Intelligent manufacturing management method and equipment based on digital twinning technology, and medium

Abstract

The invention relates to an intelligent manufacturing management method and equipment based on a digital twinning technology, and a medium. The method comprises the following steps: constructing a collaborative model, collecting real-time data based on the collaborative model, establishing a manufacturing resource intelligent configuration simulation model, and generating an initial scheduling scheme; the initial scheme is optimized through a genetic algorithm, an executable scheduling scheme is generated, and the optimization target is to minimize the maximum completion time; applying the executable scheme to a workshop, monitoring a dynamic event and collecting data; when a dynamic event occurs, generating a rescheduling scheme based on an event-driven rescheduling strategy and Monte Carlo tree search; the rescheduling scheme is transmitted to a workshop, tasks are adjusted, and meanwhile process execution data are recorded; constructing a digital twinborn model based on recorded data, analyzing dynamic event influence, identifying key nodes in combination with a fault propagation network, and optimizing a dynamic scheduling strategy; and an optimized scheduling parameter and a model updating strategy are generated by using a digital twin simulation result, and closed-loop management and real-time response are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent manufacturing, and specifically relates to an intelligent manufacturing management method, device and medium based on digital twin technology. Background Art

[0002] With the development of intelligent manufacturing technology, the traditional production management mode has been difficult to meet the requirements of real-time scheduling and resource optimization in complex production scenarios. The collaborative relationships among multiple devices, multiple tasks and multiple resources in the manufacturing system show a high degree of complexity, and the frequent occurrence of dynamic events (such as equipment failures, order changes, etc.) further increases the difficulty of production management. Therefore, how to achieve dynamic optimization and efficient collaboration of production management through intelligent technology has become a key research issue.

[0003] As a cutting-edge technology that integrates the physical world and the digital world, digital twin technology provides new ideas for the modeling, analysis and optimization of intelligent manufacturing systems. Digital twin enables the efficient management of information flow, task flow and resource flow in the manufacturing system through real-time mapping and virtual simulation of physical entities. However, traditional digital twin applications mainly focus on static modeling and monitoring analysis, lacking the ability to perform real-time optimization under the trigger of dynamic events. Especially in a complex multi-constraint production environment, how to combine digital twin technology for dynamic scheduling still needs further research.

[0004] In the actual production process, a manufacturing workshop usually involves multiple processes and devices. Each process needs to be completed on a specific device in a given order, and there is resource competition among the devices. In addition, dynamic events (such as machine failures, changes in workpiece processing time or order cancellations) will interfere with the established scheduling plan, making the original plan unable to be executed smoothly, thus affecting the overall production efficiency and resource utilization rate. Therefore, it is of great significance to construct a management method that can dynamically respond to workshop events and optimize scheduling.

[0005] To address the above challenges, the Genetic Algorithm (GA), as an intelligent optimization algorithm, has been widely applied to the workshop scheduling problem. The genetic algorithm can effectively solve the complex Job-shop Scheduling Problem (JSP) by simulating the biological evolution process, especially showing good adaptability when the optimization objective is to minimize the Makespan. However, traditional genetic algorithms show certain limitations when facing dynamic events, such as long calculation time and insufficient rescheduling ability, and are difficult to meet the requirements of real-time response.

[0006] Meanwhile, as a key link in production management, fault propagation analysis can effectively identify key nodes in manufacturing systems, providing a basis for optimizing dynamic scheduling strategies. The fault propagation network model can reveal the propagation paths and influence scopes of faults by analyzing the coupling relationships between resources and devices in the system. However, current research mainly focuses on fault analysis itself and does not combine it with scheduling strategies, lacking a complete solution from fault propagation to dynamic optimized scheduling. Summary of the Invention

[0007] The present invention provides an intelligent manufacturing management method, device and medium based on digital twin technology, aiming to solve the problems that existing intelligent manufacturing systems cannot efficiently optimize production scheduling, respond to production changes in real time and make full use of resources under dynamic event triggers.

[0008] To achieve the above object, the first aspect of the present invention provides an intelligent manufacturing management method based on digital twin technology, including the following steps: Simulate and model the collaborative relationships among various departments in the manufacturing system, and construct a collaborative model including information flow, task flow and resource flow, which is used to define the collaborative element relationships between the design department, the manufacturing department and other departments; Based on the collaborative model, collect the real-time status data of the production workshop in the manufacturing system, and establish a simulation model for intelligent allocation of workshop manufacturing resources to generate an initial scheduling plan; Optimize the initial scheduling plan using a genetic algorithm to generate an executable scheduling plan based on workshop resource constraints, where the optimization objective is to minimize the makespan; Apply the executable scheduling plan to the production workshop, monitor the occurrence of dynamic events during the production process, and collect relevant dynamic data; When a dynamic event occurs, update the collected relevant dynamic data, and based on the event-driven rescheduling strategy using the real-time feedback system status information, generate a rescheduling plan through the Monte Carlo tree search method; Transmit the generated rescheduling plan to the production workshop, adjust the operation tasks of each machine, and record the execution data of each process and the scheduling adjustment records during the production process in real time; Based on the recorded process execution data and scheduling adjustment records, construct a digital twin model of the workshop production process, which is used to simulate and analyze the impact of dynamic events on the production process, and identify key nodes in combination with the fault propagation network model to optimize the dynamic scheduling strategy; Generate optimized scheduling parameters and model update strategies using the simulation results and analysis data of the digital twin model, and apply them to subsequent production scheduling.

[0009] Furthermore, the construction method of the collaborative model includes: Analyze the collaborative relationship between the design department, manufacturing department and other departments in the design, production and testing process of the manufacturing system, and determine the functional elements of each department, including information flow, task flow and resource flow; Based on functional elements, an information flow model is established to define the data interaction relationship between departments, including the transmission of design data, feedback of production status information, and sharing of resource allocation data; Based on functional elements, a task flow model is established to describe the task allocation and dependency relationships between departments, including the input and output of design tasks, the execution path of manufacturing tasks, and the result feedback of test tasks; Based on functional elements, a resource flow model is established to describe the allocation and use of human, equipment, and material resources by each department, and to clarify the boundaries of resource flow and collaboration methods of each department; The information flow model, task flow model and resource flow model are integrated to form a collaborative model that includes the collaborative relationship between departments.

[0010] Furthermore, based on the collaborative model, real-time status data of the production workshop in the manufacturing system is collected, and a simulation model for intelligent configuration of workshop manufacturing resources is established to generate an initial scheduling plan. The method includes the following steps: Determine the manufacturing resources in the production workshop based on the collaborative model, including machines, workpieces, production lines and their status parameters; Through sensors and shop floor control systems, real-time status data of manufacturing resources is collected, including the operating status of machines, the processing sequence of workpieces, and the resource allocation of production lines; Preprocess the collected real-time status data, remove invalid data, and interpolate missing data to form a standardized data input set; Based on the task flow and resource flow relationships defined in the collaborative model and combined with the data input set of manufacturing resources, an intelligent configuration model for workshop manufacturing resources is established to describe the logical relationship between resource allocation and process scheduling. The intelligent configuration model of manufacturing resources is used to sort the processes in the task flow, and an initial scheduling plan is generated based on the resource allocation logic and process priority.

[0011] Furthermore, the method for optimizing the initial scheduling scheme using a genetic algorithm includes the following steps: Step 1: Encode the initial scheduling plan and represent the processing sequence and machine allocation of the process as chromosomes. Each chromosome consists of multiple genes, and genes represent the processing information of the process. Step 2: Based on the scheduling rules and the position swap operation, an initial population is generated. Each chromosome in the initial population represents a possible scheduling solution. Step 3: Calculate the fitness value of each chromosome in the initial population. The fitness function uses the makespan of the initial scheduling plan as the evaluation index, and the fitness value is inversely proportional to the makespan. Step 4: Based on the roulette wheel selection strategy, select chromosomes with high fitness according to the fitness values of the chromosomes to enter the next generation population. Step 5: Perform crossover operations on the selected chromosomes. Randomly select two chromosomes as parents, generate two offspring chromosomes through partially mapped crossover, and retain the excellent genes of the parents. Step 6: Perform mutation operations on the offspring chromosomes after crossover. Use the position swap method to randomly exchange the positions of two genes in the chromosome to generate new chromosomes. Step 7: Evaluate the fitness values of the newly generated chromosomes according to the fitness function, and retain the chromosomes with high fitness in the next generation population. Step 8: Repeat Steps 4 to 7 until the predetermined number of iterations is reached or the fitness value converges, and output the chromosome with the highest fitness as the optimized scheduling plan. Step 9: Decode the optimized scheduling plan to determine the processing order of each process and the resource allocation of the corresponding machines, and generate an executable scheduling plan that meets the workshop resource constraints.

