Task scheduling method and device, equipment and storage medium

Abstract

This application discloses a task scheduling method, apparatus, device, and storage medium, relating to the field of industrial production technology. The method includes: responding to a production task scheduling request; identifying all industrial scheduling entities and production constraints in production rule information, whereby production constraints characterize the constraint relationships between industrial scheduling entities during industrial production; generating a production constraint model based on each industrial scheduling entity and its corresponding production constraints, and generating candidate scheduling strategies based on the production constraint model; performing simulation evaluation on the candidate scheduling strategies, and optimizing the candidate scheduling strategies based on the simulation evaluation results to obtain a target scheduling strategy; and performing production task scheduling based on the target scheduling strategy. By applying the above technical solution, the technical problem that existing task scheduling methods often fail to guarantee the effectiveness and stability of scheduling strategies in real production environments is addressed.

Description

Technical Field

[0001] This application relates to the field of industrial production technology, and in particular to task scheduling methods, apparatus, equipment and storage media. Background Technology

[0002] In intelligent manufacturing systems, production scheduling is a core functional module of Manufacturing Execution System (MES) and Advanced Planning and Scheduling (APS). Its main task is to generate executable, optimizable, and adjustable scheduling schemes for production tasks while meeting constraints such as production process constraints, equipment capacity constraints, resource capacity constraints, logistics path constraints, and order delivery constraints. In modern industrial production scenarios such as multi-variety, small-batch, frequent order insertions, and dynamic changes in equipment status, production scheduling not only needs to generate static scheduling results but also needs to dynamically rearrange production tasks during production execution based on order changes, equipment anomalies, quality inspection feedback, and on-site operating status.

[0003] Currently, existing production task scheduling methods can use Large Language Models (LLMs) to describe the production scheduling problem as natural language requirements and leverage the semantic understanding and code generation capabilities of LLMs to directly output scheduling strategies, thereby guiding the execution of scheduled tasks. However, due to the multiple heterogeneous constraints involved in industrial production processes, such as process sequence, equipment capacity, resource capacity, logistics paths, and quality feedback, the scheduling strategies directly generated by large language models are often difficult to guarantee in terms of effectiveness and stability in real production environments, thus affecting the feasibility of industrial production. Summary of the Invention

[0004] The main objective of this application is to provide a task scheduling method, apparatus, device, and storage medium, which aims to solve the technical problem that the scheduling strategies generated by existing task scheduling methods are often difficult to guarantee the effectiveness and stability in real production environments.

[0005] To achieve the above objectives, this application proposes a task scheduling method, which includes: In response to a production task scheduling request, all industrial scheduling entities and production constraints in the production rule information are identified, wherein the production constraints are used to characterize the constraint relationships between the industrial scheduling entities in the industrial production process. A production constraint model is generated based on each of the industrial scheduling entities and the corresponding production constraints, and a candidate scheduling strategy is generated based on the production constraint model. The candidate scheduling strategies are evaluated through simulation, and the candidate scheduling strategies are optimized based on the simulation evaluation results to obtain the target scheduling strategy. Production tasks are then scheduled based on the target scheduling strategy.

[0006] In one embodiment, the step of generating a production constraint model based on each of the industrial scheduling entities and the corresponding production constraints includes: Obtain production constraint expression rules in a predefined domain-specific language, wherein the production constraint expression rules include object-level expression rules, constraint-level expression rules, event-level expression rules, and strategy-level expression rules; According to the object layer expression rules, each industrial scheduling entity is converted into an object layer statement, and the object layer statement is used to characterize the attribute features of the corresponding industrial scheduling entity. Based on the constraint layer expression rules, the event layer expression rules, and the strategy layer expression rules, each production constraint is converted into constraint layer statements, event layer statements, and strategy layer statements. The constraint layer statements are used to characterize the production constraints between each industrial scheduling entity. The event layer statements are used to characterize the triggering conditions and response rules of dynamic events in the industrial production process. The strategy layer statements are used to describe the condition information required for generating the scheduling strategy. A production constraint model is generated based on the object layer statements, the constraint layer statements, the event layer statements, and the strategy layer statements.

[0007] In one embodiment, before the step of generating a production constraint model based on each of the industrial scheduling entities and the corresponding production constraints, the method further includes: Determine the reliability of the rule source, the confidence level of entity identification, the field completeness, and the constraint consistency verification results for each of the aforementioned production constraints; The constraint confidence level corresponding to the production constraint is determined based on the reliability of the rule source, the confidence level of the entity identification, the completeness of the field, and the constraint consistency verification result. The constraint confidence level is compared with a preset confidence threshold. If the confidence level of the constraint is higher than the preset confidence threshold, the step of generating a production constraint model based on each of the industrial scheduling entities and the corresponding production constraints is performed.

[0008] In one embodiment, the step of generating candidate scheduling strategies based on the production constraint model includes: Prompt words are generated based on the production scenario description information, process constraint information, equipment capacity limitation information, optimization objectives, and the production constraint model construction strategy of the industrial production process. Based on the strategy, a prompt word is generated using a preset large language model to generate a description of the scheduling strategy concept and the corresponding executable strategy code. An initial scheduling strategy is generated based on the described scheduling strategy concept and the executable strategy code. The initial scheduling strategy is validated for executability to determine whether it conforms to the production constraint model. Based on the judgment result, the initial scheduling strategy following the production constraint model will be output as a candidate scheduling strategy.

[0009] In one embodiment, the step of performing simulation evaluation on the candidate scheduling strategy and optimizing the candidate scheduling strategy based on the simulation evaluation results to obtain the target scheduling strategy includes: The candidate scheduling strategy is input into several preset scheduling simulation environments to obtain the strategy evaluation index of the candidate scheduling strategy in the corresponding preset scheduling simulation environment; The comprehensive strategy score of the candidate scheduling strategy is determined based on the strategy evaluation index, and the performance improvement rate of the candidate scheduling strategy relative to the baseline scheduling strategy is determined based on the comprehensive strategy score. The fitness value of the candidate scheduling strategy is determined based on the performance improvement rate; Based on the fitness value, the candidate scheduling strategy is evolved to obtain the target scheduling strategy.

[0010] In one embodiment, the step of performing an evolutionary operation on the candidate scheduling strategy based on the fitness value to obtain the target scheduling strategy includes: Based on the fitness value, several parent scheduling strategies are determined from the candidate scheduling strategies; Extract the production scheduling rules from each of the parent scheduling strategies, and combine the production scheduling rules according to the rule category to generate the child scheduling strategy; The offspring scheduling strategy is determined as a new candidate scheduling strategy, and the process returns to the step of inputting the candidate scheduling strategy into several preset scheduling simulation environments until the candidate scheduling strategy meets the preset termination conditions, and the target scheduling strategy is output.

[0011] In one embodiment, after the step of scheduling production tasks based on the target scheduling strategy, the method further includes: Receive real-time feedback of on-site operating status information during the industrial production process; Based on the on-site operational status information, detect whether there are dynamic events in the industrial production process; If it exists, the scope of impact of the dynamic event is determined. The scope of impact includes a first scope and a second scope. The first scope is used to characterize the impact on a single device, a single process, or a single logistics path. The second scope is used to characterize the impact on multiple devices, multiple processes, or a group of bottleneck devices. If the scope of the event is the first scope of impact, and there are alternative equipment or alternative paths for the current production task, the production reordering type of the current production task is determined to be a local reordering method, and the production task is reordered according to the local reordering method. Alternatively, if the scope of the event is the second scope of the event, or if the dynamic event causes the delivery constraint to fail, the production rescheduling type of the current production task is determined to be the global rescheduling method, and the production task is rescheduled according to the global rescheduling method.

[0012] Furthermore, to achieve the above objectives, this application also proposes a task scheduling device, the device comprising: The constraint identification module is used to identify all industrial scheduling entities and production constraints in the production rule information in response to a production task scheduling request. The production constraints are used to characterize the constraint relationships between the industrial scheduling entities in the industrial production process. The constraint modeling module is used to generate a production constraint model based on each of the industrial scheduling entities and the corresponding production constraints, and to generate candidate scheduling strategies based on the production constraint model. The task scheduling module is used to perform simulation evaluation on the candidate scheduling strategies, optimize the candidate scheduling strategies based on the simulation evaluation results to obtain the target scheduling strategy, and schedule production tasks based on the target scheduling strategy.

[0013] In addition, to achieve the above objectives, this application also proposes a task scheduling device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the task scheduling method described above.

[0014] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and which, when executed by a processor, implements the steps of the task scheduling method described above.

[0015] This application provides a task scheduling method. In response to a production task scheduling request, the method identifies all industrial scheduling entities and production constraints within the production rule information. Production constraints characterize the constraint relationships between industrial scheduling entities during industrial production. A production constraint model is generated based on each industrial scheduling entity and its corresponding production constraints, and candidate scheduling strategies are generated according to the production constraint model. The candidate scheduling strategies are then evaluated through simulation, and optimized based on the simulation evaluation results to obtain a target scheduling strategy. Production tasks are then scheduled based on the target scheduling strategy. Because this application can generate a production constraint model based on industrial scheduling entities and their corresponding production constraints, generate candidate scheduling strategies based on the production constraint model, and then optimize the strategies based on the simulation evaluation results, thereby performing production scheduling based on the optimized target scheduling strategy, it solves the technical problem that existing task scheduling methods often struggle to guarantee the effectiveness and stability of scheduling strategies in real production environments. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating an embodiment of the task scheduling method of this application. Figure 2 The system architecture diagram provided for the task scheduling method of this application; Figure 3 This is a flowchart illustrating Embodiment 2 of the task scheduling method of this application; Figure 4 This is a flowchart illustrating Embodiment 3 of the task scheduling method of this application; Figure 5 This is a schematic diagram of the module structure of the task scheduling device according to an embodiment of this application; Figure 6 This is a schematic diagram of the device structure of the hardware operating environment involved in the task scheduling method in the embodiments of this application.

[0019] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0021] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0022] The main solution of this application embodiment is as follows: In response to a production task scheduling request, identify all industrial scheduling entities and production constraints in the production rule information. The production constraints are used to characterize the constraint relationship between each industrial scheduling entity in the industrial production process. Generate a production constraint model based on each industrial scheduling entity and its corresponding production constraints, and generate candidate scheduling strategies based on the production constraint model. Perform simulation evaluation on the candidate scheduling strategies, and optimize the candidate scheduling strategies based on the simulation evaluation results to obtain the target scheduling strategy, and perform production task scheduling based on the target scheduling strategy.

[0023] Because existing production task scheduling methods that use large language models to directly generate scheduling strategies often cannot guarantee effectiveness and stability in real production environments, they affect the feasibility of industrial production.

[0024] This application provides a solution that can generate a production constraint model based on an industrial scheduling entity and its corresponding production constraints, generate candidate scheduling strategies based on the production constraint model, optimize the strategies based on the simulation evaluation results of the candidate scheduling strategies, and perform production scheduling based on the optimized target scheduling strategy. This solves the technical problem that the scheduling strategies generated by existing task scheduling methods are usually difficult to guarantee the effectiveness and stability in real production environments.

[0025] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone; or an electronic device, task scheduling device, or an industrial production system containing task scheduling device, etc. The following description uses an industrial production system (hereinafter referred to as the system) as an example to illustrate this embodiment and the subsequent embodiments.

