MES-based digital factory flexible production scheduling system

Abstract

The invention provides a digital factory flexible production scheduling system based on MES, and relates to the technical field of digital management, and the method comprises the steps: capturing an order change instruction in real time through an integrated interface of the MES and ERP, and converting the order change instruction into a standard chemical order data stream; based on the structured data object, fusing the equipment real-time sensing data with a pre-stored process constraint matrix and dynamic material topological data, and generating a three-layer real-time resource model containing an equipment state vector, the process constraint matrix and the material topological data; layered optimization is executed based on a three-layer real-time resource model, in the first layer of optimization, an equipment state vector and a process constraint matrix are used as hard constraints, and an initial feasible solution set is generated; and the second-layer optimization executes multi-target evaluation on the initial feasible solution set to obtain a non-dominated solution set as a production scheduling plan. According to the invention, order-resource-execution full-process intelligent collaboration is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of digital management, and in particular to a flexible production scheduling system for a digital factory based on MES. Background Art

[0002] Amidst the rapid growth of the new energy vehicle industry, battery modules, as core components, face increasingly complex manufacturing challenges. With surging market demand for diversified and personalized new energy vehicles, battery module models are becoming increasingly diverse. Different types of products, such as ternary lithium, lithium iron phosphate, and energy storage batteries, need to be mixed and produced on the same production line to meet the customized orders from various vehicle manufacturers.

[0003] Traditional production scheduling methods have exposed numerous flaws when dealing with such complex situations. For one thing, dynamic order access struggles with efficient integration with ERP systems to obtain real-time order change information and quickly trigger production schedule recalculations, leading to a slow response to order changes. Furthermore, resource modeling struggles with effective integration and real-time updates of multi-dimensional resources such as equipment, processes, and materials, resulting in a mismatch between planned production schedules and actual production resources. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a flexible production scheduling system for a digital factory based on MES to realize intelligent collaboration of the entire process of order-resource-execution.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows: In a first aspect, a flexible production scheduling method for a digital factory based on MES is provided, the method comprising: Step 1: Capture order change instructions in real time through the integrated interface between MES and ERP and convert them into standardized work order data streams; Step 2: Based on the structured data object, the real-time sensor data of the equipment is integrated with the pre-stored process constraint matrix and dynamic material topology data to generate a three-layer real-time resource model including the equipment state vector, process constraint matrix and material topology data; Step 3: Perform hierarchical optimization based on the three-layer real-time resource model. The first-layer optimization uses the equipment state vector and process constraint matrix as hard constraints to generate an initial feasible solution set. The second-layer optimization performs multi-objective evaluation on the initial feasible solution set to obtain a non-dominated solution set as the production schedule. Step 4: Based on the production anomaly detection results during the production scheduling process, the equipment-level task diversion mechanism is activated. Based on the material topology data, the transportation path is abstracted into two-dimensional line segments for spatial intersection detection. Logistics avoidance instructions and process parameter adaptive adjustment instructions are generated to obtain the adjusted tasks and parameters. Step 5: Based on the adjusted tasks and parameters, verify the operator's skill compatibility, redistribute conflicting tasks, and conduct root cause analysis based on equipment fault codes and historical maintenance logs to obtain an updated task instruction set and resource status database. Step 6: Based on the updated task instruction set and resource status database, combined with the edge computing node status, the production schedule is fine-tuned in real time when the network is connected, and the optimized production schedule function is maintained based on the local resource model copy when the network is interrupted.

[0006] Furthermore, the integrated interface between MES and ERP captures order change instructions in real time and converts them into standardized work order data streams, including: Step 1.1: Capture real-time order change instructions through the integrated interface between MES and ERP; Step 1.2: Parse the key business elements in the order change instruction, including order identifier, product specifications, delivery time window, process version, and priority level; In step 1.3, the key business elements are mapped into structured data objects, which include the work order unique identifier, bill of materials fingerprint, process path code, time constraint set and optimization weight factor.

[0007] Furthermore, based on structured data objects, the real-time sensor data of the equipment is integrated with the pre-stored process constraint matrix and dynamic material topology data to generate a three-layer real-time resource model containing the equipment state vector, process constraint matrix and material topology data, including: Step 2.1: Receive the output standardized work order data stream and parse the structured data objects therein; Step 2.2: Based on the process path encoding of the structured data object, a pre-stored process constraint matrix is ​​called. The matrix includes the sequence constraints between processes, the equipment-process binding relationship, and the process parameter threshold set. Step 2.3: Based on the BOM fingerprint of the structured data object, dynamic material topology data is constructed. This data defines the physical coordinate set of the material nodes, the geometric parameters of the traversable paths between the nodes, and the real-time occupancy status of each path. Step 2.4: Fusion of the device’s real-time sensor data with the time constraint set of the structured data object, optimization of the weight factors, and generation of the device state vector. In step 2.5, the process constraint matrix, material topology data, and equipment state vector are integrated to form a three-layer real-time resource model.

[0008] Furthermore, hierarchical optimization is performed based on the three-layer real-time resource model. The first-layer optimization uses the equipment state vector and process constraint matrix as hard constraints to generate an initial feasible solution set. The second-layer optimization performs multi-objective evaluation on the initial feasible solution set to obtain a non-dominated solution set as the production schedule, including: Step 3.1: Extract the equipment state vector and process constraint matrix from the three-layer real-time resource model, and construct a constrained optimization model with process sequence as decision variables, equipment state as boundary conditions, and process constraints as hard rules. Generate an initial feasible solution set by enumerating feasible process permutations and combinations that satisfy all hard constraints. In step 3.2, a multi-objective evaluation is performed on the initial feasible solution set. For each feasible solution, weighted evaluation values ​​of the delivery date achievement rate, equipment overall efficiency, and energy consumption coefficient are calculated in parallel. By comparing the objective function values ​​of each solution pairwise, a solution set is selected in which all objective dimensions are not dominated. In step 3.3, normalize the non-dominated solution set and select the solution with the best weighted evaluation value as the final production schedule.

[0009] Furthermore, based on the production anomaly detection results during the production scheduling process, the equipment-level task diversion mechanism is activated. Based on the material topology data, the transportation path is abstracted into two-dimensional line segments for spatial intersection detection. Logistics avoidance instructions and process parameter adaptive adjustment instructions are generated, and the adjusted tasks and parameters are obtained, including: Step 4.1: Receive the production schedule execution data stream in real time and analyze the work order task sequence, equipment allocation plan, and process parameter settings contained therein; Step 4.2: Perform anomaly detection based on the real-time sensor data of the equipment state vector. Compare the actual equipment load rate with the upper load threshold defined in the process constraint matrix. If the actual value continuously exceeds the threshold, it is marked as an equipment overload anomaly. Compare the deviation between the real-time monitored value of the process parameter and the planned value. If the deviation exceeds the preset tolerance range, it is marked as a process deviation anomaly. Step 4.3: Start diversion for abnormal tasks, extract the geometric features of the original transportation path, model the new path as a two-dimensional line segment, and calculate the spatial relationship with the activated path in real time; In step 4.4, dynamic instructions are generated based on the spatial relationship detection results. When a path intersection is detected, the detour path cost is calculated based on the path load-bearing level and the work order priority weight. The non-intersection path with the lowest cost is selected to generate a logistics avoidance instruction. The process parameters are adjusted to within the threshold range through a closed-loop feedback mechanism. When there is no path intersection, the original transportation path is maintained, and adaptive calibration of the process parameters is performed to obtain an updated task instruction containing the adjusted path and parameters.