[0012] Further, the method for generating a rescheduling plan through the Monte Carlo tree search method based on an event-driven rescheduling strategy includes the following steps: Step 1: When a dynamic event occurs, monitor and collect the current system state of the workshop, including unfinished processes, the execution progress of the current task, and the available manufacturing resource status. Step 2: Based on the current system state and the characteristics of the dynamic event, initialize the Monte Carlo tree search model, and use the system initial state as the root node of the search tree. Step 3: Starting from the root node, generate the child nodes of the root node according to the priority of the workshop tasks and the sequence constraints between processes. The child nodes represent the executable next process. Step 4: Based on the upper confidence bound strategy, select the optimal child node from the child nodes of the current node and continue to expand the search tree. Step 5: When expanding to the leaf node, generate a complete path of the process scheduling through random simulation and calculate the makespan corresponding to the path. Step 6: Transmit the simulation results of the leaf node back to the root node along the search path, and update the average return value and the number of visits of each node. Step 7: Repeat Steps 4 to 6 until the set number of simulations or the search depth is reached, and use the child node with the most visits in the root node as the next operation of the current optimal process scheduling. Step 8: According to the selected next process step, set its corresponding node as the new root node, and repeat Steps 3 to 7 to generate a complete rescheduling plan.

[0013] Furthermore, the calculation formula of the upper confidence bound strategy is: Wherein, represents the child node pointed to by the current root node, represents the child node 's average return, is a fixed parameter, represents the number of simulations performed by the child node , represents the number of simulations performed by the root node of the child node , represents the total return received by the child node .

[0014] Furthermore, the dynamic events include machine failures, changes in workpiece processing times, order cancellations, and random workpiece arrivals. The generation of the rescheduling plan needs to update the current system state based on the specific type of dynamic event and redefine task priorities.

[0015] Furthermore, the method for identifying key nodes and optimizing the dynamic scheduling strategy in combination with the fault propagation network model includes the following steps: Select key parts, machines, and resources in the manufacturing system to establish a node set, where each node represents a part, device, or resource; Analyze the coupling relationships between the nodes in the manufacturing system. Based on whether there is resource sharing, dependency, or task collaboration between the nodes, construct a connection relationship matrix between the nodes, and the element values in the connection relationship matrix are assigned according to the coupling relationships; Generate a fault propagation network diagram according to the connection relationship matrix. The nodes of the network represent the resources or devices in the manufacturing system, and the edges of the network represent the coupling relationships between the nodes; Based on the fault propagation network diagram, calculate the degree, in-degree, and out-degree of each node to determine the high-degree nodes with more connection relationships in the network; Based on the degree analysis results, calculate the clustering coefficient of the high-degree nodes to identify the nodes with stronger local relevance to the surrounding nodes; On the basis of identifying the clustering nodes, calculate the average path length and network diameter of the fault propagation network diagram, analyze the positions of the key nodes in the global propagation path, and identify the nodes that have an important impact on the global fault propagation; Combined with the statistical characteristics results of the fault propagation network diagram, preferentially adjust the scheduling priorities of tasks related to key nodes and reallocate the resources required for key nodes; Using the optimized scheduling strategy, prioritize the completion of tasks related to key nodes when dynamic events occur, and generate an optimized dynamic scheduling plan.

[0016] To achieve the above object, a second aspect of the present invention provides an electronic device, including a processor and a memory. When the processor executes the computer program stored in the memory, it implements the steps of the intelligent manufacturing management method based on digital twin technology.

[0017] To achieve the above object, a third aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, it executes the steps of the intelligent manufacturing management method based on digital twin technology.

[0018] Advantages of the present invention: Compared with the prior art, an intelligent manufacturing management method, device and medium based on digital twin technology provided by the present invention realizes the full-process closed-loop management from real-time monitoring of production status to triggering of dynamic events by combining a collaborative model, a fault propagation network model and a dynamic scheduling optimization algorithm. Specifically, the present invention uses the collaborative model to clarify the relationships of information flow, task flow and resource flow in the manufacturing system, and constructs an intelligent configuration model of workshop manufacturing resources based on the real-time collected production status data; uses the genetic algorithm to optimize the initial scheduling plan and generate an efficient scheduling plan that meets the workshop resource constraints; when dynamic events occur, generates an event-driven rescheduling plan based on the Monte Carlo tree search method, and combines the fault propagation network model to identify key nodes and preferentially optimize their scheduling order, thereby improving the response efficiency of dynamic scheduling and the robustness of the scheduling plan, and effectively solving the problems of inability to optimize production scheduling in real time, inflexible handling of dynamic events and low resource utilization efficiency in the prior art. Description of the Drawings

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for description in the embodiments.

[0020] Figure 1 It is a flowchart of an intelligent manufacturing management method based on digital twin technology disclosed in an embodiment of the present invention.

[0021] Figure 2 It is a flowchart of a heuristic scheduling rule disclosed in an embodiment of the present invention.

[0022] Figure 3 It is a schematic diagram of OX crossover disclosed in an embodiment of the present invention.

[0023] Figure 4 It is a schematic diagram of a 2-exchange method disclosed in an embodiment of the present invention.

[0024] Figure 5 It is a flowchart of an IGAM algorithm disclosed in an embodiment of the present invention.

[0025] Figure 6 It is a Gantt chart of a process scheduling disclosed in an embodiment of the present invention.

[0026] Figure 7 It is a schematic diagram of a Monte Carlo tree search disclosed in an embodiment of the present invention.

[0027] Figure 8 It is a simulation flowchart of walking from the root node to the end state disclosed in an embodiment of the present invention.

[0028] Figure 9 It is a flowchart of constructing a fault propagation network model disclosed in an embodiment of the present invention. Detailed implementation manners

[0029] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0030] According to the embodiments of the present invention, it should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the following methods, in some cases, the steps shown or described can be executed in an order different from that here.

[0031] As Figure 1 shown, the present invention provides an intelligent manufacturing management method based on digital twin technology, including the following steps: Step S100: Simulate and model the collaborative relationships among various departments in the manufacturing system, and construct a collaborative model including information flow, task flow, and resource flow to define the collaborative element relationships between the design department, the manufacturing department, and other departments; Step S200: Based on the collaborative model, collect the real-time status data of the production workshop in the manufacturing system, and establish a simulation model for intelligent allocation of workshop manufacturing resources to generate an initial scheduling plan; Step S300: Optimize the initial scheduling plan using a genetic algorithm to generate an executable scheduling plan based on shop floor resource constraints. The optimization objective is to minimize the makespan. Step S400: Apply the executable scheduling plan to the production shop, monitor the occurrence of dynamic events during production, and collect relevant dynamic data. Step S500: When a dynamic event occurs, update the collected relevant dynamic data. Using the real-time feedback system status information, based on an event-driven rescheduling strategy, generate a rescheduling plan through the Monte Carlo tree search method. Step S600: Transmit the generated rescheduling plan to the production shop, adjust the operation tasks of each machine, and record in real time the execution data of each process and the scheduling adjustment records during production. Step S700: Based on the recorded process execution data and scheduling adjustment records, construct a digital twin model of the shop floor production process to simulate and analyze the impact of dynamic events on the production process, and identify key nodes in combination with the fault propagation network model to optimize the dynamic scheduling strategy. Step S800: Use the simulation results and analysis data of the digital twin model to generate optimized scheduling parameters and model update strategies, and apply them to subsequent production scheduling.

[0032] In this embodiment, as described in step S100 above, step S100 mainly involves simulating and modeling the collaborative relationships among various departments in the manufacturing system. The purpose is to construct a collaborative model that includes information flow, task flow, and resource flow, and is used to define the collaborative element relationships among the design department, manufacturing department, and other departments. Through this model, the interaction relationships and resource allocation methods among departments during the design, production, and testing processes can be clarified, providing a basis for subsequent manufacturing resource allocation and scheduling optimization.