[0026] Based on this, embodiments of this application provide a task scheduling method, referring to... Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the task scheduling method of this application.

[0027] In this embodiment, the task scheduling method includes steps S10 to S40: Step S10: In response to the production task scheduling request, identify all industrial scheduling entities and production constraints in the production rule information, wherein the production constraints are used to characterize the constraint relationships between the industrial scheduling entities in the industrial production process.

[0028] It should be noted that a production task scheduling request can be an instruction or signal triggered by an industrial production system to initiate a scheduling process. This request may include basic information such as the reason for triggering the scheduling, the scope of tasks currently to be scheduled, and the scheduling target. In actual industrial production scenarios, production task scheduling requests can be triggered in various situations, such as when a new order is placed, the production plan cycle arrives, equipment failure or quality abnormalities occur on the production site requiring rescheduling, an emergency order is inserted, or the operator manually initiates the scheduling operation. This embodiment does not impose any limitations on these situations.

[0029] It should be understood that production rule information can be a collection of various rule data related to production scheduling decisions in the industrial production process. This rule data can be stored in different industrial information systems or documents, such as process documents, MES system configurations, equipment capacity databases, production management rule bases, historical production scheduling data, and expert experience bases. Specifically, process documents can record the sequence of product processing steps and the process parameter requirements for each step; MES system configurations can record basic information about production equipment, equipment group divisions, and default scheduling parameters; equipment capacity databases can store information such as the processing capacity range of each production piece of equipment, the types of processes it supports, processing speed, changeover time, and current operating status; production management rule bases can contain production management specifications established internally by the factory, such as order priority rules, delay penalty rules, and quality control rules; historical production scheduling data can record past scheduling schemes and their execution results; and expert experience bases are structured or semi-structured data that stores the scheduling knowledge and experience of senior schedulers in the form of rules.

[0030] It should be explained that industrial scheduling entities can be physical or logical objects identified from production rule information that are related to production scheduling decisions. These can include, but are not limited to, process entities, equipment entities, resource entities, time entities, logistics path entities, quality feedback entities, and dynamic event entities. Specifically, a process entity can represent a specific processing step or manufacturing activity, such as "drilling," "etching," or "lamination" in PCB manufacturing. These process entities typically have attributes such as process name, standard processing time, selectable processing time range, and whether quality inspection is required. Equipment entities can represent production equipment that performs processing tasks, such as "Drilling Machine No. 1," "Lamination Machine No. 2," or "AOI Inspection Instrument No. 3." These equipment entities typically have attributes such as equipment identification, the set of process types they support, equipment capacity, current status (e.g., idle, processing, faulty, or under maintenance), and changeover time. Resource entities can represent other consumable or usable resources besides equipment during the production process, such as tooling fixtures, molds, operators, carriers, or raw materials. These resource entities typically have attributes such as resource identification, resource type, capacity limit, and available time window. Time entities can represent time parameters related to production scheduling, such as order delivery dates, process durations, and equipment availability periods. Logistics path entities can represent the path information for material flow between processes, including the starting workstation, target workstation, set of possible paths, transportation time, and intermediate buffer capacity. Quality feedback entities can represent rules and feedback information related to quality inspection, such as inspection processes, defect types, handling methods for inspection results (e.g., pass, rework, scrap, or manual review), and downstream process constraints after rework or scrap. Dynamic event entities can represent abnormal or changing events that may occur during production, such as order insertion events, equipment failure events, quality anomaly events, workstation congestion events, or production delay events. These dynamic event entities typically have attributes such as event type, triggering conditions, the scope of affected objects, and the suggested reordering scope (e.g., local or global).

[0031] It should also be explained that production constraints can be the constraints between various industrial scheduling entities in the industrial production process, that is, the restrictions that each industrial scheduling entity must meet in the scheduling scheme. In this embodiment, production constraints can include process sequence constraints, equipment capacity constraints, resource capacity constraints, processing time constraints, logistics path constraints, quality feedback constraints, and dynamic event constraints. Among them, process sequence constraints are used to specify the sequential dependencies between various processes in the industrial production process, such as "the etching process must be completed before the drilling process" or "the drilling process must be completed before the copper plating process". Equipment capacity constraints are used to specify the equipment that executes the processes in the industrial production process, such as "the drilling process can be completed on drilling machine 1, drilling machine 2, or drilling machine 3". Resource capacity constraints are used to specify that the number of tasks allocated to equipment, workstations, buffer areas, or human resources within the same time window cannot exceed their maximum capacity, such as "each drilling machine can process a maximum of 2 PCB boards at the same time" or "the buffer area can hold a maximum of 10 work-in-process items". Processing time constraints specify the processing time range for each step in industrial production. For example, "The standard processing time for the drilling process is 15 minutes, which can be adjusted between 12 and 18 minutes." Logistics path constraints specify the accessibility and transportation time of material flow between adjacent processes. For example, "There are two logistics paths from the drilling process to the pressing process; path A has a transportation time of 5 minutes, and path B has a transportation time of 8 minutes." Quality feedback constraints specify the handling of defects discovered during quality inspection and their impact on downstream processes. For example, "After AOI inspection discovers a defect, a rework process is triggered, and after rework, the process re-enters the inspection process." Dynamic event constraints specify the response rules after various dynamic events occur. For example, "After drilling machine 1 malfunctions, its unfinished orders are switched to drilling machine 2 or drilling machine 3 for processing."

[0032] In its implementation, upon receiving a production task scheduling request, the system can initiate the task scheduling process. First, the system can read process flow documents from the process document management system, equipment configuration information, equipment capacity data, and current equipment status from the MES system database, processing capacity parameters for each piece of equipment from the equipment capacity database, order priority rules, delay penalty rules, and quality control rules from the production management rule base, past scheduling schemes and execution results from the historical production scheduling database, and structured scheduling rules compiled from the expert experience base, thereby obtaining production rule information for the industrial production process. After acquiring the production rule information, the system can use a pre-set large language model to perform semantic parsing of the production rule information, identifying all industrial scheduling entities contained in the production rule information and the production constraints between these entities. This pre-set language model can be a pre-trained or fine-tuned large language model with natural language understanding and code generation capabilities, capable of converting unstructured text descriptions, tabular data, and rule statements in the production rule information into structured entity and constraint data.

[0033] For example, suppose a PCB flexible manufacturing plant receives a new order for 100 PCBs to be processed. The processing flow includes material preparation, etching, drilling, copper plating, lamination, and AOI inspection. After receiving the production task scheduling request triggered by this new order, the system can first read the process description "drilling after etching, copper plating after drilling" from the process document, read the equipment configuration information "drilling can be completed on drilling machine 1, drilling machine 2, or drilling machine 3" from the MES system database, and read the equipment status information "drilling machine 1 and drilling machine 2 are currently idle, drilling machine 3 is in a fault state" from the equipment capacity database. Then, the system can use a large language model to perform semantic parsing on these production rule information, identify "process" as process entity, identify "drilling machine 1, drilling machine 2, drilling machine 3" as equipment entity, identify "drilling after etching" as process sequence constraint, and identify "drilling process can be completed on drilling machine 1, drilling machine 2 or drilling machine 3" as equipment capability constraint.

[0034] Step S20: Generate a production constraint model based on each of the industrial scheduling entities and the corresponding production constraints, and generate candidate scheduling strategies based on the production constraint model.

[0035] It should be noted that the production constraint model can be a computable model expressed in structured data form, used to describe the various industrial scheduling entities in the industrial production process and the production constraint relationships between them. In this embodiment, the production constraint model can be constructed using a pre-defined Domain-Specific Language (DSL) to convert unstructured natural language rules, tabular data, and expert experience into structured data that can be directly read and processed by a computer.

[0036] It should also be noted that the production constraint model can include multiple components such as process definition, equipment definition, resource definition, process sequence constraint, equipment capacity constraint, resource capacity constraint, processing time constraint, logistics path constraint, quality feedback constraint, and dynamic event constraint.

[0037] It should be understood that a candidate scheduling strategy can be one or more selectable scheduling schemes used to guide production task scheduling, which may include executable decision rules and corresponding implementation code. In this embodiment, a complete candidate scheduling strategy can be expressed in the form of a four-tuple: "scheduling strategy idea description + executable strategy code + fitness value + applicable scenario label". The scheduling strategy idea description can use natural language to describe the core decision logic of the scheduling strategy, such as "prioritizing scheduling orders with more urgent delivery dates", "prioritizing machines with lower equipment load", or "prioritizing reducing the number of logistics route switching times", etc.; the executable strategy code can be executable program code that implements the scheduling strategy idea, such as priority calculation functions and equipment selection functions written in Python, which can be directly executed by a computer to make actual scheduling decisions; the fitness value is the comprehensive score of the candidate scheduling strategy after simulation evaluation; the applicable scenario label can be used to identify the type of production scenario to which the scheduling strategy is applicable, such as "normal production scheduling", "order rescheduling", "equipment failure detour", or "quality anomaly feedback", etc.

[0038] In its implementation, the system can acquire production constraint expression rules in a predefined domain-specific language, including object-level expression rules, constraint-level expression rules, event-level expression rules, and strategy-level expression rules. Then, the system can convert each industrial scheduling entity into object-level statements based on the object-level expression rules. These object-level statements define the attribute characteristics of at least one object among orders, processes, equipment, and resources. Next, the system can convert each production constraint into constraint-level statements based on the constraint-level expression rules. These constraint-level statements define at least one constraint relationship among process sequence, equipment capacity, resource capacity, time window, logistics accessibility, and quality feedback rules. Simultaneously, the system can convert dynamic event information into event-level statements based on the event-level expression rules. These event-level statements define the triggering conditions and response rules for at least one dynamic event among order insertion, equipment failure, quality anomaly, and workstation congestion. Furthermore, the system can convert scheduling strategy information into strategy-level statements based on the strategy-level expression rules. These strategy-level statements define at least one strategy element among priority factors, equipment selection conditions, and reordering range. Finally, the system can combine the object-level statements, constraint-level statements, event-level statements, and strategy-level statements to generate a production constraint model. After generating the production constraint model, the system can construct strategy generation prompts based on the production scenario description, process constraints, equipment capacity limitations, optimization objectives, and the production constraint model. It then uses a pre-defined large language model to generate a scheduling strategy concept description and corresponding executable strategy code based on these prompts. Subsequently, the system generates an initial scheduling strategy based on the scheduling strategy concept description and executable strategy code. It then performs an executability check on the generated initial scheduling strategy, determining whether it adheres to the process sequence constraints, equipment capacity constraints, resource capacity constraints, path reachability constraints, and dynamic event constraints in the production constraint model. Finally, the initial scheduling strategy that conforms to the production constraint model is output as a candidate scheduling strategy. Furthermore, if the initial scheduling strategy violates any constraint in the production constraint model, it is determined to be an unexecutable strategy, triggering a regeneration.

[0039] Step S30: Perform simulation evaluation on the candidate scheduling strategy, optimize the candidate scheduling strategy based on the simulation evaluation results to obtain the target scheduling strategy, and schedule production tasks based on the target scheduling strategy.

