Simulation optimization method for digital twin factory

Abstract

The invention discloses a simulation optimization method for a digital twin factory. The simulation optimization method comprises the steps of obtaining workshop equipment data and performing classified storage; generating a task priority and resource configuration rule based on the basic operation data of the equipment; constructing a three-dimensional space model by using a 3D modeling tool, and creating a digital twinborn simulation factory model in combination with real-time production data; constructing a dynamic planning state space and a decision variable, and determining the resource use time, the resource allocation position and the specific quantity of each task; according to the resource allocation scheme and the task chain dependency relationship, evaluating a risk value of a task node in combination with historical equipment fault data; and based on the influence range report of the production fault chain, evaluating a task execution sequence and adjusting resource allocation and task time arrangement. According to the method, the influence of faults on a production plan and resource waste are reduced, the stability and adaptability of a factory system are remarkably enhanced, and the reliability and fine management level of overall production are improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing and industrial automation, and in particular to a simulation optimization method for a digital twin factory. Background Art

[0002] With the accelerated advancement of the transformation of manufacturing industry towards digitalization and intelligence, the task scheduling and resource management methods of traditional factories have gradually exposed their shortcomings in coping with the modern complex production environment. In order to improve the adaptability and production efficiency of factories in dynamic environments, the application of digital simulation technology has become the key. However, there are still many technical difficulties in the construction of digital simulation factories and the optimization of task scheduling. Most of the existing task scheduling systems are based on the task allocation rules of a single process or local equipment, and cannot form a comprehensive perception of the entire factory production environment. For example, the traditional scheduling system cannot effectively integrate the physical information such as the building structure, equipment layout, and logistics routes in the factory, resulting in a lack of spatial global view support for scheduling optimization. In this context, building a digital simulation factory has become an effective way to solve this problem. However, many factories currently still lack the technical means from physical factories to digital twin factories, making it difficult to achieve global simulation and optimization of production tasks. The task chain in the modern manufacturing environment often has complex dependencies, including time dependency, resource dependency, and data dependency. It is difficult for traditional task scheduling methods to fully consider these multi-dimensional dependencies. For example, when a key device fails, its impact may be transmitted to downstream tasks through the task chain, resulting in a significant decrease in production efficiency. However, the current scheduling system lacks the ability to digitally model these complex dependencies and cannot accurately simulate the operating logic and fault propagation of the task chain. At the same time, equipment failure is a common problem in factory production. Its impact is not limited to a single device, but also propagates along the task chain, causing multiple task nodes to be affected. Traditional scheduling systems are usually unable to predict the cascading effect of failures on the task chain, and it is difficult to quantify the risk value of task nodes and the cumulative risk of equipment. Therefore, how to use digital twin factories to achieve global monitoring and dynamic scheduling of complex production tasks, optimize resource allocation in combination with task chain dependencies, and accurately assess the impact range of fault propagation risks and the risk value of task nodes are still important issues that need to be studied and solved in depth in the current technical field. Summary of the invention

[0003] In order to solve the problems existing in the above-mentioned prior art, the present invention provides a simulation optimization method for a digital twin factory, which mainly includes:

[0004] The industrial IoT sensors in the factory workshop are used to obtain workshop equipment data, and the completion time of production tasks and resource consumption are timestamped and stored in categories;

[0005] Based on the basic operation data of the equipment, frequent item sets are extracted and a table of key production process nodes is generated. Combined with the equipment operation parameters and resource consumption patterns, task priorities and resource allocation rules are determined, and a task allocation parameter table containing task priorities and resource allocation rules is generated.

[0006] Obtain physical factory architectural drawings, equipment layouts, and logistics route maps, use 3D modeling tools to build a three-dimensional space model, and combine real-time production data to create a digital twin simulation factory model;

[0007] According to the production task data in the digital twin simulation factory model, combined with the task priority rules, the task priority list and resource demand information are obtained, the dynamic programming state space and decision variables are constructed, and the resource usage time, resource allocation location and specific quantity of each task are determined;

[0008] According to the resource allocation scheme and task chain dependency, the dependency and resource allocation information of task nodes are extracted, a Bayesian network including task and equipment dependency is constructed, the impact range of fault propagation is simulated, and the risk value of task nodes is evaluated in combination with historical equipment failure data;

[0009] Based on the impact report of the production failure chain, extract the priority, resource requirements and completion time of the affected tasks, re-evaluate the task execution sequence and adjust resource allocation and task schedule.

[0010] Furthermore, the acquisition of workshop equipment data through industrial Internet of Things sensors in the factory workshop, time stamping and classified storage of production task completion time and resource consumption includes:

[0011] By installing industrial Internet of Things sensors on production equipment, logistics routes, quality inspection areas and resource consumption monitoring points in the factory workshop, the location information, cargo parameters, shelf parameters, equipment operating status, material logistics routes, resource consumption, production task completion time and quality inspection data of the equipment are collected in real time. The collected raw data are timestamped and classified for storage to build a workshop equipment data set. Cargo parameters include volume and weight, shelf parameters include volume and maximum load-bearing capacity, equipment operating status includes speed, temperature, vibration, resource consumption includes electricity and raw materials, and quality inspection data includes product qualification rate and defect rate. The basic operating data and process flow parameters of the equipment are obtained through the storage database, and the basic operating data are normalized to obtain the basic operating data table of the equipment. The basic operating data includes equipment model, operating frequency, energy consumption rate, and maintenance records. The process flow parameters include the processing time of each task, material usage and process temperature and pressure. The material status data includes the inventory, processing status and location distribution of the material.

[0012] Furthermore, based on the basic operation data of the equipment, frequent item sets are extracted and a key production process node table is generated. The task priority and resource allocation rules are determined in combination with the equipment operation parameters and resource consumption patterns, and a task allocation parameter table containing the task priority and resource allocation rules is generated, including:

[0013] Based on the basic operation data table of the equipment, support and confidence thresholds are set, and the Apriori algorithm is used to decompose the basic operation data of the equipment into frequent item sets, and the mining results are sorted. According to the priority of support and confidence, a key production process node table is obtained, including the interactive position of the equipment, logistics nodes and key production processes; according to the equipment operation parameters, including operation frequency, usage time, energy consumption and maintenance cycle, the K-Means clustering algorithm is used to calculate the similarity between devices based on the Euclidean distance, and the equipment is clustered and grouped; according to the clustering results, the resource consumption of each group of equipment is determined, and the task load of the equipment is calculated through the equipment usage time and task distribution data to determine the equipment usage pattern; according to the task completion time and task importance data, the weighted scoring method is used to calculate the task priority score, generate a task priority table, and combine the resource consumption pattern to determine the resource allocation rules required for each task, including resource allocation time, quantity and location, and generate a task allocation parameter table containing task priority and resource allocation rules.