[0010] Furthermore, based on the adjusted tasks and parameters, the operator skill matching is verified, conflicting tasks are redistributed, and root cause analysis is performed based on equipment fault codes and historical maintenance logs. This results in an updated task instruction set and resource status database, including: Step 5.1: Based on the adjusted tasks and parameters, perform operator skill compatibility verification. Extract the skill qualification code set in the task process requirements, retrieve the skill matrix, which stores each operator's qualification certificate number, skill proficiency weight, and current task load, and calculate the skill compatibility. If the compatibility falls below the preset threshold, mark it as a conflicting task. Step 5.2: Generate a reallocation plan for the conflicting tasks. Construct a three-dimensional decision space of task, equipment, and operator. Traverse the set of alternative equipment that meets the process constraints. Calculate the operator skill matching for each alternative equipment. Select the combination with the highest matching degree that meets the equipment capability constraints as the reallocation plan. Step 5.3: Perform root cause analysis based on the equipment fault code and historical maintenance logs. Analyze the pattern category corresponding to the equipment fault code, retrieve the previous occurrence records of the same fault pattern in the historical maintenance logs, calculate the probability distribution of the fault root cause, and select the root cause with the highest probability and associate it with the current task to obtain the root cause analysis results. In step 5.4, based on the reallocation plan and the root cause analysis results, the task instruction set is reconstructed, the reallocation plan is written into the task equipment allocation field, the root cause code is written into the task exception record field, and the resource status database is updated.

[0011] Furthermore, based on the updated task instruction set and resource status database, combined with the edge computing node status, the production schedule can be fine-tuned in real time when the network is connected. In the event of a network outage, the optimized production schedule function can be maintained based on the local resource model copy, including: Step 6.1: Collect network communication quality data of edge computing nodes in real time, determine network connectivity status by continuously analyzing transmission stability indicators, and output connectivity mark signals; In step 6.2, when the marking signal indicates network connectivity, the system extracts the real-time equipment occupancy status and material flow trajectory based on the updated task instruction set and resource status database. It then identifies the affected work orders that compete with the changed task for equipment resources or have process dependencies. The system then recalculates the process time windows for the affected work orders based on the equipment release time nodes and logistics buffer requirements, dynamically reconstructing the work order execution timing in the production schedule. In step 6.3, when the marking signal indicates a network outage, the three-layer real-time resource model replica stored in the edge node is activated. Based on the equipment state vector and process constraint matrix in the replica, an equipment allocation plan that maximizes processing efficiency is selected. A spatial avoidance path is generated based on the real-time occupancy status of the material topology data. Step 6.4 integrates the processing results of steps 6.2 and 6.3. In the network connectivity scenario, output a full production schedule including a timing adjustment plan. In the network interruption scenario, output an edge autonomous instruction set including device allocation instructions and path avoidance instructions.

[0012] The second aspect is a flexible production scheduling system for a digital factory based on MES, including: The acquisition module is used to capture order change instructions in real time through the integrated interface between MES and ERP and convert them into standardized work order data streams; The construction module is used to integrate real-time equipment sensor data with pre-stored process constraint matrices and dynamic material topology data based on structured data objects to generate a three-layer real-time resource model containing equipment state vectors, process constraint matrices, and material topology data. Hierarchical optimization is performed based on the three-layer real-time resource model. The first layer of optimization uses the equipment state vectors and process constraint matrices as hard constraints to generate an initial feasible solution set. The second layer of optimization performs multi-objective evaluation on the initial feasible solution set to obtain a non-dominated solution set as the production scheduling plan. The detection module is used to activate the equipment-level task diversion mechanism based on the production anomaly detection results during the production scheduling process. Based on the material topology data, the transportation path is abstracted into two-dimensional line segments for spatial intersection detection. The module generates logistics avoidance instructions and process parameter adaptive adjustment instructions to obtain the adjusted tasks and parameters. The output module is used to adjust tasks and parameters, verify operator skill matching, redistribute conflicting tasks, and conduct root cause analysis based on equipment fault codes and historical maintenance logs to obtain an updated task instruction set and resource status database; The processing module is used to fine-tune the production schedule in real time when the network is connected based on the updated task instruction set and resource status database, combined with the edge computing node status, and maintain the optimized production scheduling function based on the local resource model copy when the network is interrupted.

[0013] According to a third aspect, a computing device includes: one or more processors; The storage device is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method.

[0014] In a fourth aspect, a computer-readable storage medium stores a program, which implements the method when executed by a processor.

[0015] The above solution of the present invention includes at least the following beneficial effects: The flexible production scheduling system for digital factories based on MES achieves real-time response and conversion of order changes through efficient integration with the ERP system, builds a three-layer real-time resource model to integrate multi-dimensional production factors, generates production scheduling plans that adapt to reality with the help of hierarchical optimization, and realizes dynamic adjustment by combining anomaly detection and task diversion mechanism, operator skill matching and root cause analysis. It then relies on edge computing to ensure stable operation in the event of network anomalies, effectively solving problems such as slow order response and insufficient resource matching in complex production scenarios such as new energy vehicle battery modules, and realizing intelligent collaboration of the entire process of order-resource-execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The present invention provides a flowchart of a flexible production scheduling method for a digital factory based on MES.

[0017] Figure 2 This is a schematic diagram of a flexible production scheduling system for a digital factory based on MES provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0019] like Figure 1 As shown, an embodiment of the present invention provides a flexible production scheduling method for a digital factory based on MES, the method comprising the following steps: Step 1: Capture order change instructions in real time through the integrated interface between MES and ERP and convert them into standardized work order data streams; Step 2: Based on the structured data object, the real-time sensor data of the equipment is integrated with the pre-stored process constraint matrix and dynamic material topology data to generate a three-layer real-time resource model including the equipment state vector, process constraint matrix and material topology data; Step 3: Perform hierarchical optimization based on the three-layer real-time resource model. The first-layer optimization uses the equipment state vector and process constraint matrix as hard constraints to generate an initial feasible solution set. The second-layer optimization performs multi-objective evaluation on the initial feasible solution set to obtain a non-dominated solution set as the production schedule. Step 4: Based on the production anomaly detection results during the production scheduling process, the equipment-level task diversion mechanism is activated. Based on the material topology data, the transportation path is abstracted into two-dimensional line segments for spatial intersection detection. Logistics avoidance instructions and process parameter adaptive adjustment instructions are generated to obtain the adjusted tasks and parameters. Step 5: Based on the adjusted tasks and parameters, verify the operator's skill compatibility, redistribute conflicting tasks, and conduct root cause analysis based on equipment fault codes and historical maintenance logs to obtain an updated task instruction set and resource status database. Step 6: Based on the updated task instruction set and resource status database, combined with the edge computing node status, the production schedule is fine-tuned in real time when the network is connected, and the optimized production schedule function is maintained based on the local resource model copy when the network is interrupted.

[0020] In an embodiment of the present invention, rapid response to dynamic demand is achieved by real-time capture and conversion of order change instructions, a three-layer real-time resource model is constructed to integrate key elements such as equipment, processes, and materials, and hierarchical optimization is used to generate a production schedule that takes into account both feasibility and multi-objective benefits. Combined with anomaly detection and task diversion mechanism, operator skill matching and root cause analysis, dynamic adjustment of the production process is achieved. At the same time, edge computing is used to ensure stable operation when the network status changes, solving problems such as delayed order response, unbalanced resource matching, and insufficient response to anomalies in complex production scenarios, improving the flexibility, execution efficiency and anti-interference ability of production plans, and realizing intelligent collaboration and optimization of the entire process from order access to production execution.

[0021] In a preferred embodiment of the present invention, the above step 1 may include: Step 1.1: Capture real-time order change instructions through the integrated interface between MES and ERP; Step 1.2: Parse the key business elements in the order change instruction, including order identifier, product specifications, delivery time window, process version, and priority level; In step 1.3, the key business elements are mapped into structured data objects, which include the work order unique identifier, bill of materials fingerprint, process path code, time constraint set and optimization weight factor.

