau ] } is a process tree; let M1, M2, … Mn (n > 0) be n process trees, the relation symbol between the process trees be noted as excellent, then excellent (M1, M2, … Mn) be the process tree. Defining the order (→), selecting (×), and concurrently (| |), iterating (. + -) for a total of four relationship symbols, using the symbol ∈ to represent one of them, the relationship symbols are grouped as | = { →, x, |, ±) -for | >.

3. A presence net: a triplet o = (S, T; F') as a net of occurrence, wherein:

(1) (S, T; F') is a network without rings, scotu T ≠ Φ; s is a finite set of library sites (i.e. conditions),

referred to as a library; t is a limited set of transitions (i.e., events) that are present in or on a device>

Referred to as a transition; f is the edge set of the presence net and +>

(2)

The requirement that |. S |, is less than or equal to 1, namely the input of any library is less than or equal to 1;

(3)

x # x, i.e. no self-conflict applies to any element;

(4)

satisfy +>

Is limited.

In the presence net, x # y indicates that there is a conflict relationship between x and y, i.e., there is some common repository c in the presence net, and the path from c to x is disjoint from the path from c to y. x < y indicates that there is a causal relationship between x and y, i.e. there is a path from x to y in the presence net. x co y representing a co relationship between x and y x y satisfying x < y x # y.

4. And (4) branching a process: for a process model p = (C, A; F, M0), the branching process is a two-tuple Π = (o, h), where o = (S, T; F ') is a net occurrence, h is a slave (S, T; F')

A mapping to (C, A; F), wherein:

(1)

(2) Min (o) and M0 are in bijection relation, and Min (o) represents a minimum element set obtained according to a proper partial order relation in the net o;

(3)

and t1, t2 satisfy · t1= · t2 ^ h (t 1) = h (t 2), then t1= t2.

5. Possible extended set of branching processes: the tuple Π = (o, h) is a branching process of the process model p = (C, A; F, M0), where o = (S, T; F') is a presence net and h is a mapping. The possible extension set of the branching process Π is a set of (t, B) pairs, denoted as PE (Π), where B is a co-set of relationships (co-set) of a condition set P in Π, and t is a transition of a process model P, where:

(1) h (B) = · t, B maps to a pre-set of t in the process model p;

(2) p (e) = t and · e = B, then B does not include e.

6. Configuration: the configuration C of one presence net o = (P, T; G) is a set of transitions, where:

(1)

t' belongs to C, namely the configuration C is a closure consisting of proper partial order;

(2)

i i.e. any elements in the configuration C are non-conflicting.

For one presence net o = (P, T; G),

the local configuration of transition t is also a set of conflict-free transition sets comprising t itself and consisting of appropriate partial order, denoted as [ t ]]Satisfy t ∈ [ t ]]And t1 ∈ [ t ]]T1 satisfies [ t1]<·[t]Any t2 ∈ C having t2# t1.

7. Proper partial order relationship: the proper partial order relationship on one presence net o = (P, T; G) is the relationship between local configurations, denoted by <.

(1))<Is to

Is improved in that>

Have>

Then [ t]<·[t']；

(2)<Obtain reservations by finite expansion, i.e.

Has a [ t ]]<·[t']And Mark (t) = Mark (t '), then for [ t = Mark (t') ]]Finite expansion of [ t ]][. DELTA.E ], there is an isomorphic transformation I, such that [ t ]]⊕E<·[t']⊕I(E)。

8. Cutting and gathering: a finite configuration C of one presence net o = (P, T; G), whose Cut set, cut (C), is a set of conditions that contain only co-relations, cut (C) is defined as Cut (C) = (Min (N). U.C. -%, C.

A set of positions of the finite configuration C, h (Cut (C)) of the presence net o = (P, T; G) is a reachable marker, representing events reachable by the condition set Cut (C), and denoted h (Cut (C)) by Mark (C).

9. A truncation event: the tuple Π = (o, h) is a branching process of the process model p = (C, a; F, M0), where o = (S, T; F') is an occurrence net and h is a map. If Π contains an event e 'such that another event e is a truncation event, denoted corr (e) = e', it satisfies:

(1) Mark ([ e ]) = Mark ([ e ']), i.e., the equivalence of the situation reached through e and e';

(2) [ e '] < [ e ], local configuration [ e' ] and local configuration [ e ] are in a proper partial order relationship.

The emotional Cut ([ e ] e) reached by e]) Called truncation condition p, for which the following elements are truncated, i.e. arbitrary truncation condition

10. Anterior and posterior sets: one quadruplet p = (C, A; F, M) ₀ ) As a Petri net, let x be an element (C.U.A) to let

Then call ^· x is a preceding or input set of x, x ^· Called the postset or output set of x.