[0040] Understandably, simulation evaluation can be the process of executing candidate scheduling strategies in a simulated production environment to obtain performance metrics of the strategy under multiple preset simulation scenarios. In practical applications, simulation evaluation can simulate the operation of industrial production processes through a scheduling simulation environment. This scheduling simulation environment can include components such as an order event generator, a process state simulator, an equipment state simulator, a logistics path simulator, a dynamic event injector, a scheduling strategy executor, and a metric calculator. Specifically, the order event generator generates events such as order arrival and order insertion; the process state simulator simulates the start, execution, and completion of processes; the equipment state simulator simulates equipment idle, processing, fault, and recovery states; the logistics path simulator simulates inter-station flow and buffer occupancy; the dynamic event injector injects events such as equipment faults, quality anomalies, and workstation congestion during the simulation; the scheduling strategy executor executes scheduling tasks in the simulation environment according to the decision logic of the candidate scheduling strategies; and the metric calculator calculates various evaluation metrics after the simulation.

[0041] It is also understood that the simulation evaluation results can be multi-dimensional evaluation index data output by the simulation evaluation process, such as total production cycle, order delay time, equipment utilization rate, work-in-process quantity, number of path switching, order insertion response time, fault detour success rate, strategy execution success rate, and reordering response time, etc. This embodiment does not limit this.

[0042] It should be noted that the target scheduling strategy can be the optimal scheduling scheme for actual production task scheduling that is finally selected during the simulation evaluation and optimization process. In other words, it is the scheduling strategy with the best overall performance determined after multiple rounds of simulation evaluation, strategy screening, and evolutionary optimization of multiple candidate scheduling strategies. In this embodiment, the target scheduling strategy can include complete scheduling decision logic, including process priority calculation rules, equipment selection rules, order sorting rules, logistics path allocation rules, and rearrangement rules when dynamic events occur.

[0043] In practical applications, after obtaining candidate scheduling strategies, the system can input these strategies into several simulation environments configured for different production scenarios to obtain strategy evaluation indicators for each simulation environment. These indicators include at least one of the following: order delay time, equipment utilization rate, work-in-process inventory, path switching frequency, order insertion response time, fault detour success rate, and strategy execution success rate. Then, the system can determine a comprehensive strategy score for the candidate scheduling strategy based on these evaluation indicators, and determine the performance improvement rate of the candidate scheduling strategy relative to the baseline scheduling strategy based on the comprehensive strategy score. The baseline scheduling strategy can be a randomly selected benchmark scheduling strategy used to measure the relative performance of the candidate scheduling strategy. Next, the system can determine a fitness value reflecting the comprehensive performance level of the candidate scheduling strategy under multiple preset simulation scenarios based on the performance improvement rate. Based on this fitness value, a parent strategy is selected from the candidate scheduling strategies. Then, at least one evolutionary operation—strategy parameter mutation, strategy rule mutation, and strategy cross-combination—is performed on the parent strategy to generate a child strategy. The child strategy is then input into the simulation environment for evaluation. Based on the evaluation results, strategies whose comprehensive scores meet preset conditions are selected as the target scheduling strategy from the child strategies. The strategy parameter variation can include adjusting at least one weight parameter among the delivery urgency weight, equipment load weight, logistics route cost weight, and quality risk weight in the scheduling strategy; the strategy rule variation can include adding, deleting, or replacing rule items in the candidate strategy; the strategy cross-combination can include extracting production scheduling rule fragments from two or more parent strategies, combining them according to rule categories, and generating new child strategies. Finally, the system can convert the target scheduling strategy into an executable task and issue the executable task to the production execution system, which will then carry out actual production according to the content of the executable task. The content of the executable task can include order identifier, process identifier, target equipment, planned start time, planned end time, logistics route, and exception handling rules, etc.

[0044] In the specific implementation, refer to Figure 2 , Figure 2 The system architecture diagram provided for the task scheduling method of this application is shown below. Figure 2As shown, the industrial production system in this application may include an industrial rule cognition module, a production constraint DSL modeling module, a constraint consistency and executability verification module, a scheduling strategy generation and verification module, a scheduling simulation evaluation module, a strategy evolution optimization module, a strategy issuance and execution feedback module, and a dynamic event rescheduling and on-site status feedback module. During actual production scheduling, the industrial rule cognition module first obtains production rule information from multiple data sources such as process documents, MES system configuration, equipment capacity database, production management rule base, historical production scheduling data, and expert experience base. It then uses a large language model to perform semantic parsing of the production rule information, identifies the industrial scheduling entities within it, extracts the production constraints between each industrial scheduling entity from the identification results, and outputs the identified industrial scheduling entities and extracted production constraints to the production constraint DSL modeling module. After receiving the industrial scheduling entity and production constraints, the production constraint DSL modeling module processes them according to a preset domain-specific language syntax to generate a production constraint model. This model is then output to the constraint consistency and executability verification module, which performs a consistency check to determine if there are any issues such as sequential conflicts, lack of available equipment, equipment capacity mismatch, resource capacity conflicts, or unreachable logistics paths. If none of these issues exist, the consistency check is considered successful. Following this, the module further verifies the executability of the production constraint model, determining if it can generate a feasible scheduling plan. If so, the executability check is considered successful. In this embodiment, if either the consistency or executability check fails, the failure information is fed back to the industrial rule cognition module, which then performs semantic parsing or generates a revised constraint description. If both the consistency and executability checks pass, the production constraint model is output to the scheduling strategy generation and verification module. After receiving the production constraint model, the scheduling strategy generation and verification module can generate an initial scheduling strategy based on the production constraint model, and perform executability verification on the initial scheduling strategy to determine whether the initial scheduling strategy follows the various production constraints in the production constraint model. Then, the initial scheduling strategy that follows the production constraint model is output as a candidate scheduling strategy, and the candidate scheduling strategy is output to the scheduling simulation evaluation module.

[0045] Subsequently, the scheduling simulation evaluation module inputs the received candidate scheduling strategies into the scheduling simulation environment for simulation evaluation, obtains evaluation indicators for the candidate scheduling strategies, and outputs the evaluation indicators to the strategy evolution optimization module. Upon receiving the evaluation indicators, the strategy evolution optimization module performs at least one evolutionary operation on the candidate scheduling strategies based on the evaluation indicators, including strategy parameter mutation, strategy rule mutation, and strategy cross-combination, to generate an optimized scheduling strategy. This optimized scheduling strategy is then output as the target scheduling strategy to the strategy issuance and execution feedback module. Upon receiving the target scheduling strategy, the strategy issuance and execution feedback module converts the target scheduling strategy into an executable task and issues it to the production execution system, while simultaneously receiving on-site operational status information from the production execution system. When the on-site operational status information indicates a dynamic event occurring during industrial production, the dynamic event reordering and on-site status feedback module can detect the dynamic event based on the on-site operational status information, determine the scope of its impact, and trigger a local or global reordering based on the scope of impact. After strategy reordering, the dynamic event reordering and on-site status feedback module can reorder the reordered scheduling strategy to the production execution system through the strategy issuance and execution feedback module. Meanwhile, the dynamic event rearrangement and on-site status feedback module can feed back on-site operation status information to the industrial rule cognition module and the strategy evolution optimization module for updating the production constraint model and performing strategy evolution optimization.

[0046] This embodiment provides a task scheduling method. The method discloses a process that, in response to a production task scheduling request, identifies all industrial scheduling entities and production constraints within the production rule information. Production constraints characterize the constraint relationships between industrial scheduling entities during industrial production. A production constraint model is generated based on each industrial scheduling entity and its corresponding production constraints, and candidate scheduling strategies are generated according to the production constraint model. The candidate scheduling strategies are then evaluated through simulation, and optimized based on the simulation evaluation results to obtain a target scheduling strategy. Production task scheduling is then performed based on the target scheduling strategy. Because this embodiment can generate a production constraint model based on industrial scheduling entities and their corresponding production constraints, generate candidate scheduling strategies based on the production constraint model, and then optimize the strategies based on the simulation evaluation results of the candidate scheduling strategies, and perform production scheduling based on the optimized target scheduling strategy, it solves the technical problem that existing task scheduling methods often struggle to guarantee the effectiveness and stability of scheduling strategies in real production environments.

[0047] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 , Figure 3 This is a flowchart illustrating Embodiment 2 of the task scheduling method of this application.

[0048] In this embodiment, the step of generating a production constraint model based on each of the industrial scheduling entities and the corresponding production constraints includes: Step S21: Obtain the production constraint expression rules of the preset domain-specific language, which include object layer expression rules, constraint layer expression rules, event layer expression rules and strategy layer expression rules.

[0049] It should be noted that the predefined domain-specific language can be a formal language specifically designed for the industrial production scheduling domain, used to uniformly express production scheduling constraints and rules. In this embodiment, the predefined domain-specific language can adopt a hierarchical expression method of "object definition + constraint relationship + event triggering + execution rule," converting unstructured natural language rules, tabular data, and expert experience into structured data that computers can directly read and process. The basic syntax of the predefined domain-specific language can include four types of statements: object-level statements, constraint-level statements, event-level statements, and strategy-level statements. Each type of statement has a specific syntactic format and semantic meaning, enabling industrial production rules from different sources to be uniformly mapped into a verifiable, computable, and executable constraint model.

[0050] It should be explained that the production constraint expression rules can be specific conversion specifications defined in a predefined domain-specific language, used to convert industrial scheduling entities and production constraints into corresponding statements. In this embodiment, the production constraint expression rules can include object-level expression rules, constraint-level expression rules, event-level expression rules, and policy-level expression rules. Specifically, the object-level expression rules can be syntactic rules in a predefined domain-specific language used to standardize the definition of industrial scheduling entities, which can limit the declaration keywords, attribute fields, type system, and nested structure of industrial scheduling entities. The constraint-level expression rules can be syntactic rules in a predefined domain-specific language used to standardize the declaration of production constraints, which can limit the declaration keywords of production constraints, the referencing methods of constraint subjects and objects, the identification methods of constraint types, and the expression methods of constraint strength. The event-level expression rules can be syntactic rules in a predefined domain-specific language used to standardize the description of dynamic events, which can limit the definition keywords of dynamic events, the enumeration values ​​of event types, the expression syntax of triggering conditions, and the declaration methods of response rules. The policy layer expression rules can be the syntax rules in a predefined domain-specific language used to standardize the condition configuration method for generating scheduling policies. They can limit the declaration keywords of the condition information required for generating scheduling policies, the definition method of priority factors, the expression method of device selection conditions, and the configuration method of reordering range, etc.

[0051] In its implementation, after identifying industrial scheduling entities and extracting their corresponding production constraints, the system can obtain predefined domain-specific language rules for expressing production constraints, including object-level, constraint-level, event-level, and strategy-level rules. Specifically, the object-level rules define the template format for object-level statements corresponding to various types of industrial scheduling entities (including process entities, equipment entities, resource entities, order entities, etc.), including the name, data type, mandatory / unmandable fields, and value range of each field. The constraint-level rules define the template format for constraint-level statements corresponding to various types of production constraints (including process sequence constraints, equipment capacity constraints, resource capacity constraints, etc.), including constraint expression methods and constraint parameter format requirements. The event-level rules define the template format for event-level statements corresponding to various types of dynamic events (such as order insertion events, equipment anomaly events, quality anomaly events, etc.), including event type definition format, trigger condition expression method, and response rule structure. The strategy-level rules define the template format for strategy-level statements corresponding to scheduling strategy-related information, including the syntax structure of priority rules, the expression method of equipment selection conditions, and the definition method of scenario conditions.