[0022] In an embodiment of the present invention, order change instructions are captured through the integrated interface between MES and ERP, which can respond to dynamic changes in orders in real time; key business elements are analyzed to accurately extract core order information; and they are mapped into structured data objects containing unique work order identifiers, etc., thereby achieving standardization and structuring of order information, improving the system's processing efficiency and accuracy for order changes, and ensuring that production scheduling can be adjusted in a timely manner based on the latest order information.

[0023] In an embodiment of the present invention, the specific steps include: Step 1.1: To establish collaborative interaction between the MES and ERP systems, an integrated interface must first be constructed. This interface can establish a data channel based on industry-standard communication protocols (such as OPC UA, RESTful API, and other methods that are compatible with the company's existing system architecture). When order change events are triggered within the ERP system due to operations such as order addition, modification, and cancellation, the MES can use this integrated interface to actively monitor or receive push notifications from the ERP system, accurately capturing real-time order change instructions. This process requires ensuring the reliability and real-time nature of data transmission across the interface. By setting up a heartbeat mechanism and data validation rules (such as MD5 checksums and field integrity checks), it is ensured that order change instructions are not lost or misplaced during cross-system transmission.

[0024] Step 1.2: After receiving the order change instruction, the instruction content must be deeply disassembled to extract key business elements that have a direct impact on production scheduling decisions. Specifically, using string parsing and XML / JSON structured data parsing technologies (flexibly selected based on the instruction data format output by the ERP system), the order identifier (used to uniquely mark the order and distinguish different production tasks), product specifications (clarifying production object characteristics such as the battery module model, material system, and dimensional parameters), delivery time window (defining the time period within which the order must be completed and delivered, which serves as the time constraint for production scheduling), process version (linking the corresponding product's production process and parameter standards to ensure production process compliance), and priority level (used to determine production scheduling priority when multiple orders are running in parallel) are identified and extracted field by field. During the parsing process, regular expression matching and predefined field mapping tables are used to ensure the accuracy of element extraction and convert unstructured or semi-structured order change instructions into...

[0025] In step 1.3, based on the extracted key business elements, standardized structured data objects that can be recognized and called by various modules of the system are constructed to realize the transformation of order information into production scheduling and execution logic. In the specific operation, according to the preset enterprise production data model specifications, each business element is matched with the fields and data types that the system can recognize: Work order unique identification: generated using UUID (Universally Unique Identifier) ​​or enterprise-defined coding rules to ensure that the work order is uniquely marked throughout the entire production scheduling process for task tracking and status association; BOM fingerprint: Hash operations are performed on the material components corresponding to product specifications (such as the battery cells, housings, connectors, and other material details of ternary lithium / lithium iron phosphate battery modules) (or feature codes are generated based on material coding rules) to form fingerprint information that can be quickly retrieved and verified for material matching, which is used for subsequent material topology data association and completeness checks; Process path coding: This code is used to identify the production process sequence of the corresponding product in the enterprise process knowledge base. This clarifies the process execution sequence, equipment usage sequence, and other process logic from raw material input to finished product output, providing a basis for equipment allocation and process scheduling during production scheduling. Time constraint set: The delivery time window is broken down into time node constraints such as the lower limit of the work order production start time and the upper limit of the production completion time. Combined with the standard working time data of each process, it is converted into time dimension constraints for production scheduling; Optimize weight factors: Based on the order priority level and combined with the company's production strategy (such as delivery urgency, customer importance, profit contribution, etc.), through preset weight calculation rules (such as hierarchical scoring method and linear weighting method), the order is assigned weight coefficients for different target dimensions (delivery time achievement rate, equipment efficiency, etc.) in production scheduling optimization (such as multi-objective evaluation stage) to guide the optimization direction of the production scheduling plan.

[0026] In a preferred embodiment of the present invention, the above step 2 may include: Step 2.1: Receive the output standardized work order data stream and parse the structured data objects therein; Step 2.2: Based on the process path encoding of the structured data object, a pre-stored process constraint matrix is ​​called. The matrix includes the sequence constraints between processes, the equipment-process binding relationship, and the process parameter threshold set. Step 2.3: Based on the BOM fingerprint of the structured data object, dynamic material topology data is constructed. This data defines the physical coordinate set of the material nodes, the geometric parameters of the traversable paths between the nodes, and the real-time occupancy status of each path. Step 2.4: Fusion of the device’s real-time sensor data with the time constraint set of the structured data object, optimization of the weight factors, and generation of the device state vector. In step 2.5, the process constraint matrix, material topology data, and equipment state vector are integrated to form a three-layer real-time resource model.

[0027] In an embodiment of the present invention, by parsing the standardized work order data stream, the structured information on the order side is accurately taken over, and the production task benchmark is anchored for resource modeling; the pre-stored process constraint matrix is ​​called to solidify the process logic, equipment adaptation relationship and parameter boundaries as model hard constraints to ensure production scheduling compliance and avoid process conflicts; a dynamic material topology is constructed based on the bill of materials fingerprint to map the material space layout and logistics path status in real time; the equipment sensor data is integrated with the order time and optimization factor to generate an equipment state vector to dynamically reflect the matching degree between the equipment's real-time capability and order demand, so that the production scheduling decision fits the actual production capacity; finally, the three-layer elements are integrated to form a three-dimensional real-time resource model of "process-material-equipment", which comprehensively integrates the constraints and status of all production factors, provides an accurate and dynamic resource base for hierarchical optimization of production scheduling, and enables the production scheduling plan to shift from "based on empirical assumptions" to "based on real-time resource precise adaptation", improves the feasibility of production scheduling, and helps digital factories cope with the complex resource coordination needs of multi-variety mixed flow production, and enhances the order-resource-execution synergy efficiency.

[0028] In an embodiment of the present invention, the specific steps include: In step 2.1, the system must establish a data receiving channel to obtain the standardized work order data stream output in step 1. This data stream can be transmitted from the MES and ERP integration module to the data preprocessing unit of the production scheduling system via a message queue (such as RabbitMQ or Kafka, depending on the enterprise information architecture), a file transfer protocol (FTP / SFTP), or direct API data push.

[0029] Specifically, based on the defined structured data object format (including fields such as the work order unique identifier and bill of materials fingerprint), the specific content of each field is identified and extracted through field matching, keyword extraction, and other means. For example, for the "work order unique identifier" field, its format compliance is verified according to the enterprise coding specifications (such as the "order type + timestamp + serial number" rule); for the "bill of materials fingerprint", its associated material composition information is confirmed to be accurate by comparing it with the enterprise material master data dictionary. This parsing process must ensure the integrity and accuracy of the data, providing a reliable input foundation for the subsequent resource modeling process, converting the business needs of the order side into structured data objects that can be recognized and calculated by the system, and realizing the connection between order information and the production scheduling resource association logic.

[0030] In step 2.2, based on the structured data object, its "Process Path Code" field is the key index for accessing the pre-stored process constraint matrix. The company must pre-build corresponding process constraint matrices for different products and process versions in the process knowledge base of the production scheduling system.