For the analysis of the static process model, the traditional method mainly uses the method of the reachability tree or the reachability graph, and the main problem of the method is the explosion of the state space. To address this problem, escapza discloses a method of full prefix expansion in An Improvement of McMillan's Unfolding Algorithm [ J ]. Lncs, 1996. Through the full prefix expansion of the process model, the process model can be expanded into a branch process pi = (o, h) containing a truncation event, wherein o = (S, T; F') is a presence net, h is a mapping function, for example, for a process model p = (C, a; F, M0), the corresponding branch process is continuously expanded by using a possible expansion set of the branch process until the branch process is expanded to the truncation condition, and finally, a branch process containing the truncation event is obtained.

The technical scheme adopted by the invention is as follows:

a behavior equivalence process tree generation method based on complete finite prefix expansion comprises the following steps:

A. expanding the process model into a branch process containing a truncation event by using a full prefix expansion method;

B. extracting an activity relation group meeting a reconfigurable condition from the branch process;

C. according to a preset conflict judgment condition, when the extracted activity relation groups are judged to have no conflict, executing step D;

D. selecting a group of activity relation groups with the highest priority from the extracted activity relation groups for reconstruction according to a preset priority judgment rule;

and repeating the steps A-D until all the activity relation groups are reconstructed to obtain the corresponding process tree.

The steps A-D are executed once, only one group of activity relation groups are reconstructed, and some activity relations can be found after other activity relation groups are reconstructed, so that the steps A-D need to be executed again after each group of activity relation groups is reconstructed. And after each reconstruction, judging whether the activity number of the process model is greater than 1, if so, jumping to A, and if the activity number is equal to 1, finishing surface reconstruction, wherein the result obtained by reconstruction is the process tree.

By the above method, the non-BSPM model equivalent to the process tree can be converted into an equivalent process tree.

Further, in the step B, the method for determining whether the activity relationship group satisfies the reconfigurable relationship includes:

b1: extracting a transition set from the branch process; the set of transitions is the set of events in the process model.

B2: and respectively distinguishing the relationship of each transition group in the transition set according to the definition of a preset reconfigurable relationship, and if the distinction is passed, judging that the activity relationship group corresponding to the transition group meets the reconfigurable relationship.

Further, the reconfigurable relations include a reconfigurable sequential relation, a reconfigurable iterative relation, a reconfigurable selection relation and a reconfigurable concurrent relation.

Further, the determination condition of the reconfigurable sequential relationship is as follows:

for process model p = (C, A; F, M0), Π = (o, h) is a branch process corresponding to process model p, which includes a truncation event, wherein o = (S, T; F') is an occurrence net, h is a mapping function from a transition set to an active set, C is a condition set, A is an active set,

referred to as an activity, M0 is an initial state of p, S is a pool set, T is a transition set, and->

Referred to as a transition, is asserted>

For arbitrary->

The activities mapped into the process model are a1 and a2 respectively; when t1 and t2 meet the following conditions, the activities a1 and a2 have a reconfigurable sequential relationship:

1. the configuration of t1 differs from the configuration of t2 only by t2: [ t1] = [ t2] - { t2}; .

2. Configuration [ t1]The set of activities mapped into the process model does not contain the activities of event t2 mapped into the process model:

3. if there is T2'∈ T, satisfying T2' ≠ T2 ^ h (T2 ') = h (T2), then for all T1' satisfying [ T1'] ≠ T2' } = [ T2'] there must be h (T1') = h (T1);

4. t ∈ T does not exist, corr (T) = T1 is satisfied unless h (T) = h (T1) Λ [ T2] = [ T ] - { T };

5. for all T epsilon T, when T ≠ T1 ^ T ≠ T2, the relationship between the event T and the event T1 on the expanded net is the same as the relationship between the event T and the event T2.

Further, the determination condition of the reconfigurable iterative relationship is as follows:

referred to as an activity, M0 is an initial state of p, S is a pool set, T is a set of transitions, and ` according to a predetermined pattern>

Referred to as a transition, is asserted>

For arbitrary->

The activities mapped into the process model are a1 and a2, respectively, i.e., h (t 1) = a1 and h (t 2) = a2; when t1 and t2 meet the following conditions, reconfigurable iterative relationships exist between activities a1 and a 2: />

1. the configuration of t1 and the configuration of t2 differ only by t2: [ t1] = [ t2] - { t2};

2. t2 is a truncation event and the configuration of its corresponding event corr (t 2) differs from the configuration of t1 by only corr (t 2), i.e. [ t1] = [ corr (t 2) ] - { corr (t 2) };

3. if h (T2) ≠ h (corr (T2)), then a1 and a2 have a reconfigurable iterative relationship, otherwise, only for all T ∈ T, and T ≠ T1 ^ T ≠ T2, and T1, T2 have no reconfigurable relationship, a1 and a2 have a reconfigurable iterative relationship.

Further, the condition for determining the reconfigurable selection relationship is:

for process model p = (C, A; F, M0), Π = (o, h) is a branch process corresponding to process model p that includes a truncation event, wherein o = (S, T; F') is a presence net, h is a mapping function from a transition set to an active set, C is a condition set, A is an active set,

Referred to as a transition, is asserted>

For arbitrary->

Its activity mapped into the process model is a1, a2, i.e. h (t 1) = a1, h (t 2) = a2; when t1 and t2 meet the following conditions, activities a1 and a2 have a reconfigurable selection relationship:

1. t1 conflicts with t2;

2. the occurrence conditions of t1 and t2 are the same;

3. the postamble of t1 and the postamble of t2 are both empty or intersect.