[0014] Furthermore, obtaining the physical factory building drawings, equipment layout diagrams and logistics route maps, using 3D modeling tools to build a three-dimensional space model, and combining real-time production data to create a digital twin simulation factory model includes:

[0015] Obtain the architectural drawings, equipment layout and logistics route map of the physical factory, and use the 3D modeling tool Unity to build a three-dimensional spatial model of the factory, including the workshop structure, equipment distribution, product library, raw material library and logistics channel; embed the floor space, working area and connection location of all equipment, including power connection, data interface, material entry and exit points into the model, and generate a geometric model of each device; collect real-time production data from the production line, input Simulink and connect with the three-dimensional spatial model, simulate the execution status of production tasks, and display the operating status of each device in real time. The real-time production data includes equipment status, production task progress and equipment The interaction between equipment and devices, equipment status includes on or off, running speed, and energy consumption; production task progress includes task start time, task completion time, and processing progress; the digital twin platform GEPredix is ​​used to combine the three-dimensional space model with the real-time data access interface to create a digital twin simulation factory model; real-time production data is used to build a dynamic process simulation model, simulate the execution process of production tasks in the factory, set production environment parameters, and calculate task scheduling and resource allocation in real time through simulation to obtain dynamic process simulation results including task completion time and equipment utilization; production environment parameters include equipment failure, production task priority, and resource constraints.

[0016] Furthermore, according to the production task data in the digital twin simulation factory model, combined with the task priority rules, the task priority list and resource demand information are obtained, the dynamic programming state space and decision variables are constructed, and the resource usage time, resource allocation location and specific quantity of each task are determined, including:

[0017] Through the production task data in the digital twin simulation factory model, the completion time, required resource type and quantity of each task are extracted; according to the characteristics of the task, the priority rules are defined, where tasks with shorter completion time or higher resource requirements have higher priority; according to the execution requirements of the task, the resource requirement information of each task is obtained, and the completion time and resource requirement information of each task are arranged in order of priority to obtain a task priority list. The resource requirement information includes equipment, workstations, and raw materials; according to the task priority list, the state space and decision variables of the dynamic programming model are determined. The state space includes the resource allocation at each moment, and the decision variable is the resource allocation decision of the task; the scheduling efficiency is evaluated by calculating the resource usage time, total resource consumption and task completion time of each task, and the resource scheduling efficiency is set as the objective function. According to the dynamic programming model, a bottom-up dynamic programming method is used to gradually deduce from the initial state of the task to the final state of the task, and the optimal configuration of resources at each moment is calculated. In the calculation process, the resource usage efficiency of each task at different time points is evaluated to determine whether the resource constraints are met, and the resource allocation plan for each task is obtained. The resource allocation plan is determined, including the resource usage time, location and quantity of each task.

[0018] Furthermore, according to the resource allocation scheme and task chain dependency, the dependency and resource allocation information of the task nodes are extracted, a Bayesian network including the task and device dependency is constructed, the impact range of fault propagation is simulated, and the risk value of the task nodes is evaluated in combination with historical equipment failure data, including:

[0019] According to the resource allocation plan, obtain each task node, resource allocation location, time and equipment information, and identify the sequential dependencies between tasks; based on each task node in the task chain and its dependencies, construct a Bayesian network structure, set the task node as a random variable, and the state of each device as its parent node, to obtain the structure of the Bayesian network, in which the node represents the task and the device, and the edge represents the dependency between the task and the device; obtain historical equipment failure data, task completion time and equipment failure records, and use the maximum likelihood estimation method to set a conditional probability table for each node to represent the conditional probability of equipment failure or task failure; for each node in the task chain, obtain the probability of the task being affected, and incorporate it into the calculation of the Bayesian network to determine the conditional probability of each task and device state. The affected tasks include delayed task completion and task failure; if a device or task node fails, use Bayesian The network performs forward propagation to simulate the process of fault extension from the node to the dependent nodes; the probability of subsequent task delay or failure caused by a single fault is calculated for each task node, and the degree of impact of each subsequent task node is obtained; the impact of fault propagation in the task chain is calculated through the Bayesian network, the fault propagation probability of each task node is obtained, and the impact range of the fault propagating from the initial node to the task chain is determined; according to the fault propagation probability of each task node, the cumulative probability of each task node in the task chain being affected by equipment failure is evaluated, and the risk value of each task node and the cumulative risk value of shared equipment are obtained. All task nodes in the task chain are ranked by risk value to determine the task nodes most seriously affected by fault propagation; based on the fault propagation analysis results, the risk values ​​of task nodes, the cumulative risk values ​​of shared equipment and the impact range of fault propagation are integrated to generate an impact range report for the production fault chain.

[0020] It also includes, based on the fault propagation probability of each task node, evaluating the cumulative probability of each task node in the task chain being affected by the device failure, and obtaining the risk value of each task node and the cumulative risk value of the shared device, specifically including:

[0021] Obtain historical dependency data between tasks, including resource sharing, time delay, and data interaction. Use recurrent neural networks for model training, build a task dependency weight prediction model, and predict the task dependency weights between nodes. Obtain the dependency relationship and initial failure probability P of each task node in the task chain. 0 , using the fault propagation impact formula of the task node Calculate the fault propagation probability P of node i at time t+1 i (t+1), where α is the fault self-recovery attenuation factor of the node itself, β is the fault propagation amplification factor, which indicates the degree of impact of external faults on the node, obtained by fitting historical data, N(i) represents the set of all upstream dependent tasks of node i, and γ ijis the task dependency weight between nodes i and j, indicating the influence of task j on i. For each node in the task chain, starting from time t = 0, the fault propagation influence formula is used for multiple iterations to calculate the fault propagation probability P of the node. i (t), record the fault propagation increment ΔP of the node in each iteration i (t) = P i (t+1)-P i (t), and determine whether a stable state is reached based on a preset increment threshold; if Then the iteration is stopped and the final fault propagation probability of each node in the stable state is obtained, where ε is the iterative convergence threshold; according to the weighted value of the fault propagation probability and the importance of the equipment, the task node risk value evaluation formula R is used i =P i (final) W i , calculate the risk value of the device sharing node in the task chain, where W i is the contribution weight of node i to the production chain, indicating the importance of the task to the overall production chain, which can be calculated by the historical task completion rate or resource consumption ratio. i (final) represents the final fault propagation probability of the task being affected by device failure or other task failure after the task node i reaches a stable state in the fault propagation model; for the node set G(e) sharing the device, the cumulative value of the device risk is calculated Determine the risk value R of each node i and the cumulative risk value R of shared equipment e .

[0022] Furthermore, the report on the impact range of the production fault chain is used to extract the priority, resource requirements and completion time of the affected tasks, re-evaluate the task execution sequence and adjust resource allocation and task schedule, including:

[0023] Based on the impact range report of the production fault chain, the priority, resource requirements and task completion time of each affected task are extracted. Based on the priority and resource requirements of the tasks, the execution order of the tasks is re-evaluated to obtain the task execution sequence table; based on the re-determined task execution order, the task execution order is optimized using a genetic algorithm, and the optimization objective function is set to minimize the fault recovery time, maximize the resource utilization rate and minimize the task completion time. The tasks are dynamically adjusted and rearranged to obtain the optimized production task execution sequence table, including the re-execution order of the tasks, resource allocation and task completion time; the optimized task execution sequence table is input into the multi-task dynamic scheduling simulation environment to simulate the environment of multi-task concurrent execution, evaluate the resource competition, equipment usage and scheduling conflicts during the task execution process, and verify the effectiveness of the optimized task execution sequence table based on the task completion time, resource utilization rate and equipment idle time; if the effectiveness is lower than expected, the task execution sequence is adjusted again until the execution time, resource allocation and priority of all task nodes are met.