[0031] In specific implementation, the mapping relationship between process path coding and knowledge base index (such as establishing a coding-matrix ID association table) is used to accurately locate the process constraint matrix of the corresponding product. Three types of core constraints are predefined in this matrix: Sequential constraints between processes: The execution order of production processes is clearly defined in the form of a directed graph or sequence list (for example, in battery module production, the battery cell must be placed in the shell before the welding process can proceed), ensuring that the process sequence conforms to the process logic during production scheduling. Equipment-process binding relationship: defines the equipment set that can be used for each process (for example, the injection process can only be performed on specific injection equipment), as well as the equipment's adaptability to the process (for example, whether the equipment's production rate and accuracy meet the process requirements). This provides a range of options and capacity constraints for subsequent equipment allocation. Process parameter threshold set: defines the allowable fluctuation range of key parameters in each process (for example, the welding temperature must be controlled within "250℃±5℃" and the injection volume must be within "100ml±2ml"), which serves as the basis for parameter monitoring and abnormality judgment during the production process.

[0032] During the calling process, the system deeply associates the process constraint matrix with the process path code of the current work order, so that the production scheduling decisions strictly follow the enterprise process standards, avoid production quality problems caused by process logic conflicts, and lay a solid foundation for the feasibility of the production scheduling plan from the process dimension.

[0033] In step 2.3, for the “Bill of Materials Fingerprint” of the structured data object, it is necessary to construct the corresponding dynamic material topology data to reflect the spatial distribution, flow path and real-time status of the materials in the production process.

[0034] Based on the material composition information associated with the bill of materials fingerprint and combined with the company's production layout (such as the location of material storage areas in the workshop and the distribution of workstations on the production lines), each material node (such as battery cell storage racks, casing buffer areas, and temporary storage locations for work-in-progress products) is assigned a set of physical coordinates (using the workshop coordinate system, with units of meters or millimeters used for precise location marking). At the same time, the system identifies the accessible paths between material nodes (such as AGV transport channels and manual handling routes), defines their geometric parameters (path length, width, turning radius, etc.), and collects the occupancy status of the paths in real time (such as whether they are occupied by other material transport tasks and the current load weight), generating dynamically updated path information.

[0035] For example, in a battery module production workshop, material topology data must include the coordinates and width of the transport path from the warehouse to the production line buffer, as well as the real-time number and load of AGVs operating along this path. This approach integrates the static spatial layout and dynamic flow status of materials into topological data, providing a visual and computable spatial model for subsequent logistics planning and material integrity checks during production scheduling. This addresses the issue of blind material flow path planning and conflicts with actual space in traditional production scheduling.

[0036] In step 2.4, the generation of the equipment state vector requires the integration of two parts of data: one is the order-side constraints such as the "time constraint set" and "optimization weight factor" in the structured data object; the other is the real-time sensor data of the equipment (equipment operation data collected through the Internet of Things gateway, such as load rate, energy consumption, current process execution progress, etc.).

[0037] During specific operations, the system's data fusion module receives sensor data uploaded by the device's IoT gateway in real time and extracts key indicators relevant to production scheduling decisions (such as whether the device is currently idle, how long it has been running, and the real-time load percentage). Simultaneously, it analyzes the "time constraint set" of structured data objects to clarify the time requirements of the work order for the equipment (e.g., the corresponding process must be started within 2 hours and completed within 4 hours). Based on the "optimization weight factor," it determines the priority bias for efficiency, energy consumption, and other goals during equipment scheduling (e.g., high-priority orders must be delivered first, and energy consumption constraints can be appropriately relaxed).

[0038] By fusing data from multiple sources (for example, combining real-time equipment load with time constraints to assess the equipment's available capacity while meeting time requirements), a device state vector is generated. This vector, in the form of a multidimensional array or feature list, dynamically reflects the equipment's real-time capabilities (such as its ability to accept new tasks and the expected completion time after accepting them) and its compatibility with order requirements (such as whether the equipment's time margin meets the work order's delivery window).

[0039] In step 2.5, based on the above four sub-steps, the system integrates the process constraint matrix, dynamic material topology data, and equipment state vector through the resource model integration unit to build a three-layer real-time resource model of "process-material-equipment".

[0040] Specifically, the process constraint matrix serves as the "logical framework" to define the boundaries of production scheduling process rules; dynamic material topology data serves as the "spatial carrier," providing the spatial layout and path status of material flow; and the equipment state vector serves as the "executor," reflecting the equipment's real-time production capacity and task acceptance capabilities. These three elements are deeply integrated through data interfaces and associated indexes (e.g., linking equipment-process relationships in the process constraint matrix to equipment state vectors, and linking material nodes in material topology data to material requirements in process paths). This creates a three-dimensional resource model that encompasses production logic constraints, spatial material status, and real-time equipment capabilities.

[0041] In a preferred embodiment of the present invention, the above step 3 may include: Step 3.1: Extract the equipment state vector and process constraint matrix from the three-layer real-time resource model, and construct a constrained optimization model with process sequence as decision variables, equipment state as boundary conditions, and process constraints as hard rules. Generate an initial feasible solution set by enumerating feasible process permutations and combinations that satisfy all hard constraints. In step 3.2, a multi-objective evaluation is performed on the initial feasible solution set. For each feasible solution, weighted evaluation values ​​of the delivery date achievement rate, equipment overall efficiency, and energy consumption coefficient are calculated in parallel. By comparing the objective function values ​​of each solution pairwise, a solution set is selected in which all objective dimensions are not dominated. In step 3.3, normalize the non-dominated solution set and select the solution with the best weighted evaluation value as the final production schedule.

[0042] In an embodiment of the present invention, a constraint optimization model is constructed by extracting core resource constraints, and an initial feasible solution is generated by enumeration, thereby building a solid foundation for the "compliance" of production scheduling and ensuring that the plan complies with equipment capabilities and process logic. Based on multi-objective evaluation, core indicators such as delivery time, equipment efficiency, and energy consumption are calculated in parallel, and non-dominated solution sets are screened through dominance relationships, taking into account multiple objectives such as production efficiency and cost control, and adapting to the multi-dimensional benefit demands of the enterprise. The non-dominated solution sets are normalized and optimized, which not only resolves the conflict of multi-objective decision-making, but also accurately anchors the optimal solution through standardized operations. The overall process upgrades production scheduling from "experience-driven" to "data-driven", balancing delivery, efficiency, and energy consumption while ensuring process and equipment constraints, and providing digital factories with scientific, adaptive, and competitive production scheduling solutions.

[0043] In an embodiment of the present invention, the specific steps include: Step 3.1: Accurately extract the equipment state vector and process constraint matrix from the three-layer real-time resource model. The equipment state vector captures the equipment's current operating status, including information such as whether the equipment is idle, the progress of tasks currently being executed, and the equipment load rate. This information forms the equipment's capacity boundary for undertaking production tasks. The process constraint matrix defines the rules that must be followed during the production process, such as the strict sequence between processes (for example, in battery production, cell filling must be completed before packaging can begin), the binding relationship between specific processes and equipment (for example, high-precision welding processes can only be performed by equipment with specific laser welding capabilities), and the thresholds for process parameters in each process (for example, the temperature of a coating process must be controlled between 200°C and 220°C). Based on this extracted information, a constrained optimization model is constructed. This model uses the process sequence as the decision variable, determining the execution order of each production task; the equipment state as the boundary condition, ensuring that the equipment cannot undertake tasks beyond its capacity; and the process constraints as hard rules to ensure compliance with the production process.

[0044] Conduct feasible solution enumeration. Based on the established constraint optimization model, use the algorithm to enumerate all possible combinations of processes that meet the equipment state boundary conditions and process constraints.

[0045] For example, for a production task that includes processes A, B, and C, combined with the equipment status (assuming that equipment 1 can execute processes A and B, and equipment 2 can execute processes B and C) and process constraints (process A must be before process B), we enumerate multiple possible process-equipment matching and execution order combinations such as "equipment 1 (process A) → equipment 1 (process B) → equipment 2 (process C)" and "equipment 1 (process A) → equipment 2 (process B) → equipment 2 (process C)". Then, we check one by one whether these combinations have process conflicts (such as whether they violate the sequence of processes) and whether the equipment is capable of taking over (such as whether the equipment load rate is within an acceptable range). We filter out all combinations that meet the conditions and generate an initial feasible solution set.