Further, the conditions for determining the reconfigurable concurrency relationship are as follows:

initial situation called an activity, with M0 being pStatus, S is a pool set, T is a transition set>

Referred to as a transition, is asserted>

For arbitrary->

1、t1 co t2；

2. neither t1 nor t2 is a truncation event;

3. the occurrence conditions of t1 and t2 are the same;

4. the existence of T ∈ T ≠ T1 ^ T ≠ T2 satisfies

Or for any T epsilon T ≠ T1 ^ T ≠ T2, there is ^ T ≠ T>

Further, in step C, the predetermined conflict determination condition is:

Referred to as a transition, is combined with a signal processing unit>

For arbitrary->

Its activity mapped into the process model is a1, a2, i.e. h (t 1) = a1, h (t 2) = a2; for any activity a1, a2 ∈ A, when t1, t2 satisfy the following condition, there is no conflict between the activity relation groups:

1. a1 and a2 have no relation or only one relation;

2. when a1 and a2 have a reconfigurable sequential relationship, the relationship between a1 and a2 and any a ∈ p.A ≠ a1 ^ a ≠ a2 only has the reconfigurable sequential relationship between the activity a and the a1 or only has the reconfigurable sequential relationship between the activity a2 and the a;

3. when a1 and a2 have a reconfigurable iterative relationship, no relationship can exist between a1 and a2 and any a epsilon p.A ^ a ≠ a1 ^ a ≠ a2;

4. when a1 and a2 have a reconfigurable selective relationship or a reconfigurable concurrent relationship, the relationship between a1 and a2 and any a ∈ p.A ^ a ≠ a1 ^ a ≠ a2 can only have a selective relationship or a concurrent relationship.

Further, the predetermined priority determination rule in step D specifically includes: and judging the priority of the activity relation group according to the type of the reconfigurable relation to which the activity relation group belongs. Namely, each reconstruction selects a group of reconstructions with the highest priority of the types (the four types are reconfigurable sequential relationship, reconfigurable iterative relationship, reconfigurable selective relationship and reconfigurable concurrent relationship) to which the active relationship group belongs.

The sequence of reconstructing each activity relationship group directly influences the reduction degree of the activity and relationship of the original process model and the complexity of the reconstructed process model. And taking the type to which the activity relation group belongs as the basis of the reconstruction sequence, so that the generated process tree can be logical-smooth.

Further, the priority of the type of the reconfigurable relationship is as follows:

the reconfigurable sequential relationship has a highest priority, followed by a reconfigurable iterative relationship, the priorities of the reconfigurable selection relationship and the reconfigurable concurrency relationship being determined based on the number of active elements involved in the active relationship in the behavioral structure of the process model.

The reconfiguration priority designed by the invention can ensure the consistency of the logic of the process tree and the original process model and the simplification of the process tree structure.

The terms "X1", "X2", "X'" and the like (X is t, a, F, and the like) herein do not denote any data, but merely distinguish the corresponding parameters, and have no other special meaning.

In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:

through the behavior equivalence process tree generation method based on the complete finite prefix expansion, the process model of the non-BSPM model can be converted into the behavior equivalence process tree. The generated process tree structure is high in simplification degree and highly consistent with the logic of the original process model (high in accuracy), and no equivalent reconfigurable activity information is lost. In addition, the invention has higher equivalent behavior process tree generation efficiency.

Drawings

The invention will now be described, by way of example, with reference to the accompanying drawings, in which:

FIG. 1 is a process model classification that is equivalently convertible to a process tree.

FIG. 2 is one embodiment of a BSPM model being converted into a process tree.

FIG. 3 is one embodiment of a non-BSPM model conversion to a process tree.

FIG. 4 one embodiment of a process model generated by the PLG.

FIG. 5 is a statistical result of the time and corresponding number of activities of the process tree extracted in the PLG-generated process model.

Fig. 6 is the average activity time consumption statistics for the RD group experiments.

FIG. 7 is an embodiment of a partial ED group to generate a process tree, where (a) is an experimental model, (b) is a process tree equivalent to the experimental model in (a), and (c) is a reconstructed process model according to (b).

FIG. 8 is a process model that cannot be converted to a process tree, (a) is an experimental model, and (b) is an expanded mesh corresponding to (a).

FIG. 9 is another process model that cannot be converted into a process tree, (a) is an experimental model, (b) is an expanded mesh corresponding to (a), (c) is a process model reconstructed according to (b), and (d) is an expanded mesh corresponding to (c).

FIGS. 10 to 13 show four embodiments of activity relationship determination, wherein (a) is an original Petri net and (b) is an expanded net corresponding to (a).

Fig. 14 is an example of the activity relation priority determination.

Detailed Description

All of the features disclosed in this specification, or all of the steps in any method or process so disclosed, may be combined in any combination, except combinations of features and/or steps that are mutually exclusive.