[0024] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0025] The present invention provides a simulation optimization method for a digital twin factory. The present invention collects real-time data through industrial Internet of Things sensors, and combines factory building drawings, equipment layout diagrams and logistics route maps to generate a three-dimensional digital twin model, thereby realizing accurate visualization of the factory production environment and providing comprehensive global view support for scheduling optimization. Through frequent item set mining and task priority rules, the task priority can be dynamically adjusted, and resource allocation can be optimized in combination with a dynamic programming model, which significantly improves resource utilization efficiency and the rationality of task scheduling. The present invention can also accurately evaluate the risk value of task nodes and the cumulative risk of equipment by analyzing the dependencies and resource allocation status in the task chain, and effectively predict the cascading impact of equipment failures on the task chain, thereby realizing quantitative analysis of the impact of task chain failure propagation, and providing support for scheduling optimization and risk avoidance. The present invention can flexibly adjust the task execution order and resource allocation scheme in a dynamic production environment, optimize production plans, reduce equipment idleness and resource conflicts, and improve production efficiency. This invention comprehensively uses digital twins, industrial Internet of Things and risk assessment technology to significantly improve the accuracy and flexibility of task scheduling, enhance the adaptability and production efficiency of the production system, improve the overall production reliability and refined management level, and provide strong guarantees for the efficient operation of modern smart factories. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of a simulation optimization method of a digital twin factory of the present invention;

[0027] Figure 2It is a schematic diagram of a simulation optimization method of a digital twin factory of the present invention. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0029] like Figure 1-2 In this embodiment, a simulation optimization method of a digital twin factory may specifically include:

[0030] Step S101, obtain workshop equipment data through industrial Internet of Things sensors in the factory workshop, and timestamp and classify the production task completion time and resource consumption.

[0031] Through the industrial Internet of Things sensors installed in the production equipment, logistics routes, quality inspection areas and resource consumption monitoring points in the factory workshop, the location information, cargo parameters, rack parameters, equipment operation status, material logistics routes, resource consumption, production task completion time and quality inspection data of the equipment are collected in real time. The collected raw data are timestamped and classified for storage to build a workshop equipment data set. The cargo parameters include volume and weight, the rack parameters include volume and maximum load-bearing capacity, the equipment operation status includes speed, temperature, vibration, resource consumption includes electricity and raw materials, and quality inspection data includes product qualification rate and defect rate. The basic operation data and process flow parameters of the equipment are obtained through the storage database, and the basic operation data are normalized to obtain the basic operation data table of the equipment. The basic operation data includes equipment model, operation frequency, energy consumption rate, maintenance record, process flow parameters include processing time, material usage and process temperature and pressure of each task, and material status data includes material inventory, processing status and location distribution.

[0032] For example, in a factory workshop, industrial IoT sensors are installed to collect real-time data related to equipment and production. For example, on processing center A, the location information of the equipment is collected as workshop 1 area B, and the equipment operation status includes a speed of 1200 revolutions per minute, a temperature of 75 degrees Celsius, and a vibration amplitude of 0.02 mm. At the same time, the logistics line sensor shows that the equipment is processing a product numbered P001. The material logistics route shows that the material starts from warehouse W1, passes through logistics nodes L2 and L3, and finally arrives at processing center A. After processing, it is sent to quality inspection area C2. The resource consumption monitoring point records that the equipment consumes 2.5 kWh of electricity and 1.2 kg of raw materials during processing. The sensors in the quality inspection area show that the qualified rate of this batch of products P001 is 95% and the defect rate is 5%. In addition, the cargo parameters and rack parameters of the goods in transportation are collected. The cargo parameters include a volume of 0.5 cubic meters and a weight of 80 kilograms, and the rack parameters include a volume of 1 cubic meter and a maximum load of 100 kilograms. On this basis, all the collected data are time-stamped, such as the processing start time is 2020-10-16-10:15:00, the completion time is 2020-10-16-10:35:00, and the data are classified and stored in the workshop equipment data set. The basic operation data of the equipment is obtained by storing the database, and the model of machining center A is MCX-500, the operating frequency is 50 Hz, the energy consumption rate is 1.8 kWh / hour, and the last maintenance record is 2020-07-15. Normalized according to the highest standard of 2.5 kWh / hour, after normalization, the energy consumption rate in the basic operation data table is normalized to 0.72, and the process parameters show that the processing time of task P001 is 20 minutes, the material usage is 1.2 kg, the processing temperature is 75 degrees Celsius, and the pressure is 50 MPa. The material status data further shows that the current inventory is 500 pieces, of which 50 pieces are in processing and 450 pieces are in warehouse W1. The location distribution shows that the number of materials on the logistics route is 10 pieces.

[0033] Step S102, based on the basic equipment operation data, extract frequent item sets and generate a key production process node table, combine the equipment operation parameters and resource consumption patterns, determine the task priority and resource allocation rules, and generate a task allocation parameter table containing task priority and resource allocation rules.

[0034] Based on the basic operation data table of the equipment, support and confidence thresholds are set, and the Apriori algorithm is used to decompose the basic operation data of the equipment into frequent item sets. The mining results are sorted, and the key production process node table is obtained according to the priority of support and confidence, including the interactive position of the equipment, logistics nodes and key production processes. According to the equipment operation parameters, including operation frequency, usage time, energy consumption and maintenance cycle, the K-Means clustering algorithm is used to calculate the similarity between devices based on the Euclidean distance, and the equipment is clustered and grouped. According to the clustering results, the resource consumption of each group of equipment is determined, and the task load of the equipment is calculated through the equipment usage time and task distribution data to determine the equipment usage mode. According to the task completion time and task importance data, the task priority score is calculated using the weighted scoring method to generate a task priority table. Combined with the resource consumption mode, the resource configuration rules required for each task are determined, including the resource allocation time, quantity and location, and a task allocation parameter table containing task priority and resource configuration rules is generated.

[0035] For example, according to the basic operation data table of the equipment, the support threshold is set to 40%, the confidence threshold is set to 65%, and the basic operation data is analyzed using the Apriori algorithm. The decomposed frequent item set contains the combination of equipment model ZX-750, operating frequency of 60 Hz, and energy consumption rate of 2.2 kWh / h, with a support of 45% and a confidence of 70%. After sorting, according to the priority of support and confidence, a key production process node table is obtained, in which the main interaction position of the equipment ZX-750 is area D, the key logistics node is M3-M4, and the key production process is steel stamping task R101. Next, according to the equipment operation parameters, including operating frequency, usage time, energy consumption and maintenance cycle, the K-Means clustering algorithm is used to group the equipment, and the similarity is measured by Euclidean distance. It is obtained that the ZX-750 equipment and another equipment DF-450 are clustered into a group. The average operating frequency of this group of equipment is 55 Hz, the average energy consumption rate is 2.3 kWh / h, and the maintenance cycle is 50 days. Another group of equipment, including LK-200 and TY-310, has an average operating frequency of 35 Hz, an energy consumption rate of 1.6 kWh / h, and a maintenance cycle of 80 days. Based on the clustering results, the resource consumption of the equipment is further calculated, and the total energy consumption of the group where ZX-750 is located is 450 kWh / day, and the total usage time is 90 hours / day. The task distribution data shows that each device in this group of equipment undertakes 12 tasks per day on average, of which the task load of ZX-750 equipment is 18 tasks, which is an ultra-high load equipment, and its usage mode is determined to be intermittent high-intensity operation. According to the task completion time and task importance, the task priority is calculated using the weighted scoring method. Among them, the completion time of stamping task R101 is 25 minutes, the importance weight is 0.9, and the calculated priority score is 22.5. The completion time of welding task R102 is 15 minutes, the importance weight is 0.7, and the priority score is 10.5. The generated task priority table shows that R101 takes precedence over R102. Combined with the resource consumption pattern, the resource configuration rules assigned to task R101 are ZX-750 equipment, running time 09:00-11:30, raw material location warehouse Q2, and resource quantity 20. The resource configuration rules assigned to task R102 are DF-450 equipment, running time 13:00-14:00, raw material location warehouse Q3, and resource quantity 12. The task allocation parameter table finally generated contains the priorities and resource configuration rules of tasks R101 and R102, which are used to guide task scheduling and resource allocation.

[0036] Step S103, obtain the physical factory building drawings, equipment layout drawings and logistics route maps, use 3D modeling tools to build a three-dimensional space model, and create a digital twin simulation factory model in combination with real-time production data.

[0037] Obtain the architectural drawings, equipment layout diagrams, and logistics route maps of the physical factory, and use the 3D modeling tool Unity to build a three-dimensional spatial model of the factory, including the workshop structure, equipment distribution, product library, raw material library, and logistics channel. Embed the floor space, work area, and connection locations of all equipment, including power connections, data interfaces, and material entry and exit points, into the model to generate a geometric model of each device. Collect real-time production data from the production line, input Simulink to connect with the three-dimensional spatial model, simulate the execution status of production tasks, and display the operating status of each device in real time. Real-time production data includes equipment status, production task progress, and interactions between equipment. Equipment status includes on or off, operating speed, and energy consumption. The progress of production tasks includes task start time, task completion time, and processing progress. Use the digital twin platform GEPredix to combine the three-dimensional spatial model with the real-time data access interface to create a digital twin simulation factory model. Use real-time production data to build a dynamic process simulation model, simulate the execution process of production tasks in the factory, set production environment parameters, and calculate task scheduling and resource allocation in real time through simulation to obtain dynamic process simulation results including task completion time and equipment utilization. Production environment parameters include equipment failure, production task priority, and resource constraints.

[0038] For example, by obtaining architectural drawings, equipment layout diagrams and logistics route maps, a 3D spatial model including workshop structure, equipment distribution and logistics channels was constructed using the 3D modeling tool Unity. The factory contains two main workshops, of which workshop 1 has a size of 50 meters × 30 meters and a height of 10 meters, and workshop 2 has a size of 40 meters × 20 meters and a height of 8 meters. The equipment layout diagram shows that workshop 1 is equipped with 10 equipment, including a ZX-750 stamping machine, which covers an area of ​​4 square meters and a working area of ​​6 square meters. The material entry and exit points are located on the north side of the equipment, the power connection is located on the east side of the equipment, and the data interface is located on the west side of the equipment. Workshop 2 is equipped with 8 equipment, including a DF-450 welding machine, which covers an area of ​​3 square meters and a working area of ​​5 square meters. The material entry and exit points are located on the south side of the equipment, the power connection is located on the west side of the equipment, and the data interface is located on the north side of the equipment. Special product warehouse and raw material warehouse areas are also set up, which are used to store finished products and raw materials required for production respectively. The product warehouse is located in the northeast corner of workshop 1, with an area of ​​20 square meters, while the raw material warehouse is set in the southwest corner of workshop 2, with an area of ​​15 square meters. After embedding this information into the three-dimensional space model, the geometric model of each device is generated to accurately represent its footprint and working area. Real-time production data is collected from the production line, input into Simulink, docked with the three-dimensional space model, simulated the execution status of the production task, and displayed the equipment operation status in real time. Real-time data shows that the punching machine ZX-750 is currently in operation, with an operating speed of 1200 times / hour and an energy consumption of 2.5 kWh, and the welding machine DF-450 is currently in standby state with an energy consumption of 0.5 kWh. In addition, the real-time progress of production task P301 shows that its start time is 2025-10-16-09:00:00, the current processing progress is 75%, the estimated completion time is 2020-10-16-09:30:00, and task P301 will be automatically transferred to the welding machine DF-450 after completion. Through the digital twin platform GEPredix, the three-dimensional space model is combined with the real-time data access interface to create a digital twin simulation factory model, which displays the operating status of all equipment in the workshop, the progress of production tasks and the interaction between equipment in real time and synchronously. Based on real-time production data, a dynamic process simulation model is constructed to simulate the execution process of tasks in the factory, such as setting production environment parameters in the simulation environment, including the failure probability of the punching machine ZX-750 is 0.02%, the priority task weight of the welding machine DF-450 is 1.5, and the resource limit is that only 50 pieces of raw materials are currently available. Through simulation calculation, the actual completion time of task P301 is 202-10-16-09:35:00, and the completion time of task P302 is 2020-10-16-10:00:00. The dynamic process simulation results also show that the utilization rate of the punching machine ZX-750 is 90%, and the utilization rate of the welding machine DF-450 is 60%.After building the digital twin simulation factory model, users can independently adjust various parameters of the model, including but not limited to process sequence, equipment configuration, task priority, resource allocation, etc.

[0039] Step S104, based on the production task data in the digital twin simulation factory model and combined with the task priority rules, obtain the task priority list and resource demand information, construct the dynamic programming state space and decision variables, and determine the resource usage time, resource allocation location and specific quantity of each task.

[0040] Through the production task data in the digital twin simulation factory model, the completion time, required resource type and quantity of each task are extracted. According to the characteristics of the task, the priority rules are defined, where tasks with shorter completion time or higher resource requirements have higher priority. According to the execution requirements of the task, the resource requirement information of each task is obtained, and the completion time and resource requirement information of each task are arranged in priority order to obtain the task priority list. The resource requirement information includes equipment, workstations, and raw materials. According to the task priority list, the state space and decision variables of the dynamic programming model are determined. The state space includes the resource allocation at each moment, and the decision variable is the resource allocation decision of the task. The scheduling efficiency is evaluated by calculating the resource usage time, total resource consumption and task completion time of each task. The resource scheduling efficiency is set as the objective function. According to the dynamic programming model, a bottom-up dynamic programming method is used to gradually deduce from the initial state of the task to the final state of the task, and the optimal configuration of resources at each moment is calculated. In the calculation process, the resource usage efficiency of each task at different time points is evaluated to determine whether the resource constraints are met, and the resource allocation plan for each task is obtained. The resource allocation plan is determined, including the resource usage time, location and quantity of each task.