[0046] In a specific example, suppose a work order includes processes X, Y, and Z. The process constraint matrix stipulates that process X must precede process Y, which in turn must precede process Z. Furthermore, process X can only be performed by machine M, process Y can only be performed by machine N, and process Z can be performed by either machine N or machine P. The machine state vector indicates that machine M is currently idle, machine N is currently executing a task but is expected to be released in one hour, and machine P is idle. Then, according to the constrained optimization model, feasible process permutations are enumerated, such as "machine M (process X) → machine N (process Y, to be executed in one hour) → machine N (process Z)" and "machine M (process X) → machine N (process Y, to be executed in one hour) → machine P (process Z)." After verification that these permutations satisfy the machine state boundaries and process constraints, they are included in the initial feasible solution set.

[0047] In step 3.2, in order to select a high-quality solution that takes into account multi-objective benefits from the initial feasible solution set, it is necessary to carry out multi-objective evaluation and non-dominated solution set screening in an orderly manner. The specific steps are as follows: The first step is to determine the multi-objective evaluation dimensions and weights. For each feasible solution in the initial feasible solution set, the weighted evaluation values ​​of the three core objectives of delivery time achievement rate, equipment comprehensive efficiency and energy consumption coefficient are clearly calculated in parallel. The delivery time achievement rate is used to measure the punctuality of order delivery. It needs to be calculated in combination with the delivery time window of the work order (for example, the order requires delivery within 5 days, whether the completion time of the order in the production schedule is within this time window) and the time matching of the production schedule; the equipment comprehensive efficiency considers the equipment's utilization rate, capacity utilization rate, etc., and reflects the utilization level of equipment resources. For example, during the execution of the production schedule, the actual working time of the equipment is calculated to account for the ratio of the planned working time, as well as the ratio of the actual output of the equipment to the theoretical maximum output. The energy consumption coefficient is used to calculate the total energy consumption of the equipment under the production schedule, which is related to the production cost and green manufacturing requirements, such as the electricity consumption and gas consumption of the equipment during operation. These three types of goals need to be based on the company's production strategy. For example, if the company currently focuses on quickly delivering market orders, or focuses on controlling production costs, different optimization weight factors can be assigned by pre-setting weights (such as using a hierarchical scoring method to give a higher weight to delivery-priority companies for delivery achievement rates; using a linear weighting method to comprehensively consider the importance of each goal to the company), so that the evaluation results are in line with the company's actual production needs.

[0048] The second step is to perform non-dominated solution set screening. The multi-objective evaluation values ​​calculated for each feasible solution are compared pairwise. If there exists a solution X that outperforms solution Y in at least one of the three objective dimensions of delivery time achievement rate, equipment overall efficiency, and energy consumption coefficient, and is not inferior to Y in all other dimensions, then solution X is said to "dominate" solution Y. By traversing all feasible solutions in the initial feasible solution set, a solution set is screened for all objective dimensions that are not dominated by other solutions, namely the non-dominated solution set.

[0049] In this specific embodiment, assume that there are three initial feasible solutions: feasible solution A, feasible solution B, and feasible solution C. Feasible solution A has a delivery time achievement rate of 90%, an overall equipment efficiency of 80%, and an energy consumption coefficient of 0.7; feasible solution B has a delivery time achievement rate of 85%, an overall equipment efficiency of 85%, and an energy consumption coefficient of 0.6; and feasible solution C has a delivery time achievement rate of 88%, an overall equipment efficiency of 82%, and an energy consumption coefficient of 0.65. Comparing feasible solutions A and B, feasible solution A is superior to B in terms of delivery time achievement rate (90%>85%), and B is superior to A in terms of equipment comprehensive efficiency (85%>80%) and energy consumption coefficient (0.6<0.7). The two have advantages in each other's dimensions and cannot dominate each other. Comparing feasible solutions A and C, A is superior to C in terms of delivery time achievement rate (90%>88%), and C is superior to A in terms of equipment comprehensive efficiency (82%>80%) and energy consumption coefficient (0.65<0.7), and they cannot dominate each other. Comparing feasible solutions B and C, B is superior to C in terms of delivery time achievement rate (85%<88%, here B is inferior to C), equipment comprehensive efficiency (85%>82%), and energy consumption coefficient (0.6<0.65) in terms of equipment comprehensive efficiency and energy consumption coefficient, but is inferior to C in terms of delivery time achievement rate, so they cannot dominate each other. Therefore, these three feasible solutions are retained in the non-dominated solution set, forming a high-quality candidate pool for subsequent production scheduling decisions. Through this screening, a set of solutions with greater multi-objective benefit advantages is focused on, laying a solid foundation for the final determination of the production scheduling plan.

[0050] In step 3.3, in order to determine the production schedule that best meets the company's production needs from the non-dominated solution set, the non-dominated solution set needs to be normalized and weighted to select the best solution. The operation is as follows: First, normalization is performed. Because the delivery deadline achievement rate, equipment overall efficiency, and energy consumption coefficient have different dimensions and value ranges (for example, the delivery deadline achievement rate is usually expressed as a percentage, while the energy consumption coefficient may be a specific energy consumption value), in order to more fairly and accurately evaluate different feasible solutions across multiple objectives, the multi-objective evaluation values ​​of each feasible solution in the non-dominated solution set must be normalized. Using a specific normalization algorithm (such as linear normalization, which maps each objective value to the interval [0, 1]), the dimensionality effect is eliminated, making the evaluation values ​​of different objectives comparable. For example, for the delivery time achievement rate, if the achievement rate of a feasible solution is 90%, and the maximum value of all feasible solutions under this goal is 95% and the minimum value is 80%, then it is normalized to 0.67 using the formula (90%-80%) ÷ (95%-80%); for the energy consumption coefficient, if the energy consumption coefficient of the feasible solution is 0.6, and the maximum value under this goal is 0.8 and the minimum value is 0.4, then it is normalized to (0.8-0.6) ÷ (0.8-0.4) = 0.5 (here it is assumed that the smaller the energy consumption coefficient, the better, and reverse normalization is used). And so on, the normalization of other goals such as the overall efficiency of the equipment is completed.

[0051] Next, a weighted selection process is performed. After normalization, the normalized evaluation values ​​of each feasible solution are weighted and summed, combining the company's preset optimization weighting factors for each objective (e.g., if a company prioritizes delivery, a higher weight is assigned to the delivery deadline achievement rate; if a company prioritizes cost, a higher weight is assigned to the energy consumption coefficient). The weighted evaluation value is calculated. The solution with the optimal weighted evaluation value (i.e., the one that best aligns with the company's production strategy after considering multiple objectives) is selected and finalized as the production schedule.

[0052] In this example, assume that the non-dominated solution set contains two feasible solutions: solution D and solution E. After normalization, feasible solution D has a normalized delivery time achievement ratio of 0.8, a normalized equipment overall efficiency of 0.7, and a normalized energy consumption coefficient of 0.6. Feasible solution E has a normalized delivery time achievement ratio of 0.7, a normalized equipment overall efficiency of 0.8, and a normalized energy consumption coefficient of 0.5. In the company's production strategy, a weight of 0.4 is assigned to the delivery time achievement ratio, a weight of 0.3 to the equipment overall efficiency, and a weight of 0.3 to the energy consumption coefficient. Therefore, the weighted evaluation value of feasible solution D is 0.8 × 0.4 + 0.7 × 0.3 + 0.6 × 0.3 = 0.32 + 0.21 + 0.18 = 0.71; the weighted evaluation value of feasible solution E is 0.7 × 0.4 + 0.8 × 0.3 + 0.5 × 0.3 = 0.28 + 0.24 + 0.15 = 0.67. Because feasible solution D has a higher weighted evaluation value and is more consistent with the company's production strategy, feasible solution D is selected as the final production schedule. This will clarify the specific plan guiding production execution and ensure that the production schedule achieves optimal corporate benefits while balancing multiple objectives.

[0053] In a preferred embodiment of the present invention, the above step 4 may include: Step 4.1: Receive the production schedule execution data stream in real time and analyze the work order task sequence, equipment allocation plan, and process parameter settings contained therein; Step 4.2: Perform anomaly detection based on the real-time sensor data of the equipment state vector. Compare the actual equipment load rate with the upper load threshold defined in the process constraint matrix. If the actual value continuously exceeds the threshold, it is marked as an equipment overload anomaly. Compare the deviation between the real-time monitored value of the process parameter and the planned value. If the deviation exceeds the preset tolerance range, it is marked as a process deviation anomaly. Step 4.3: Start diversion for abnormal tasks, extract the geometric features of the original transportation path, model the new path as a two-dimensional line segment, and calculate the spatial relationship with the activated path in real time; In step 4.4, dynamic instructions are generated based on the spatial relationship detection results. When a path intersection is detected, the detour path cost is calculated based on the path load-bearing level and the work order priority weight. The non-intersection path with the lowest cost is selected to generate a logistics avoidance instruction. The process parameters are adjusted to within the threshold range through a closed-loop feedback mechanism. When there is no path intersection, the original transportation path is maintained, and adaptive calibration of the process parameters is performed to obtain an updated task instruction containing the adjusted path and parameters.

[0054] In an embodiment of the present invention, multiple values ​​are achieved by constructing a complete process of "accurate identification of anomalies - diversion path modeling - dynamic instruction generation": first, the production scheduling execution data is analyzed in real time, and anomalies are detected based on the dual dimensions of equipment status and process constraints to solve the problems of delayed anomaly discovery and fuzzy judgment in traditional production, and accurately locate equipment overload and process deviations; then, the abnormal tasks are diverted, the path characteristics are extracted and the spatial relationship is modeled and detected, and the detour path is dynamically planned based on the path load and work order priority to avoid logistics conflicts and improve the resilience of the production system; finally, instructions are generated by scenario, and the cost is weighed to select the optimal path and closed-loop parameter adjustment when there is intersection. When there is no intersection, the path is maintained and the parameters are adaptively calibrated to ensure the comprehensiveness of the abnormal response and avoid excessive adjustment and waste of resources.

[0055] In an embodiment of the present invention, the specific steps include: In step 4.1, to carry out subsequent anomaly detection and task adjustment, it is necessary to first analyze the production schedule execution data flow and clarify the production baseline information, as follows: The system continuously receives production scheduling execution data streams through a real-time data interaction interface with the production execution system. This data stream carries dynamic execution information at the production site and requires structured analysis to extract key production factors: Work order task sequence analysis: This function identifies the sequence of production processes within each work order from the data stream, as well as the corresponding product model, production stage, and other information. For example, for a battery production work order, the task sequence is analyzed as "cell loading → cell welding → cell packaging → finished product testing." This clarifies the timing of production execution and provides a foundation for subsequent tracking of task execution progress and identifying anomalies.

[0056] Equipment allocation plan analysis: Extract the production equipment number and type assigned to each process, as well as the planned start and end times for the equipment required to perform the process. For example, the "cell welding" process is analyzed to be performed by welding equipment numbered M-002, with a planned start time of 9:00 AM and a planned end time of 9:30 AM. This defines the equipment's workload baseline, which can be used to subsequently determine whether the equipment load is abnormal.

[0057] Process parameter setting value analysis: Obtain the target values ​​of the process parameters corresponding to each process step, such as the welding temperature must reach 250°C, the welding pressure must be maintained at 0.5MPa, etc. At the same time, extract the allowable tolerance range of the process parameters, such as the welding temperature tolerance of ±5°C, as the benchmark threshold for subsequent judgment of whether the process parameters are abnormal.

[0058] Through the above analysis operation, the dynamically changing production scheduling execution data flow is converted into structured correlation information of "work order-process-equipment-process parameters", providing clear "plan baseline value" and "constraint boundary" for subsequent anomaly detection links, ensuring that anomaly detection has a clear reference basis.

[0059] In step 4.2, to promptly detect deviations in the production execution process, it is necessary to combine the real-time sensor data of the equipment with the process constraints to conduct anomaly detection, as follows: The system uses IoT sensors deployed on production equipment to collect real-time sensor data of the equipment state vector, which mainly contains two key information: The actual equipment load rate is calculated by counting the number of tasks actually completed by the equipment per unit time and comparing it with the equipment's rated capacity. For example, if equipment M-002 is rated to complete 10 welding tasks per hour and has actually completed 12 tasks within the current hour, its actual load rate is calculated to be 120%, reflecting the equipment's actual workload.

[0060] Real-time monitoring values ​​of process parameters: Use temperature sensors, pressure sensors, etc. to obtain the actual values ​​of process parameters when the equipment executes the process in real time. For example, the real-time welding temperature of the M-002 welding equipment when performing the "battery core welding" process is monitored to be 260℃.

[0061] Based on the "planned baseline value" and "constraint boundary" obtained through analysis, two-dimensional anomaly detection is performed: Equipment overload anomaly detection: Compares the actual equipment load rate with the predefined "load threshold upper limit" in the process constraint matrix. For example, if the upper load threshold for equipment M-002 in the "cell welding" process is 110%, if its actual load rate (e.g., 120%) continuously exceeds this threshold and is determined to have exceeded the limit within a set time window (e.g., the load rate exceeds the threshold for three consecutive minutes), the equipment is marked as "equipment overload anomaly" and the corresponding abnormal equipment (M-002) and abnormal process (cell welding) are associated, allowing for precise location of the source of the anomaly.

[0062] Process Deviation Anomaly Detection: Calculates the deviation between the real-time monitored process parameter value and the planned value obtained from step 4.1. For example, for soldering temperature, the planned value is 250°C, while the real-time monitored value is 260°C, resulting in a calculated deviation of +10°C. If this deviation exceeds the preset tolerance (e.g., ±5°C), it is flagged as a "Process Deviation Anomaly" and the specific parameter type (soldering temperature) and deviation value (+10°C) are recorded, providing detailed information for subsequent exception handling.

[0063] Through the above-mentioned "real-time data comparison benchmark value, constraint boundary" operation, accurate identification and classification marking of equipment overload anomalies and process deviation anomalies can be achieved.

[0064] In step 4.3, when a production anomaly is detected, to prevent the impact of the anomaly from expanding, it is necessary to initiate diversion of the abnormal task and detect logistics path conflicts, as follows: Abnormal task diversion triggers: Based on the abnormal information marked in step 4.2, such as the "Equipment Overload Exception" caused by overload on equipment M-002, the task affected by the abnormality is precisely located, namely the "Battery Cell Welding" process currently or about to be executed by M-002 and subsequent tasks that depend on it. The device-level task diversion mechanism is then activated. These tasks originally assigned to M-002 are dispatched to backup equipment (such as welding equipment numbered M-003), or the execution sequence of the tasks is adjusted. This relieves the workload of the abnormal equipment and prevents the abnormality from further impacting the overall production schedule.