Any feature disclosed in this specification (including any accompanying claims, abstract) may be replaced by alternative features serving equivalent or similar purposes, unless expressly stated otherwise. That is, unless expressly stated otherwise, each feature is only an example of a generic series of equivalent or similar features.

A. the process model is expanded into a branch process containing truncation events using a full prefix expansion method.

B. And extracting the activity relation group which satisfies the reconfigurability from the branch process. So-called activity relationship sets, i.e., combinations of activities associated in the process model, such as a sequential relationship between two activities of adjacent levels, a parallel/concurrent/selective relationship between two activities of the same level, etc. The reconfigurable activity relationship group can be judged by the definition of the reconfigurable relationship, that is, by comparing the relationship of the activity relationship component with the defined reconfigurable relationship, whether the activity relationship group is the reconfigurable activity relationship group or not can be determined, or further, which kind of reconfigurable activity relationship is determined.

The reconfigurable relations comprise a reconfigurable sequential relation, a reconfigurable iterative relation, a reconfigurable selection relation and a reconfigurable concurrent relation. The corresponding definitions are as follows:

1. reconfigurable sequential relationship judgment conditions: for process model p = (C, A; F, M0), Π = (o, h) is a branch process corresponding to process model p, which includes a truncation event, wherein o = (S, T; F') is an occurrence net, h is a mapping function from a transition set to an active set, C is a condition set, A is an active set,

Referred to as a transition, is asserted>

For arbitrary +>

Its activity mapped into the process model is a1, a2, i.e. h (t 1) = a1, h (t 2) = a2; when t1 and t2 satisfy the following conditions, a1 and a2 have a reconfigurable sequential relationship, which is denoted as → (a 1, a 2).

(1) T1= [ t2] - { t2}, i.e., the configuration of t1 differs from the configuration of t2 by only t2;

(2)

i.e. local configuration t1]The set of activities mapped into the process model p does not contain the activities that event t2 maps into the process model p.

(3) If there is T2'∈ T, satisfying T2' ≠ T2 ^ h (T2 ') = h (T2), for all T1' satisfying [ T1'] ≠ T2' } = [ T2'] there must be h (T1') = h (T1).

(4) There is no T ∈ T, corr (T) = T1 is satisfied unless h (T) = h (T1) Λ [ T2] = [ T ] - { T }.

(5) For all T ∈ T, when T ≠ T1 ^ T ≠ T2, the relationship between event T and event T1 on the unfolded net is the same as that between event T and event T2, i.e. T # T1 ^ T # T2 or T co T1 ^ T co T2 or T > T1 ^ T > T2 or T < T1 ^ T < T2.

When a reconfigurable sequential relationship exists between a1 and a2, a2 must occur after activity a1 occurs. When t1 and t2 satisfy the four conditions (1), (2), (3) and (4) of definition 1, a1 and a2 can be determined to be in a sequential relationship, but only when (5) is satisfied at the same time, a1 and a2 can be determined to have a reconfigurable sequential relationship. Condition (1) of definition 1 guarantees that t2 can occur directly after t1, determining that a2 can be reached directly from a 1; condition (2) ensures that a2 does not occur before a 1; condition (3) condition (4) determines that what happens before a2 must be a1, while excluding the possibility that a2 has an iterative relationship with a 1; the condition (5) determines that the reconstruction of a1 and a2 does not affect other relations; when t1 and t2 meet the condition of definition 1, the judgment conditions of the reconfigurable iteration relationship, the selection relationship and the concurrency relationship are not met.

The sequential relationship is ordered, with the meaning of existence → (a 1, a 2) and → (a 2, a 1) being different. The iterative relationships of all activities can be extracted as long as the sequential relationships are traversed once, but after a plurality of activities are reconstructed, new activities collect direct predecessors of old activities, and new sequential relationships may appear.

For example, in fig. 10 (b), event a and event b, [ a ] = { a }, [ b ] = { a, b }, and a and b satisfy five conditions defining 1 at the same time, so that a and b have a reconfigurable sequential relationship, i.e., → (a, b), and similarly → (b, c) and → (c, d). Therefore → (a, b) after reconstruction → (b, c) cannot be reconstructed, and therefore the process tree corresponding to fig. 10 (a) is → (→ (a, b), → (c, d)).

2. Reconfigurable iterative relationship judgment conditions: for process model p = (C, a; F, M0), Π = (o, h) is a branch process corresponding to process model p that includes a truncation event, where o = (S, T; F') is a presence net and h is a mapping function.

h (t 1) = a1 and h (t 2) = a2, and when t1 and t2 meet the following conditions, a1 and a2 have a reconfigurable iterative relationship, wherein a1 occurs first and is marked as ∈ (a 1 and a 2), and a2 occurs first and is marked as ∈ (a 2 and a 1).

