A petri net scheduling method and system based on dead zone boundary search and resource control constraints

Abstract

The application provides a Petri net scheduling method and system based on dead zone boundary search and resource control constraints, and the method comprises the following steps: constructing a time Petri net to make it react to multi-workpiece parallel processing work, and determining the emission rule of the place transition in the time Petri net; running the time Petri net by using a single-step better control strategy according to the emission rule; adding the constraint condition of parallel processing to the control strategy based on the resource utilization rate during multi-workpiece parallel processing; and substituting the control strategy into a dead zone search scheduling algorithm to determine a high-quality scheduling scheme, and scheduling the procedures of multi-workpiece parallel processing according to the high-quality scheduling scheme. The system of the application is based on the above method. The application not only combines the single-step optimal control strategy and the greedy emission logic of the emission rule to perform a depth-first search, and compared with most Petri nets, the scheduling result can be quickly obtained, and the application further proposes an optimization mode based on the dead zone boundary compared with the general depth-first algorithm, and the search efficiency is further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of scheduling optimization, in particular to a Petri net scheduling method and system based on dead zone boundary search and resource control constraints. BACKGROUND

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute the prior art.

[0003] In the field of production and manufacturing, it is inevitable to schedule multiple workpieces in the same parallel process under combined equipment or complex production processes. However, with the increase of processing procedures and the limitation of parallel manufacturing equipment, traditional scheduling algorithms often fail to fully utilize the advantages of parallel processing, resulting in low production efficiency and resource waste. In addition, traditional scheduling algorithms often use heuristic methods or rule-based methods, which cannot fully consider the mutual influence between multiple workpieces and the overall processing efficiency.

[0004] To solve this problem, discrete event system simulation and time-based Petri net methods are widely used. Discrete event system simulation is a method of simulating the dynamic behavior of discrete events, which evaluates and optimizes the performance of the system by simulating the occurrence and processing of events in the system. Time Petri net is a powerful mathematical tool for expressing the behavior of discrete event systems. It can describe the concurrency and synchronization behavior of the system and provide rich analysis tools such as state space search and performance evaluation. However, in existing technologies, the research on scheduling problems of multiple workpieces in the same parallel process is still relatively limited.

[0005] There are still limited works in existing technologies that directly use such time Petri nets for search algorithms to solve processing problems. For example, the patent publication CN112073984A discloses a system scheduling method based on Petri net and heuristic search, which models the constituent elements in the manufacturing system using Petri net to obtain the Petri net model of the system; initializes the system resources in the Petri net model; uses the A-Star search algorithm based on cost estimation to search the Petri net model and obtain the partial reachable graph of the Petri net model; establishes a target function for system scheduling with the minimum system running cost; based on the partial reachable graph of the Petri net model and the target function, uses the improved fruit fly optimization algorithm to optimize the path in the Petri net model, and outputs the path that meets the minimum target function as the system scheduling path. Although this patent document uses A-star heuristic function to perform full graph pruning strategy for ordinary library time delay Petri net model, it cannot solve the problem of improving the processing efficiency of multiple workpieces.

[0006] Therefore, it is necessary to propose a new Petri net scheduling method, which combines time Petri net simulation of discrete event system and rule-based technology to solve the scheduling problem of multiple workpieces in the same parallel process, so as to improve the production efficiency. SUMMARY

[0007] The application provides a Petri net scheduling method and system based on dead zone boundary search and resource control constraints, which not only combines a single-step optimal control strategy and a greedy firing logic of firing rules for depth-first search, but also constructs a time Petri net, so that compared with most Petri nets, the scheduling result can be quickly obtained, and compared with general depth-first algorithms, an optimization mode based on a dead zone boundary is proposed, and the search efficiency is further improved.

[0008] The technical scheme for achieving the object of the application is as follows:

[0009] In one aspect, the application provides a Petri net scheduling method based on dead zone boundary search and resource control constraints, comprising the following steps:

[0010] The time Petri net is constructed to react to the parallel processing work of multiple workpieces, and the firing rules of the place transitions in the time Petri net are determined;

[0011] The time Petri net is run according to the firing rules and the single-step optimal control strategy;

[0012] The constraint conditions of parallel processing are added to the control strategy based on the resource utilization rate of parallel processing of multiple workpieces;

[0013] The control strategy is substituted into the dead zone search scheduling algorithm to determine a high-quality scheduling scheme, and the workpieces are scheduled according to the high-quality scheduling scheme.