[0052] For example, the system can read the production constraint expression rules in a predefined domain-specific language from internal storage. In these production constraint expression rules, the object-level expression rules specify that the object-level statement template for process entities includes five fields: process_id (process identifier), process_name (process name), standard_processing_time (standard processing time), optional_processing_time_range (optional processing time range), and quality_check_required (whether quality inspection is required); the object-level statement template for equipment entities includes five fields: machine_id (equipment identifier), supported_process_set (set of supported process types), capacity (equipment capacity), current_status (current status), and setup_time (setup time). The constraint-level expression rules specify that the constraint-level statement template for process sequence constraints includes three fields: previous_process (previous process identifier), next_process (subsequent process identifier), and constraint_type (constraint type); the constraint-level statement template for equipment capacity constraints includes two fields: machine_id (equipment identifier) ​​and supported_process_set (set of supported processes). The event layer expression rules define the event layer statement template for dynamic events, including four fields: event_type (event type), trigger_condition (trigger condition), affected_scope (affected object scope), and rescheduling_scope (rescheduling scope). The strategy layer expression rules define the strategy layer statement template, including four fields: strategy_id (strategy number), priority_rule (priority rule), machine_selection_rule (machine selection rule), and applicable_scenario (applicable scenario).

[0053] Step S22: Convert each of the industrial scheduling entities into object layer statements according to the object layer expression rules. The object layer statements are used to characterize the attribute features of the corresponding industrial scheduling entities.

[0054] It should be noted that object-level statements can be structured statements written in a predefined domain-specific language to define the attribute characteristics of industrial scheduling entities. These statements can describe the basic information and attribute characteristics of various industrial scheduling entities, with each entity corresponding to one or more object-level statements. For example, the object-level statements corresponding to a process entity may include fields such as process identifier, process name, standard processing time, and optional processing time range; the object-level statements corresponding to a device entity may include fields such as device identifier, set of supported process types, device capacity, current status (idle, processing, faulty, or under maintenance), and changeover time; the object-level statements corresponding to a resource entity may include fields such as resource identifier, resource type, capacity limit, and available time window; and the object-level statements corresponding to an order entity may include fields such as order identifier, product type, delivery time, order priority, quantity, and process route.

[0055] It should also be noted that the attribute characteristics can be the basic attributes and status information of each industrial scheduling entity. For example, the attribute characteristics of a process entity can include, but are not limited to, process identifier, process name, standard processing time, optional processing time range, and quality inspection requirements; the attribute characteristics of an equipment entity can include, but are not limited to, equipment identifier, supported process set, equipment capacity, current status, and changeover time; the attribute characteristics of an order entity can include, but are not limited to, order identifier, product type, delivery time, priority, quantity, and process route.

[0056] In its implementation, after acquiring the production constraint expression rules, the system can convert each industrial scheduling entity into object-level statements based on the object-level expression rules. Specifically, for each industrial scheduling entity, the system searches for the corresponding statement template in the object-level expression rules according to the entity's type (e.g., process entity, equipment entity, resource entity, or order entity), and extracts the specific values ​​corresponding to each field in the statement template from the industrial scheduling entity's attribute information to generate the entity's object-level statement according to the statement template format. For example, after identifying a process entity as a process type, the system can search for the object-level statement template corresponding to the process. This template may contain fields such as process number, process name, standard processing time, optional processing time range, and quality inspection requirements. Then, the system can extract the specific values ​​of each field from the process entity's attribute information and fill them into the corresponding positions in the template, thereby generating the object-level statement for that process entity.

[0057] Step S23: Based on the constraint layer expression rules, the event layer expression rules, and the strategy layer expression rules, convert each of the production constraints into constraint layer statements, event layer statements, and strategy layer statements. The constraint layer statements are used to characterize the production constraints between each of the industrial scheduling entities. The event layer statements are used to characterize the triggering conditions and response rules of dynamic events in the industrial production process. The strategy layer statements are used to describe the condition information required for the generation of scheduling strategies.

[0058] It should be explained that constraint layer statements can be structured statements written in a predefined domain-specific language to characterize the production constraint relationships between various industrial scheduling entities. In this embodiment, constraint layer statements can be used to define various production constraints such as process sequence, equipment capacity, resource capacity, time window, logistics accessibility, and quality feedback rules. For example, constraint layer statements corresponding to process sequence constraints may include fields such as preceding process identifier, subsequent process identifier, and constraint type (e.g., mandatory, optional, or rework); constraint layer statements corresponding to equipment capacity constraints may include fields such as equipment identifier, supported process type set, and capacity limit; constraint layer statements corresponding to resource capacity constraints may include fields such as resource identifier, resource type, capacity limit, and available time window; and constraint layer statements corresponding to logistics path constraints may include fields such as starting process identifier, target process identifier, available path set, transportation time, and buffer capacity.

[0059] It should also be explained that event-level statements can be structured statements written in a predefined domain-specific language to characterize the triggering conditions and response rules of dynamic events in the industrial production process. Dynamic events can be sudden anomalies or changes that occur during industrial production and require system response, such as order insertion events, equipment malfunction events, quality anomaly events, workstation congestion events, and production delay events.

[0060] Understandably, triggering conditions can be specific conditions or judgment criteria required to trigger the occurrence of a dynamic event. For example, the triggering condition for an equipment failure event can be "abnormal equipment operating status signal" or "error code in equipment self-inspection report"; the triggering condition for a quality anomaly event can be "test results exceeding the preset quality acceptance standard range" or "defect type belonging to the preset defect category requiring processing"; the triggering condition for a workstation congestion event can be "the current work-in-process quantity in the buffer reaches the capacity limit" or "the length of the equipment waiting queue exceeds the preset threshold"; and the triggering condition for a production delay event can be "the actual completion time of the process exceeds the planned completion time by more than the preset time deviation tolerance". Correspondingly, response rules can be the processing logic and operational procedures that the system should execute after a dynamic event occurs. For example, the response rule for a device failure event could be "reassign the unfinished tasks on the faulty device to alternative devices and update the logistics path"; the response rule for a quality anomaly event could be "determine the handling method (such as rework, re-inspection, scrapping, or manual review) based on the defect type and adjust the affected subsequent processes"; and the response rule for an order insertion event could be "find a time window that can be inserted into the existing production schedule and readjust conflicting orders based on the priority and process requirements of the urgent orders".

[0061] It should also be explained that the policy layer statements can be structured statements written in a predefined domain-specific language to describe the conditional information required for generating the scheduling policy. The conditional information required for generating the scheduling policy can include various parameters, rules, and constraints on which the scheduling policy is based, such as priority factors and their weight settings, device selection conditions, path selection conditions, reordering rules, and applicable scenarios. The priority factors can include at least two of the following: order delivery urgency, process criticality, equipment load, changeover penalty, logistics route cost, and quality risk. Equipment selection criteria specify the judgment standards used when selecting target equipment, such as "prioritize the equipment with the lowest current load" or "prioritize the equipment with the shortest changeover time." Route selection criteria specify the judgment standards used when selecting logistics routes, such as "prioritize the route with the shortest transportation time" or "prioritize the route with the largest buffer capacity." Reordering rules specify the specific strategies for reordering after a dynamic event occurs, such as "local reordering only adjusts affected orders and their subsequent processes" or "global reordering recalculates all incomplete orders." Applicable scenarios specify the production scenario types to which this strategy layer statement applies, such as "normal production scheduling scenario," "order insertion and reordering scenario," "equipment failure detour scenario," or "quality anomaly feedback scenario."

[0062] In its implementation, the system can generate production constraint statements according to the constraint layer expression rules, including: generating process sequence constraint statements for process sequence constraints, generating equipment capacity constraint statements for equipment capacity constraints, generating resource capacity constraint statements for resource capacity constraints, generating time window constraint statements for time window constraints, generating logistics path constraint statements for logistics path constraints, and generating quality feedback constraint statements for quality feedback constraints. Simultaneously, the system can generate corresponding event layer statements for order insertion events, equipment failure events, quality anomaly events, workstation congestion events, and production delay events, based on the dynamic event definition methods specified by the event layer expression rules. These event layer statements can include the event type field, trigger condition field, affected object field, and reordering range field for the dynamic event. Furthermore, the system can generate strategy layer statements containing process priority calculation rules, equipment selection rules, path selection rules, reordering rules after dynamic event triggering, and strategy application scenario information, based on the scheduling strategy generation condition information definition methods specified by the strategy layer expression rules. Finally, the system can combine the generated object layer statements, constraint layer statements, event layer statements, and strategy layer statements to generate a complete production constraint model.

[0063] Step S24: Generate a production constraint model based on the object layer statements, the constraint layer statements, the event layer statements, and the strategy layer statements.

[0064] In this embodiment, after generating object-level statements, constraint-level statements, event-level statements, and strategy-level statements, the system can organize these statements according to a preset model structure to form a complete and structured production constraint model.

[0065] Furthermore, to improve the accuracy of the production constraint model and the reliability of the scheduling strategy, before the step of generating the production constraint model based on each of the industrial scheduling entities and the corresponding production constraints, the method further includes: determining the rule source reliability, entity identification confidence, field completeness, and constraint consistency verification result of each of the production constraints; determining the constraint confidence of the corresponding production constraint based on the rule source reliability, entity identification confidence, field completeness, and constraint consistency verification result; comparing the constraint confidence with a preset confidence threshold; and if the constraint confidence is higher than the preset confidence threshold, performing the step of generating the production constraint model based on each of the industrial scheduling entities and the corresponding production constraints.

[0066] It should be noted that rule source reliability can be a quantitative indicator used to measure the credibility of the source of production rule information. In industrial production systems, production rule information can come from various data sources, such as process documents, MES system configurations, equipment capability databases, production management rule bases, and expert experience bases. These different types of data sources typically have different levels of reliability. For example, MES system configurations and equipment capability databases are usually system data that has undergone rigorous review and verification, and their reliability is relatively high; process documents may have some deviations due to untimely version updates, and their reliability is moderate; expert experience bases rely on the accuracy of individual experience, and their reliability may vary considerably. In this embodiment, rule source reliability can be scored for the source data source of each production rule. Different weight scores can be assigned to these different data sources. For example, MES system configurations and equipment capability databases can be assigned a higher source reliability score, process documents can be assigned a medium score, and expert experience rules can be assigned a lower score.

[0067] It should also be noted that entity recognition confidence score can be used as a metric to quantify the accuracy of the large language model in identifying industrial scheduling entities from production rule information. In practical applications, when the large language model performs semantic parsing of production rule information, it can generate a confidence score for each identified entity to reflect the model's degree of confidence in the recognition results.

[0068] It should be noted that field completeness can be a quantitative indicator used to measure whether the field information contained in a production constraint is complete. It can be determined based on the ratio of the number of necessary fields that have been filled in a production constraint to the total number of necessary fields. For example, if a process sequence constraint contains a main process and an object process but lacks a constraint type, then the field completeness of the constraint is two-thirds (i.e., 2 / 3), indicating that the constraint information is incomplete and needs further supplementation or confirmation.