(1) T1= t 2-t 2, i.e. the configuration of t1 and the configuration of t2 differ by only t2;

(2) [ t1] = [ corr (t 2) ] - { corr (t 2) }, i.e., t2 is a truncation event, and the configuration of its corresponding event corr (t 2) and the configuration of t2 differ by only corr (t 2);

(3) If h (t 2) ≠ h (corr (t 2)), a1 and a2 have a reconfigurable iteration relation, and a1 occurs first;

(4) If h (T2) = h (corr (T2)), only if T ∈ T and T ≠ T1 ^ T ≠ T2 for all T ∈ T, if there is no reconfigurable relationship between T and T1, T2, there is a reconfigurable iterative relationship between a1 and a2, and a2 occurs first.

When a reconfigurable iterative relationship exists between a1 and a2, whether a2 occurs or not can be selected by activating a1, and a1 can occur again after a2 occurs. When t1 and t2 satisfy two conditions (1) and (2) of definition 2, it can be determined that a1 and a2 are in an iterative relationship, but only when (3) and (4) are satisfied, it can be determined that a1 and a2 have a reconfigurable iterative relationship.

Condition (1) of definition 2 guarantees that t2 can occur directly after t1, determining that a2 can be reached directly from a 1; the condition (2) excludes the possibility that t1 and t2 have an order relationship, and determines that a1 can be directly reached from a 2. Condition (3) determines that a reconfigurable iterative relationship exists between a1 and a2, and a1 occurs first; the condition (4) excludes the condition that a1 and a2 are not reconfigurable, and determines that a reconfigurable iterative relationship exists between a1 and a2, and a2 occurs first; when t1 and t2 satisfy the condition of definition 2, the judgment conditions of the reconfigurable iteration relationship, the selection relationship and the concurrency relationship are not satisfied any more. The iterative relationship is ordered, and the existence of ℃. (a 1, a 2) differs in meaning from ℃. (a 2, a 1). The priority of the reconfigurable iterative relationship is lower than that of the other three relationships, so that the reconfigurable iterative relationship can be judged only after the other relationships are judged. The reconfigurable iterative relationship can be extracted once through traversal, but when a plurality of activities are reconfigured, new activities collect direct predecessors of old activities, and new iterative relationships may appear.

For example, in fig. 11 (b), event a2 and event c, [ a2] = { a1, c, a2}, [ c ] = { a1, c }, [ corr (a 2) ] = [ a1] = { a1}, and a2 and b satisfy the two conditions (1) (2) that define 2 but do not satisfy the condition (3) or (4), and there is no reconfigurable iterative relationship between a and c. However, when a and b are reconstructed, x (a, b) and c satisfy the condition of definition 2, and there is a reconfigurable iterative relationship therebetween, so that the process tree corresponding to fig. 11 (a) is ∈ (× (a, b), c).

3. Reconfigurable selection relation judgment conditions: for process model p = (C, a; F, M0), Π = (o, h) is a branch process corresponding to process model p that includes a truncation event, where o = (S, T; F') is a presence net, and h is a mapping function.

h (t 1) = a1, h (t 2) = a2, and when t1 and t2 satisfy the following conditions, a1 and a2 have a reconfigurable selective relationship and are marked as x (a 1, a 2) or x (a 2, a 1). />

(1) t1# t2, i.e., the collision of t1 with t2;

(2) T 1-t 2, i.e., the occurrence conditions of t1 and t2 are the same;

(3)

or>

I.e., the last set of t1 and the last set of t2 are both empty or intersect.

When a1 and a2 have a reconfigurable selection relationship, a1 and a2 conflict, one happens and the other does not happen, and after a1 or a2 happens, the reachable activities are consistent. When t1 and t2 satisfy the condition (1) defined in definition 3, a1 and a2 can be determined to be in the selection relationship, but only when the conditions (2) and (3) are satisfied, the reconfigurable selection relationship between a1 and a2 can be determined.

Condition (1) of definition 3 guarantees that t1 and t2 are conflicting, one occurs and the other does not; condition (2) determines that the event experienced when t1 is reached is the same as when t2 is reached; condition (3) determines that a1 is consistent with the activity that can be reached after a 2.

When t1 and t2 satisfy the condition of definition 3, the determination conditions of the reconfigurable selection relationship, selection relationship and concurrency relationship are not satisfied any more. The selection relationship is disordered, and the meanings of x (a 1, a 2) and x (a 2, a 1) are consistent. The reconfigurable selection relation can extract the selection relation of all activities only by one traversal, but after a plurality of activities are reconfigured, new activities collect direct predecessors of old activities, and new selection relations may appear.

For example, event a and event b in fig. 12 (b), which satisfy (1) a # b; (2)

(3) h (a) # h (b) = { S2}, and thus there is x (a, b). Since f and e satisfy the condition (1) of definition 3 but do not satisfy the remaining two conditions, f and e have a selective relationship but cannot be reconstructed. When d and f are reconstructed, → (d, f) is obtained, then e and → (d, f) satisfy the three conditions defined in definition 3, and there is a selective relationship between e and → (d, f).