[0014] Based on one aspect, in one embodiment of the application, the time Petri net is constructed to react to the parallel processing work of multiple workpieces, and the time Petri net comprises the following steps:

[0015] The place of the time Petri net is defined, the place comprises a workpiece place and a machine place, the workpiece place represents the number of workpieces waiting for processing on a workbench, each workpiece place has at least one mark, and each mark represents a workpiece process to be processed; the machine place represents the number of available machines, each machine place has at least one mark, and each mark represents a processing process;

[0016] The transition of the time Petri net is defined, the transition represents that a workpiece is taken out from the workpiece place and sent to a machine in the machine place for processing; or the transition comprises that a machine completes a process and sends the workpiece to a next machine for the next process or completes the processing of the workpiece.

[0017] Defining the arc of the timed Petri net, the arc includes input arc and output arc, the input arc represents the transition from the place, represents the precondition of the transition, the output arc represents the place from the transition, represents the change of the token in the place after the execution of the transition.

[0018] Based on one aspect, in an embodiment of the application, the determination of the firing rule of the place transition in the timed Petri net comprises:

[0019] Triggering the transition when the precondition is met;

[0020] If the token has been used, other transitions cannot use the token;

[0021] Different priorities are assigned to different transitions;

[0022] Only the transition within a certain time range can be executed;

[0023] In a specific case, a transition is allowed to preempt the place of other transitions.

[0024] Based on one aspect, in an embodiment of the application, the timed Petri net is run according to the firing rule using the single-step optimal control strategy, comprising:

[0025] The place, transition and arc of the timed Petri net are designed into a model, and a state vector is built in the model;

[0026] The depth-first search is used for the state vector, and only one action is selected as the transition of the next state for each state of the state vector;

[0027] During the transition process, the global clock corresponding to the model is changed as little as possible when the transition action is selected, that is, the single-step optimal new transition action is generated to fire.

[0028] Based on one aspect, in an embodiment of the application, the constraint condition of parallel processing is added to the control strategy based on the resource utilization rate during the parallel processing of multiple workpieces, comprising:

[0029] The place, transition and arc of the timed Petri net are designed into a model, and the overall processing route is obtained according to the model;

[0030] The constraint condition is configured for the similar or same function places in the overall processing route;

[0031] The resource production priority and the resource optimal allocation priority are inferred from the resource allocation relationship of different transitions in the model;

[0032] Based on the resource production priority and the resource optimal allocation priority, the place prone to resource allocation imbalance in the model is limited;

[0033] Mark the library which is easy to cause but easy to cause resource allocation uneven as bad mark.

[0034] In an embodiment of the present application, the control strategy is substituted into the dead zone search scheduling algorithm to determine a high-quality scheduling scheme, based on one aspect, comprising:

[0035] The mark in the current search is taken out from the model, it is judged whether it is a target state, if it is a target state, the search result is returned; otherwise, the next step is performed;

[0036] The priority of the transition in the model is set, the last global time increment is used as the sorting method by default, and the next state possible in the state corresponding to the transition is sorted;

[0037] The addition of constraint conditions and the judgment of dead zones are performed on the sorting, if the new state enters the dead zone or is constrained, expansion is not performed; otherwise, the most preferred mark is taken out for expansion according to the priority;

[0038] It is judged whether the new state has no enabled transition, if not, the above process is repeated to increase the depth for searching; if yes, the mark is returned and is placed in the dead zone;

[0039] It is judged whether the mark can continue to expand or return, if the mark cannot continue to return, it is indicated that the target state is unreachable; if the mark can continue to expand, the search outside the dead zone after the previous step of selecting and updating the dead zone is returned.

[0040] In the present application, the rate of the dead zone boundary search depends on the bad mark distribution and the processing length, and great advantages can be achieved in search efficiency, the workpiece entity in the model search algorithm is marked as different names and different coloring is performed on the target, and the processing equipment (non-resource library) and operation (transition) passed by the workpiece are collected in the model calculation.

[0041] In the present application, the causes of the quality decline of the model strategy include that the limited robot resources cause the processing to take one route as the dominant. In the present application, the resource allocation constraints of the model include that the remaining processing time of the workpiece in the above-mentioned one route and the remaining processing position are taken as the evaluation and the time delay compensation is added according to the route. The state constraint directly affects the increase of the dead lock amount for the algorithm, and the search oriented to the dead zone can make up for the deficiency in this aspect, forming a complement.

[0042] On the other hand, the present application provides a Petri net scheduling system based on dead zone boundary search and resource control constraints, comprising:

[0043] A construction unit constructs a model and determines the firing rule of the place transition in the time Petri net, and the time Petri net reflects the parallel processing work of multiple workpieces;

[0044] a running unit in which a control strategy is run, the running unit uses a single-step better control strategy to run the timed Petri net according to the firing rule;

[0045] a control unit provided with a control strategy, the model adds a constraint condition of parallel processing to the control strategy based on resource utilization rate during parallel processing of multiple workpieces;

[0046] a scheduling unit which substitutes the control strategy into a dead zone search scheduling algorithm to determine a high-quality scheduling scheme, and the model schedules the processes of parallel processing of multiple workpieces according to the high-quality scheduling scheme.