[0069] It should also be noted that the constraint consistency check result can be a judgment result used to measure whether there is a logical conflict between the production constraint and the existing production constraints. It can be expressed in the form of a consistency score (with a value range of 0 to 1) or in the form of a Boolean value of "pass / fail". Among them, when using the consistency score form, constraints that do not conflict with existing constraints can be set with a higher consistency score, and constraints that conflict with existing constraints can be set with a lower consistency score.

[0070] In this embodiment, the constraint consistency verification result can be determined by comparing the current production constraints with existing process routes, equipment capabilities, and resource capacity constraints. For example, if an existing constraint stipulates that "a certain piece of equipment is currently in a fault state," while the current constraint stipulates that "the equipment can perform processing tasks," then there is a conflict between the two constraints, the consistency verification fails, and the consistency score is 0.

[0071] It should be explained that constraint confidence can be a quantitative indicator used to comprehensively measure the reliability and trustworthiness of production constraints. It can be obtained by weighting the reliability of rule sources, entity identification confidence, field completeness, and constraint consistency verification results.

[0072] It should also be explained that the preset confidence threshold can be a pre-set benchmark score used to determine the reliability of production constraints. This score can be set according to the actual needs of the industrial production scenario; for example, it could be set to 0.8. In practical applications, when the confidence score of a production constraint is higher than the preset confidence threshold, it indicates that the constraint has high reliability and can be used to generate a production constraint model. When the confidence score is lower than or equal to the preset confidence threshold, it indicates that the constraint's reliability is insufficient and requires manual verification or re-analysis before further processing.

[0073] In its implementation, after the industrial rule recognition module identifies each production constraint, the system can determine the rule source reliability, entity identification confidence, field completeness, and constraint consistency verification result for each production constraint. Specifically, the system first identifies the source data source of each production rule and determines the rule source reliability of each production rule based on the preset reliability weights corresponding to different data sources. Simultaneously, the system can obtain the identification confidence of each industrial scheduling entity from the semantic parsing results of the large language model, and determine the necessary field set for each type of constraint based on its constraint type. Then, it checks the number of filled necessary fields in the production constraint and uses the ratio of the number of filled necessary fields to the total number of necessary fields as the field completeness of the production constraint. Furthermore, the system can compare the current production constraint with existing production constraints that have passed verification to determine if there are any logical conflicts. If the current production constraint does not conflict with any existing constraints, the constraint consistency verification result can be set to a higher consistency score (e.g., 1.0); if conflicts exist, the constraint consistency verification result can be set to a lower consistency score (e.g., 0.3) based on the severity of the conflict. Subsequently, the system multiplies the rule source reliability, entity identification confidence, field completeness, and constraint consistency verification results by their respective weighting coefficients, and sums the products to obtain the constraint confidence of the production constraint. The system then compares the constraint confidence of each production constraint with a preset confidence threshold. For production constraints with a confidence level higher than the preset threshold, the system marks them as reliable constraints and executes the step of generating a production constraint model based on each industrial scheduling entity and its corresponding production constraint. For production constraints with a confidence level lower than or equal to the preset threshold, the system marks them as constraints awaiting confirmation and outputs them to a manual confirmation process for manual review and confirmation by operators. Finally, the confirmed production constraints are re-entered into the system to generate the production constraint model.

[0074] Understandably, this embodiment comprehensively evaluates the reliability of each production constraint through four dimensions: reliability of rule source, confidence of entity identification, completeness of fields, and consistency verification results. This expands the assessment of production constraint quality from a single dimension to a multi-dimensional comprehensive evaluation, improving the accuracy and comprehensiveness of constraint selection. This, in turn, contributes to improving the accuracy of the production constraint model and the reliability of the scheduling strategy. Furthermore, by comparing the comprehensive constraint confidence with a pre-set confidence threshold, this embodiment enables the automated identification and interception of low-quality constraints. This prevents inaccurate or incomplete constraints caused by unreliable rule sources, semantic recognition errors, missing information fields, or logical conflicts from entering subsequent modeling stages, further enhancing the reliability of the generated scheduling strategy.

[0075] Further, the step of generating candidate scheduling strategies based on the production constraint model includes: Step S25: Generate prompt words based on the production scenario description information, process constraint information, equipment capacity limitation information, optimization objectives, and the production constraint model construction strategy of the industrial production process.

[0076] It should be explained that production scenario description information can be a set of data used to describe the environmental characteristics of the current industrial production process. This can include scenario-level information such as the current order size, equipment size, product type, order urgency distribution, and production cycle requirements. For example, in a PCB flexible manufacturing scenario, production scenario description information could include information such as "the current batch contains 100 PCB orders," "the product type is multilayer board," "order delivery time is between 3 and 7 days," and "there are 3 urgent orders that need to be completed within 2 days."

[0077] It should also be explained that process constraint information can be a collection of process rules and technical requirements that must be followed between various processes in industrial production. This information can originate from process documents and process flow designs, including the sequential relationship between processes, inter-process dependencies, and process parameter requirements. For example, in PCB manufacturing, process constraint information may include rules such as "etching must be completed before drilling," "drilling must be completed before copper plating," and "lamination temperature must be controlled between 180°C and 200°C."

[0078] It should be understood that equipment capacity limitation information can be a collection of constraints on the processing capacity of various production equipment in the industrial production process. This information can originate from the MES system configuration and equipment capacity database, including the types of processes each piece of equipment supports, the equipment's processing speed or processing time, the equipment's capacity limit (the maximum number of tasks that can be processed simultaneously), the equipment's current operating status (idle, processing, faulty, or under maintenance), and the equipment's changeover time. For example, in PCB manufacturing, equipment capacity limitation information could include information such as "the drilling process can be completed on drilling machine 1, drilling machine 2, or drilling machine 3," "drilling machines 1 and 2 are currently idle, while drilling machine 3 is faulty," and "each drilling machine can process a maximum of 2 PCB boards at the same time."

[0079] It should also be understood that the optimization objective can be a data expression of one or more optimization directions expected to be achieved during industrial production scheduling. For example, minimizing order delay time, maximizing equipment utilization, minimizing work-in-process inventory, minimizing the number of path switching, minimizing order insertion response time, maximizing fault detour success rate, etc. This embodiment does not impose any restrictions on this. In practical applications, the optimization objective can be expressed as a combination of multiple objectives in the form of a weighted sum, for example, "the optimization objective has a weight of 0.4 for order delay time, a weight of 0.3 for equipment utilization, and a weight of 0.3 for work-in-process inventory."

[0080] It should be noted that the strategy generation prompts can be natural language descriptions or structured texts that conform to the input format requirements of a large language model, formed by uniformly organizing production scenario description information, process constraint information, equipment capacity limitation information, optimization objectives and production constraint models. They can be used to guide the preset large language model to understand the scheduling problem and generate corresponding scheduling strategies.

[0081] In its implementation, after obtaining a validated production constraint model, the system can retrieve scenario description information of the current industrial production process from the MES system or order management system, and extract process constraint information and equipment capacity limitation information from the production constraint model, while also obtaining the optimization objective. Then, the system can organize the production scenario description information, process constraint information, equipment capacity limitation information, optimization objective, and the generated production constraint model into a complete strategy to generate prompts according to a preset prompt template.

[0082] Step S26: Generate a scheduling strategy concept description and the corresponding executable strategy code based on the strategy prompt words using a preset large language model.

[0083] It should be noted that the pre-set large language model can be a large-scale pre-trained language model with natural language understanding and code generation capabilities. Specifically, it can be a general commercial large language model or a special large language model that is fine-tuned based on industrial production scheduling corpus.

[0084] It should be understood that the description of scheduling strategy ideas can be natural language text used to describe the core decision logic of the scheduling strategy. It can be used to express the high-level decision thinking of the scheduling strategy, such as "prioritize scheduling orders with more urgent delivery dates", "prioritize selecting machines with lower equipment load", "prioritize reducing logistics route switching", etc.

[0085] It should also be understood that executable strategy code can be computer program code that implements the scheduling strategy concept. It can be written in a predefined programming language (such as Python) to implement specific process priority calculations, equipment selection calculations, and path switching logic calculations. In practical applications, executable strategy code can be directly executed by a computer to perform actual scheduling decision calculations, such as calculating the priority score of each process to be scheduled, selecting target equipment, and target logistics paths.

[0086] In its implementation, after constructing the strategy generation prompts, the system can input these prompts into a pre-defined large language model. Upon receiving the prompts, the pre-defined large language model performs semantic understanding and reasoning on the production scenario description, process constraints, equipment capacity limitations, optimization objectives, and production constraint models within the prompts. Based on the semantic understanding and reasoning results, it generates a scheduling strategy concept description, which is output in natural language text format to describe the high-level decision-making logic of the scheduling strategy. Building upon this scheduling strategy concept description, the pre-defined large language model can further use a pre-defined programming language to write executable strategy code corresponding to the description, used to implement specific process priority calculation logic, equipment selection logic, and path switching logic.

[0087] Step S27: Generate an initial scheduling strategy based on the scheduling strategy concept description and the executable strategy code.

[0088] Understandably, the initial scheduling strategy can be a complete strategy scheme used to guide the scheduling of production tasks, generated by a pre-defined large language model based on the concept of scheduling strategy and executable strategy code. In this embodiment, since the initial scheduling strategy is a scheduling strategy directly generated by the pre-defined large language model, it may violate the production constraint model. Therefore, it needs to be screened through subsequent executability verification to generate candidate scheduling strategies.

[0089] In practical applications, the system can combine the scheduling strategy concept description and the corresponding executable strategy code into a single scheduling strategy. Each scheduling strategy can be expressed as a four-tuple: "scheduling strategy concept description + executable strategy code + fitness value + applicable scenario label." During the initial generation phase, the fitness value and applicable scenario label can be temporarily empty or set to default values. Subsequently, the system can perform the above combination operation on multiple sets of "scheduling strategy concept description + executable strategy code" generated from a pre-defined large language model to generate multiple initial scheduling strategies. Each initial scheduling strategy represents a different scheduling decision logic.

[0090] Step S28: Perform an executability check on the initial scheduling strategy to determine whether the initial scheduling strategy follows the production constraint model.

[0091] It should be noted that executability verification can be a systematic check of the initial scheduling strategy to determine whether it conforms to the constraints defined in the production constraint model. In practical applications, the system can perform executability verification by comparing and verifying the various decision rules in the scheduling strategy (including process priority rules, equipment selection rules, path allocation rules, etc.) with the various constraints in the production constraint model (including process sequence constraints, equipment capacity constraints, resource capacity constraints, logistics path constraints, etc.) to check for any violations of constraints. If any rule in the scheduling strategy violates any constraint in the production constraint model, the initial scheduling strategy can be determined to have failed the executability verification; if all rules in the scheduling strategy conform to all constraints in the production constraint model, the initial scheduling strategy can be determined to have passed the executability verification.

[0092] Step S29: Based on the judgment result, the initial scheduling strategy that follows the production constraint model is output as a candidate scheduling strategy.

[0093] It should be noted that if the initial scheduling strategy passes the executability check, it means that the initial scheduling strategy conforms to the production constraint model and can be effectively executed in a real production environment. In this case, it can be output as a candidate scheduling strategy to directly enter the subsequent simulation evaluation stage. If the initial scheduling strategy fails the executability check, it means that the initial scheduling strategy violates the production constraint model. In this case, the regeneration process of the initial scheduling strategy can be triggered. For example, it can return to the strategy generation prompt and re-enter the preset large language model to generate a new scheduling strategy.