4. Reconfigurable concurrency relationship judgment conditions: for process model p = (C, a; F, M0), Π = (o, h) is a branch process corresponding to process model p that includes a truncation event, where o = (S, T; F') is a presence net, and h is a mapping function.

h (t 1) = a1, h (t 2) = a2, and when t1 and t2 satisfy the following conditions, a1 and a2 have a reconfigurable concurrency relationship and are marked as | | (a 1, a 2) or | | (a 2, a 1).

(1)t1 co t2；

(2) Neither t1 nor t2 is a truncation event;

(3) T1 to t1, t2 to t2, t1 to t2, i.e., the occurrence conditions of t1 and t2 are the same;

(4) The existence of T ∈ T ≠ T1 ^ T ≠ T2 satisfies

If T does not satisfy the former, then for any T ∈ T ≠ T1 ≠ T2, there is ≠ whether or not>

When a1 and a2 have reconfigurable concurrency, a1 and a2 can occur simultaneously, and after a1 or a2 occurs, the next activity can be reached. When t1 and t2 satisfy the condition (1) defined in definition 3, a1 and a2 can be determined to be in a concurrent relationship, but only when the conditions (2), (3) and (4) are satisfied, a1 and a2 can be determined to have a reconfigurable concurrent relationship.

Condition (1) of definition 4 guarantees that the occurrence conditions of t1 and t2 do not conflict; condition (2) excludes the possibility that t1 or t2 has a selective relationship with other events; condition (3) determines that the event experienced upon reaching t1 is the same as that upon reaching t2; condition (4) determines that after both t1 and t2 occur, the next event can be reached.

When t1 and t2 satisfy the condition of definition 4, the determination conditions of the reconfigurable selection relationship, the selection relationship and the concurrency relationship are not satisfied any more. The concurrency relationship is disordered, and the meaning of | | (a 1, a 2) and | | (a 2, a 1) is consistent. The reconfigurable concurrency relationship can extract the concurrency relationship of all activities only through one traversal, but after a plurality of activities are reconfigured, new activities collect direct predecessors of old activities, and new concurrency relationships may appear.

For example, event b and event c in FIG. 13 (b), which satisfy (1) b co c; (2) neither b nor c is a truncation event; (3) [ b ] - { b } = [ c ] - { c } = { a }; (4) b · e = { S5 }. Λ c · e = { S6}, so there is | | (b, c). D and e satisfy conditions (1), (2) and (4) of definition 4, but do not satisfy condition (3), and thus d and e have a selective relationship but cannot be reconstructed. When b and c are reconstructed to obtain | | (b, c), then e and → (d, f) satisfy the four conditions defined by 4, and there is a concurrent relationship between b and c.

C. And D, according to a preset conflict judgment condition, when judging that no conflict exists between the extracted activity relation groups, executing.

If there is a conflict between the activity relationship groups, the activity (the elements in the activity relationship group) cannot be reconstructed, and the process model for plan conversion is not the process model equivalent to the process tree behavior.

D. And reconstructing the extracted activity relation group according to a preset rule, wherein the reconstructed process tree is the process tree which is obtained, namely the process tree equivalent to the process model behavior.

Example two

The computer-implemented process of the above-mentioned processes a-D is:

inputting: process model p = (C, A; F, M0);

and (3) outputting: a relationship matrix RM;

fair reference definition 1, subsequent judgment is as follows

For the explanation of each parameter in the flow, refer to the explanation section of technical terms above.

For the relationship matrix, it is necessary to determine whether the activities in the relationship matrix conflict, and if so, the activities cannot be reconstructed, that is, the process model is not a process model equivalent to the process tree behavior.

5. Determination conditions as to whether or not the active relationship conflicts (i.e., the above-described predetermined conflict determination conditions): for the process model p = (C, A; F, M0), the Relation Matrix (RM) is a relation matrix corresponding to the process model p, and when any a1, a2 belongs to A, t1 and t2 meet the following conditions, the activity relations in the relation matrix do not conflict.

(1) a1 and a2 have no relation or only one relation;

(2) When → (a 1, a 2) is present, the relationship that can exist between a1 and a2 and any a ∈ p.A ≠ a1 ≠ a2 is only → (a, a 1) or → (a 2, a);

(3) When ∈ (a 1, a 2) exists, no relationship can exist between a1 and a2 and any a ∈ p.A ≠ a1 ≠ a2;

(4) When x (a 1, a 2) or | (a 1, a 2) is present, there is only a selective or concurrent relationship between a1 and a2 and any a ∈ p.A ^ a ≠ 2.

In step D, the reconstruction for each activity relationship has a definition of priority (i.e., a reconstruction rule). The most preferred is the sequential relationship, after the sequential relationship is reconstructed, new non-sequential relationships are particularly easy to extract, and the new non-sequential relationships may conflict with currently extracted non-sequential relationships, but on the contrary, after the activities of two non-sequential relationships are reconstructed, a part of structural information may be lost, because the reconstruction of the concurrency, selection and iteration relationships may merge arrows respectively pointing to the two activities into one, so that new conflicts are not generated between the reconstructed new activities and the new relationships extracted again, and even conflicts between the activities before reconstruction and the activities of other original non-sequential relationships may disappear, so that the sequential relationship is preferentially reconstructed.