[0047] Based on another aspect, in an embodiment of the present application, the building unit defines a place of the timed Petri net, the place includes a workpiece place and a machine place, the workpiece place represents the number of workpieces waiting for processing on a workbench, each workpiece place has at least one token, and each token represents a process of a workpiece to be processed; the machine place represents the number of available machines, each machine place has at least one token, and each token represents a processing process;

[0048] The building unit defines a transition of the timed Petri net, the transition represents taking out a workpiece from the workpiece place and sending it to a machine in the machine place for processing; or, the transition includes that a machine finishes processing a process and sends it to a next machine for processing of a next process or finishes processing of the workpiece;

[0049] The building unit defines an arc of the timed Petri net, the arc includes an input arc and an output arc, the input arc represents a direction from the place to the transition, representing a precondition of the transition, and the output arc represents a direction from the transition to the place, representing a change of the token in the place after execution of the transition.

[0050] Based on another aspect, in an embodiment of the present application, the building unit determines a firing rule of the place transition of the timed Petri net, including:

[0051] triggering the transition when the precondition is met;

[0052] if the token has been used, other transitions cannot use the token;

[0053] allocating different priorities to different transitions;

[0054] the transition can be executed only within a specific time range;

[0055] allowing a transition to preempt the place of other transitions in a specific case.

[0056] Based on another aspect, in an embodiment of the present application, the running unit uses depth-first search for state vectors in the model, and only selects one action for each state of the state vector as the transition to the next state;

[0057] The running unit selects the transition action in the transition process, and the model corresponding global clock changes minimally, i.e., generates a new transition action of one-step optimality.

[0058] Based on another aspect, in an embodiment of the present application, the control unit adds the constraint condition of parallel processing to the control strategy based on resource utilization rate during multi-workpiece parallel processing, including:

[0059] The library, transition and arc of the time Petri net are designed as a model, and the overall processing route is obtained according to the model;

[0060] The constraint condition is configured for the similar or same function of the library in the overall processing route;

[0061] The resource production priority and the resource optimal allocation priority are inferred from the resource allocation relationship of different transitions in the model;

[0062] Based on the resource production priority and the resource optimal allocation priority, the library prone to cause uneven resource allocation is limited in the model;

[0063] The library prone to cause uneven resource allocation is marked as a bad mark.

[0064] Based on another aspect, in an embodiment of the present application, the scheduling unit substitutes the control strategy into the dead zone search scheduling algorithm to determine a high-quality scheduling scheme, including:

[0065] The mark in the current search is taken out from the model, and it is judged whether it is a target state. If it is a target state, the search result is returned. Otherwise, the next step is performed;

[0066] The priority of the transition in the model is set, and the last global time increment is used as the sorting method by default to sort the possible next states under the corresponding state of the transition;

[0067] The addition of constraint conditions and the judgment of dead zones are performed on the sorting. If the new state enters the dead zone or is constrained, it is not expanded. Otherwise, the most priority mark is taken out for expansion according to the priority;

[0068] It is judged whether the new state has no transition that can be enabled. If not, the above process is repeated to increase the depth for searching. If yes, the mark is returned and is placed in the dead zone;

[0069] If the mark can not continue to back off, it means that the target state is unreachable, and if the mark can continue to expand, return to the previous step to select the update dead zone and then search outside the dead zone.

[0070] Compared with the prior art, the present application has the beneficial effects that:

[0071] 1、The present application combines the single-step optimal control strategy and the greedy transmission logic of the transmission rule to perform the depth-first search, and compared with most Petri nets, the scheduling result can be quickly output.

[0072] 2、The present application proposes an optimization mode based on the dead zone boundary compared with the general depth-first algorithm, and further improves the search efficiency.

[0073] 3、The present application combines the resource allocation constraint in the control strategy, predicts the unreasonable scheduling scheme, and further makes the model search the scheduling result with higher quality in advance. BRIEF DESCRIPTION OF DRAWINGS

[0074] Figure 1 A Petri net scheduling method flow chart based on the dead zone boundary search and resource control constraint is provided for the present application.

[0075] Figure 2 A schematic diagram of the mark state information provided for the present application is provided.

[0076] Figure 3 A schematic diagram of the state transition realized by the Petri net association matrix provided for the present application is provided.

[0077] Figure 4 A schematic diagram of the resource constraint based on the path time difference and the limited priority provided for the present application is provided.