[0094] It should be understood that this embodiment organizes production scenario description information, process constraint information, equipment capacity limitation information, optimization objectives, and production constraint models into structured strategy generation prompts. This enables the pre-defined large language model to comprehensively and accurately understand the complete background and various constraints of the current scheduling problem, thereby guiding the large language model to generate a scheduling strategy adapted to the specific production scenario. This avoids the large language model generating a scheduling scheme that does not match the actual situation due to a lack of scenario understanding. Furthermore, this embodiment performs executability verification on the initial scheduling strategy, preventing unexecutable scheduling strategies from entering subsequent simulation evaluation stages and thus avoiding the waste of computational resources.

[0095] This embodiment discloses the production constraint expression rules for obtaining a predefined domain-specific language. These rules include object-level expression rules, constraint-level expression rules, event-level expression rules, and strategy-level expression rules. Each industrial scheduling entity is converted into an object-level statement based on the object-level expression rules. Each production constraint is then converted into constraint-level statements, event-level statements, and strategy-level statements based on the constraint-level, event-level, and strategy-level expression rules. A production constraint model is generated based on these statements. Since this embodiment can convert unstructured industrial knowledge into a structured production constraint model that can be directly parsed and computed by a computer, it achieves automated conversion from unstructured industrial rules to a computable constraint model. This improves the automation level and efficiency of production scheduling strategy production, thereby enhancing the efficiency of production scheduling.

[0096] Based on the first and / or second embodiments of this application, in the third embodiment of this application, the content that is the same as or similar to the above embodiments can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 4 , Figure 4 This is a flowchart illustrating the third embodiment of the task scheduling method of this application.

[0097] In this embodiment, step S30 includes steps S31 to S34: Step S31: Input the candidate scheduling strategy into several preset scheduling simulation environments and obtain the strategy evaluation index of the candidate scheduling strategy in the corresponding preset scheduling simulation environment.

[0098] It should be noted that the preset scheduling simulation environment can be a pre-built virtual simulation system used to simulate the operation of industrial production processes. It can simulate various activities in industrial production processes, such as order arrival, process execution, equipment operation, logistics flow, and dynamic event occurrence. In practical applications, the preset scheduling simulation environment can be configured for different production scenarios. For example, it can build various simulation scenarios for different order sizes (e.g., 50 orders, 100 orders, 200 orders), different equipment sizes (e.g., 5 machines, 10 machines, 20 machines), or different order urgency levels (e.g., 80% ordinary orders and 20% urgent orders).

[0099] It should be noted that strategy evaluation metrics can be quantitative evaluation data used to measure the performance of candidate scheduling strategies in the simulation environment. These metrics may include total production cycle, order delay time, equipment utilization rate, work-in-process inventory, path switching frequency, order insertion response time, fault detour success rate, strategy execution success rate, and rescheduling response time. Specifically, order delay time can be the sum or average of the actual completion time exceeding the delivery time of an order; equipment utilization rate can be the proportion of busy equipment time to the total simulation time, reflecting the efficiency of equipment resource utilization; work-in-process inventory can be the time average of the number of incomplete processes during the simulation, reflecting the degree of work-in-process accumulation during production; path switching frequency can be the number of times the processing path changes due to rescheduling during order execution; order insertion response time can be the time difference between the arrival time of a new order and the generation time of the rescheduling strategy; fault detour success rate can be the proportion of fault events where alternative equipment or alternative paths are successfully assigned to the total number of fault events; and strategy execution success rate can be the proportion of successfully executed scheduling tasks to the total number of issued scheduling tasks.

[0100] In its implementation, after obtaining candidate scheduling strategies, the system can input these strategies into multiple scheduling simulation environments configured for different production scenarios, and execute the simulation process in each environment. Specifically, the system first initializes the order set, equipment set, process route, logistics path, and event queue. Then, it reads the candidate scheduling strategies and calculates the priority of currently executable processes according to the decision logic within the strategies. Next, it selects the target equipment and logistics path based on equipment availability and path constraints, advances the simulation time, and updates the order status, equipment status, process completion status, and logistics status. In the event of order insertion, faults, or quality anomalies, the corresponding rescheduling process can be triggered. After the simulation, the system uses an indicator calculator to statistically analyze various evaluation indicators to obtain the strategy evaluation indicators corresponding to the candidate scheduling strategies.

[0101] Step S32: Determine the comprehensive strategy score of the candidate scheduling strategy based on the strategy evaluation index, and determine the performance improvement rate of the candidate scheduling strategy relative to the baseline scheduling strategy based on the comprehensive strategy score.

[0102] It should be understood that the comprehensive strategy score can be a single quantitative score used to measure the overall performance of candidate scheduling strategies, calculated based on multiple strategy evaluation indicators. In this embodiment, the system can obtain the comprehensive strategy score by normalizing each strategy evaluation indicator and then weighting and summing them according to preset weight coefficients. The corresponding calculation formula is as follows:

[0103] In the formula, A comprehensive strategy score is given to the candidate scheduling strategies; The time efficiency score can be calculated based on order delay time and order insertion response time. Resource utilization score, which can be calculated based on equipment utilization and load balancing; Cost control scores can be calculated based on work-in-process inventory, number of path changes, and changeover setup costs. The response capability score can be calculated based on the fault detour success rate and rearrangement response time. To perform stability scoring, it can be calculated based on the success rate of strategy execution and the magnitude of plan changes; to These are the weighting coefficients corresponding to the evaluation indicators of each strategy.

[0104] In practical applications, the weight coefficients of various strategy evaluation indicators can be set according to the optimization priorities of the actual production scenario. For example, when the factory is currently facing the main problem of order delays, the weight of the order delay time indicator can be increased; when the factory is currently facing the main problem of insufficient equipment capacity, the weight of the equipment utilization rate indicator can be increased.

[0105] It should be noted that the baseline scheduling strategy can be a reference scheduling strategy used as a performance comparison benchmark in a preset scheduling simulation environment. It can be a randomly selected scheduling strategy, a benchmark strategy generated using simple rules (such as First-Come First-Served (FCFS), Earliest Due Date (EDD), etc.), or an existing scheduling strategy currently in use in the production environment. This embodiment does not impose any restrictions on this. In this embodiment, the baseline scheduling strategy can be used to provide a relative performance comparison benchmark for candidate scheduling strategies, allowing the calculation of the performance improvement rate of the candidate scheduling strategy relative to the baseline scheduling strategy to eliminate the impact of absolute score differences between different simulation scenarios on strategy evaluation.

[0106] It should be understood that the performance improvement rate can be the relative improvement percentage of the candidate scheduling strategy compared to the baseline scheduling strategy in terms of the overall strategy score. In this embodiment, the performance improvement rate can be obtained by subtracting the overall strategy score of the baseline scheduling strategy from the overall strategy score of the candidate scheduling strategy, and then dividing by the overall strategy score of the baseline scheduling strategy. The corresponding calculation formula is as follows:

[0107] In the formula, For performance improvement rate, The candidate scheduling strategies are evaluated as a comprehensive strategy score within a predefined scheduling simulation environment. The baseline scheduling strategy is evaluated as a comprehensive strategy score within the same preset scheduling simulation environment.

[0108] In practical applications, a performance improvement rate greater than 0 indicates that the candidate scheduling strategy is superior to the baseline scheduling strategy; a performance improvement rate less than 0 indicates that the candidate scheduling strategy is inferior to the baseline scheduling strategy; and a performance improvement rate equal to 0 indicates that the candidate scheduling strategy and the baseline scheduling strategy have comparable performance. In this embodiment, the performance improvement rate can be used to uniformly measure the relative superiority or inferiority of candidate scheduling strategies under different simulation scenarios.

[0109] In its implementation, after obtaining the policy evaluation metrics of candidate scheduling strategies under various preset scheduling simulation environments, the system normalizes each policy evaluation metric to unify their value ranges to the same interval. Subsequently, the system performs a weighted summation of the normalized metrics based on preset weight coefficients to obtain the comprehensive policy score of the candidate scheduling strategy. It then obtains the comprehensive policy score of the baseline scheduling strategy under the same preset scheduling simulation environment. Finally, it subtracts the comprehensive policy score of the baseline scheduling strategy from the comprehensive policy score of the candidate scheduling strategy and divides it by the comprehensive policy score of the baseline scheduling strategy to obtain the performance improvement rate of the candidate scheduling strategy in that simulation scenario.

[0110] Step S33: Determine the fitness value of the candidate scheduling strategy based on the performance improvement rate.

[0111] It should be noted that the fitness value can be a quantitative measure of the overall performance of a candidate scheduling strategy across multiple preset simulation scenarios. It can be obtained by averaging the performance improvement rates of the candidate scheduling strategy in each simulation environment. For example, the system can use the average performance improvement rate of the candidate scheduling strategy across multiple simulation environments as the fitness value of that candidate scheduling strategy. The corresponding calculation formula is as follows:

[0112] In the formula, Candidate scheduling strategies fitness value, To preset the number of scheduling simulation environments, Individual candidate scheduling strategies In the preset scheduling simulation environment The performance improvement rate compared to the baseline scheduling strategy.

[0113] It should also be noted that the fitness value of the candidate scheduling strategy can not only measure the optimization effect of the candidate scheduling strategy in terms of production efficiency, resource utilization and dynamic response capability, but also promote the evolution process to prioritize the retention of high-quality scheduling strategies with stable performance in a variety of complex industrial scenarios, thereby helping to improve the robustness and generalization ability of the entire system.

[0114] Step S34: Perform an evolutionary operation on the candidate scheduling strategy based on the fitness value to obtain the target scheduling strategy.

[0115] It should be noted that evolutionary operations can simulate the biological evolution mechanism by performing parameter mutation, rule crossover, strategy selection, and scenario adaptation on candidate scheduling strategies. Through multiple generations of iteration, the candidate scheduling strategy group gradually evolves, ultimately obtaining the target scheduling strategy with optimal overall performance. In this embodiment, evolutionary operations can include strategy parameter mutation, strategy rule mutation, strategy crossover combination, and scenario label adaptation. Specifically, strategy parameter mutation refers to applying a perturbation to the priority weight vector of candidate scheduling strategies, adjusting the weight values ​​of each priority factor; strategy rule mutation refers to adding, deleting, or replacing rule items in candidate scheduling strategies; strategy crossover combination refers to selecting two or more high-scoring strategies, extracting effective rule fragments from each strategy, and combining them to generate new offspring strategies; scenario label adaptation refers to classifying and labeling strategies according to different production scenarios, establishing a mapping relationship between strategies and production scenarios.