An activity with an iterative relationship does not generate a relationship with other activities, so the iterative relationship is prioritized behind the sequential relationship.

The remainder is to determine the priority of the selection relationship and the concurrency relationship. Obviously, the priority of both the selection relation and the concurrent relation is after the sequential relation. How to determine the priority relationship between the two is dependent on the behavior structure of the process model.

For example, in the process model of FIG. 14, the selection relationship between activity b and activity c takes precedence, and the concurrency relationship between process model (b) activity b and activity d takes precedence. How to judge the priority relationship of them, taking the process model in fig. 14 as an example, the three relationships of | (b, d), | (c, d) and × (b, c) have related activity elements, the priority needs to be judged, the activity d has the same relationship with b, c at the same time, the relationship is as good as 1,b, another relationship is as good as 2 between c, and the related activity elements having the relationship as good as 2 are the least, so the relationship between b, c is preferred.

In one embodiment, the computer-implemented process for determining the priority of an activity relationship is:

inputting: the process model p = (C, A; F, M0), and the corresponding relation matrix is RM (see the above embodiment);

and (3) outputting: judging the result and returning to a preferentially reconstructed relation group RL;

/>

by the method, a group of activities can be reconstructed to obtain a new process model. By continuously reconstructing the process model, when the process model has only one activity left, the name of the activity is the prefix expression of the process tree corresponding to the process model p.

In one embodiment, a computer-implemented method of process tree generation is:

inputting: process model procedure p

And (3) outputting: process tree expression

The method comprises the following steps:

/>

the pseudo code for the process model reconstruction algorithm delAFromP is shown below.

Inputting: the process model p = (C, A; F, M0), the activities that a1, a2 need to reconstruct, is the relationship between a1 and a2

And (3) outputting: reconstructed Process model p'

The method comprises the following steps:

the method deletes a certain group of activities a1 and a2 in the original process model, adds a reconstructed new activity newA, and the new activity newA is named by prefix expressions of the activities a1 and a 2. Then according to the updated activity set, some conditions which do not influence the relationship between the activities are deleted. Secondly, it is decided which flow relations to keep according to the updated activity set and condition set, note that although the new activity newA collects all the flow relations of the old activities a1 and a2, these flow relations still need to be judged whether to need to be kept. Finally, the new process model is returned.

EXAMPLE III

This example discloses the test procedure applied to the method of the present invention. According to this embodiment, the feasibility and effectiveness of the method of the above embodiment can be verified, and the efficiency of the generation can be directly observed.

The present embodiment uses a Process Log Generator (PLG) to generate one hundred different Process models, as shown in fig. 4, the example model includes 4 basic structures used in the analysis of the present invention, and the Process model generated by PLG is denoted as RD group; and constructing ten non-BSPM models equivalent to the process tree, recording the models as ED groups, and testing the two groups of models as test cases. In order to ensure the confidence of the test result, the unification of hardware during the test of the two groups of data needs to be ensured. In this example, an Intel (R) Core (TM) i5-6200U CPU@2.30GHz 2.40GHz PC with 8G RAM was selected as the process carrier.

The test procedure was designed as follows: (1) Visualizing the process models by using an open-source graphic visualization tool Graphviz for each group of process models; (2) According to the process model activity relation extraction method provided by the invention, the corresponding process tree is extracted and visualized; (3) Comparing the process model with the process tree to determine the accuracy of the process tree; (4) And counting the time for extracting the process tree and the activity number of the corresponding process model.

For the accuracy of the behavior equivalence process tree generation method, the process model and the corresponding process tree can be compared in a visualization mode. As shown in FIG. 7, the process tree generated by the method of the present invention is consistent with the logic of the original process model.

For the RD group of experimental models, the time for extracting the process tree and the number of activities corresponding to the process model are counted, and the result is shown in fig. 5. Fig. 6 shows the average activity time consumption statistics of the RD group experimental model, and it can be seen from the figure that the efficiency of the algorithm for converting the BSPM into the process tree is high, and the average time consumption to each activity tends to be 0.5 and also tends to be stable.

Fig. 7 shows the transformation of part of 10 experimental models in ED set experimental data, in which, (a) is the original Petri net, (b) is the process model equivalent to the behavior of the original Petri net, and (c) is the new Petri net constructed from the process tree. It can be seen that for most models, after they are converted into a process tree, their overall behavior structure can be easily observed. After a new Petri network is constructed according to the process tree, the front and rear Petri networks (namely 7a and 7 c) are compared, the new Petri network can be found out to simplify partial structures of the original Petri network, so that a plurality of useless library and flow relations on the behavior structure are reduced, the redundancy of an original process model is reduced, and the visualization effect is improved.

Example four

This example separately discloses two process models that are not suitable for the method of the present invention to convert into a behavior equivalence process tree.