[0078] Figure 5 A flow chart of the dead zone boundary search algorithm with the control strategy provided for the present application is provided. DETAILED DESCRIPTION

[0079] The present application will be described in detail below in combination with the embodiments shown in the drawings, but it should be noted that these embodiments are not limitations of the present application, and equivalent transformations or substitutions of the functions, methods or structures made by those skilled in the art according to these embodiments are within the protection scope of the present application.

[0080] Example 1:

[0081] Please refer to Figure 1 , Figure 1A flowchart of a Petri net scheduling method based on dead zone boundary search and resource control constraints provided in an embodiment of the present invention. The Petri net scheduling method based on dead zone boundary search and resource control constraints provided in an embodiment of the present invention includes:

[0082] Step 1: Construct a time Petri net to reflect the parallel processing of multiple workpieces and determine the emission rules of the library transition in the time Petri net.

[0083] The above-mentioned construction of a timed Petri net to reflect the parallel processing of multiple workpieces includes defining places in the timed Petri net, defining transitions in the timed Petri net, and defining arcs in the timed Petri net. The places in the timed Petri net are defined, and the places include workpiece places and machine places. Workpiece places represent the number of workpieces waiting for processing on a workbench, and each workpiece place has at least one tag, each tag representing a process for processing a workpiece. Machine places represent the number of available machines, and each machine place has at least one tag, each tag representing a process. Transitions in the timed Petri net are defined, and a transition represents the removal of a workpiece from a workpiece place and its transfer to a machine in a machine place for processing. Alternatively, a transition may involve the previous machine completing a process and transferring it to the next machine for the next process or completing the current workpiece. Arcs in the timed Petri net are defined, and the arcs include input arcs and output arcs. Input arcs indicate the transition from a place to a transition and represent the precondition for the transition. Output arcs indicate the transition from a transition to a place and represent the change in the place tag after the transition is executed.

[0084] The emission rules of the place transition in the above-mentioned time-determined Petri net include: triggering the transition when the precondition is met; if the current identifier is already used, other transitions cannot use the same identifier; assigning different priorities to different transitions; only transitions within a specific time range can be executed; and allowing a transition to preempt the place of other transitions under specific circumstances.

[0085] It should be noted that, when constructing the time Petri net simulation system emission logic, the embodiment of the present invention includes: (1) constructing a state vector, which contains comprehensive information of the designed model, including resource location, number of control library resources, remaining time for resource processing, remaining time for action completion, and corresponding system clock information; the basic identifier can be expressed as M = {P, T Token ,Time,T res}, where set P is the set of token numbers in the place, T Token To record the set of clocks in the token, Time is the time to reach this mark, T res is the remaining completion time of the corresponding action; the resource location and the number of resources in the control library are recorded in the set P, and the remaining time for resource processing is recorded in TToken The Petri net state generation relationship is expressed by the incidence matrix, and the corresponding information transformation relationship is filled in between the states as the actual firing logic.

[0086] It should be noted that the scheduling method is more applied to a physical model of a manufacturing system taking the S3PR net as a prototype, wherein a place represents a module for implementing processing and a resource control place for controlling capacity, a transition represents a control operation and the like related to scheduling actions, and a token in a processing place represents an actual workpiece. Figure 2 The schematic diagram of the state information provided by the embodiment of the present application is shown in the following figure, Figure 2 M i The identification information is represented by M, Time represents a system clock, T res All transition residual clocks are represented by T Token The residual processing time of all resources in each resource place is represented by T Figure 3 The schematic diagram of the state transition between Petri nets provided by the embodiment of the present application is shown in the following figure.

[0087] Step two, running the time Petri net according to the firing rule using a single-step optimal control strategy.

[0088] The scheduling method of the embodiment of the present application is based on a Petri search framework, and the main part of the scheduling method adopts a depth-first search, and the state with the minimum global time in a single step is selected between different depths to update. The above-mentioned running of the time Petri net according to the firing rule using a single-step optimal control strategy includes that the places, transitions and arcs of the time Petri net are designed into a model, and a state vector is built in the model; the depth-first search is used for the state vector, and only one action is selected for each state of the state vector as the transition of the next state; in the transition process, the transition action is selected, and the model corresponding to the minimum global clock change is generated to fire the single-step optimal new transition action.

[0089] In actual application, the depth-first search is used for the state vector, and only one action is selected for each state as the transition of the next state; in the transition process, the transition action is selected, and the single-step optimal (minimum global clock change) new state transition action is fired.

[0090] Step three, add the constraint condition of parallel processing to the control strategy based on the resource utilization rate of multi-workpiece parallel processing.