[0041] For example, in a digital twin simulation factory, production task data is extracted, where the completion time of task Q1 is 3 hours, and the required resources include 1 punching machine with equipment model XT-800, workstation X3, and 8 raw material steel plates. The completion time of task Q2 is 2 hours, and the required resources include 1 welding machine with equipment model RW-450, workstation X1, and 12 raw material welding rods. The completion time of task Q3 is 4 hours, and the required resources include 1 cutting machine with equipment model CT-200, workstation X2, and 5 raw material aluminum plates. According to the characteristics of the tasks, priority rules are defined, in which tasks with shorter task completion time or higher resource requirements have higher priority. According to the calculation, task Q2 has the highest priority due to its short completion time and moderate resource requirements, followed by task Q1 with higher resource requirements, and finally task Q3. According to these rules, a task priority list is generated: task Q2>task Q1>task Q3. According to the task priority list, the state space and decision variables of the dynamic programming model are constructed, where the state space includes the resource allocation at each moment, and the decision variables are the resource allocation decisions at each moment. For example, task Q2 is preferentially allocated to RW-450 welding machine, station X1 and 12 welding rods, and is planned to be completed within 1-2 hours. Task Q1 is then allocated to XT-800 punching machine, station X3 and 8 steel plates, and is planned to be completed within 2-5 hours. Task Q3 is finally allocated to CT-200 cutting machine, station X2 and 5 aluminum plates, and is planned to be completed within 5-9 hours. In order to evaluate the scheduling efficiency, the resource scheduling efficiency is set as the objective function, and the goal is to minimize the ratio of total resource consumption to task completion time, where resource scheduling efficiency = total resource consumption during total task completion time / total task completion time. Through the dynamic programming model, the bottom-up dynamic programming method is used to gradually deduce from the initial state of the 0th hour to the final state of the 9th hour, calculate the resource allocation at each moment, and evaluate the resource utilization efficiency of each task at different time points. The efficiency of task Q2 is 2 hours / 12 welding rods = 0.1667, the efficiency of task Q1 is 3 hours / 8 steel plates = 0.375, and the efficiency of task Q3 is 4 hours / 5 aluminum plates = 0.8. Therefore, the resource allocation plan is obtained as follows: the resource usage time of task Q2 is the 1st to 2nd hour, the position is station X1, using RW-450 welding machine and 12 welding rods, the resource usage time of task Q1 is the 2nd to 5th hour, the position is station X3, using XT-800 punching machine and 8 steel plates, and the resource usage time of task Q3 is the 5th to 9th hour, the position is station X2, using CT-200 cutting machine and 5 aluminum plates.

[0042] Step S105, according to the resource allocation scheme and task chain dependency, extract the dependency and resource allocation information of the task nodes, build a Bayesian network including the task and equipment dependency, simulate the impact range of fault propagation, and evaluate the risk value of the task nodes in combination with historical equipment failure data.

[0043] According to the resource allocation plan, obtain each task node, resource allocation location, time and equipment information, and identify the sequential dependencies between tasks. Based on each task node in the task chain and its dependencies, construct a Bayesian network structure, set the task node as a random variable, and the state of each device as its parent node to obtain the structure of the Bayesian network, where the node represents the task and the device, and the edge represents the dependency between the task and the device. Obtain historical equipment failure data, task completion time and equipment failure records, and use the maximum likelihood estimation method to set a conditional probability table for each node to represent the conditional probability of equipment failure or task failure. For each node in the task chain, obtain the probability of the task being affected, and incorporate it into the calculation of the Bayesian network to determine the conditional probability of each task and device state. The affected tasks include delayed task completion and task failure. If a device or task node fails, use the Bayesian network for forward propagation to simulate the process of the failure extending from the node to the dependent nodes. Calculate the probability of subsequent task delays or failures caused by a single failure for each task node, and obtain the degree of impact of each subsequent task node. The impact of fault propagation in the task chain is calculated through the Bayesian network, the fault propagation probability of each task node is obtained, and the impact range of the fault propagating from the initial node to the task chain is determined. According to the fault propagation probability of each task node, the cumulative probability of each task node in the task chain being affected by equipment failure is evaluated, and the risk value of each task node and the cumulative risk value of shared equipment are obtained. All task nodes in the task chain are ranked by risk value to determine the task nodes most seriously affected by fault propagation. Based on the fault propagation analysis results, the risk value of the task node, the cumulative risk value of the shared equipment, and the impact range of fault propagation are integrated to generate an impact range report for the production fault chain.

[0044] For example, in a production process of a factory, according to the resource allocation plan, the resource allocation location, time and equipment information of the task node are obtained, including task T1 assigned to the equipment ZX-800 in workshop A, the task time is 9:00-10:30, and the location is station B1, task T2 assigned to the equipment DF-300 in workshop B, the task time is 10:30-12:00, and the location is station C2, and task T3 assigned to the equipment RT-200 in workshop C, the task time is 12:00-13:30, and the location is station D3. The sequential dependency relationship between tasks is that T1 needs to be completed before T2 can be started, and T2 needs to be completed before T3 can be started. Based on these task nodes and their dependencies, a Bayesian network structure is constructed, in which task nodes T1, T2, and T3 are random variables, equipment ZX-800, DF-300, and RT-200 are their parent nodes, and edges represent the dependencies between tasks and equipment, such as task T1 directly depends on the running status of equipment ZX-800, and task T2 depends on the status of DF-300. From the historical data of the factory, the fault records of the equipment and the completion time of the task are obtained. For example, the historical data of equipment ZX-800 shows that its average failure probability is 5%, while the historical data of task T1 shows that when the equipment fails, the probability of task failure is 80% and the probability of delay is 50%. The conditional probability table for each node is set using the maximum likelihood estimation method, and the conditional probability table for task T1 is obtained as follows: if the equipment ZX-800 operates normally, the probability of successful completion of task T1 is 95%, and if the equipment ZX-800 fails, the probability of failure of task T1 is 80%. Similarly, the conditional probabilities of tasks T2 and T3 are generated based on the historical data of DF-300 and RT-200, respectively. Through the Bayesian network, the probability of each node in the task chain being affected by the equipment failure is calculated. If the device ZX-800 fails at 9:30, the Bayesian network performs forward propagation, simulating the failure to extend from ZX-800 to T1, and the failure probability of task T1 is calculated to be 80% and the delay probability is 50%. Since the delay of T1 directly affects the startup time of T2, the delay probability of T2 is calculated to be 40%, and the probability of failure of T2 is 30%. Similarly, this impact is propagated to T3 through T2, and the delay probability of T3 is calculated to be 20% and the failure probability is 10%. By calculating the impact of fault propagation, the fault propagation probability of each node in the task chain is obtained. For example, the fault propagation probability of task T1 is 80%, that of task T2 is 40%, and that of task T3 is 20%. Based on these probabilities, the risk value of each task node is further calculated. The risk value formula is risk value = fault propagation probability × task importance weight. If the importance weight of task T1 is 0.8, T2 is 0.6, and T3 is 0.5, then the risk value of T1 is 64%, T2 is 24%, and T3 is 10%.For the risk accumulation value of the shared device ZX-800, the sum of the risk values ​​of all task nodes that depend on the device is calculated, and the risk accumulation value of ZX-800 is 64%. According to the risk value sorting, the task node most seriously affected by fault propagation is determined to be T1, followed by T2 and T3. Based on the fault propagation analysis results, the risk value of the task node, the risk accumulation value of the device, and the impact range of fault propagation are integrated to generate an impact range report of the production fault chain. The report lists in detail the propagation path when the fault occurs at 9:30, from ZX-800 to T1, and then to T2 and T3, and indicates the delay probability, failure probability and risk value of each node.

[0045] Among them, according to the fault propagation probability of each task node, the cumulative probability of each task node in the task chain being affected by equipment failure is evaluated, and the risk value of each task node and the cumulative risk value of the shared equipment are obtained.

[0046] Obtain historical dependency data between tasks, including resource sharing, time delay, and data interaction, use recurrent neural network to train the model, build a task dependency weight prediction model, and predict the task dependency weights between nodes. Obtain the dependency relationship and initial failure probability P of each task node in the task chain 0 , using the fault propagation impact formula of the task node Calculate the fault propagation probability P of node i at time t+1 i (t+1), where α is the fault self-recovery attenuation factor of the node itself, β is the fault propagation amplification factor, which indicates the degree of impact of external faults on the node, obtained by fitting historical data, N(i) represents the set of all upstream dependent tasks of node i, and γ ij is the task dependency weight between nodes i and j, indicating the impact strength of task j on i. For each node in the task chain, starting from time t = 0, the fault propagation impact formula is used for multiple iterations to calculate the fault propagation probability P of the node. i (t), record the fault propagation increment ΔP of the node in each iteration i (t) = P i (t+1)-P i (t), and judge whether a stable state is reached based on the preset increment threshold. Then the iteration is stopped and the final fault propagation probability of each node in the stable state is obtained, where ε is the iterative convergence threshold. According to the weighted value of the fault propagation probability and the importance of the equipment, the task node risk value evaluation formula R is used i =P i (final) W i , calculate the risk value of the device sharing node in the task chain, where W iis the contribution weight of node i to the production chain, indicating the importance of the task to the overall production chain, which can be calculated by the historical task completion rate or resource consumption ratio. i (final) represents the final fault propagation probability of the task being affected by device failure or other task failure after the task node i reaches a stable state in the fault propagation model. For the node set G(e) sharing the device, the cumulative value of the device risk is calculated Determine the risk value R of each node i and the cumulative risk value R of shared equipment e .

[0047] For example, in a production task chain of a factory. Tasks T1, T2, and T3 have a dependency relationship, where task T1 is the initial task, and task T2 is triggered after completion, and task T2 is triggered after completion. Task T1 depends on device A, and tasks T2 and T3 share device B. Obtain historical dependency data between tasks, including resource sharing status, time delay, and data interaction. The resource sharing status is that tasks T1 and T2 use device A, T1 occupies 40% of the running time of device A, and T2 occupies 60% of the running time of device A. Tasks T2 and T3 share device B, and T2's running time on device B accounts for 50%, and T3 accounts for 50%. The time delay is that the completion time of task T1 has a direct impact on the start time of T2. Historical data shows that task T1 is delayed by 1 hour, and task T2 is delayed by an average of 0.8 hours. Similarly, task T2 is delayed by 1 hour, and task T3 is delayed by an average of 0.6 hours. The data interaction is that the output data volume of task T1 is 70% of the total input data volume of task T2. The output data volume of task T2 is 80% of the total input data volume of task T3. Using a recurrent neural network, we build a task dependency weight prediction model and predict that the dependency weight of task T2 on task T1 is 0.6, and the dependency weight of task T3 on task T2 is 0.8. Based on historical data, we get the initial failure probability of each task node, where the initial failure probability of task T1 is 10%, task T2 is 8%, and task T3 is 5%. Using the failure propagation impact formula of the task node Calculate the fault propagation probability P of node i at time t+1 i (t+1), where α is the fault self-recovery attenuation factor of the node itself, β is the fault propagation amplification factor, which indicates the degree of impact of external faults on the node, obtained by fitting historical data, N(i) represents the set of all upstream dependent tasks of node i, and γ ij is the task dependency weight between nodes i and j, indicating the influence of task j on i. If α = 0.9, β = 0.7, starting from time t = 0, the fault propagation influence formula is used for multiple iterations to calculate the fault propagation probability of each task node. In the first iteration, the fault propagation probability P of task T2 is calculated based on the initial fault probability and dependency relationship.2 (1) = 0.108, P 3 (1) = 0.0866. Record the fault propagation increment ΔP of each iteration i (t) = P i (t+1)-P i (t), the increment of task T2 is 0.028. The iteration continues until the fault propagation increment of each node is less than the preset convergence threshold ε = 0.01. In the fourth round of iteration, the fault propagation increment of all task nodes is lower than the threshold, the iteration is stopped, and the fault propagation probability in the stable state is obtained, which is 10% for task T1, 12% for task T2, and 9% for task T3. According to the fault propagation probability of the task node and the node importance weight, the risk value of each task node is calculated. If the contribution weights of tasks T1, T2, and T3 are 0.7, 0.6, and 0.5 respectively, then according to the task node risk value evaluation formula R i =P i (final) W i , we get the risk values ​​of the task nodes as R1=0.07, R2=0.072, R3=0.045. For the node set {T2, T3} of the shared device B, we use the formula Calculate the cumulative risk value of the equipment

[0048] R2+R3=0.072+0.045=0.117. Therefore, the risk value of each node in the task chain and the cumulative risk value of the shared device are obtained. Task T2 has the highest risk value, and the cumulative risk value of device B is 0.117, which is the device with the highest risk in the entire task chain.

[0049] Step S106, based on the impact scope report of the production fault chain, extract the priority, resource requirements and completion time of the affected tasks, re-evaluate the task execution sequence and adjust the resource allocation and task schedule.

[0050] Based on the impact range report of the production fault chain, the priority, resource requirements and task completion time of each affected task are extracted. Based on the priority and resource requirements of the task, the execution order of the task is re-evaluated to obtain the task execution sequence table. Based on the re-determined task execution order, the task execution order is optimized using a genetic algorithm. The optimization objective function is set to minimize the fault recovery time, maximize the resource utilization rate and minimize the task completion time. The tasks are dynamically adjusted and rearranged to obtain the optimized production task execution sequence table, including the re-execution order of the task, resource allocation and task completion time. The optimized task execution sequence table is input into the multi-task dynamic scheduling simulation environment to simulate the environment of multi-task concurrent execution, evaluate the resource competition, equipment usage and scheduling conflicts during the task execution process, and verify the effectiveness of the optimized task execution sequence table based on the task completion time, resource utilization rate and equipment idle time. If the effectiveness is lower than expected, the task execution order is adjusted again until the execution time, resource allocation and priority of all task nodes are met.

[0051] For example, in the production fault chain impact range report of a factory, tasks V1, T2, T3 and T4 are affected by the failure of equipment A, among which the priority of task V1 is 0.9, and the resource requirements include station X1, equipment ZX-800 and 5 steel plates, and the task completion time is expected to be 2 hours. The priority of task V2 is 0.8, and the resource requirements include station X2, equipment DF-500 and 10 welding rods, and the task completion time is expected to be 3 hours. The priority of task V3 is 0.7, and the resource requirements include station X3, equipment CT-300 and 3 aluminum plates, and the task completion time is expected to be 4 hours. The priority of task V4 is 0.6, and the resource requirements include station X4, equipment RT-200 and 2 copper pipes, and the task completion time is expected to be 2 hours. According to the priority and resource requirements of the tasks, the execution order of the tasks is re-evaluated. The results show that task V1 takes precedence over T2, T2 takes precedence over T3, and T3 takes precedence over T4, and the preliminary task execution order table is obtained: T1>T2>T3>T4. Based on the re-determined task execution order, the genetic algorithm is used to optimize the task execution order, and the optimization objective function is set to minimize the fault recovery time, maximize the resource utilization, and minimize the task completion time. The initial population contains different combinations of task execution orders, such as [T1, T2, T3, T4] and [T2, T1, T4, T3]. Through crossover and mutation operations, a new task order is generated, and the fault recovery time, resource utilization, and task completion time under each order are calculated. The best task execution sequence table obtained after optimization is [T1, T3, T2, T4], where task V1 is executed on workstation X1 and equipment ZX-800 from 8:00 to 10:00, task V3 is executed on workstation X3 and equipment CT-300 from 10:00 to 14:00, task V2 is executed on workstation X2 and equipment DF-500 from 14:00 to 17:00, and task V4 is executed on workstation X4 and equipment RT-200 from 17:00 to 19:00. The optimized task execution sequence table is input into the multi-task dynamic scheduling simulation environment to simulate the concurrent execution of multiple tasks and evaluate the resource competition, equipment usage and scheduling conflicts during task execution. The simulation results show that the resource utilization rate is 92%, the idle time of the equipment ZX-800 and DF-500 is 0 hours and 1 hour respectively, the actual completion time of tasks V1 and T3 is consistent with expectations, but task V2 is slightly delayed in resource competition, the delay time is 0.5 hours, and task V4 is completed on time. According to the simulation results, the effectiveness of the optimized task execution sequence table meets the expected indicators, the fault recovery time is reduced by 15%, the resource utilization rate is increased by 12%, and the task completion time is reduced by 10%.If the simulation results do not meet the expected goals, the task execution sequence table will be further adjusted. For example, in this optimization, if it is found that the delay of task V2 causes the start time of task V4 to be postponed, the resource scheduling can be optimized by increasing the priority of T2 or changing its equipment allocation until the simulation results meet the completion time, resource allocation and priority requirements of all tasks, ultimately ensuring that the factory's production efficiency reaches the optimal level.

[0052] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the concept of the present application. The above exemplary features are replaced with the technical features with similar functions disclosed in the present application (but not limited to) to form a technical solution.

Claims

1. A simulation optimization method for a digital twin factory, characterized in that: The method comprises: The industrial IoT sensors in the factory workshop are used to obtain workshop equipment data, and the completion time of production tasks and resource consumption are timestamped and stored in categories; Based on the basic operation data of the equipment, frequent item sets are extracted and a table of key production process nodes is generated. Combined with the equipment operation parameters and resource consumption patterns, task priorities and resource allocation rules are determined, and a task allocation parameter table containing task priorities and resource allocation rules is generated. Obtain physical factory building drawings, equipment layout diagrams, and logistics route maps, use 3D modeling tools to build a three-dimensional space model, and combine real-time production data to create a digital twin simulation factory model; According to the production task data in the digital twin simulation factory model, combined with the task priority rules, the task priority list and resource demand information are obtained, the dynamic programming state space and decision variables are constructed, and the resource usage time, resource allocation location and specific quantity of each task are determined; According to the resource allocation scheme and task chain dependency, the dependency and resource allocation information of task nodes are extracted, a Bayesian network including task and equipment dependency is constructed, the impact range of fault propagation is simulated, and the risk value of task nodes is evaluated in combination with historical equipment failure data; Based on the impact scope report of the production failure chain, extract the priority, resource requirements and completion time of the affected tasks, re-evaluate the task execution sequence and adjust resource allocation and task schedule.

2. The method according to claim 1, wherein: The method obtains workshop equipment data through industrial Internet of Things sensors in the factory workshop, timestamps the completion time of production tasks and resource consumption, and stores them in categories, which is characterized by: By installing industrial Internet of Things sensors on production equipment, logistics routes, quality inspection areas and resource consumption monitoring points in the factory workshop, the location information, cargo parameters, shelf parameters, equipment operating status, material logistics routes, resource consumption, production task completion time and quality inspection data of the equipment are collected in real time. The collected raw data are timestamped and classified for storage to build a workshop equipment data set. Cargo parameters include volume and weight, shelf parameters include volume and maximum load-bearing capacity, equipment operating status includes speed, temperature, vibration, resource consumption includes electricity and raw materials, and quality inspection data includes product qualification rate and defect rate. The basic operating data and process flow parameters of the equipment are obtained through the storage database, and the basic operating data are normalized to obtain the basic operating data table of the equipment. The basic operating data includes equipment model, operating frequency, energy consumption rate, and maintenance records. The process flow parameters include the processing time of each task, material usage and process temperature and pressure. The material status data includes the inventory, processing status and location distribution of the material.

3. The method according to claim 1, wherein: Based on the basic operation data of the equipment, frequent item sets are extracted and a key production process node table is generated. The task priority and resource allocation rules are determined in combination with the equipment operation parameters and resource consumption patterns, and a task allocation parameter table containing the task priority and resource allocation rules is generated, which is characterized by: Based on the basic operation data table of the equipment, support and confidence thresholds are set, and the Apriori algorithm is used to decompose the basic operation data of the equipment into frequent item sets, and the mining results are sorted. According to the priority of support and confidence, a key production process node table is obtained, including the interactive position of the equipment, logistics nodes and key production processes; according to the equipment operation parameters, including operation frequency, usage time, energy consumption and maintenance cycle, the K-Means clustering algorithm is used to calculate the similarity between devices based on the Euclidean distance, and the equipment is clustered and grouped; according to the clustering results, the resource consumption of each group of equipment is determined, and the task load of the equipment is calculated through the equipment usage time and task distribution data to determine the equipment usage pattern; according to the task completion time and task importance data, the weighted scoring method is used to calculate the task priority score, generate a task priority table, and combine the resource consumption pattern to determine the resource allocation rules required for each task, including resource allocation time, quantity and location, and generate a task allocation parameter table containing task priority and resource allocation rules.

4. The method according to claim 1, wherein: The method obtains physical factory building drawings, equipment layout diagrams and logistics route maps, uses 3D modeling tools to build a three-dimensional space model, and combines real-time production data to create a digital twin simulation factory model, which is characterized by: Obtain the architectural drawings, equipment layout and logistics route map of the physical factory, and use the 3D modeling tool Unity to build a three-dimensional spatial model of the factory, including the workshop structure, equipment distribution, product library, raw material library and logistics channel; embed the floor space, working area and connection location of all equipment, including power connection, data interface, material entry and exit points into the model, and generate a geometric model of each device; collect real-time production data from the production line, input Simulink and connect with the three-dimensional spatial model, simulate the execution status of production tasks, and display the operating status of each device in real time. The real-time production data includes equipment status, production task progress and equipment The interaction between equipment and devices, equipment status includes on or off, running speed, and energy consumption; production task progress includes task start time, task completion time, and processing progress; the digital twin platform GEPredix is ​​used to combine the three-dimensional space model with the real-time data access interface to create a digital twin simulation factory model; real-time production data is used to build a dynamic process simulation model, simulate the execution process of production tasks in the factory, set production environment parameters, and calculate task scheduling and resource allocation in real time through simulation to obtain dynamic process simulation results including task completion time and equipment utilization; production environment parameters include equipment failure, production task priority, and resource constraints.

5. The method according to claim 1, wherein: According to the production task data in the digital twin simulation factory model, combined with the task priority rules, the task priority list and resource demand information are obtained, the dynamic programming state space and decision variables are constructed, and the resource usage time, resource allocation location and specific quantity of each task are determined, which is characterized by: Through the production task data in the digital twin simulation factory model, the completion time, required resource type and quantity of each task are extracted; according to the characteristics of the task, the priority rules are defined, where tasks with shorter completion time or higher resource requirements have higher priority; according to the execution requirements of the task, the resource requirement information of each task is obtained, and the completion time and resource requirement information of each task are arranged in order of priority to obtain a task priority list. The resource requirement information includes equipment, workstations, and raw materials; according to the task priority list, the state space and decision variables of the dynamic programming model are determined. The state space includes the resource allocation at each moment, and the decision variable is the resource allocation decision of the task; the scheduling efficiency is evaluated by calculating the resource usage time, total resource consumption and task completion time of each task, and the resource scheduling efficiency is set as the objective function. According to the dynamic programming model, a bottom-up dynamic programming method is used to gradually deduce from the initial state of the task to the final state of the task, and the optimal configuration of resources at each moment is calculated. In the calculation process, the resource usage efficiency of each task at different time points is evaluated to determine whether the resource constraints are met, and the resource allocation plan for each task is obtained. The resource allocation plan is determined, including the resource usage time, location and quantity of each task.

6. The method according to claim 1, wherein: According to the resource allocation scheme and task chain dependency, the dependency and resource allocation information of the task nodes are extracted, a Bayesian network including the task and equipment dependency is constructed, the impact range of fault propagation is simulated, and the risk value of the task nodes is evaluated in combination with historical equipment failure data, which is characterized by: According to the resource allocation plan, obtain each task node, resource allocation location, time and equipment information, and identify the sequential dependencies between tasks; based on each task node in the task chain and its dependencies, construct a Bayesian network structure, set the task node as a random variable, and the state of each device as its parent node, to obtain the structure of the Bayesian network, in which the node represents the task and the device, and the edge represents the dependency between the task and the device; obtain historical equipment failure data, task completion time and equipment failure records, and use the maximum likelihood estimation method to set a conditional probability table for each node to represent the conditional probability of equipment failure or task failure; for each node in the task chain, obtain the probability of the task being affected, and incorporate it into the calculation of the Bayesian network to determine the conditional probability of each task and device state. The affected tasks include delayed task completion and task failure; if a device or task node fails, use Bayesian The network performs forward propagation to simulate the process of fault extension from the node to the dependent nodes; the probability of subsequent task delay or failure caused by a single fault is calculated for each task node, and the degree of impact of each subsequent task node is obtained; the impact of fault propagation in the task chain is calculated through the Bayesian network, the fault propagation probability of each task node is obtained, and the impact range of the fault propagating from the initial node to the task chain is determined; according to the fault propagation probability of each task node, the cumulative probability of each task node in the task chain being affected by equipment failure is evaluated, and the risk value of each task node and the cumulative risk value of shared equipment are obtained. All task nodes in the task chain are ranked by risk value to determine the task nodes most seriously affected by fault propagation; based on the fault propagation analysis results, the risk values ​​of task nodes, the cumulative risk values ​​of shared equipment and the impact range of fault propagation are integrated to generate an impact range report for the production fault chain.

7. The method according to claim 6, wherein: According to the fault propagation probability of each task node, the cumulative probability of each task node in the task chain being affected by the device failure is evaluated to obtain the risk value of each task node and the cumulative risk value of the shared device, which is characterized by: Obtain historical dependency data between tasks, including resource sharing, time delay, and data interaction. Use recurrent neural networks for model training, build a task dependency weight prediction model, and predict the task dependency weights between nodes. Obtain the dependency relationship and initial failure probability P0 of each task node in the task chain, and use the failure propagation impact formula of the task node. Calculate the fault propagation probability P of node i at time t+1 i (t+1), where α is the fault self-recovery attenuation factor of the node itself, β is the fault propagation amplification factor, which indicates the degree of impact of external faults on the node, obtained by fitting historical data, N(i) represents the set of all upstream dependent tasks of node i, and γ ij is the task dependency weight between nodes i and j, indicating the influence of task j on i. For each node in the task chain, starting from time t = 0, the fault propagation influence formula is used for multiple iterations to calculate the fault propagation probability P of the node. i (t), record the fault propagation increment ΔP of the node in each iteration i (t) = P i (t+1)-P i (t), and determine whether a stable state is reached based on a preset increment threshold; if Then the iteration is stopped and the final fault propagation probability of each node in the stable state is obtained, where ε is the iterative convergence threshold; according to the weighted value of the fault propagation probability and the importance of the equipment, the task node risk value evaluation formula R is used i =P i (final) W i , calculate the risk value of the device sharing node in the task chain, where W i is the contribution weight of node i to the production chain, indicating the importance of the task to the overall production chain, which can be calculated by the historical task completion rate or resource consumption ratio. i (final) represents the final fault propagation probability of the task being affected by device failure or other task failure after the task node i reaches a stable state in the fault propagation model; for the node set G(e) sharing the device, the cumulative value of the device risk is calculated Determine the risk value R of each node i and the cumulative risk value R of shared equipment e .

8. The method according to claim 1, wherein: The report on the impact range of the production fault chain extracts the priority, resource requirements and completion time of the affected tasks, re-evaluates the task execution sequence and adjusts the resource allocation and task schedule, and is characterized by: Based on the impact range report of the production fault chain, the priority, resource requirements and task completion time of each affected task are extracted. Based on the priority and resource requirements of the tasks, the execution order of the tasks is re-evaluated to obtain the task execution sequence table; based on the re-determined task execution order, the task execution order is optimized using a genetic algorithm, and the optimization objective function is set to minimize the fault recovery time, maximize the resource utilization rate and minimize the task completion time. The tasks are dynamically adjusted and rearranged to obtain the optimized production task execution sequence table, including the re-execution order of the tasks, resource allocation and task completion time; the optimized task execution sequence table is input into the multi-task dynamic scheduling simulation environment to simulate the environment of multi-task concurrent execution, evaluate the resource competition, equipment usage and scheduling conflicts during the task execution process, and verify the effectiveness of the optimized task execution sequence table based on the task completion time, resource utilization rate and equipment idle time; if the effectiveness is lower than expected, the task execution sequence is adjusted again until the execution time, resource allocation and priority of all task nodes are met.

Images