[0065] Path geometry extraction and modeling: For tasks involving material transport route adjustments due to task diversion, the geometric features of the original transport path are extracted based on the dynamic material topology data of the production workshop. For example, the original path "battery cells transported from the material storage area to the M-002 welding equipment" includes the starting coordinates (the material storage area's coordinates in the workshop coordinate system are X1, Y1), the ending coordinates (the coordinates of the M-002 equipment are X2, Y2), and the path width (assuming 1.5 meters). Furthermore, based on the new equipment allocation after the task diversion, the new material transport path, such as "battery cells transported from the material storage area to the M-003 welding equipment," is abstractly modeled as a two-dimensional line segment. This line segment starts at the material storage area's coordinates (X1, Y1) and ends at the M-003 equipment's coordinates (X3, Y3), forming line segment L for subsequent spatial relationship calculations.

[0066] Real-time calculation of spatial relationships: This involves calling the "activated path set" recorded in the dynamic material topology data of the production workshop, which is the logistics path information currently being used by other material transportation tasks. For example, the transportation path currently being used by the AGV from the "finished product storage area to the inspection area" is obtained, which is abstracted as line segment L1 in the workshop coordinate system (starting point coordinates X4, Y4, end point coordinates X5, Y5). The spatial relationship between the newly modeled path segment L and these activated path segments (such as L1) is then calculated in real time. By determining whether the line segments intersect, are parallel, or separate, it is determined whether there is a risk of "path intersection" between the new path and the existing activated path, providing a basis for the subsequent generation of dynamic instructions.

[0067] In step 4.4, based on the detection results of the path spatial relationship, dynamic instructions are generated by scenario to ensure the continuous and compliant execution of production tasks. The details are as follows: Scenario 1: Path intersection is detected (e.g., line segment L intersects line segment L1) Detour Path Cost Calculation and Instruction Generation: The system uses a pre-built path attribute database to obtain the load-bearing ratings of each logistics path within the workshop (for example, path P-001 has a load-bearing limit of 5 tons and can transport heavier materials; path P-002 has a load-bearing limit of 3 tons and is suitable for light loads), as well as the priority weights of each work order (for example, the current "cell welding" work order is urgent and has a high priority weight). Based on this information, a "path cost function" is constructed, which comprehensively considers factors such as the detour path's distance, its compatibility with the path's load-bearing rating, and its matching with the work order's priority, to calculate the costs of different detour paths. From the various possible non-intersecting paths, the path with the lowest cost is selected, such as path P-001. A "logistics avoidance instruction" is generated, stating, "The AGV should adjust its transport route and transport the cells from the material storage area to the M-003 welding equipment via path P-001." This ensures that material transportation avoids intersection conflicts and ensures smooth logistics.

[0068] Closed-loop adjustment of process parameters: A closed-loop feedback control mechanism is used to address abnormal process deviations. For example, if the monitored welding temperature is 260°C, which is 250°C higher than the planned value and the temperature tolerance is ±5°C, the system automatically adjusts parameters such as the welding equipment's heating power based on the temperature deviation (deviation +10°C) through a closed feedback loop, gradually lowering the welding temperature until it is within the threshold range specified by the process constraint matrix (e.g., 250°C ±5°C), ensuring process parameter compliance.

[0069] Scenario 2: No path intersection is detected (for example, line segment L is separated from the active path segment) Transport path maintenance and instruction generation: When it is detected that the new path does not intersect with the existing active path, the originally planned transport path (such as line segment L) is maintained unchanged, and a task diversion instruction is directly generated. The instruction content is "the AGV car follows the original planned path to transport the battery cells to the M-003 welding equipment and perform the 'battery cell welding' process after task diversion", ensuring that material transportation is carried out as planned.

[0070] Adaptive process parameter calibration: Adaptive control algorithms are used to automatically calibrate process parameter deviations. For example, a fuzzy control algorithm dynamically adjusts the equipment's process parameter settings based on the deviation between the real-time monitored process parameter and the planned value (e.g., a temperature deviation of +7°C). For example, the system gradually adjusts the heating time and current of welding equipment, automatically calibrating the process parameters to within the threshold range to ensure they meet production requirements.

[0071] In a preferred embodiment of the present invention, the above step 5 may include: Step 5.1: Based on the adjusted tasks and parameters, perform operator skill compatibility verification. Extract the skill qualification code set in the task process requirements, retrieve the skill matrix, which stores each operator's qualification certificate number, skill proficiency weight, and current task load, and calculate the skill compatibility. If the compatibility falls below the preset threshold, mark it as a conflicting task. Step 5.2: Generate a reallocation plan for the conflicting tasks. Construct a three-dimensional decision space of task, equipment, and operator. Traverse the set of alternative equipment that meets the process constraints. Calculate the operator skill matching for each alternative equipment. Select the combination with the highest matching degree that meets the equipment capability constraints as the reallocation plan. Step 5.3: Perform root cause analysis based on the equipment fault code and historical maintenance logs. Analyze the pattern category corresponding to the equipment fault code, retrieve the previous occurrence records of the same fault pattern in the historical maintenance logs, calculate the probability distribution of the fault root cause, and select the root cause with the highest probability and associate it with the current task to obtain the root cause analysis results. In step 5.4, based on the reallocation plan and the root cause analysis results, the task instruction set is reconstructed, the reallocation plan is written into the task equipment allocation field, the root cause code is written into the task exception record field, and the resource status database is updated.

[0072] In an embodiment of the present invention, through skill matching verification and conflicting task redistribution, accurate adaptation of operators and tasks is achieved, thus avoiding low production efficiency or quality problems caused by skill mismatch and improving the rationality of human resource allocation; with the help of equipment fault codes and historical maintenance logs for root cause analysis, the source of the fault can be accurately located, providing a basis for subsequent maintenance and prevention of similar faults, and reducing the interference of equipment failures on production; at the same time, based on the redistribution plan and the root cause analysis results, the task instruction set is reconstructed and the resource status database is updated, ensuring the real-time synchronization of production task information and resource status, and ensuring the coordination and efficiency of all links in the production system.

[0073] In an embodiment of the present invention, the specific steps include: In step 5.1, based on the adjusted task and parameters, key process requirements are extracted from the task instructions to identify the skill qualification codes required to perform the task. For example, to perform high-precision electronic component soldering, the "Advanced Electronic Soldering Skill Qualification Code E-001" is required.

[0074] Retrieves a pre-built skills matrix stored in the system database. This matrix includes each operator's qualification certificate number, the proficiency weight of each skill (a value between 0 and 1 representing mastery, with higher values ​​indicating greater proficiency), and the current task load (reflecting the proportion of working time consumed). For example, operator Li Si holds a welding qualification certificate numbered Z-003, with a welding skill proficiency weight of 0.8 and a current task load of 0.5.

[0075] Based on the skills matrix, for each operator, the skill qualifications required for the task are combined with their skill proficiency weight and the current task load, and the skill match is calculated using pre-set rules. The skill proficiency weight is first multiplied by (1 - current task load), and then corrected based on the precise match of the qualification certificate to obtain the match value.

[0076] The calculated skill matching degree is compared with a preset threshold (e.g., 0.7) set by the company based on production accuracy, task complexity, etc. If an operator's matching degree falls below the threshold, the task is marked as a conflicting task, paving the way for subsequent reallocation, ensuring production quality and efficiency and avoiding errors caused by skill mismatch.

[0077] In step 5.2, with the conflicting tasks as the core, sort out the process requirements, time constraints, material requirements, etc. of the tasks, and integrate the set of alternative equipment that can execute the tasks (which must meet process adaptability, such as a specific painting task requires equipment with corresponding spraying functions), as well as the operator's skills, load, and other information to build a three-dimensional decision space of task-equipment-operator.

[0078] Alternative equipment that meets the process constraints is screened from the equipment library to form a set. For each alternative equipment, an operator with qualified operation qualifications and a current task load that is not saturated (e.g., less than 0.8) is selected from the skill matrix.