[0116] In its implementation, after determining the fitness values ​​of each candidate scheduling strategy, the system can perform evolutionary operations on these strategies based on the fitness values ​​to obtain the target scheduling strategy. These evolutionary operations can include strategy parameter mutation, strategy rule mutation, strategy cross-combination, and scenario label adaptation. In strategy parameter mutation, the system can apply perturbations to the priority weight vectors of the candidate scheduling strategies, adjusting the weight values ​​of each priority factor. The perturbation amount can be determined based on the changes in the evaluation metrics of the candidate scheduling strategies in the previous round of simulation evaluation. For example, when simulation results show high order delays, the delivery urgency weight can be increased; when equipment utilization is low, the equipment load balancing weight can be increased; and when path switching is frequent, the logistics path cost weight can be increased. In policy rule mutation, the system can identify inefficient rules in candidate scheduling policies based on simulation evaluation results and add, delete, or replace them. For example, if a policy has a high failure rate in equipment failure scenarios, rules such as "alternative equipment priority" or "path detour cost" can be added; if a rule does not affect the score in the long term or causes excessive policy complexity, it can be deleted; if a rule performs poorly in a specific scenario, alternative rules can be generated using a large language model. In policy cross-combination, the system can select multiple candidate scheduling policies with high fitness values ​​as parent scheduling policies, extract effective rule fragments from each parent scheduling policy, and then combine the extracted rule fragments according to rule categories to generate child scheduling policies. For example, the system can extract delivery priority rules from candidate scheduling policy A, extract equipment selection rules from candidate scheduling policy B, and then combine the two to generate a new child scheduling policy C. In scenario tag adaptation, the system can form a family of strategies based on different production scenarios, including normal production scheduling strategy, order insertion and rescheduling strategy, equipment failure detour strategy, and quality anomaly feedback strategy, and establish a mapping relationship between strategies and production scenarios, thereby realizing adaptive scheduling and dynamic strategy switching under different industrial scenarios.

[0117] It should be noted that this embodiment independently evaluates candidate scheduling strategies by inputting them into multiple preset scheduling simulation environments configured for different production scenarios. This allows the performance evaluation of candidate scheduling strategies to be no longer limited to a single scenario, thus comprehensively measuring the adaptability of candidate scheduling strategies under different production conditions. Furthermore, this embodiment calculates the performance improvement rate of candidate scheduling strategies relative to the baseline scheduling strategy, and uses the average performance improvement rate of candidate scheduling strategies across multiple training scenarios as the fitness value. This allows the fitness value to comprehensively reflect the ability of candidate scheduling strategies to maintain stable performance in various complex industrial scenarios, thereby guiding the evolution process to prioritize the retention of high-quality scheduling strategies that perform stably across multiple scenarios, ultimately improving the robustness and generalization ability of the scheduling strategies.

[0118] Further, step S34 includes: Step S341: Determine several parent scheduling strategies from the candidate scheduling strategies based on the fitness values.

[0119] It should be noted that the parent scheduling strategy can be a candidate scheduling strategy selected as the genetic basis for generating the next generation of scheduling strategies during the strategy evolution and optimization process. In this embodiment, the system can select a parent scheduling strategy from the candidate scheduling strategy population based on the fitness value of each candidate scheduling strategy. A candidate scheduling strategy with a higher fitness value has a higher probability of being selected as a parent scheduling strategy, while a candidate scheduling strategy with a lower fitness value has a lower probability of being selected as a parent scheduling strategy.

[0120] Step S342: Extract the production scheduling rules from each of the parent scheduling strategies, and combine the production scheduling rules according to the rule category to generate the child scheduling strategy.

[0121] Understandably, production scheduling rules can be rule items within a scheduling strategy used to guide specific scheduling decisions. They can be extracted from the executable strategy code of the parent scheduling strategy or identified from the scheduling strategy concept description of the parent scheduling strategy. In this embodiment, production scheduling rules can be categorized by function into process priority rules, equipment selection rules, path allocation rules, and exception reordering rules. Specifically, process priority rules are used to calculate the priority order of each process to be scheduled, determining which processes should be prioritized for production; equipment selection rules are used to select the target equipment for executing a process from multiple candidate equipment, determining the processing equipment to which the process should be assigned; path allocation rules are used to select the material flow path from multiple candidate logistics paths, determining the process flow path for work-in-process; and exception reordering rules are used to determine the scope and method of reordering when dynamic events occur.

[0122] It is also understandable that rule categories can be a classification system based on the functional attributes and decision objects of production scheduling rules, such as process priority rule categories, equipment selection rule categories, path allocation rule categories, and abnormal reordering rule categories.

[0123] It should be noted that a child scheduling strategy can be an individual strategy generated from a parent scheduling strategy through evolutionary operations, serving as a new generation of candidate scheduling strategies. It can inherit superior rule fragments from its parent strategy and obtain new rule combination methods through rule combinations between different parent strategies. In practical applications, after a child scheduling strategy is determined as a new candidate scheduling strategy, it can re-enter the simulation evaluation phase to perform performance evaluation and fitness calculations, and participate in subsequent evolutionary iterations, forming a cyclical optimization process of "evaluation-selection-evolution-re-evaluation".

[0124] Step S343: Determine the offspring scheduling strategy as a new candidate scheduling strategy, and return to the step of inputting the candidate scheduling strategy into several preset scheduling simulation environments until the candidate scheduling strategy meets the preset termination conditions, and output the target scheduling strategy.

[0125] In its implementation, after determining the fitness values ​​of each candidate scheduling strategy, the system sorts these strategies from highest to lowest fitness value and designates one or more top-ranked strategies as parent scheduling strategies. Subsequently, the system extracts production scheduling rules from the executable strategy code of each parent strategy and categorizes these rules according to their type. After rule extraction and categorization, the system combines the production scheduling rules from each parent strategy according to their type to generate child scheduling strategies. Specifically, the system can extract process priority rules and path allocation rules from parent scheduling strategy A, and equipment selection rules and exception reordering rules from parent scheduling strategy B, combining the extracted rules into a complete child scheduling strategy according to their type. Then, the system can use the evolved offspring scheduling strategies as new candidate scheduling strategies, re-perform simulation evaluation and evolution process, and determine whether the current candidate scheduling strategy meets the preset termination conditions after each round of evolution. If the current candidate scheduling strategy meets the preset termination conditions, the evolution iteration stops, and the candidate scheduling strategy with the highest fitness value among the current candidate scheduling strategies is output as the target scheduling strategy. The preset termination conditions may include reaching a preset number of iterations, the strategy score improving below a threshold for multiple consecutive rounds, the order delay time being lower than the target value, the order insertion response time or fault detour success rate meeting the set requirements, or the strategy generation time exceeding the production scheduling allowable time limit, etc.

[0126] It should be noted that in this embodiment, the parent scheduling strategy is selected from the candidate scheduling strategies based on the fitness value. This makes it more likely that the superior strategy with a high fitness value will pass its rule fragments to the offspring. The offspring scheduling strategy is then determined as a new candidate scheduling strategy and returned to the simulation evaluation stage for iteration. This allows the scheduling strategy population to continuously evolve towards a better direction in multiple iterations, thereby improving the reliability of the generated scheduling strategy.

[0127] Furthermore, after the step of scheduling production tasks based on the target scheduling strategy, the method further includes: Step S51: Receive the real-time feedback of the on-site operating status information during the industrial production process.

[0128] It should be understood that on-site operational status information can be real-time operational status data fed back by production equipment, testing equipment, and logistics systems during the actual execution of the industrial production process. Examples include equipment operating status, process completion status, logistics status, and quality inspection results. Specifically, equipment operating status can refer to the current operating condition of each production piece of equipment, such as whether it is idle, processing, faulty, or under maintenance; process completion status can refer to the progress of each production process, such as whether a particular process in an order has been completed, is in progress, or has not yet started; logistics status can refer to the flow of materials between processes, such as the current location of work-in-process, the current quantity of work-in-process in the buffer zone, and whether there is congestion in the logistics path; and quality inspection results can refer to the output of quality inspection equipment after inspecting finished or semi-finished products, such as whether the product passed inspection, has defects requiring rework, or needs to be scrapped.

[0129] Step S52: Detect whether there are dynamic events in the industrial production process based on the on-site operation status information.

[0130] It is understood that dynamic events can be abnormal or changing situations that occur suddenly during the execution of an industrial production process and require a system response, such as order insertion events, equipment malfunction events, quality malfunction events, workstation congestion events, and production delay events. This embodiment does not limit these events.

[0131] Step S53: If it exists, determine the scope of influence of the dynamic event. The scope of influence includes a first scope and a second scope. The first scope is used to characterize the impact on a single device, a single process, or a single logistics path. The second scope is used to characterize the impact on multiple devices, multiple processes, or a group of bottleneck devices.

[0132] It should also be understood that the scope of an event's impact can refer to the extent and degree to which a dynamic event affects an industrial production process. This can include the affected objects (such as specific orders, processes, and equipment), the time span of the impact, and the severity of the impact. For example, a drilling machine malfunction may only affect the unfinished drilling processes on that machine and subsequent processes that depend on these drilling processes, thus having a relatively limited scope of impact. However, a malfunction of a critical bottleneck device (such as the only pressing machine) may cause all orders awaiting pressing to come to a standstill, resulting in a much larger scope of impact.

[0133] Step S54: If the scope of the event is the first scope of the event and there are alternative equipment or alternative paths for the current production task, determine the production reordering type of the current production task as a local reordering method, and reorder the production task according to the local reordering method.

[0134] Alternatively, in step S55: if the scope of the event is the second scope of the event, or if the dynamic event causes the delivery constraint to fail, determine the production rescheduling type of the current production task as the global rescheduling method, and reschedule the production task according to the global rescheduling method.

[0135] It should be noted that production rescheduling types can be scheduling adjustment methods determined based on the scope of impact of dynamic events, used to address these events. These can include partial rescheduling and global rescheduling. Partial rescheduling refers to adjusting only the portion of production tasks affected by the dynamic event, such as adjusting the production schedules of affected orders and their subsequent processes while keeping the schedules of other unaffected orders and processes unchanged. Global rescheduling refers to rescheduling all unfinished production tasks in the entire industrial production process, i.e., recalculating the scheduling order, equipment allocation, and logistics paths of all unfinished orders. In practical applications, partial rescheduling is typically suitable for scenarios where the impact of the dynamic event is relatively small and there are alternative equipment or alternative paths; global rescheduling is typically suitable for scenarios where the impact of the dynamic event is large, involves bottleneck equipment groups, or causes widespread failure of delivery constraints.

[0136] In practical implementation, edge devices (such as production equipment, testing equipment, and logistics systems) can continuously collect and report on-site operational status information during production execution. After receiving the on-site operational status information from the edge devices in real time, the system can compare the on-site operational status information with a preset normal operating status benchmark and detect whether there are dynamic events in the current production process by judging whether the operational status information is abnormal or changes. If no dynamic events are detected, the system can continue to monitor the subsequently received on-site operational status information; if a dynamic event is detected, the system can determine the scope of impact of the dynamic event based on the detection results. Then, the system can determine whether to use a local rescheduling or a global rescheduling approach for production rescheduling based on the scope of impact of the dynamic event. Specifically, if the dynamic event only affects a single device, a single process, or a single logistics path, and there are alternative devices or alternative paths for the current production task, a local rescheduling can be triggered; if the dynamic event affects multiple devices, multiple processes, or a bottleneck device group, or if the dynamic event causes a large-scale failure of delivery constraints, a global rescheduling can be triggered. After determining the production rescheduling type, the system can reschedule production tasks accordingly. For example, for local rescheduling, the system can adjust only the orders affected by dynamic events and their subsequent processes, while keeping the scheduling plans of other unaffected orders and processes unchanged. For global rescheduling, the system can recalculate the scheduling order, equipment allocation, and logistics paths for all currently unfinished orders. After completing the rescheduling, the system can send the rescheduled scheduling strategy to the production execution system for execution and feed back the dynamic event information and rescheduling results to the industrial rule cognition module, the production constraint DSL modeling module, and the scheduling simulation evaluation module for updating the constraint model and subsequent strategy evolution optimization.

[0137] It should be noted that this embodiment receives real-time feedback of on-site operational status information during industrial production, automatically detects dynamic events based on this information, and then determines the production rescheduling type based on the impact range of the dynamic events. This allows for the selection of appropriate rescheduling strategies, enabling local rescheduling for events with small impact ranges, where only the affected portion is adjusted, thus reducing interference with ongoing production tasks and computational overhead. Conversely, global rescheduling for events with large impact ranges comprehensively optimizes all unfinished tasks, ensuring the global optimality of the scheduling scheme under major anomalies.

[0138] This embodiment discloses a method for inputting candidate scheduling strategies into several preset scheduling simulation environments to obtain strategy evaluation metrics for the candidate scheduling strategies in the corresponding preset scheduling simulation environments; determining a comprehensive strategy score for the candidate scheduling strategy based on the strategy evaluation metrics, and determining the performance improvement rate of the candidate scheduling strategy relative to the baseline scheduling strategy based on the comprehensive strategy score; determining the fitness value of the candidate scheduling strategy based on the performance improvement rate; and performing an evolutionary operation on the candidate scheduling strategy based on the fitness value to obtain the target scheduling strategy. This embodiment allows for a comprehensive measurement of the adaptability of candidate scheduling strategies under different production conditions by independently evaluating them in multiple preset scheduling simulation environments configured for different production scenarios. Furthermore, this embodiment calculates the performance improvement rate of the candidate scheduling strategy relative to the baseline scheduling strategy and uses the average performance improvement rate of the candidate scheduling strategy in multiple training scenarios as the fitness value. This allows the fitness value to comprehensively reflect the ability of the candidate scheduling strategy to maintain stable performance in various complex industrial scenarios, thereby improving the robustness and generalization ability of the scheduling strategy.

[0139] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the task scheduling method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0140] This application also provides a task scheduling device, please refer to... Figure 5 The task scheduling device includes: The constraint identification module 10 is used to identify all industrial scheduling entities and production constraints in the production rule information in response to a production task scheduling request. The production constraints are used to characterize the constraint relationships between the industrial scheduling entities in the industrial production process. The constraint modeling module 20 is used to generate a production constraint model based on each of the industrial scheduling entities and the corresponding production constraints, and to generate candidate scheduling strategies based on the production constraint model. The task scheduling module 30 is used to perform simulation evaluation on the candidate scheduling strategies, optimize the candidate scheduling strategies based on the simulation evaluation results to obtain the target scheduling strategy, and schedule production tasks based on the target scheduling strategy.

[0141] The task scheduling apparatus provided in this application, employing the task scheduling method described in the above embodiments, can solve the technical problem that the scheduling strategies generated by existing task scheduling methods are often difficult to guarantee in real production environments in terms of effectiveness and stability. Compared with the prior art, the beneficial effects of the task scheduling apparatus provided in this application are the same as those of the task scheduling method provided in the above embodiments, and other technical features in the task scheduling apparatus are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0142] This application provides a task scheduling device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the task scheduling method in the first embodiment described above.

[0143] The following is for reference. Figure 6 The diagram illustrates a structural schematic of a task scheduling device suitable for implementing embodiments of this application. The task scheduling device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, personal digital assistants (PDAs), tablet computers (PADs), portable media players (PMPs), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 6 The task scheduling device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0144] like Figure 6 As shown, the task scheduling device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in the read-only memory 1002 or a program loaded from the storage device 1003 into the random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the task scheduling device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the task scheduling device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows task scheduling devices with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0145] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0146] The task scheduling device provided in this application, employing the task scheduling method described in the above embodiments, can solve the technical problem of task scheduling. Compared with the prior art, the beneficial effects of the task scheduling device provided in this application are the same as those of the task scheduling method described in the above embodiments, and other technical features of the task scheduling device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0147] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0148] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0149] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the task scheduling method described in the above embodiments.

[0150] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, radio frequency (RF), etc., or any suitable combination thereof.

[0151] The aforementioned computer-readable storage medium may be included in the task scheduling device; or it may exist independently and not be assembled into the task scheduling device.

[0152] The aforementioned computer-readable storage medium carries one or more programs. When the one or more programs are executed by the task scheduling device, the task scheduling device: in response to a production task scheduling request, identifies all industrial scheduling entities and production constraints in the production rule information, wherein the production constraints are used to characterize the constraint relationships between various industrial scheduling entities in the industrial production process; generates a production constraint model based on each industrial scheduling entity and its corresponding production constraints, and generates candidate scheduling strategies based on the production constraint model; performs simulation evaluation on the candidate scheduling strategies, optimizes the candidate scheduling strategies based on the simulation evaluation results to obtain a target scheduling strategy, and performs production task scheduling based on the target scheduling strategy.

[0153] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0154] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0155] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0156] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described task scheduling method. This solves the technical problem that existing task scheduling methods often fail to guarantee the effectiveness and stability of scheduling strategies in real production environments. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the task scheduling method provided in the above embodiments, and will not be repeated here.

[0157] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the task scheduling method described above.

[0158] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A task scheduling method, characterized in that, The method includes: In response to a production task scheduling request, all industrial scheduling entities and production constraints in the production rule information are identified, wherein the production constraints are used to characterize the constraint relationships between the industrial scheduling entities in the industrial production process. A production constraint model is generated based on each of the industrial scheduling entities and the corresponding production constraints, and a candidate scheduling strategy is generated based on the production constraint model. The candidate scheduling strategies are evaluated through simulation, and the candidate scheduling strategies are optimized based on the simulation evaluation results to obtain the target scheduling strategy. Production tasks are then scheduled based on the target scheduling strategy.

2. The method as described in claim 1, characterized in that, The step of generating a production constraint model based on each of the industrial scheduling entities and the corresponding production constraints includes: Obtain production constraint expression rules in a predefined domain-specific language, wherein the production constraint expression rules include object-level expression rules, constraint-level expression rules, event-level expression rules, and strategy-level expression rules; According to the object layer expression rules, each industrial scheduling entity is converted into an object layer statement, and the object layer statement is used to characterize the attribute features of the corresponding industrial scheduling entity. Based on the constraint layer expression rules, the event layer expression rules, and the strategy layer expression rules, each production constraint is converted into constraint layer statements, event layer statements, and strategy layer statements. The constraint layer statements are used to characterize the production constraints between each industrial scheduling entity. The event layer statements are used to characterize the triggering conditions and response rules of dynamic events in the industrial production process. The strategy layer statements are used to describe the condition information required for generating the scheduling strategy. A production constraint model is generated based on the object layer statements, the constraint layer statements, the event layer statements, and the strategy layer statements.

3. The method as described in claim 1, characterized in that, Before the step of generating a production constraint model based on each of the industrial scheduling entities and the corresponding production constraints, the method further includes: Determine the reliability of the rule source, the confidence level of entity identification, the field completeness, and the constraint consistency verification results for each of the aforementioned production constraints; The constraint confidence level corresponding to the production constraint is determined based on the reliability of the rule source, the confidence level of the entity identification, the completeness of the field, and the constraint consistency verification result. The constraint confidence level is compared with a preset confidence threshold. If the confidence level of the constraint is higher than the preset confidence threshold, the step of generating a production constraint model based on each of the industrial scheduling entities and the corresponding production constraints is performed.

4. The method as described in claim 1, characterized in that, The step of generating candidate scheduling strategies based on the production constraint model includes: Prompt words are generated based on the production scenario description information, process constraint information, equipment capacity limitation information, optimization objectives, and the production constraint model construction strategy of the industrial production process. Based on the strategy, a prompt word is generated using a preset large language model to generate a description of the scheduling strategy concept and the corresponding executable strategy code. An initial scheduling strategy is generated based on the described scheduling strategy concept and the executable strategy code. Perform an executability check on the initial scheduling strategy to determine whether the initial scheduling strategy conforms to the production constraint model; Based on the judgment result, the initial scheduling strategy following the production constraint model will be output as a candidate scheduling strategy.

5. The method as described in claim 1, characterized in that, The step of performing simulation evaluation on the candidate scheduling strategies and optimizing the candidate scheduling strategies based on the simulation evaluation results to obtain the target scheduling strategy includes: The candidate scheduling strategy is input into several preset scheduling simulation environments to obtain the strategy evaluation index of the candidate scheduling strategy in the corresponding preset scheduling simulation environment; The comprehensive strategy score of the candidate scheduling strategy is determined based on the strategy evaluation index, and the performance improvement rate of the candidate scheduling strategy relative to the baseline scheduling strategy is determined based on the comprehensive strategy score. The fitness value of the candidate scheduling strategy is determined based on the performance improvement rate; Based on the fitness value, the candidate scheduling strategy is evolved to obtain the target scheduling strategy.

6. The method as described in claim 5, characterized in that, The step of performing an evolutionary operation on the candidate scheduling strategy based on the fitness value to obtain the target scheduling strategy includes: Based on the fitness value, several parent scheduling strategies are determined from the candidate scheduling strategies; Extract the production scheduling rules from each of the parent scheduling strategies, and combine the production scheduling rules according to the rule category to generate the child scheduling strategy; The offspring scheduling strategy is determined as a new candidate scheduling strategy, and the process returns to the step of inputting the candidate scheduling strategy into several preset scheduling simulation environments until the candidate scheduling strategy meets the preset termination conditions, and the target scheduling strategy is output.

7. The method according to any one of claims 1 to 6, characterized in that, Following the step of scheduling production tasks based on the target scheduling strategy, the method further includes: Receive real-time feedback of on-site operating status information during the industrial production process; Based on the on-site operational status information, detect whether there are dynamic events in the industrial production process; If it exists, the scope of influence of the dynamic event is determined. The scope of influence includes a first scope and a second scope. The first scope is used to characterize the impact on a single device, a single process, or a single logistics path. The second scope is used to characterize the impact on multiple devices, multiple processes, or a group of bottleneck devices. If the scope of the event is the first scope of impact, and there are alternative equipment or alternative paths for the current production task, the production reordering type of the current production task is determined to be a local reordering method, and the production task is reordered according to the local reordering method. Alternatively, if the scope of the event is the second scope of the event, or if the dynamic event causes the delivery constraint to fail, the production rescheduling type of the current production task is determined to be the global rescheduling method, and the production task is rescheduled according to the global rescheduling method.

8. A task scheduling device, characterized in that, The device includes: The constraint identification module is used to identify all industrial scheduling entities and production constraints in the production rule information in response to a production task scheduling request. The production constraints are used to characterize the constraint relationships between the industrial scheduling entities in the industrial production process. The constraint modeling module is used to generate a production constraint model based on each of the industrial scheduling entities and the corresponding production constraints, and to generate candidate scheduling strategies based on the production constraint model. The task scheduling module is used to perform simulation evaluation on the candidate scheduling strategies, optimize the candidate scheduling strategies based on the simulation evaluation results to obtain the target scheduling strategy, and schedule production tasks based on the target scheduling strategy.

9. A task scheduling device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the task scheduling method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the task scheduling method as described in any one of claims 1 to 7.

Images