[0091] The above-mentioned adding of the constraint condition of parallel processing to the control strategy based on the resource utilization rate of multi-workpiece parallel processing comprises: the places, transitions and arcs of the time Petri net are designed into a model, and the overall processing route is obtained according to the model; the constraint condition is configured to the similar or same function places in the overall processing route; the resource production priority and the resource optimal allocation priority are obtained by reasoning from the resource allocation relationship of different transitions in the model; the places prone to occur but easy to cause uneven resource allocation in the model are limited based on the resource production priority and the resource optimal allocation priority; and the places prone to occur but easy to cause uneven resource allocation are marked as bad marks.

[0092] In actual application, the overall processing route is obtained according to the actual physical model; the constraint condition is configured to the similar or same function places in the processing route, and the resource production priority and the resource optimal allocation priority are obtained by reasoning from the resource allocation relationship; the resource places prone to occur but easy to cause uneven resource allocation are limited based on the priority, and this state is regarded as a bad mark, that is, TokenN T represents the remaining processing time of the resources in the ordinary resource processing place, T TokenL T represents the remaining processing time of the resources in the place prone to occur but easy to cause the solution quality to decrease, T Trackfix T represents the time difference of the process compared with another process.

[0093] In a complex physical model, the transition action is easily limited by resource allocation, resulting in uneven allocation of control actions, and then the local convergence is easily generated by using the above-mentioned greedy search idea, which is usually manifested in the physical model as that the processing route is not fully utilized due to the influence of the robot arm control resource, and a period of idling is generated. Figure 4 The schematic diagram of the resource constraint based on the path time difference and the limited priority provided by the embodiment of the present application, Figure 4 In the above-mentioned schematic diagram, a represents a robot arm a, Ta, Ta1 and Ta2 are controlled by the robot arm a; b represents a robot arm b, Tb, Tb1 are controlled by the robot arm b, and the model itself is prone to generate the control strategy of Tb1 priority processing. The constraint condition directly adds the resource constraint condition by obtaining information from the path angle, and the constraint condition can predict and process the local convergence solution caused by the limited robot arm resource.

[0094] ​The adjustment of the control strategy based on resource utilization analysis can retain fast search and improve the quality of search results for the local convergence of the search results, and the parallel processing constraint directly restricts the idling of resources by directly restricting the remaining processing time, thereby balancing the resource utilization efficiency of each core position.

[0095] Step four, substituting the control strategy into the dead zone search scheduling algorithm to determine a high-quality scheduling scheme, and scheduling the processes of the parallel processing of multiple workpieces according to the high-quality scheduling scheme.

[0096] Please refer to Figure 5 , Figure 5 The flowchart of the dead zone search scheduling algorithm provided by the embodiment of the present application, Figure 5 The graph search method in the flowchart is compatible with the priority strategy and the constraint strategy, and is based on a backtracking method, records the state set in the dead zone as a boundary for backtracking in the search, and can be used for different intelligent methods and improve search efficiency.

[0097] The dead zone search scheduling algorithm provided by the embodiment of the present application determines a high-quality scheduling scheme, including:

[0098] Step (1), taking out the identifier in the current search from the model, judging whether it is a target state, if yes, returning the search result, otherwise, proceeding to the next step;

[0099] Step (2), setting a priority for the transition in the model, and using the lowest global time increment as the sorting method by default, and sorting the next state that may exist in the state corresponding to the transition;

[0100] Step (3), adding a constraint condition to the sorting and judging the dead zone, if the new state enters the dead zone or is constrained, not expanding, otherwise, taking out the most preferred identifier for expansion according to the priority;

[0101] Step (4), judging whether the new state has no enabled transition, if not, repeating step (3) to search in depth, if yes, returning the identifier and placing it in the dead zone;

[0102] Step (5), judging whether the identifier can continue to expand or backtrack, if the identifier cannot continue to backtrack, it means that the target state is unreachable, if the identifier can continue to expand, returning to step (1) to select and update the dead zone and search outside the dead zone.

[0103] In the embodiment of the present application, the rate of the dead zone boundary search depends on the bad mark distribution and the processing length, and great advantages can be achieved in search efficiency, the model search algorithm internally marks the workpiece entity as different names and performs different coloring on the target, and the model calculation collects the processing equipment (non-resource library) and operation (transition) passed by the workpiece.

[0104] In the embodiment of the present application, the causes of the model strategy quality decline include that the limited mechanical arm resources cause the processing to take a route as the dominant. In the embodiment of the present application, the resource allocation constraints of the model include that the state constraint is taken as the dominant, the remaining processing time of the workpiece in a route and the remaining processing position are taken as the evaluation, and the time delay compensation is added according to the route. The state constraint directly affects the algorithm to increase the amount of deadlock, and the search facing the dead zone can make up for the deficiency in this aspect, forming a complement.