1. There are a variety of relationships between a certain set of activities.

As shown in FIG. 8, 8 (a) is an original Petri net, and FIG. 8 (b) is an expanded net of the original Petri net. In the case of fig. 8, the expanded mesh in (b) is obtained by expanding the original model, and it can be detected on the expanded mesh that events T1 and T0, T1 and T2 satisfy the reconfigurable selective relationship determination condition, events T6_1 and T5 satisfy the reconfigurable parallel relationship determination condition, events T6_1 and T2_2, and events T5 and T0_2 satisfy the reconfigurable iterative relationship determination condition, and at this time, events x (T0, T1), × (T1, T0), × (T1, T2), × (T2, T1), | (T5, T6), | (T6, T5), | (T5, T0), | (T6, T2) can be extracted. Among them, it can be found that the four relationships ∈ (T5, T0) with × (T0, T1), × (T1, T0), | (T5, T6), | (T6, T5) conflict, and the four relationships ∈ (T6, T2) with × (T1, T2), × (T2, T1), | (T5, T6), | (T6, T5) conflict, and thus the case in fig. 8 cannot be converted into a process tree.

2. Relationships between a certain set of activities do not belong to a basic relationship.

Fig. 9 (a) is an original Petri net, fig. 9 (b) is an expanded net of the original Petri net, fig. 9 (c) is a model after active reconstruction, and fig. 9 (d) is an expanded net of (c). In case two, the expanded mesh (b) is detected, and it can be found that the events T3 and T2_1, and T1 and T0_1 satisfy the reconfigurable iterative relationship determination condition, and at this time ∈ (T0, T1) and ∈ (T2, T3) can be extracted. Oc (T0, T1) and oc (T2, T3) do not conflict, and when they are reconstructed, fig. 9 (c) is obtained, and when they are expanded, the expanded mesh of fig. 9 (d) is obtained, and at this time, a new relationship cannot be detected on the expanded mesh (d), but all the activities on fig. 9 (c) have not been reconstructed into one, so the case in fig. 9 cannot be converted into a process tree.

The invention is not limited to the foregoing embodiments. The invention extends to any novel feature or any novel combination of features disclosed in this specification, and to any novel method or process steps or any novel combination of steps disclosed.

Claims

1. A behavior equivalence process tree generation method based on complete finite prefix expansion is characterized by comprising the following steps:

B. extracting an activity relation group meeting a reconfigurable condition from the branch process; the method for judging whether the activity relation group meets the reconfigurable relation comprises the following steps:

b1: extracting a transition set from the branch process;

b2: according to the definition of a preset reconfigurable relationship, the relationship of each transition group in the transition set is distinguished, if the discrimination is passed, the activity relationship group corresponding to the transition group is judged to meet the reconfigurable relationship;

C. according to a preset conflict judgment condition, when the extracted activity relation groups are judged to have no conflict, executing D; the preset conflict judgment condition is as follows:

Referred to as a transition, is asserted>

For arbitrary->

1. a1 and a2 have no relation or only one relation;

4. when a1 and a2 have a reconfigurable selection relationship or a reconfigurable concurrency relationship, the relationship which can exist between a1 and a2 and any a epsilon p.A ^ a ≠ a1 ^ a ≠ a2 is only the selection relationship or the concurrency relationship;

D. selecting a group of activity relation groups with the highest priority from the extracted activity relation groups for reconstruction according to a preset priority judgment rule; the predetermined priority judgment rule is specifically as follows: judging the priority of the activity relation group according to the type of the reconfigurable relation to which the activity relation group belongs;

2. The behavioral equivalence process tree generation method according to claim 1, wherein the reconfigurable relationships include reconfigurable sequential relationships, reconfigurable iterative relationships, reconfigurable selection relationships, and reconfigurable concurrency relationships.

3. The behavior equivalence process tree generation method of claim 2, wherein the reconfigurable ordering relationship is determined by:

when t1 and t2 meet the following conditions, the activities a1 and a2 have a reconfigurable sequence relationship:

2. Configuration [ t1]]The set of activities mapped into the process model does not contain the activities mapped into the process model by event t2:

4. The behavior equivalence process tree generation method of claim 2, wherein the reconfigurable iterative relationship is determined by:

when t1 and t2 meet the following conditions, the activities a1 and a2 have a reconfigurable iterative relationship:

5. The behavior equivalence process tree generation method of claim 2, wherein the reconfigurable selection relationship is determined by:

when t1 and t2 meet the following conditions, activities a1 and a2 have a reconfigurable selection relationship:

1. t1 conflicts with t2;

2. the occurrence conditions of t1 and t2 are the same;

3. the postamble of t1 and the postamble of t2 are both empty or intersect.

6. The behavior equivalence process tree generation method of claim 2, wherein the reconfigurable concurrency relationship is determined by:

1、t1 co t2；

2. neither t1 nor t2 is a truncation event;

3. the occurrence conditions of t1 and t2 are the same;

4. the existence of T ∈ T ≠ T1 ^ T ≠ T2 satisfies

Or for any T epsilon T ≠ T1 ^ T ≠ T2, there is a ^>

7. The behavior equivalence process tree generation method according to claim 2, wherein the type of the reconfigurable relationship has a priority of: