Method and system for dynamic production scheduling optimization based on human-computer game cooperation

Abstract

The application discloses a dynamic production scheduling optimization method and system based on human-computer game cooperation, relates to the technical field of production scheduling, and has the scheduling optimization steps: S1, a hierarchical distributed multi-agent proxy scheduling optimization platform is constructed, the platform comprises a perception layer, a decision layer and an execution layer, and specifically comprises a disturbance monitoring agent DMA, a digital twin agent DTA, a human expert agent HEA, a machine optimization agent MOA, a game coordination agent GCA and an execution agent PEA; S2, the disturbance monitoring agent DMA is arranged on the perception layer and is used for detecting equipment failure and task insertion disturbance events, and a digital twin environment is constructed through the digital twin agent DTA. The application gives equal importance to 'production stability optimization' and 'time efficiency optimization', scheduling personnel keep the stability of the manufacturing system through experience-driven emergency intervention and guidance, machines keep the efficiency robustness of the manufacturing system through simulation prediction of time efficiency, and a compromise optimization scheduling scheme is dynamically generated through human-computer alternating dynamic game.

Description

Technical Field

[0001] This invention relates to the field of production scheduling technology, specifically to a dynamic production scheduling optimization method and system based on human-machine game-theoretic collaboration. Background Technology

[0002] Against the backdrop of the transformation of modern manufacturing towards intelligence and flexibility, dynamic production scheduling, as a core link to ensure efficient production operation, faces challenges from complex disturbances such as equipment failures and urgent task insertions. Traditional static scheduling methods, such as rule-based scheduling strategies, allocate resources based solely on preset priority rules, making it difficult to cope with sudden changes in the production process, resulting in low production efficiency and serious resource waste. Although mathematical programming methods can solve the optimal scheduling scheme to a certain extent, the computational complexity increases exponentially with the scale of the problem, failing to meet the real-time requirements of dynamic scheduling. Typically, the shortest time and lowest cost are used as the single optimization objective, ignoring the implicit costs of tooling changes, tool switching, and program overloading caused by frequent adjustments to production routes in discrete manufacturing scenarios, leading to decreased production stability. To address this, we propose a dynamic production scheduling optimization method and system based on human-machine game collaboration. Summary of the Invention

[0003] To address the aforementioned technical problems, this technical solution provides a dynamic production scheduling optimization method and system based on human-machine game-theoretic collaboration, thus resolving the aforementioned issues.

[0004] To achieve the above objectives, the technical solution adopted in this invention is: a dynamic production scheduling optimization method based on human-machine game theory collaboration, wherein the scheduling optimization steps are as follows:

[0005] S1. Construct a hierarchical distributed multi-agent agent scheduling and optimization platform. The platform includes a perception layer, a decision layer, and an execution layer, specifically including a disturbance monitoring agent (DMA), a digital twin agent (DTA), a human expert agent (HEA), a machine optimization agent (MOA), a game-theoretic coordination agent (GCA), and an execution agent (PEA).

[0006] S2. Set up a monitoring agent DMA in the perception layer to detect equipment failures and task insertion disturbances. Construct a digital twin environment through the digital twin agent DTA, output a candidate device set and tooling and program switching cost matrix, and complete state synchronization.

[0007] S3. Design a visual interface on the virtual terminal of the digital twin agent (DTA) corresponding to the decision-making level. The human expert agent (HEA) performs drag-and-drop intervention operations based on the device status and experience to maintain device stability and minimize switching costs.

[0008] S4, Machine Optimization Agent (MOA) integrates a scheduling simulation engine based on ECP and MWR rules, supporting Monte Carlo perturbation simulation and maximum delay time calculation, for evaluating the time robustness of manual solutions;

[0009] S5. The game-coordination agent GCA coordinates the allocation order of the disturbed processes. The scheduling optimization platform cyclically calls the human expert agent HEA and the machine optimization agent MOA to perform alternating optimization until all affected processes are adjusted, and a balanced scheduling scheme is generated.

[0010] S6. The execution agent PEA generates executable instructions for the final scheduling scheme. By adjusting the timing instructions of the logistics system, it controls the fixture switching and program loading tasks, and feeds back the actual execution status to the monitoring agent DMA to update the platform knowledge base, thereby realizing closed-loop scheduling control.

[0011] Preferably, in step S2, real-time monitoring of equipment failures and emergency order insertion disturbances is based on DMA. This real-time monitoring step uses sensors distributed across key system nodes to collect and synthesize time information into useful information, including equipment failure and emergency order insertion disturbance information. The DMA is then updated in a designated database, triggering a three-level response mechanism. This response mechanism includes:

[0012] Generate a set of affected processes, freeze the affected processes, and mark the abnormal status;

[0013] A virtual debugging environment is built using a digital twin agent (DTA) to generate a set of candidate devices and a switching cost matrix.

[0014] Synchronize the real-time load rate of the equipment, remaining processing capacity, and process constraint rules.

[0015] Preferably, the intervention operation in step S3 is handled by a human expert agent (HEA). The HEA receives a set of candidate affected processes and a list of corresponding candidate devices through a visual interface. The HEA first selects an affected process and generates an adjustment plan based on the lowest conversion cost and experience.

[0016] Preferably, the maximum delay time calculation in step S4 is processed by the Machine Optimization Agent (MOA). The MOA injects the manually adjusted scheme and random disturbances into the priority scheduling simulation model to simulate the adjusted maximum delay time.

[0017] The ECP rule is: prioritize the machine that can complete the processing task earliest;

[0018] The MWR rule is: prioritize processing workpieces with more remaining operations, and calculate the overall priority using the following formula:

[0019] ;

[0020] Where S is the priority score, ω1 is the priority weight based on the ECP rule, ω2 is the priority weight based on the MWR rule, and S E S is the priority score based on ECP rules. M This is a priority score based on the MWR rule.

[0021] Preferably, the equilibrium scheduling scheme in step S5 is generated by the Game Coordination Agent (GCA). The Game Coordination Agent (GCA) performs continuous game optimization by sequentially assigning the affected processes to the participants and cyclically executing steps S3 and S4 to calculate the stability and robustness of the scheme.

[0022] Preferably, the stability calculation formula is:

[0023] ;

[0024] CO ijp Let CO be the planned completion time of the j-th operation for workpiece i. ijr Let be the actual completion time of the j-th operation of workpiece i, qi be the total number of operations of workpiece i, Oi be the number of operations in workpiece i that did not require machine tool replacement, excluding the influence of faulty machine tools, and SM be the stability calculation result; MS P The planned completion time is used to standardize the denominator;

[0025] The formula for calculating the robustness target is:

[0026] ;

[0027] MS R MS represents the actual processing completion time, the total actual time spent on the entire production process. P The planned completion time is the original total expected time of the schedule, and RM is the robustness result obtained from the calculation.

[0028] After calculating and locking the final scheduling scheme, the PEA agent automatically generates tooling preparation instructions, triggering the equipment's fixture replacement and CNC program loading control signals.

[0029] Preferably, the method for detecting equipment failures and task insertion disturbances in step S2 involves pre-setting monitoring indicators, including abnormal transmission rate indicators, channel error count indicators, and transmission timeout indicators, and reading the disturbance monitoring agent DMA status register value in real time through the embedded system register.

[0030] Preferably, in step S2, based on the acquired DMA channel status and device bus signal data, feature analysis and anomaly modeling are performed to achieve real-time monitoring and output structured data.

[0031] Preferably, in step S2, a digital twin model of the equipment and production line is constructed based on a 3D modeling and simulation engine. Real-time monitoring data drives dynamic simulation and interaction. When an anomaly is detected, the digital twin agent DTA environment filters a set of candidate equipment based on the equipment status and evaluates the costs from the dimensions of tooling disassembly, installation, debugging, and program modification and transmission, forming a tooling and program switching cost matrix.

[0032] A dynamic production scheduling optimization system based on human-machine game-theoretic collaboration, comprising:

[0033] The disturbance sensing data synchronization module is configured to build a digital twin environment through the digital twin agent DTA, output a candidate device set and tooling / program switching cost matrix, and complete state synchronization;

[0034] The stability module is configured to perform drag-and-drop intervention operations based on device status and experience, prioritizing maintaining device stability and minimizing switching costs.

[0035] The simulation scheduling module is configured for multi-objective optimization analysis and robustness evaluation of scheduling schemes;

[0036] The platform's core scheduling module is configured to control the interaction order through game theory coordination agents, achieving a dynamic trade-off between stability and efficiency through collaborative optimization.

[0037] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0038] This invention utilizes DMA monitoring at the perception layer combined with digital twins to detect equipment faults and task disturbances in real time, quickly outputting candidate equipment and cost matrices to provide accurate data for decision-making. A visual interface allows schedulers to intervene flexibly based on experience, balancing equipment stability and cost control. The MOA-integrated scheduling simulation engine scientifically evaluates the robustness of manual solutions through Monte Carlo simulation and maximum delay time calculation. A game-theoretic coordination and alternating optimization mechanism ensures that scheduling schemes reach global equilibrium under complex disturbances. PEA transforms scheduling schemes into executable instructions, updating the knowledge base with execution status feedback to achieve a closed-loop scheduling system, driving continuous system optimization and upgrades to adapt to changing production environments. It prioritizes both "optimal production stability" and "optimal time efficiency." Schedulers maintain the stability of the manufacturing system through experience-driven emergency intervention, while the machine maintains the robustness of manufacturing system efficiency through simulation-based time efficiency prediction. A compromise-based optimized scheduling scheme is dynamically generated through a dynamic game between humans and machines. Attached Figure Description

[0039] Figure 1 This is a flowchart of the scheduling optimization steps of the present invention;

[0040] Figure 2 This is a framework diagram of the scheduling optimization system of the present invention;

[0041] Figure 3 This is the original scheduling Gantt chart;

[0042] Figure 4 This is a Gantt chart for scheduling the method of the present invention;

[0043] Figure 5 This invention provides a Gantt chart for the traditional dynamic scheduling method.

[0044] Figure 6 This is a flowchart of the human-computer game interaction process of the present invention;

[0045] Figure 7 This is a diagram of the human-computer interaction scheduling (Gantt chart) interface. Detailed Implementation

[0046] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0047] Reference Figure 1 As shown, the dynamic production scheduling optimization method based on human-machine game collaboration has the following scheduling optimization steps:

[0048] S1. Construct a hierarchical distributed multi-agent agent scheduling and optimization platform. The platform includes a perception layer, a decision layer, and an execution layer, specifically including a disturbance monitoring agent (DMA), a digital twin agent (DTA), a human expert agent (HEA), a machine optimization agent (MOA), a game-theoretic coordination agent (GCA), and an execution agent (PEA).

[0049] S2. Set up a monitoring agent DMA in the perception layer to detect equipment failures and task insertion disturbances. Construct a digital twin environment through the digital twin agent DTA, output a candidate device set and tooling and program switching cost matrix, and complete state synchronization.

[0050] S3. Design a visual interface on the virtual terminal of the digital twin agent (DTA) corresponding to the decision-making level. The human expert agent (HEA) performs drag-and-drop intervention operations based on the device status and experience to maintain device stability and minimize switching costs.

[0051] S4, Machine Optimization Agent (MOA) integrates a scheduling simulation engine based on ECP and MWR rules, supporting Monte Carlo perturbation simulation and maximum delay time calculation, for evaluating the time robustness of manual solutions;

[0052] S5. The game-coordination agent GCA coordinates the allocation order of the disturbed processes. The scheduling optimization platform cyclically calls the human expert agent HEA and the machine optimization agent MOA to perform alternating optimization until all affected processes are adjusted, and a balanced scheduling scheme is generated.

[0053] S6. The execution agent PEA generates executable instructions for the final scheduling scheme, controls fixture switching and program loading tasks, and feeds back the actual execution status to the monitoring agent DMA to update the platform knowledge base, thereby realizing closed-loop scheduling control.

[0054] The ECP rule is the earliest completion rule, which prioritizes the machine that can complete the processing task earliest.

[0055] The MWR rule is the rule of having the most remaining processes, which prioritizes processing workpieces with more remaining processes.

[0056] The monitoring agent DMA in the perception layer of this application captures disturbances such as equipment failures and task insertions in real time. The virtual environment constructed by the digital twin agent DTA can not only synchronize the physical equipment status, but also output a candidate equipment set and cost matrix, providing accurate data support for subsequent decision-making and improving the system's response speed to dynamic changes. The visualization interface of the decision layer enables schedulers to flexibly intervene based on experience. Combined with the machine optimization agent MOA to evaluate the time robustness of the solution, it combines human experience with the advantages of algorithms. The game coordination agent GCA generates equilibrium solutions through alternating optimization, taking into account both efficiency and stability and avoiding the limitations of single decision-making. The execution agent PEA transforms the scheduling solution into actual instructions, and the execution status is fed back to the monitoring agent DMA to update the knowledge base, forming a complete closed loop. This allows the platform to continuously learn and optimize, adapt to various uncertainties in the production process, and continuously improve the scientificity and reliability of the scheduling strategy.

[0057] In step S2, real-time monitoring of equipment failures and emergency order insertion disturbances is based on DMA. This real-time monitoring step uses sensors distributed across key system nodes to collect and synthesize time information into useful data, including equipment failure and emergency order insertion disturbance information. The DMA is then updated in the designated database, triggering a three-level response mechanism. This response mechanism includes:

[0058] Generate a set of affected processes, freeze the affected processes, and mark the abnormal status;

[0059] A virtual debugging environment is built using a digital twin agent (DTA) to generate a set of candidate devices and a switching cost matrix.

[0060] Synchronize the real-time load rate of the equipment, remaining processing capacity, and process constraint rules.

[0061] This application's DMA real-time monitoring of equipment failures and emergency order insertions can quickly pinpoint affected processes, generate a clear set of affected processes, and freeze related workflows, preventing abnormal states from spreading to the entire production chain and reducing disorder and chaos caused by disturbances. Through status marking, both the scheduling system and humans can intuitively identify "breakpoints" in the production chain, providing clear guidance for subsequent resource allocation and solution adjustments, and shortening the problem location time.

[0062] By using DTA to build a virtual debugging environment, the device switching process can be simulated based on real-time data. A set of candidate devices and a tooling and program switching cost matrix can be generated in advance. This can quantify cost indicators such as debugging time and program modification amount for different device replacement schemes, assisting in the selection of the optimal alternative path and avoiding downtime losses caused by blind switching.

[0063] In step S3 of this application, the visualization interface presents information in the form of a graphical Kanban board. The left panel displays the set of affected processes, presenting the process name, current progress, estimated completion time, urgency level, etc. in a Gantt chart style. The right panel displays a list of candidate equipment, including equipment ID, current load rate, compatibility score with the process, estimated cost of tooling and program switching, etc. Each element supports color marking, such as red marking high-risk processes and yellow marking high-cost switching equipment, which makes it easy for experts to quickly identify key information.

[0064] Human experts acting as agents for HEA select the highest priority or most impactful processes from the affected process set. During the selection process, the system provides a process impact analysis report, such as the prediction of the chain reaction of the delay in the process to subsequent processes. After selecting the process, the experts, based on their extensive production experience and combined with the switching cost data in the candidate equipment list, comprehensively consider the historical failure rate of the equipment, ease of maintenance, and implicit factors of compatibility with the existing production process to formulate an adjustment plan. When a critical piece of equipment fails, the experts may choose a backup device that, although with a slightly higher switching cost, is more stable and can be quickly integrated into the production line, thus avoiding greater losses caused by frequent subsequent adjustments.

[0065] After the adjustment plan is generated, the system performs a preliminary feasibility verification, checks for issues such as equipment resource conflicts and process timing contradictions, and provides a verification report. Experts can simulate the implementation effect of the plan in a visual interface and view the operating status and capacity changes of the production line after the process adjustment through digital twin agent (DTA) technology. This helps experts further optimize the plan to ensure that it meets both the goal of minimizing conversion costs and actual production needs.

[0066] In step S4, the maximum delay time calculation is processed by the Machine Optimization Agent (MOA). The MOA injects the manually adjusted scheme and random disturbances into the priority scheduling simulation model to simulate the adjusted maximum delay time.

[0067] The ECP rule is: prioritize the machine that can complete the processing task earliest;

[0068] The MWR rule is: prioritize processing workpieces with more remaining operations, and calculate the overall priority using the following formula:

[0069] ;

[0070] Where S is the priority score, ω1 is the priority weight based on the ECP rule, ω2 is the priority weight based on the MWR rule, and S E S is the priority score based on ECP rules. M This is a priority score based on the MWR rule.

[0071] In step S5, the equilibrium scheduling scheme is generated through Game Coordination Agent (GCA). GCA performs continuous game optimization by sequentially assigning the affected processes to the participants and cyclically executing steps S3 and S4, thus calculating the stability and robustness of the scheme.

[0072] The stability calculation formula is as follows:

[0073] ;

[0074] CO ijp Let CO be the planned completion time of the j-th operation for workpiece i. ijr Let be the actual completion time of the j-th operation of workpiece i, qi be the total number of operations of workpiece i, Oi be the number of operations in workpiece i that did not require machine tool replacement, excluding the influence of faulty machine tools, and SM be the stability calculation result; MS P The planned completion time is used to standardize the denominator;

[0075] The formula for calculating the robustness target is:

[0076] ;

[0077] MS R MS represents the actual processing completion time, the total actual time spent on the entire production process. P The planned completion time is the original total expected time of the schedule, and RM is the robustness result obtained from the calculation.

[0078] After calculating and locking the final scheduling scheme, the PEA agent automatically generates tooling preparation instructions, triggering the equipment's fixture replacement and CNC program loading control signals.

[0079] The method for detecting equipment failures and task insertion disturbances in step S2 uses pre-defined monitoring indicators, including abnormal transmission rate indicators, channel error count indicators, and transmission timeout indicators, and reads the disturbance monitoring agent DMA status register value in real time through embedded system registers.

[0080] In step S2, based on the acquired DMA channel status and device bus signal data, feature analysis and anomaly modeling are performed to achieve real-time monitoring and output structured data.

[0081] In step S2, a digital twin model of the equipment and production line is built based on a 3D modeling and simulation engine. Real-time monitoring data drives dynamic simulation and interaction. When an anomaly is detected, the digital twin agent DTA environment filters a set of candidate equipment based on the equipment status and evaluates the costs from the dimensions of tooling disassembly, installation, debugging, and program modification and transmission, forming a tooling and program switching cost matrix.

[0082] Reference Figure 2 As shown, a dynamic production scheduling optimization system based on human-machine game theory collaboration is included. The scheduling optimization system comprises:

[0083] The disturbance sensing data synchronization module is configured to build a digital twin environment through the digital twin agent DTA, output a candidate device set and tooling / program switching cost matrix, and complete state synchronization;

[0084] The stability module is configured to perform drag-and-drop intervention operations based on device status and experience, prioritizing maintaining device stability and minimizing switching costs.

[0085] The simulation scheduling module is configured for multi-objective optimization analysis and robustness evaluation of scheduling schemes;

[0086] The platform's core scheduling module is configured to control the interaction order through game theory coordination agents, achieving a dynamic trade-off between stability and efficiency through collaborative optimization.

[0087] To verify the effectiveness of the method of the present invention, a flexible workshop scheduling scenario was selected as a test case (listed in Table 1). The scenario includes 3 CNC machine tools (M1-M3) to handle 3 workpieces (workpiece 1-workpiece 3). Each workpiece includes 3 processes, involving milling, drilling and other processes.

[0088] The process route parameters are shown in Table 1, and the equipment switching cost is shown in Table 2. M / T represents the candidate machine tool / corresponding completion time (unit time).

[0089] Table 1 below shows the process route parameters for a flexible workshop with 3 workpieces and 3 machine tools.

[0090] Process Processing time (per unit of time) Processing time (per unit of time) Processing time (per unit of time) <![CDATA[Workpiece 1 - Process 1 (O 1，1 )]]> 4 3 / <![CDATA[Workpiece 1 - Process 2 (O 1，2 )]]> 2 / 4 <![CDATA[Workpiece 1 - Process 3 (O 1，3 )]]> 3 2 2 <![CDATA[Workpiece 2 - Process 1 (O 2，1 )]]> / 2 4 <![CDATA[Workpiece 2 - Process 2 (O 2，1 )]]> 4 5 3 <![CDATA[Workpiece 2 - Process 3 (O 2，1 )]]> 3 4 3 <![CDATA[Workpiece 3 - Process 1 (O 3，1 )]]> 2 3 2 <![CDATA[Workpiece 3 - Process 2 (O 4，1 )]]> / 4 2 <![CDATA[Workpiece 3 - Process 3 (O 5，1 )]]> / 2 4

[0091] Table 2 below shows the equipment switching cost table.

[0092] Switch type Tooling changeover time (per unit of time) Tool calibration time (per unit of time) Program reload time (per unit of time) M1→M2 2 1 1 M1→M3 3 2 1 M2→M3 2 1 1 M2→M1 3 1 1 M3→M2 2 2 1 M3→M1 3 2 1

[0093] S1, Appendix Figure 3 As shown in the Gantt chart, at t = 2, machine tool M3 suddenly requires maintenance, lasting for 6 units of time, affecting operation 2O of workpiece 2. 2，2 Process 3 O for workpiece 2 2，3 Process 3 O for workpiece 1 1，3 .

[0094] S2, the artificial intervention stage, will O 2，2 The work process is assigned to the dispatcher, who chooses to keep M3 unchanged based on experience and equipment switching costs.

[0095] S3. Priority scheduling optimization involves injecting random disturbances into the system and recommending a solution based on computer simulation: O... 1，3 Assigned to M1.

[0096] S4, Game Theory Trade-offs and Decisions, O 2，3 The work process is assigned to the scheduler, who, based on experience and equipment changeover costs, chooses to keep M3 as the processing equipment. The final compromise is shown in the attached document. Figure 4 As shown in the attached diagram, traditional dynamic scheduling... Figure 5 As shown in the attached diagram, the human-computer interactive game process is as follows: Figure 6 As shown, the human-computer interaction scheduling (Gantt chart) interface is as follows: Figure 7 As shown, the verification results are compared in Table 3.

[0097] Table 3 below compares the verification results.

[0098] index Traditional dynamic scheduling Method of the present invention Average number of switching times / workpiece 2 1 Stability Indicators 0.3857 0.2125 Robustness index 0.9 0.5

[0099] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A dynamic production scheduling optimization method based on human-machine game-theoretic collaboration, characterized in that, The scheduling optimization steps are as follows: S1. Construct a hierarchical distributed multi-agent agent scheduling and optimization platform. The platform includes a perception layer, a decision layer, and an execution layer, specifically including a disturbance monitoring agent (DMA), a digital twin agent (DTA), a human expert agent (HEA), a machine optimization agent (MOA), a game-theoretic coordination agent (GCA), and an execution agent (PEA). S2. Set up a disturbance monitoring agent (DMA) in the perception layer to detect equipment failures and task insertion disturbance events. Construct a digital twin environment through a digital twin agent (DTA), output a candidate device set and a tooling and program switching cost matrix, and complete state synchronization. S3. Design a visual interface on the virtual terminal of the digital twin agent (DTA) corresponding to the decision-making level. The human expert agent (HEA) performs drag-and-drop intervention operations based on the device status and experience to maintain device stability and minimize switching costs. S4, Machine Optimization Agent (MOA) integrates a scheduling simulation engine based on ECP and MWR rules, supporting Monte Carlo perturbation simulation and maximum delay time calculation, for evaluating the time robustness of manual solutions; S5. The game-coordination agent GCA coordinates the allocation order of the disturbed processes. The scheduling optimization platform cyclically calls the human expert agent HEA and the machine optimization agent MOA to perform alternating optimization until all affected processes are adjusted, and a balanced scheduling scheme is generated. S6. The execution agent PEA generates executable instructions for the final scheduling scheme. By adjusting the timing instructions of the logistics system, it controls the fixture switching and program loading tasks, and feeds back the actual execution status to the monitoring agent DMA to update the platform knowledge base, thereby realizing closed-loop scheduling control.

2. The dynamic production scheduling optimization method based on human-machine game theory collaboration according to claim 1, characterized in that, In step S2, real-time monitoring of equipment failures and emergency order insertion disturbances is based on DMA. This real-time monitoring step uses sensors distributed across key system nodes to collect and synthesize time information into useful data, including equipment failure and emergency order insertion disturbance information. The DMA is then updated in the designated database, triggering a three-level response mechanism. This response mechanism includes: Generate a set of affected processes, freeze the affected processes, and mark the abnormal status; A virtual debugging environment is built using a digital twin agent (DTA) to generate a set of candidate devices and a switching cost matrix. Synchronize the real-time load rate of the equipment, remaining processing capacity, and process constraint rules.

3. The dynamic production scheduling optimization method based on human-machine game theory collaboration according to claim 1, characterized in that, In step S3, the intervention operation is handled by the Human Expert Agent (HEA). The HEA receives a set of candidate affected processes and a list of corresponding candidate equipment through a visual interface. The HEA first selects an affected process, generates an adjustment plan based on the lowest conversion cost and experience, and then the human expert specifies that the corresponding process should be processed on the corresponding machine tool.

4. The dynamic production scheduling optimization method based on human-machine game theory collaboration according to claim 1, characterized in that, In step S4, the maximum delay time calculation is processed by the Machine Optimization Agent (MOA). The MOA injects the manually adjusted scheme and random disturbances into the priority scheduling simulation model to simulate the adjusted maximum delay time. The ECP rule is: prioritize the machine that can complete the processing task earliest; The MWR rule is: prioritize processing workpieces with more remaining operations, and calculate the overall priority using the following formula: ； Where S is the priority score, ω1 is the priority weight based on the ECP rule, ω2 is the priority weight based on the MWR rule, and S E S is the priority score based on ECP rules. M This is a priority score based on the MWR rule.

5. The dynamic production scheduling optimization method based on human-machine game theory collaboration according to claim 1, characterized in that, In step S5, the equilibrium scheduling scheme is generated through Game Coordination Agent (GCA). GCA performs continuous game optimization by sequentially assigning the affected processes to the participants and cyclically executing steps S3 and S4, thus calculating the stability and robustness of the scheme.

6. The dynamic production scheduling optimization method based on human-machine game theory collaboration according to claim 5, characterized in that, The stability calculation formula is as follows: ； CO ijp Let CO be the planned completion time of the j-th operation for workpiece i. ijr Let be the actual completion time of the j-th operation of workpiece i, qi be the total number of operations of workpiece i, Oi be the number of operations in workpiece i that did not require machine tool replacement, excluding the influence of faulty machine tools, and SM be the stability calculation result; MS P The planned completion time is used to standardize the denominator; The formula for calculating the robustness target is: ； MS R MS represents the actual processing completion time, the total actual time spent on the entire production process. P The planned completion time is the original total expected time of the schedule, and RM is the robustness result obtained from the calculation. After calculating and locking the final scheduling scheme, the PEA agent automatically generates tooling preparation instructions, triggering the equipment's fixture replacement and CNC program loading control signals.

7. The dynamic production scheduling optimization method based on human-machine game theory collaboration according to claim 1, characterized in that, The method for detecting equipment failures and task insertion disturbances in step S2 uses pre-defined monitoring indicators, including abnormal transmission rate indicators, channel error count indicators, and transmission timeout indicators, and reads the disturbance monitoring agent DMA status register value in real time through embedded system registers.

8. The dynamic production scheduling optimization method based on human-machine game theory collaboration according to claim 1, characterized in that, In step S2, data is collected based on the DMA channel status and device bus signal data. Through feature analysis and anomaly modeling, real-time monitoring and output of structured data are achieved.

9. The dynamic production scheduling optimization method based on human-machine game-theoretic collaboration according to claim 1, characterized in that, In step S2, a digital twin model of the equipment and production line is built based on a 3D modeling and simulation engine. Real-time monitoring data drives dynamic simulation and interaction. When an anomaly is detected, the digital twin agent (DTA) filters a set of candidate equipment based on the equipment status and evaluates the costs from the dimensions of tooling disassembly, installation, debugging, and program modification and transmission, forming a tooling and program switching cost matrix.

10. A dynamic production scheduling optimization system based on human-machine game-theoretic collaboration, used to implement the dynamic production scheduling optimization method based on human-machine game-theoretic collaboration as described in claim 1, characterized in that, The scheduling optimization system includes: The disturbance sensing data synchronization module is configured to build a digital twin environment through the digital twin agent DTA, output a candidate device set and tooling / program switching cost matrix, and complete state synchronization; The stability module is configured to perform drag-and-drop intervention operations based on device status and experience, prioritizing maintaining device stability and minimizing switching costs. The simulation scheduling module is configured for multi-objective optimization analysis and robustness evaluation of scheduling schemes; The platform's core scheduling module is configured to control the interaction order through game theory coordination agents, achieving a dynamic trade-off between stability and efficiency through collaborative optimization.

Images