[0105] The embodiment of the present application is aimed at the existing manufacturing system processing scheduling algorithm, which defines a search algorithm with a minimum search cost, so that it has more implementation value; secondly, the idea of greedy launch is used in the front end of the search algorithm, which allows the part of the unfinished processing to move after processing, reduces the time property deadlock situation; in addition, the algorithm proposes a constraint mode of processing route based on the processing characteristics, which ensures the quality of the scheduling result while quickly calculating.

[0106] The embodiment of the present application combines single-step optimization and greedy launch logic to perform depth-first search, which can quickly output scheduling results compared to most algorithms. Compared with the general depth-first algorithm, the embodiment of the present application proposes an optimization method based on the dead zone edge, which can further improve the search efficiency, and is also an effective solution to the deadlock problem in model design. The embodiment of the present application combines resource allocation constraints, which can predict unreasonable scheduling schemes that cause the quality of the solution in the greedy search to decline, and then search for higher quality scheduling results in advance. The embodiment of the present application proposes a core search framework of the greedy algorithm, which supports compatibility of different constraint conditions and priorities for further optimization of the algorithm.

[0107] Embodiment 2:

[0108] Based on the Petri net scheduling method based on dead zone boundary search and resource control constraint disclosed in embodiment 1, the embodiment of the application provides a Petri net scheduling system based on dead zone boundary search and resource control constraint, which comprises a construction unit, a running unit, a control unit and a scheduling unit, wherein: the construction unit constructs a model and determines the firing rule of the place transition in the time Petri net, and the time Petri net reacts to the parallel processing of multiple workpieces. The running unit runs with a control strategy, and the running unit uses the single-step optimal control strategy to run the time Petri net according to the firing rule. The control unit is provided with a control strategy, and the model adds the constraint condition of parallel processing to the control strategy based on the resource utilization rate during the parallel processing of multiple workpieces. The scheduling unit substitutes the control strategy into the dead zone search scheduling algorithm to determine an optimal scheduling scheme, and the model schedules the process of the parallel processing of multiple workpieces according to the optimal scheduling scheme.

[0109] In actual application, the construction unit of the embodiment of the application defines the places of the time Petri net, the places include workpiece places and machine places, the workpiece places represent the number of workpieces waiting for processing on the workbench, each workpiece place has at least one token, and each token represents a process of a workpiece to be processed; the machine places represent the number of available machines, each machine place has at least one token, and each token represents a processing process.

[0110] The construction unit of the embodiment of the application defines the transitions of the time Petri net, the transitions represent that a workpiece is taken out from the workpiece place and sent to a machine in the machine place for processing; or the transitions include that a machine finishes processing a process and sends the workpiece to a next machine for processing a next process or finishes processing the workpiece.

[0111] The construction unit of the embodiment of the application defines the arcs of the time Petri net, the arcs include input arcs and output arcs, the input arcs represent that the places point to the transitions, represent the precondition of the transitions, and the output arcs represent that the transitions point to the places, represent that the tokens in the places change after the execution of the transitions.

[0112] The construction unit of the embodiment of the application determines the firing rule of the place transition in the time Petri net, which comprises: triggering the transition when the precondition is met; if the token has been used, other transitions cannot use the token; different priorities are allocated to different transitions; only the transitions within a specific time range can be executed; and one transition is allowed to preempt the places of other transitions under specific conditions.

[0113] The running unit of the embodiment of the application uses the depth-first search for the state vector in the model, and only selects one action as the transition of the next state for each state of the state vector; in the transition process, the running unit selects the transition action, so that the corresponding global clock of the model changes the least, that is, a single-step optimal new transition action is generated to be fired.

[0114] The control unit of the embodiment of the application adds the constraint condition of parallel processing to the control strategy based on resource utilization ratio during multi-workpiece parallel processing, including: the place, transition and arc of the time Petri net are designed as a model, and the overall processing route is obtained according to the model; the constraint condition is configured to the similar or same function place in the overall processing route; the resource production priority and the resource optimal allocation priority are obtained by reasoning the resource allocation relationship of different transitions in the model; the place prone to occur but easy to cause uneven resource allocation in the model is limited based on the resource production priority and the resource optimal allocation priority; and the place prone to occur but easy to cause uneven resource allocation is marked as a bad mark.