[0033] There are complex collaborative relationships among various departments (such as the design department, manufacturing department, and testing department) in the manufacturing system during the product design, production, and testing processes. First, it is necessary to analyze the core functional elements of each department, including: Information flow: Describes the data transfer relationships among departments. For example, the design drawings generated by the design department need to be transferred to the manufacturing department, and the production status information of the manufacturing department needs to be fed back to the testing department. Task flow: Defines the task allocation and interdependent relationships among departments. For example, after the task of the design department is completed, it triggers the production task of the manufacturing department. Resource flow: Describes the demand and allocation relationships of each department for resources such as manpower, equipment, and materials.

[0034] In the collaborative model, the information flow is a crucial part for defining data interaction between departments. Specifically: Determine the information transmission path, such as the process of design data being transmitted from the design department to the manufacturing department; Define the information transmission rules, including data format, transmission period, and data verification method; Determine the feedback mechanism, such as the manufacturing department promptly feeding back production status information to the design department or the test department.

[0035] The task flow model is used to describe the task allocation and dependency relationships between departments. Its establishment process includes: clarifying the task inputs, outputs of each department, and the dependency order between tasks; Establishing a task triggering mechanism, such as the start of a manufacturing task depending on the completion of a design task; Defining the priority and time constraints for task execution to ensure efficient collaboration between tasks.

[0036] The resource flow model is used to describe the resource allocation and flow situation of each department in the manufacturing system. The specific steps are: listing the types of resources required by each department (such as equipment, manpower, materials, etc.); Determining the resource allocation rules and boundary conditions, such as the design department having priority in using a certain type of equipment; Analyzing the dynamic changes of resources, such as the idle or faulty situation of equipment during the manufacturing process.

[0037] After completing the modeling of the information flow, task flow, and resource flow, the three are integrated to construct a unified collaborative model. The collaborative model is represented by a multi-layer network structure, where: the information flow is the path of data interaction; the task flow is the logic of task allocation and execution; the resource flow is the rules of resource allocation and scheduling.

[0038] The constructed collaborative model needs to be verified and adjusted to ensure that it accurately reflects the collaborative relationship in the manufacturing system. The verification content includes: whether the data transmission is accurate; whether the task dependencies are logically reasonable; whether the resource allocation meets the boundary conditions.

[0039] In this embodiment, as described in step S200 above, through this step, the real-time status of each resource in the production workshop is clarified, providing accurate input information for scheduling optimization and generating a preliminary scheduling plan.

[0040] Based on the collaborative model constructed in step S100, first clarify the key manufacturing resources and status parameters to be monitored. Manufacturing resources include but are not limited to machines, workpieces, production lines, and their related parameters; Status data includes the real-time working status of resources, task execution progress, and current availability. The specific process is as follows: Based on the definition of the resource flow in the collaborative model, clarify the machine status, workpiece position, and production line load conditions to be collected. For example, it is necessary to collect whether the machine is in a running state, the progress of the workpiece in the production process, and the overall utilization rate of the production line.

[0041] Through data acquisition devices such as sensors and industrial Internet of Things (IoT) devices deployed in the workshop, the above-mentioned resource status data can be obtained in real time. For example, the operating status of a machine can be collected by vibration sensors, production progress information can be recorded by barcode scanners, and the load data of the production line can be collected by PLC (programmable logic controller).

[0042] Use the manufacturing execution system (MES) or the workshop control system to integrate the real-time data stream of the sensors and upload it to the central data processing system through wireless or wired communication.

[0043] Since the data collected in real time may contain noise, outliers, or missing data, it is necessary to preprocess the data to ensure the accuracy and consistency of the data. The preprocessing includes the following steps: Data cleaning: Remove noise data and invalid data. For example, if the status of a certain machine is abnormal due to sensor failure, the abnormal data should be removed.

[0044] Data completion: Fill in the missing data using interpolation or historical data matching methods. For example, the missing machine status data for a certain time period can be interpolated based on the trend of adjacent time periods.

[0045] Data standardization: Uniformly format the data from different sources to ensure the consistency of the data range, unit, and format of each status parameter.

[0046] After preprocessing, a standardized real-time data set is generated as the input of the intelligent configuration model. Using the processed real-time status data and combining the definitions of the resource flow and task flow in the collaborative model, an intelligent manufacturing resource configuration model is established. The intelligent configuration model describes the allocation relationship between resources and tasks and includes the following content: Association relationship between tasks and resources: Clearly define the resources required for each task. For example, the first process of a certain workpiece needs to be processed on a specific machine, and the second process needs to be assigned to another machine to complete.

[0047] Resource constraint conditions: Define the availability constraints of resources, such as the maximum load of a certain machine, and specific workpieces must be processed in sequence, etc.

[0048] Task execution order: Determine the priority of tasks according to the task flow model. For example, a certain workpiece needs to complete the previous process before starting the subsequent process.

[0049] Based on the intelligent manufacturing resource configuration model, a preliminary scheduling plan is generated to ensure that resource allocation and task scheduling meet the constraint conditions of the current production status. The specific process is as follows: Allocate each task to the available resources that meet the constraint conditions. For example, allocate a certain process of a specific workpiece to an idle machine.

[0050] Determine the execution order of each process on resources in combination with the task priority. For example, arrange the processing order of processes on different resources according to the sequence constraints of workpiece processing.

[0051] Allocate a time window for each process to ensure that there are no conflicts in resource usage. For example, a certain machine needs to schedule the time for the next task after the previous task is completed.

[0052] After the initial scheduling plan is generated, it is necessary to verify the plan to ensure its feasibility. The verification content includes: checking whether there are resource allocation conflicts. For example, whether a certain machine is arranged to execute overlapping tasks of multiple processes. Verifying whether the task execution order meets the dependency relationships in the task flow. For example, whether the second process of a certain workpiece starts after the first process is completed. Ensuring that the scheduling plan is consistent with the collected real-time status data. For example, whether the machine assigned to a certain task is currently idle. After verification, the generated initial scheduling plan is used as the input for the next optimization.

[0053] In this embodiment, as described in step S300 above, the improved genetic algorithm searches for the optimal solution of the scheduling plan by simulating the biological evolution process and combining operations such as selection, crossover, and mutation. The following is the specific detailed implementation process.

[0054] In this step, first, it is necessary to mathematically model the static Job-shop scheduling problem (static JSP). Assume that the transfer time between the same machines for each workpiece is the same, and there is no priority between workpieces. The optimization goal is to minimize the makespan. . Its mathematical model is as follows: The objective function is: (1) The constraint conditions are: (2) Where ; (3) Where ; ; (4) Where ; (5) Where ;

[0055] (6) Where ; Among them, represents the objective function, represents the makespan; represents job the completion time of processing on machine ; represents job the start time of processing on machine ; represents job the start time of processing on machine ; represents job the processing time on machine ; represents job the transfer time from machine to machine ; represents the machine number; represents the job number; represents the total number of machines; represents the total number of jobs; represents that job needs to be processed on the th machine first, and then on the th machine; is the set of process sequences, indicating that jobs must be processed in sequence; represents a sufficiently large positive number; represents a binary variable (0 or 1). If job is prior to job in processing on machine , then , otherwise it is 0; represents the start time of job on machine ; represents the start time of job on machine ;

[0056] Equation (1) indicates that the objective function is to minimize the makespan; Inequality (2) constrains that the makespan must be greater than or equal to the completion time of each job, that is, the makespan is the completion time of the last process in the processing; Inequality (3) constrains the sequential processing order of processes; Inequalities (4) and (5) constrain that only one process can be carried out on one machine at the same time; Inequality (6) constrains that the start time of each job is after time 0.

[0057] The genetic algorithm is an intelligent optimization algorithm that draws on the biological evolution mechanism and searches for the optimal solution to a problem through population iteration. Its specific process includes the following steps: Population initialization: Generate a set of random chromosomes (initial solutions).

[0058] Fitness evaluation: Calculate the fitness value of each chromosome to measure its quality.

[0059] Selection operation: Select high-quality individuals based on the fitness value to enter the next generation.

[0060] Crossover operation: Generate new individuals by exchanging part of the genes of the chromosomes.

[0061] Mutation operation: Randomly mutate the genes of the chromosomes to increase the diversity of the population.

[0062] Termination condition: When the preset number of generations or convergence condition is reached, output the optimal solution.

[0063] The optimization process of the genetic algorithm based on static JSP is as follows: (1) Chromosome encoding Encoding in the genetic algorithm represents the potential solutions to the problem as chromosomes, which will be created, evaluated, selected, crossed, and mutated during the evolution process of the genetic algorithm. Encoding is a very crucial step in the genetic algorithm because it directly affects the performance and efficiency of the algorithm. When choosing an encoding method, the characteristics of the problem and the requirements of the algorithm need to be considered. The choice of encoding affects the design of the crossover and mutation operations because these operations must maintain the legality of the chromosomes, that is, the newly generated chromosomes are still a valid solution to the problem. Therefore, the encoding method needs to be closely combined with the characteristics of the problem and the operations of the genetic algorithm. In this paper, for the JSP problem, the permutation encoding based on operations is the most direct choice because it can directly represent the execution order of tasks and has the characteristics of simple decoding, easy programming, and high flexibility compared with other encoding methods. Most importantly, under this encoding method, any gene sequence can represent a feasible schedule.

[0064] The following explains the permutation encoding based on operations: For the static JSP problem, the length of the gene sequence of each chromosome is m×n, and each gene represents an operation of a workpiece. Then, in a chromosome, the serial number of a workpiece can only appear m times, and the serial number of the workpiece operation is determined by the order in which the workpiece serial number appears in the gene sequence. Taking a 3×3 JSP case as an example for detailed explanation, the workpieces are represented by the numbers 1, 2, and 3. Suppose a chromosome is generated as [1 3 3 1 2 2 3 1 2]. The first occurrence of the number 3 represents the first operation of workpiece 3, the second occurrence of the number 3 represents the second operation of workpiece 3, and so on.

[0065] The decoding process follows specific rules to convert the encoded chromosomes into a specific scheduling plan, which is presented in the form of a Gantt chart. In this paper, a process insertion method is used for decoding, which can ensure that each process is scheduled as early as possible without disturbing the arrangement of other processes, thus effectively utilizing machine resources.

[0066] The following is an explanation of the process insertion decoding: According to the encoding method described above, a 3×3 JSP case is used for detailed explanation. Suppose a chromosome [1 3 3 1 2 2 3 1 2] needs to be decoded. Convert this gene sequence into a process sequence, , denoted as sequence. The specific algorithm flow of decoding is as shown in Table 1 below, where the parameter M is the set of machines: Table 1: Specific algorithm flow table for decoding Decoding is completed to obtain the start processing time and completion time of each process; draw the Gantt chart Gantt_chart and obtain the makespan.

[0067] (2) Population initialization The initial solution has a certain impact on the performance of the final algorithm. A good initial solution can greatly improve the solution quality and convergence speed of the algorithm. Therefore, for the JSP problem, this method uses the PR heuristic scheduling rule and combines several widely used scheduling rules to generate the initial population of the genetic algorithm.

[0068] First, the following is an explanation of the PR heuristic scheduling rule, Figure 2 which shows the flowchart of the rule execution: Step 1: The first process of all workpieces is processed by the corresponding machine. If multiple processes are assigned to the same machine, the workpiece with the longest processing time of this process is given priority; if multiple processes have the same processing time, the workpiece with the longest remaining processing time is given priority.

[0069] Step 2: The next process of all workpieces is processed by the corresponding machine. If multiple processes are assigned to the same machine, the workpiece with the earliest completion time of the previous process is given priority; if the previous processes of these processes to be assigned have the same completion time, the workpiece with the longest processing time of this process is given priority.

[0070] Step 3: Repeat Step 2 until the allocation of all processes is completed.

[0071] Furthermore, the following is an explanation of four scheduling rules that meet the assumed conditions: Shortest processing time first (SPT): The workpiece with the shortest processing time in the queue of processes to be processed has the highest priority.

[0072] Longest Processing Time First (LPT): The job with the longest processing time in the queue of jobs to be processed has the highest priority.

[0073] Shortest Remaining Processing Time First (SRPT): The job with the shortest remaining processing time among the corresponding jobs in the operation queue has the highest priority.

[0074] Longest Remaining Processing Time First (LRPT): The job with the longest remaining processing time among the corresponding jobs in the operation queue has the highest priority.

[0075] Based on the PR scheduling rule and the above four scheduling rules, the detailed operation steps of the hybrid scheduling rule (DRM) for initializing the population are as follows: Five chromosome generation methods, namely PR+swap, SPT+swap, LPT+swap, SRPT+swap, and LRPT+swap, are used to generate chromosomes with a population size of 20%. The swap method means swapping two positions in the chromosome genes. For example, PR+swap means first generating a chromosome (i.e., a scheduling plan) using the PR rule, and then mutating this chromosome using the swap method to generate more chromosomes. The population initialization method based on DRM ensures both high-quality population individuals and the randomness of individuals, which helps the genetic algorithm converge faster.

[0076] (3) Fitness function In each iteration of the genetic algorithm, all chromosomes need to be evaluated by the fitness function to determine their performance. Individuals with high fitness have a greater chance of participating in crossover and mutation, thus generating a new population. By continuously selecting individuals with high fitness, the algorithm can effectively approach the optimal solution. The fitness function is generally a transformation of the objective function, used to measure the quality of individuals. From the objective function (1) of the above static JSP model, the fitness function can be obtained as: (7) Where, is the fitness function; represents the objective function; It can be seen from the above formula that the larger the makespan, the smaller the fitness value; conversely, the larger.

[0077] (4) Selection operation In this embodiment, the roulette wheel selection method is adopted. The core idea of this method is to determine the probability of an individual being selected according to its fitness by simulating a roulette wheel game. This method can effectively balance the selection probability of individuals while maintaining the diversity of the population. The roulette wheel selection is described as follows: Step 1: Calculate the fitness of each chromosome according to the fitness function (7) , and thus calculate the sum of the fitness of the population : (8) Among them, is the population size.

[0078] Step 2: Calculate the selection probability of chromosome on the roulette wheel : (9) Among them, is the chromosome number.

[0079] Step 3: Calculate the cumulative probability of chromosome on the roulette wheel : That is, the sum of the individual selection probabilities of the first chromosomes: (10) Among them, represents the selection probability of chromosome (i.e., the fitness ratio).

[0080] Step 4: Randomly generate a number , ; Step 5: If , then select the th chromosome; if , then select the first chromosome.

[0081] (5) Crossover operation The crossover operation in the genetic algorithm simulates the reproduction process in biological inheritance. It is a mechanism that combines partial characteristics of two parent individuals to produce offspring. The crossover operation is one of the core steps in the genetic algorithm. It helps to explore new solution spaces, increase the diversity of the population, and contribute to the spread of excellent genes. Among many crossover operators, in this embodiment, a crossover operator OX based on process coding is used, which can well inherit the excellent characteristics of the parent generation. The OX crossover operator algorithm is described below: The two parent chromosomes are denoted as and , and the two offspring chromosomes generated by crossover are denoted as and .

[0082] Step 1: Randomly select the start and end positions of several genes in the parent chromosomes and . The selected positions of the two chromosomes are the same, and are respectively called gene segments and .

[0083] Step 2: Retain until . Traverse the workpieces in from the beginning, remove the workpiece labels that are included in and are the first occurrences, and insert the remaining part in the relative order into .

[0084] Step 3: Retain until . Traverse the workpieces in from the beginning, remove the workpiece labels that are included in and are the first occurrences, and insert the remaining part in the relative order into .

[0085] Figure 3 Fig. shows a case of OX crossover.

[0086] (6) Mutation operation The mutation operation in the genetic algorithm is a process of randomly adjusting the individual genes. This adjustment can be either a minor change in the gene value or a complete randomization. It mimics gene mutations in the biological world, introduces new genetic elements into the population, and helps the algorithm escape from local optima and broaden the search scope. Common mutation operations in the genetic algorithm include single-point mutation, multi-point mutation, Gaussian mutation, etc. In this embodiment, the 2-exchange mutation method based on position swapping is adopted, and this method is described below: For a chromosome in the population, the length of the gene sequence is . Randomly generate two different integers in , and swap the data at the positions of these two integers to generate a new chromosome. Figure 4 Fig. shows a case of 2-exchange.

[0087] In summary, it can be understood that to prevent premature convergence and find a better solution, this embodiment uses an improved genetic algorithm (IGAM). This method generates 2n offspring by crossing two parent generations n times, and then selects the two different individuals with the highest fitness from the parent generation and these 2n offspring to form a new generation population. This generation method not only retains the excellent characteristics of the parent generation but also ensures the richness of the offspring population. As shown in Figure 5 , the IGAM genetic algorithm is described below: N represents the population size, represents the crossover rate in the crossover algorithm, represents the mutation rate in the mutation algorithm.

[0088] Step 1: Generate N chromosomes according to the DRM population initialization method; Step 2: Calculate the fitness value of the individual according to the fitness function; Step 3: Determine whether the termination condition is satisfied. If it is satisfied, output the current solution; otherwise, go to Step 4; Step 4: Select the next generation population to be iterated according to the roulette wheel selection strategy; Step 5: Randomly generate a decimal number in the interval [0, 1] . If , randomly select two individuals from the current population as parents, then perform the crossover operation on the two parents n times, and then select the two best individuals from the parents and the generated offspring as the offspring; otherwise, go to Step 6; Step 6: Randomly generate a decimal number in the interval [0, 1] . If , randomly select two individuals from the current population as parents, then perform the mutation operation on the two parents, and then use the generated offspring chromosomes as the offspring; otherwise, directly use the two parent chromosomes as the offspring; Step 7: Repeat Steps 5 and 6 N / 2 times to generate a new generation population, and go to Step 3.

[0089] Through the above steps, the improved genetic algorithm is used to optimize the initial scheduling scheme, generating an efficient executable scheduling scheme based on the workshop resource constraints, ensuring a reasonable task execution order and minimizing the makespan, laying a foundation for subsequent dynamic event processing and scheduling adjustment.

[0090] To verify the effectiveness of the IGAM algorithm proposed in this embodiment, the present invention conducts experiments based on 8 JSP standard example data, and compares the solution results of the IGA algorithm with randomly initialized population and the IGAM algorithm. The algorithm conducts 10 independent experiments, and takes the best one as the final result. The relevant experimental settings are as follows: Programming language: Python Basic parameters: Population size N: 40; Crossover probability : 0.8; Mutation probability : 0.1; Maximum number of iterations n*m; Transfer time is 3 for all.

[0091] Iteration termination condition: When the maximum number of iterations is reached, the iteration terminates.

[0092] Table 2: Running results of the IGAM algorithm

[0093] As can be seen from the comparison of the two different algorithms in Table 2 above, the maximum processing time obtained by the IGAM algorithm is smaller than that of the IGA algorithm, indicating that under the condition of the same number of iterations, the solution quality obtained by using the IGAM algorithm is generally higher than that of the IGA algorithm. The experiment proves that the population initialization method proposed by the present invention can enable the genetic algorithm to obtain better solutions. Taking the FT06 standard example as an example, Figure 6 is the Gantt chart for the operation scheduling of 6 workpieces and 6 machines when the maximum processing time is 67. The IGAM can obtain the optimal solution to the problem approximately in the second generation. To sum up, the population initialization method proposed by the present invention for the static JSP scheduling problem enables the genetic algorithm to have good search ability.

[0094] In this embodiment, as described in step S400 above, the goal of this step is to ensure the implementation of the scheduling plan in actual production and at the same time provide real-time data support for subsequent dynamic scheduling.

[0095] The optimized scheduling plan generated in step S300 includes the processing sequence of each operation, the resource allocation of the corresponding machine, and the time arrangement. The following is the specific implementation process: Task allocation: According to the optimized scheduling plan, allocate the processing tasks of each operation to the corresponding machine and clarify its processing time window.

[0096] Start execution: Send the allocated tasks to the equipment in the production line through the shop floor control system (such as the MES system), and the equipment executes each task according to the preset time sequence.

[0097] Execution process monitoring: Use the sensors installed on the production equipment and workpieces to record the execution status of the tasks in real time, including the equipment operation status, workpiece progress, production completion time, etc.

[0098] Through the implementation of the scheduling plan, the production workshop starts to execute the optimized scheduling plan.

[0099] During the production process, due to the complexity of the workshop environment, various dynamic events may occur, interfering with the established scheduling plan. The following are common dynamic events including: machine failures, changes in workpiece processing time, order cancellations, arrival of random workpieces, etc., among which: Trigger monitoring method for machine failures: Real-time monitor the health status of the equipment through the built-in state sensors of the equipment (such as vibration sensors, temperature sensors). Once the equipment stops running or shows abnormal status, trigger a dynamic event.

[0100] Trigger monitoring method for changes in workpiece processing time: Real-time collect the processing time of the workpiece through the time recording system of the production equipment, compare it with the preset time, and trigger an event when the deviation exceeds the threshold.

[0101] Trigger monitoring method for order cancellation: Receive real-time updates on order status through the MES system. When an order is cancelled, relevant dynamic events are triggered.

[0102] Trigger monitoring method for the arrival of random workpieces: Monitor the generation of new orders in real-time through the order management system. When a new task is detected, an event is triggered.

[0103] To respond to dynamic events, it is necessary to collect dynamic data during the production process in real-time. These data include: Equipment status data: The operating status of the equipment (such as whether it is running normally, running time, fault status, etc.).

[0104] Workpiece status data: The actual start time, completion time, and processing progress of each process.

[0105] Production line status data: The overall load situation of the production line, including equipment idle rate, utilization rate, etc.

[0106] The specific data collection methods are as follows: Use built-in sensors in the equipment (such as pressure sensors, vibration sensors) to collect the operating status of the equipment; collect the processing progress of workpieces through barcode scanning equipment.

[0107] Collect task assignment and execution status through the shop floor control system or the MES system.

[0108] Upload the collected data to the central data platform through industrial Internet of Things (IoT) devices for subsequent analysis and processing. The occurrence of dynamic events usually has the following impacts on the established scheduling plan: Equipment failures may cause tasks to not be completed on time, thus delaying the execution of subsequent processes. The extension or shortening of processing time will disrupt the original time arrangement, resulting in unreasonable resource allocation. Cancelled tasks may lead to the idleness of relevant equipment and resources, and uncompleted tasks need to be reallocated. The addition of new tasks will increase resource requirements and may cause resource competition with existing tasks.

[0109] Update the task completion status based on the real-time collected workpiece status data, and adjust the time arrangement of subsequent tasks in the scheduling plan. Identify conflict problems in resource allocation by monitoring equipment status and load conditions, and adjust the resource allocation order. When a dynamic event is detected, transfer the relevant data to the dynamic scheduling module to trigger the subsequent rescheduling process.

[0110] Through the above steps, step S400 has completed the implementation of the optimized scheduling plan, the monitoring and collection of dynamic events, providing accurate real-time data support for subsequent dynamic scheduling optimization (step S500), ensuring the continuity and efficiency of the production process in a dynamic environment.

[0111] In this embodiment, as described in step S500 above, this step aims to address the interference of dynamic events on the production process, adjust the scheduling plan, and ensure the continuity and efficiency of the production system.

[0112] The occurrence of dynamic events will cause the established scheduling plan to fail to meet the actual production requirements, so it is necessary to respond in a timely manner and update the system state. Dynamic events include the following types: Machine failure: A certain machine cannot continue to run, and relevant tasks need to be rearranged to other available machines.

[0113] Change in workpiece processing time: The actual processing time of a certain process exceeds or is shorter than the planned time, and the time arrangement of subsequent processes needs to be adjusted.

[0114] Order cancellation: The order for which tasks have been assigned is cancelled by the customer, and relevant tasks need to be deleted from the scheduling plan.

[0115] Random workpiece arrival: New workpiece tasks need to be inserted into the current scheduling plan and compete for resources with existing tasks.

[0116] The content of the system state update includes: the task list of unfinished processes currently, the available status of each machine (idle or busy), the processing progress of workpieces (completed processes, processes to be completed), and the list of resources or tasks affected by dynamic events. Through real-time data collection and event-triggering mechanisms, the system inputs dynamic events and their related data into the rescheduling module.

[0117] After a dynamic event occurs, an event-driven rescheduling strategy is adopted to reschedule unfinished tasks. The core features of event-driven rescheduling include: triggering rescheduling immediately when a dynamic event occurs, without waiting for the time interval of periodic scheduling. Rescheduling only the tasks affected by dynamic events and their associated tasks, reducing the computational complexity brought by global scheduling. Based on the original scheduling plan, incrementally adjusting the affected part of the processes to avoid causing large disturbances to the entire production plan.

[0118] The specific process is as follows: Collect information related to dynamic events (such as the number of the faulty machine, new tasks, etc.). Determine the set of affected processes and their constraints (such as task dependencies, resource availability). According to the latest system state and constraints, re-plan the scheduling of the affected processes.

[0119] In this embodiment, an optimized scheduling plan is generated through the Monte Carlo Tree Search method (MCTS). It should be noted that: The Monte Carlo Method is a numerical calculation method based on probability and statistics theory. It simulates the random process of practical problems through random sampling, and then makes statistical estimates of the solutions to the problems. Its basic idea is to use random numbers to simulate various possibilities of practical problems and obtain approximate solutions to the problems through a large number of repeated simulations.

[0120] The Monte Carlo method can generate samples by constructing a Markov chain and use these samples to approximately estimate the expected value and optimal strategy in Markov Decision Processes (abbreviated as MDPs). Applying the Monte Carlo method to MDPs can significantly improve the accuracy and efficiency of decision-making, reduce the computational complexity, and is suitable for dealing with high-dimensional problems. As Figure 7 shown, the specific process is as follows: Take the preliminary resource allocation plan as the root node of the search tree. The node represents the current scheduling state, and the edge pointing to the child node represents the action. Its iterative search is divided into the following four steps: (1) Selection: Starting from the root node, use the Upper Confidence Bound (UCB) strategy to select the optimal child node and expand a new child node, representing a new scheduling plan. The formula is as follows: (11) (12) Among them, represents the child node pointed to by the current root node, that is, the possible state in the next step of the current state, represents the average return of the child node , represents the number of simulations performed by the child node , represents the number of simulations performed by the root node of the child node , represents the total return received by the child node . is a fixed parameter. When is small, the tree search tends to select the action with the largest value; when is large, it tends to explore, that is, it tends to select , that is, the action with fewer simulation times. Generally, take .

[0121] (2) Expansion: If the selected node is a leaf node, expand a new child node, select according to the selection strategy, and incorporate it into the search tree.

[0122] (3) Simulation: Starting from the newly expanded node, conduct a complete random simulation until reaching the end of the search tree.

[0123] (4) Back propagation: Propagate the results of the simulation along the path back to the root node to update the statistical information of each node, namely and .

[0124] In MCTS, walking from the root node to the end state is called a simulation. When the iteration stops, the action with the most visited root node or the largest evaluation value will be returned, and the node pointed to by the selected action will be set as the new root node for the next iteration.

[0125] Each simulation starts from the root node and goes all the way to the leaf nodes of the entire tree. A schematic diagram of a simulation process is shown in Figure 8.

[0126] After the rescheduling plan is generated, it needs to be verified to ensure the feasibility and stability of the plan: The specific verification includes: checking whether the rescheduling plan meets all resource constraints and task dependencies, verifying the execution effect of the plan in actual production through a simulation model, such as whether tasks can be completed on schedule, and if conflicts or deficiencies occur in the simulation test, further adjusting parameters and optimizing. Finally, the verified rescheduling plan is transmitted to the production workshop to replace the original scheduling plan and execute subsequent production tasks.

[0127] Through the above steps, step S500 has completed the whole process from dynamic event triggering to generating a rescheduling plan based on MCTS, realizing efficient scheduling optimization in a dynamic environment and providing guarantee for the stable operation of subsequent production.

[0128] In this embodiment, as described in step S600 above, the rescheduling plan generated and verified in step S500 needs to be quickly transmitted to each device and control system in the production workshop. Through the workshop manufacturing execution system (MES), the operation tasks of each machine and the relevant resource allocation information are transmitted to the production line. The specific content includes processing tasks, start and end times, and resource requirements. The workshop control system will dynamically adjust the task execution order and resource allocation according to the task priorities and allocation rules. At the same time, after receiving the new task, the device side will feedback a confirmation signal to the central system through the communication module to ensure the smooth implementation of the rescheduling plan.

[0129] The occurrence of dynamic events affects the original scheduling plan, and the implementation of the rescheduling plan requires adjusting the operation tasks of the machines: When a machine needs to stop the current task due to a fault or other dynamic events, the system will interrupt the task, re-plan and schedule it to other available machines. For some tasks that have been completed, their data will be recorded and marked as interrupted tasks.

[0130] According to the rescheduling plan, the tasks affected by dynamic events are reallocated to idle machines. At the same time, the unaffected machines continue to execute tasks according to the original plan to avoid the shutdown of the entire production system.

[0131] After the rescheduling plan is determined, critical tasks are preferentially started to ensure the maximization of resource utilization and the continuity of overall production.

[0132] During the implementation of the rescheduling plan, it is necessary to record the actual execution of each process in real time. The main records include the actual start time and completion time of each process, compare them with the planned time, and analyze the deviations. Monitor the running status of the machines (such as idle, running or faulty) and their load conditions. Record the quantity and quality of the completed tasks and compare them with the preset goals.

[0133] Data collection mainly relies on sensors on the equipment (such as status sensors, time counters) and the MES system, and transmits the data to the central database for storage through industrial Internet of Things (IoT) devices to form a complete data chain.

[0134] The process of implementing rescheduling needs to track all scheduling adjustment records to ensure that the scheduling changes are clearly traceable. The main contents of the records include: recording the types of dynamic events that trigger rescheduling, such as machine failures, order cancellations, etc.; detailed records of the original scheduling plan before adjustment and the rescheduling plan after adjustment, including the changes in the machines and time for task allocation; recording the time from the triggering of the dynamic event to the completion of rescheduling, and analyzing the response speed of the scheduling system.

[0135] To ensure the effective implementation of scheduling adjustments, it is necessary to establish a complete execution closed-loop mechanism. Specific measures include: the scheduling system needs to verify the feasibility of the rescheduling plan before issuing it, and the key adjustment contents need to be confirmed by the workshop operators. Monitor the execution of scheduling adjustments in real time to ensure that all adjustment tasks are completed as planned. After each process is completed, the equipment end feeds back the task status to the scheduling system to update the scheduling records, forming a complete closed loop from plan generation to execution completion.

[0136] The process execution data and scheduling adjustment records logged during production provide the basic data for subsequent optimization and simulation. First, clean and preprocess the collected data, remove outliers, and fill in missing data to ensure data quality. Then, evaluate the execution effect of the scheduling plan based on the recorded data, including key indicators such as production efficiency, resource utilization rate, and response speed to dynamic events. Finally, update the digital twin model of the workshop in combination with this data to improve its simulation and prediction capabilities for dynamic events.

[0137] In this embodiment, as described in step S700 above, the digital twin model is a virtual mapping of the workshop production process. Its core is to construct a dynamic and interactive simulation model using real-time data and historical data in actual production for analyzing and optimizing the production process. The specific steps are as follows: The process execution data (such as task start time, completion time, machine utilization rate, etc.) and scheduling adjustment records (such as task assignment changes, reasons for triggering dynamic events, etc.) collected in step S600 are used as the input of the digital twin model. The data is processed through cleaning, filling, and formatting to ensure its integrity and consistency.

[0138] By inputting dynamic data such as the task flow, resource flow, and information flow of the workshop into the simulation system, a dynamic simulation model of the workshop covering equipment, workpieces, and processes is constructed. Combining the real-time data collected by IoT devices and sensors, the digital twin model is dynamically updated to enable it to reflect the state changes of the production workshop in real time.

[0139] Realtime display the operating status of the production workshop, including task progress, equipment load, resource allocation, etc. By analyzing the recorded deviation data, identify the bottlenecks and inefficient links in the current production scheduling. Simulate the impact of dynamic events (such as machine failures, new tasks added) to provide a reference for subsequent scheduling adjustments.

[0140] To analyze the impact of dynamic events on the production system and its propagation path, it is necessary to combine the fault propagation network model. This model identifies key nodes and their impacts on the system by describing the dependency relationships between various nodes (such as equipment, workpieces, tasks) in the production system. The specific construction steps are as follows: As Figure 9 shown, the core elements of the production workshop (such as equipment, workpieces, tasks) are used as the nodes of the fault propagation network. For example, in the production of the NIO ES8, the core parts of the power system (such as the battery pack, vehicle controller) can be selected as nodes.

[0141] Generate an adjacency matrix based on the dependency relationships between nodes (such as task sequence, resource sharing) to define the direct connection relationships between nodes. For example, if task A depends on task B to start, the corresponding position in the matrix is 1.

[0142] Use complex network modeling tools (such as Gephi) to visualize the adjacency matrix as a fault propagation network, showing the coupling relationships between nodes and potential fault propagation paths.

[0143] By analyzing the fault propagation network model, identify the key nodes in the production system. Key nodes are those that have a greater impact on the operation of the production system, and their failures may lead to a significant decline in the operating efficiency of the entire system. Specific analysis methods include: node degree analysis, clustering coefficient analysis, average path length analysis, simulation analysis of the impact of dynamic events, etc.; the above analyses will be specifically described in detail below: Node degree analysis: Calculate the degree (number of connected neighbors), in-degree (number of being affected by other nodes), and out-degree (number of affecting other nodes) of each node. Nodes with a higher degree are usually key nodes, indicating that they have greater interactivity in the network.

[0144] Clustering coefficient analysis: Calculate the clustering coefficient of each node, reflecting the coupling degree between its surrounding neighbor nodes. Nodes with a higher clustering coefficient may be local fault propagation centers.

[0145] Average path length analysis: Measure the shortest path length between nodes in the fault propagation network. If the path from a certain node to other nodes is short, then this node has a high efficiency in fault propagation and may be a weak link in the system.

[0146] Through the above analysis, determine the key nodes (such as tasks with high out-degree, key equipment, etc.), providing a basis for the subsequent optimization of dynamic scheduling strategies.

[0147] Simulation analysis of the impact of dynamic events: Use the digital twin model and the fault propagation network model to conduct simulation analysis on the impact of dynamic events in the production process. The following are the specific steps: Introduce dynamic events in the digital twin model, such as random workpiece arrival, machine failure, etc., and observe their impact on the production process. After simulating the event trigger, observe the queuing of unfinished tasks, resource conflicts, and task delays, etc.

[0148] Combined with the fault propagation network model, analyze the paths and scopes of the propagation of dynamic events from key nodes to other nodes. For example, the failure of a certain key equipment may cause multiple tasks connected to it to be delayed simultaneously.

[0149] Compare the key indicators (such as the makespan, equipment utilization rate) of the original scheduling scheme and the scheduling scheme after simulation adjustment, and evaluate the optimization effect of the scheduling adjustment.

[0150] According to the results of the simulation analysis, optimize the dynamic scheduling strategy to improve the robustness and response speed of the system. The optimization strategies include the following: When a dynamic event occurs, prioritize the scheduling of tasks related to critical nodes to avoid the cascading effect of their failures on the entire production system.

[0151] For resource conflicts, improve resource utilization efficiency by adjusting task priorities or adding backup resources.

[0152] Dynamically adjust scheduling rules (such as task priority rules and equipment allocation rules) so that the system can adapt to the changing production environment.

[0153] Combine event-driven rescheduling and periodic scheduling to balance real-time response capabilities and global optimization capabilities, and balance the stability and flexibility of production.

[0154] Apply the optimized scheduling parameters and model update strategies to subsequent production scheduling to achieve closed-loop optimization of dynamic event handling. The specific output content includes: adjusted task allocation and time planning. Optimize the parameters of the digital twin model based on the latest data to improve its prediction ability for dynamic events.

[0155] Through the above steps, step S700 realizes the whole process from recording data to digital twin model construction, fault propagation analysis, dynamic event simulation, and scheduling optimization, providing the production system with efficient and sustainable dynamic scheduling optimization capabilities.

[0156] In this embodiment, as described in step S800 above, this step aims to continuously optimize the scheduling plan through simulation analysis and feedback of historical data to improve the efficiency and robustness of the production system.

[0157] In step S700, based on the recorded process execution data and scheduling adjustment records, a digital twin model of the workshop has been constructed and combined with a fault propagation network for simulation analysis. This model can dynamically reflect the operating state of the production system, identify potential problems, and test the execution effects of different scheduling plans through simulation. Specifically, it includes: evaluating the execution time, resource utilization rate, and task delay of each process through the simulation results. Determining the scope and degree of influence of dynamic events on critical nodes and the overall production process through the fault propagation network. Simulating the effects of different scheduling rules in the production environment and analyzing their impacts on the makespan and resource utilization rate.

[0158] Combined with simulation analysis using digital twin models, optimized scheduling parameters are generated to improve production scheduling. The process of generating optimized scheduling parameters includes: Reducing overall production time by adjusting task sequence and resource allocation; Optimizing task allocation to reduce equipment idleness and overload; Prioritizing key node tasks to reduce the chain reaction caused by dynamic events; Redefining task priorities based on simulation results. For example, prioritizing tasks that have a greater impact on key nodes; Reallocating machine resources to improve utilization. For example, assigning tasks to machines with higher idle rates to avoid resource conflicts; Optimizing the start and end times of processes to ensure smooth transitions between tasks and reduce waiting times. Using scheduling rules that have proven effective in simulation, such as shortest processing time first (SPT) and longest remaining processing time first (LRPT), new scheduling parameters are generated. Using machine learning methods, optimal scheduling parameters are predicted based on historical data and simulation results.

[0159] As the production environment changes and data accumulates, the digital twin model needs to be continuously updated to maintain its simulation accuracy and predictive capabilities. The content and steps of the model update include: adjusting the parameters of the digital twin model according to the latest production data (such as equipment status, task execution time, etc.) to ensure that the model can accurately reflect the current production environment. Update the dependencies and resource constraints between processes to adapt to new task requirements and resource allocation. Combined with the feedback from simulation analysis, optimize the logic of the digital twin model. For example, enhance the model's ability to respond to dynamic events (such as machine failures and order changes). Introduce the long-term trend analysis function of historical data to predict dynamic events that may occur in the future and further improve the model's early warning capabilities. After the update is completed, the accuracy and applicability of the model are verified through simulation testing to ensure that it can guide subsequent production scheduling. The optimized scheduling parameters and updated digital twin model will be applied to subsequent production scheduling to improve the performance of the overall production system. The specific application process is as follows: Combining the optimized scheduling parameters with the model constraints, a new scheduling solution is generated to ensure that it meets the current production environment and task requirements. The new solution must achieve an optimal balance between task allocation, resource utilization, and time scheduling.

[0160] During the execution of the new scheduling plan, the digital twin model monitors production status and adjusts task and resource allocation in real time to ensure smooth production. When dynamic events occur, the optimized scheduling strategy is combined to quickly respond and adjust the production plan.

[0161] The execution data of the new scheduling plan is fed back to the digital twin model to further verify the optimization effect and accumulate data for the next round of optimization, forming a closed-loop optimization process of "simulation analysis - parameter optimization - scheduling implementation - data feedback".

[0162] Finally, the core outputs of step S800 include: information such as task priorities, resource allocations, and time schedules, which are used to guide the generation of subsequent scheduling plans. Rules for parameter adjustment and methods for logic optimization of the digital twin model to ensure that the model can continuously adapt to changes in the production environment. An optimization evaluation report generated based on simulation analysis and actual execution effects, including indicators such as improved production efficiency and shortened dynamic event response times.

[0163] Through the above steps, step S800 has completed the generation of optimized scheduling parameters and model update strategies using the simulation results and analysis data of the digital twin model, applied them to subsequent production scheduling, and constructed a dynamically optimized intelligent manufacturing closed-loop system.

[0164] According to another aspect of the embodiments of the present application, an electronic device is also provided, including a processor and a memory. The processor is used to implement the steps of the method when executing the computer program stored in the memory.

[0165] In the above embodiments of the present invention, the descriptions of the respective embodiments have their own emphases. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0166] In several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of units or modules can be in electrical or other forms.

[0167] In addition, the functional units in each embodiment of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0168] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs.

[0169] The foregoing are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An intelligent manufacturing management method based on digital twin technology, characterized in that The steps include: Simulate and model the collaborative relationships between departments in the manufacturing system, and build a collaborative model that includes information flow, task flow, and resource flow to define the collaborative element relationships between the design department, manufacturing department, and other departments; Based on the collaborative model, real-time status data of the production workshop in the manufacturing system is collected, and a simulation model for intelligent configuration of workshop manufacturing resources is established to generate an initial scheduling plan; Genetic algorithms are used to optimize the initial scheduling plan and generate an executable scheduling plan based on workshop resource constraints, where the optimization goal is to minimize the maximum completion time. Apply the executable scheduling plan to the production workshop, monitor the occurrence of dynamic events in the production process, and collect relevant dynamic data; When dynamic events occur, the system updates the collected dynamic data and uses real-time feedback of system status information to generate a rescheduling plan based on an event-driven rescheduling strategy using the Monte Carlo tree search method. The generated rescheduling plan is then delivered to the production workshop to adjust the work tasks of each machine and record the execution data and scheduling adjustment records of each process in real time. Based on recorded process execution data and scheduling adjustment records, a digital twin model of the workshop production process is constructed to simulate and analyze the impact of dynamic events on the production process. This model is combined with a fault propagation network model to identify key nodes and optimize dynamic scheduling strategies. Utilize the simulation results and analysis data of the digital twin model to generate optimized scheduling parameters and model update strategies, which are then applied to subsequent production scheduling.

2. The intelligent manufacturing management method based on digital twin technology according to claim 1, wherein The collaborative model construction methods include: Analyze the collaborative relationship between the design department, manufacturing department and other departments in the design, production and testing process of the manufacturing system, and determine the functional elements of each department, including information flow, task flow and resource flow; Based on functional elements, an information flow model is established to define the data interaction relationship between departments, including the transmission of design data, feedback of production status information, and sharing of resource allocation data; Based on functional elements, a task flow model is established to describe the task allocation and dependency relationships between departments, including the input and output of design tasks, the execution path of manufacturing tasks, and the result feedback of test tasks; Based on functional elements, a resource flow model is established to describe the allocation and use of human, equipment, and material resources by each department, and to clarify the boundaries of resource flow and collaboration methods of each department; The information flow model, task flow model and resource flow model are integrated to form a collaborative model that includes the collaborative relationship between departments.

3. The intelligent manufacturing management method based on digital twin technology according to claim 1, characterized in that, The method for collecting real-time status data of a production workshop in a manufacturing system based on a collaborative model, establishing a simulation model for intelligent configuration of workshop manufacturing resources, and generating an initial scheduling plan includes the following steps: Determine the manufacturing resources in the production workshop based on the collaborative model, including machines, workpieces, production lines and their status parameters; Through sensors and shop floor control systems, real-time status data of manufacturing resources is collected, including the operating status of machines, the processing sequence of workpieces, and the resource allocation of production lines; Preprocess the collected real-time status data, remove invalid data, and interpolate missing data to form a standardized data input set; Based on the task flow and resource flow relationships defined in the collaborative model, combined with the data input set of manufacturing resources, an intelligent configuration model of workshop manufacturing resources is established to describe the logical relationship between resource allocation and process arrangement; Use the intelligent configuration model of manufacturing resources to sort the processes in the task flow, and generate an initial scheduling plan according to the resource allocation logic and process priorities.

4. The intelligent manufacturing management method based on digital twin technology according to claim 1, wherein, The method of optimizing the initial scheduling plan using the genetic algorithm includes the following steps: Step 1: Encode the initial scheduling plan, represent the processing order of processes and machine allocation as chromosomes, each chromosome consists of multiple genes, and the genes represent the processing information of processes; Step 2: Generate an initial population based on scheduling rules combined with position swap operations. Each chromosome in the initial population represents a possible scheduling plan; Step 3: Calculate the fitness value of each chromosome in the initial population. The fitness function takes the makespan of the initial scheduling plan as the evaluation index, and the fitness value is inversely proportional to the makespan; Step 4: Based on the roulette wheel selection strategy, select chromosomes with high fitness according to the fitness values of the chromosomes to enter the next generation population; Step 5: Perform crossover operations on the selected chromosomes. Randomly select two chromosomes as parents, generate two offspring chromosomes through partially mapped crossover, and retain the excellent genes of the parents; Step 6: Perform mutation operations on the offspring chromosomes after crossover. Use the position swap method to randomly exchange the positions of two genes in the chromosome to generate new chromosomes; Step 7: Evaluate the fitness value of the newly generated chromosomes according to the fitness function, and retain the chromosomes with high fitness in the next generation population; Step 8: Repeat steps 4 to 7 until the predetermined number of iterations is reached or the fitness value converges, and output the chromosome with the highest fitness as the optimized scheduling plan; Step 9: Decode the optimized scheduling plan to determine the processing order of each process and the resource allocation of the corresponding machine, and generate an executable scheduling plan that meets the workshop resource constraints.

5. The intelligent manufacturing management method based on digital twin technology according to claim 1, wherein, The method of generating a rescheduling plan through the Monte Carlo tree search method based on an event-driven rescheduling strategy includes the following steps: Step 1: When a dynamic event occurs, monitor and collect the current system state of the workshop, including unfinished processes, the execution progress of the current task, and the available manufacturing resource status; Step 2: Initialize the Monte Carlo tree search model based on the current system state and the characteristics of the dynamic event, and use the system initial state as the root node of the search tree; Step 3: Starting from the root node, generate the child nodes of the root node according to the priorities of workshop tasks and the sequential constraints between processes. The child nodes represent the executable next processes; Step 4: Based on the upper confidence bound strategy, select the optimal child node from the child nodes of the current node to continue expanding the search tree; Step 5: When expanding to the leaf node, generate a complete path of process scheduling through random simulation and calculate the makespan corresponding to the path; Step 6: Propagate the simulation results of the leaf node back along the search path to the root node, and update the average return value and visit count of each node; Step 7: Repeat Steps 4 to 6 until the set number of simulation times or search depth is reached. Select the child node with the most visit times in the root node as the next operation for the current optimal process scheduling. Step 8: Based on the selected next process, set its corresponding node as the new root node, and repeat Steps 3 to 7 to generate a complete rescheduling plan.

6. The intelligent manufacturing management method based on digital twin technology according to claim 5, characterized in that, The calculation formula of the upper confidence bound strategy is as follows: Among them, represents the child node pointed to by the current root node, represents the child node 's average return, is a fixed parameter, represents the number of simulations performed by the child node and represents the number of simulations performed by the root node of the child node , represents the total return received by the child node .

7. The intelligent manufacturing management method based on digital twin technology according to claim 5, characterized in that, The dynamic events include machine failures, changes in workpiece processing times, order cancellations, and random workpiece arrivals. The generation of the rescheduling plan needs to update the current system state and redefine task priorities based on the specific type of dynamic event.

8. The intelligent manufacturing management method based on digital twin technology according to claim 1, wherein The method for identifying key nodes in combination with the fault propagation network model and optimizing the dynamic scheduling strategy includes the following steps: Select key parts, machines, and resources in the manufacturing system to establish a node set, where each node represents a part, device, or resource. Analyze the coupling relationships between the nodes in the manufacturing system. Based on whether there is resource sharing, dependence, or task collaboration between the nodes, construct a connection relationship matrix between the nodes, and the element values in the connection relationship matrix are assigned according to the coupling relationships. Generate a fault propagation network diagram according to the connection relationship matrix. The nodes of the network represent the resources or devices in the manufacturing system, and the edges of the network represent the coupling relationships between the nodes. Based on the fault propagation network diagram, calculate the degree, in-degree, and out-degree of each node, and determine the high-connectivity nodes with more connection relationships in the network. Based on the degree analysis results, calculate the clustering coefficient of the high-connectivity nodes to identify the nodes with stronger local relevance to the surrounding nodes. On the basis of identifying the clustering nodes, calculate the average path length and network diameter of the fault propagation network diagram, analyze the positions of the key nodes in the global propagation path, and identify the nodes that have an important impact on the global fault propagation. Combined with the statistical characteristic results of the fault propagation network diagram, preferentially adjust the scheduling priorities of the tasks related to the key nodes and reallocate the resources required by the key nodes. Using the optimized scheduling strategy, preferentially complete the tasks related to the key nodes when dynamic events occur, and generate an optimized dynamic scheduling plan.

9. An electronic device, characterized in that, It includes a processor and a memory. When the processor executes the computer program stored in the memory, it realizes the steps of the intelligent manufacturing management method based on the digital twin technology according to any one of Claims 1 to 8.

10. A computer-readable storage medium, on which a computer program is stored, characterized in that, When the computer program is run by the processor, it executes the steps of the intelligent manufacturing management method based on the digital twin technology according to any one of Claims 1 to 8.

Images