[0115] The scheduling unit of the embodiment of the application determines a high-quality scheduling scheme by substituting the control strategy into the dead zone search scheduling algorithm, including: taking out the mark in the current search from the model, judging whether it is a target state, if it is a target state, returning the search result; otherwise, proceeding to the next step; setting the priority of the transition in the model, using the lowest global time increment as the sorting method by default, and sorting the next state that may exist in the state corresponding to the transition; adding the constraint condition and judging the dead zone for the sorting, if the new state enters the dead zone or is constrained, not expanding; otherwise, taking out the most preferred mark for expansion; judging whether the new state has no enabled transition, if not, repeating the above process to increase the depth for searching; if yes, returning the mark and placing it in the dead zone; judging whether the mark can continue to expand or retreat, if the mark cannot continue to retreat, it means that the target state is unreachable; if the mark can continue to expand, returning to the previous step to select and update the dead zone and then searching outside the dead zone.

[0116] The embodiment of the application combines single-step optimal and greedy launch logic to perform depth-first search, and compared with most algorithms, can quickly output scheduling results. Compared with general depth-first algorithms, the embodiment of the application proposes an optimization method based on the edge of the dead zone, which can further improve the search efficiency, and is also an effective solution to the deadlock problem in model design. The embodiment of the application combines resource allocation constraints, and can predict unreasonable scheduling schemes that cause the solution quality to decrease in the greedy search, and then search for higher-quality scheduling results in advance. The embodiment of the application proposes a core search framework of the greedy algorithm, which supports compatibility of different constraint conditions and priorities for further optimization of the algorithm.

[0117] The embodiment of the application not only combines single-step optimal control strategy and greedy launch logic of launch rules to perform depth-first search, and compared with most Petri nets, can quickly output scheduling results, but also constructs a time Petri net, which proposes an optimization method based on the boundary of the dead zone compared with general depth-first algorithms, and further improves the search efficiency.

[0118] The above detailed description merely illustrates feasible implementation manners of the present application, and is not intended to limit the protection scope of the present application. Equivalent implementation manners or changes made without departing from the spirit of the present application shall be included in the protection scope of the present application.

[0119] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be regarded as exemplary and non-limiting, the scope of the present application is defined by the appended claims rather than the above description, and all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any reference signs in the claims should not be regarded as limiting the claims involved.

[0120] In addition, it should be understood that although the present specification is described in terms of embodiments, not every embodiment contains only one independent technical solution, and the specification is described in this way only for the sake of clarity, and those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be properly combined to form other embodiments that those skilled in the art can understand.

Claims

1. A Petri net scheduling method based on dead zone boundary search and resource control constraints, characterized in that: include: Construct a timed Petri net to reflect the parallel processing of multiple workpieces and determine the emission rules of place transitions in the timed Petri net; Run the temporal Petri net using a single-step better control strategy based on the emission rule; Adding parallel processing constraints to the control strategy based on resource utilization during multi-workpiece parallel processing; Substitute the control strategy into the dead zone search scheduling algorithm to determine the best scheduling solution, and schedule the multi-workpiece parallel processing process according to the best scheduling solution; The temporal Petri net is constructed to reflect the multi-workpiece parallel processing work, including: Define the places of the temporal Petri net. The places include workpiece places and machine places. The workpiece places represent the number of workpieces waiting to be processed on the workbench. Each workpiece place has at least one tag, and each tag represents a process for processing the workpiece. The machine places represent the number of available machines. Each machine place has at least one tag, and each tag represents a processing process. Define the transitions of the time Petri net. A transition means taking a workpiece from the workpiece library and sending it to a machine in the machine library for processing. Alternatively, a transition includes the previous machine completing a process and sending it to the next machine for the next process or completing the processing of the current workpiece. Define arcs of the time Petri net. The arcs include input arcs and output arcs. The input arc indicates that the place points to the transition and indicates the precondition of the transition. The output arc indicates that the transition points to the place and indicates that the label in the place changes after the transition is executed. The method of running the temporal Petri net using a single-step optimal control strategy according to the emission rule includes: The places, transitions and arcs of the time Petri net are designed as a model, and a state vector is built into the model; Use depth-first search on the state vector, and for each state in the state vector, select only one action as the transition to the next state; During the transition process, when selecting the transition action, the global clock change corresponding to the model is minimized, that is, a new transition action with the best single step is generated and emitted; Among them, the constraints of parallel processing are added to the control strategy based on the resource utilization during multi-workpiece parallel processing, including: The places, transitions and arcs of the time Petri net are designed into a model, and the overall processing route is obtained according to the model; The basis for configuring constraints on libraries with similar or identical functions in the overall processing route; The resource production priority and the optimal resource allocation priority are obtained by inferring the resource allocation relationship for different changes in the model; Based on resource production priority and resource optimal allocation priority, the model restricts the places that are prone to occur but are prone to cause uneven resource distribution; Mark the places that are prone to occur but may cause uneven resource distribution as bad identifiers; Substituting the control strategy into the dead zone search scheduling algorithm to determine a high-quality scheduling solution includes: Extract the identifier currently being searched from the model and determine whether it is the target state. If so, return the search result; otherwise, proceed to the next step. Set the priority for the transitions in the model. By default, the lowest global time increment is used as the sorting method to sort the possible next states of the corresponding state of the transition. Add constraints to the sorting and determine the dead zone. If the new state enters the dead zone or is constrained, it will not be expanded. Otherwise, the highest priority identifier will be taken out for expansion according to the priority. Determine whether the new state has no enabled transitions. If not, repeat the above process with increasing depth to search. If yes, roll back the flag and place it in the dead zone. Determine whether this identifier can continue to expand or roll back. If the identifier cannot continue to roll back, it means that the target state is unreachable; if the identifier can continue to expand, return to the previous step, select update dead zone, and then return to search outside the dead zone.

2. A Petri net scheduling method based on dead zone boundary search and resource control constraints according to claim 1, characterized in that: The emission rules for determining place transitions in the timed Petri net include: Trigger the transition when the precondition is met; If this logo has been used, other changes cannot use this logo; Assign different priorities to different transitions; Changes can only be implemented within a specific time frame; Under certain circumstances, a transition is allowed to seize the library place of other transitions.

3. A Petri net scheduling system based on dead zone boundary search and resource control constraints, characterized in that: include: A construction unit, wherein the construction unit constructs a model and determines an emission rule of a place transition in a time Petri net, wherein the time Petri net reflects the parallel processing of multiple workpieces; An operation unit, wherein a control strategy is executed in the operation unit, and the operation unit executes a temporal Petri net using a single-step optimal control strategy according to an emission rule; A control unit, wherein the control unit is provided with a control strategy, wherein the model adds a constraint condition of parallel processing to the control strategy based on resource utilization during parallel processing of multiple workpieces; a scheduling unit, wherein the scheduling unit substitutes the control strategy into the dead zone search scheduling algorithm to determine a high-quality scheduling solution, and the model schedules the process of parallel processing of multiple workpieces according to the high-quality scheduling solution; The construction unit defines the places of the temporal Petri net, which include workpiece places and machine places. The workpiece places represent the number of workpieces waiting to be processed on the workbench, and each workpiece place has at least one tag, each tag representing a process for processing the workpiece; the machine places represent the number of available machines, and each machine place has at least one tag, each tag representing a processing process; The building block defines the transition of the time Petri net, where the transition represents taking a workpiece from the workpiece library and sending it to a machine in the machine library for processing; or, the transition includes the previous machine completing a process and sending it to the next machine for the next process or completing the processing of the current workpiece; The construction unit defines arcs of the time Petri net, which include input arcs and output arcs. The input arc indicates that the place points to the transition and indicates the precondition of the transition. The output arc indicates that the transition points to the place and indicates that the label in the place changes after the transition is executed. The operation unit operates the temporal Petri net using a single-step optimal control strategy according to the emission rule, including: The places, transitions and arcs of the time Petri net are designed as a model, and a state vector is built into the model; Use depth-first search on the state vector, and for each state in the state vector, select only one action as the transition to the next state; During the transition process, when selecting the transition action, the global clock change corresponding to the model is minimized, that is, a new transition action with the best single step is generated and emitted; The control unit adds the constraint conditions of parallel processing to the control strategy based on the resource utilization rate during the parallel processing of multiple workpieces, including: The places, transitions and arcs of the time Petri net are designed into a model, and the overall processing route is obtained according to the model; The basis for configuring constraints on libraries with similar or identical functions in the overall processing route; The resource production priority and the optimal resource allocation priority are obtained by inferring the resource allocation relationship for different changes in the model; Based on resource production priority and resource optimal allocation priority, the model restricts the places that are prone to occur but are prone to cause uneven resource distribution; Mark the places that are prone to occur but may cause uneven resource distribution as bad identifiers; The scheduling unit substitutes the control strategy into the dead zone search scheduling algorithm to determine a high-quality scheduling solution, including: Extract the identifier currently being searched from the model and determine whether it is the target state. If so, return the search result; otherwise, proceed to the next step. Set the priority for the transitions in the model. By default, the lowest global time increment is used as the sorting method to sort the possible next states of the corresponding state of the transition. Add constraints to the sorting and determine the dead zone. If the new state enters the dead zone or is constrained, it will not be expanded. Otherwise, the highest priority identifier will be taken out for expansion according to the priority. Determine whether the new state has no enabled transitions. If not, repeat the above process with increasing depth to search. If yes, roll back the flag and place it in the dead zone. Determine whether this identifier can continue to expand or roll back. If the identifier cannot continue to roll back, it means that the target state is unreachable; if the identifier can continue to expand, return to the previous step, select update dead zone, and then return to search outside the dead zone.

Images