[0079] For each operator corresponding to the alternative device, calculate the skill matching again according to the rules in step 5.1.

[0080] Compare the skill compatibility of operators for each alternative device and select the combination with the highest compatibility that meets the device's capacity constraints (e.g., the compatibility between device operation difficulty and operator skills, and the compatibility between work intensity and operator physical strength and time load). This is then used as the reallocation solution. For example, a conflicting task originally assigned to device A and operator Wang Wu could be assigned to device B and operator Zhao Liu after adjustment, ensuring a smooth production process.

[0081] In step 5.3, when the device triggers a fault code, the system analyzes the fault code and identifies the corresponding fault mode category. For example, fault code F-002 corresponds to "device circuit short circuit fault mode."

[0082] Based on the failure mode category, the historical maintenance log database is searched for the previous multiple (e.g., five) failures of the same mode to form a historical failure collection. The log records the time of occurrence, device number, repair measures, root cause, etc. of the past failures in detail.

[0083] Count the frequency of each root cause in the historical fault collection and calculate its probability distribution in the total record. For example, among five failures with the same mode, "circuit component aging" occurs three times with a probability of 0.6, while "external power supply abnormality" occurs two times with a probability of 0.4.

[0084] The root cause with the highest screening probability (such as "circuit component aging") is associated with the current task to obtain the root cause analysis results, providing a basis for formulating repair plans and optimizing equipment maintenance strategies, quickly resolving faults and reducing production impacts.

[0085] In step 5.4, based on the reallocation plan (clarifying the new execution equipment and operator) and the root cause analysis results (determining the root cause of the fault, such as "circuit component aging causing equipment failure"), sort out the fields that need to be modified in the task instruction set, such as the task equipment allocation field and the task exception record field (where the root cause code and description need to be written).

[0086] According to the system's pre-set data format and specifications, the reallocation information and root cause analysis results are written to the corresponding fields in the task instruction set. For example, the original device code is replaced with the new one, and the root cause code G-001 and the description "failure due to circuit component aging" are added to the exception record field to complete the reconstruction.

[0087] The reconstructed task instruction set information is synchronously written to the resource status database. This database records real-time equipment status (e.g., whether it's repaired or idle), operator task load (e.g., Zhao Liuyin adds a new task, which requires an update), and task execution status (the conflicting task's new status is "pending"), ensuring the database accurately reflects the latest resource status.

[0088] In a preferred embodiment of the present invention, the above step 6 may include: Step 6.1: Collect network communication quality data of edge computing nodes in real time, determine network connectivity status by continuously analyzing transmission stability indicators, and output connectivity mark signals; In step 6.2, when the marking signal indicates network connectivity, the system extracts the real-time equipment occupancy status and material flow trajectory based on the updated task instruction set and resource status database. It then identifies the affected work orders that compete with the changed task for equipment resources or have process dependencies. The system then recalculates the process time windows for the affected work orders based on the equipment release time nodes and logistics buffer requirements, dynamically reconstructing the work order execution timing in the production schedule. In step 6.3, when the marking signal indicates a network outage, the three-layer real-time resource model replica stored in the edge node is activated. Based on the equipment state vector and process constraint matrix in the replica, an equipment allocation plan that maximizes processing efficiency is selected. A spatial avoidance path is generated based on the real-time occupancy status of the material topology data. Step 6.4 integrates the processing results of steps 6.2 and 6.3. In the network connectivity scenario, output a full production schedule including a timing adjustment plan. In the network interruption scenario, output an edge autonomous instruction set including device allocation instructions and path avoidance instructions.

[0089] In an embodiment of the present invention, intelligent adaptation and dynamic optimization of production scheduling are achieved by accurately determining the network status. When the network is connected, based on the latest task and resource data, the association between affected work orders is deeply mined, the timing is accurately reconstructed, and the guarantee plan fits the actual production situation; when the network is interrupted, the local resource model copy is activated, and autonomous instructions are quickly generated with efficiency and space avoidance as the guide to maintain production resilience. Regardless of the network status, an adaptation plan can be output to solve the problem of out-of-control production scheduling under network fluctuations. Through the fine timing adjustment of connected scenarios and the autonomous guarantee of interrupted scenarios, a balance between production continuity and plan optimization is achieved, allowing production plans to operate stably and intelligently in a complex network environment, helping to promote efficient collaboration in the entire production process.

[0090] In an embodiment of the present invention, the specific steps include: Step 6.1: To achieve accurate perception of network status and provide a basis for subsequent production scheduling strategy adjustments, the specific process is as follows: First, network communication quality data is collected in real time through edge computing nodes. The data includes but is not limited to data transmission rate, packet loss rate, communication delay time and other indicators that can reflect network transmission performance.

[0091] Secondly, the collected network communication quality data is continuously analyzed, focusing on evaluating transmission stability indicators. For example, by monitoring whether the packet loss rate within a certain time window is within the preset normal threshold range and whether the communication delay is within the acceptable fluctuation range, the current network connectivity status can be determined.

[0092] Based on the above determination, a corresponding connectivity signal is output. If network transmission is stable and all indicators meet normal communication requirements, a "network connected" signal is output. If network transmission is abnormal, such as packet loss exceeding a threshold or excessive communication delay, a "network disconnected" signal is output.

[0093] In step 6.2, when the network is connected, the production schedule needs to be dynamically adjusted to adapt to the latest production dynamics in real time. The specific process is as follows: First, based on the updated task instruction set and resource status database, key information is extracted from it, including the current real-time occupancy status of each device (such as whether the device is executing a task, the progress of the executed task, etc.) and the flow trajectory of materials in the workshop (such as the path and time node for transporting materials from one workstation to another).

[0094] Next, based on the extracted information, the affected work order set associated with the change task is identified. Specifically, if a work order and the change task require the use of the same equipment, indicating a competition for equipment resources; or if the execution of a work order's process requires the completion of the change task's process, indicating a process dependency, then the work order is included in the affected work order set.

[0095] Then, for the affected work order set, combined with the release time nodes of the relevant equipment (that is, the time when the equipment can be put into new tasks after completing the current task) and the logistics buffer requirements required for materials during transportation (such as material temporary storage time, time required for transportation route switching, etc.), the time window of each process of each affected work order is recalculated to clarify the earliest start time and latest end time of each process.

[0096] Finally, based on the recalculated process time window, the original work order execution timing relationship in the production schedule is dynamically reconstructed, and the execution order and time schedule of the affected work orders are adjusted to ensure that the production schedule can match the current equipment status and material flow conditions, thereby improving the feasibility and accuracy of the plan.

[0097] In step 6.3, when the network is interrupted, to ensure the continuity of the production process, it is necessary to rely on edge nodes to implement localized production scheduling decisions. The specific process is as follows: First, once a marker signal indicating "network interruption" is received, a three-layer real-time resource model copy pre-stored in the edge node is immediately activated. This copy contains key resource information such as equipment state vector, process constraint matrix, and material topology data, which can provide data support for localized decision-making.

[0098] Secondly, based on the equipment state vectors in the model replica (reflecting the equipment's idle state, load capacity, etc.) and the process constraint matrix (specifying the compatibility between processes and equipment, process parameter limits, etc.), a device allocation plan that maximizes processing efficiency is selected while satisfying all process constraints. For example, tasks are prioritized to equipment that is currently idle and most compatible with the task's process requirements, reducing task wait times and improving overall processing efficiency.

[0099] At the same time, based on the real-time occupancy status recorded by the material topology data in the model copy (such as which transportation paths are being used, the cache status of material nodes, etc.), combined with the workshop space layout, spatial avoidance paths are generated for material transportation to avoid path conflicts during material transportation and ensure the smooth flow of materials.

[0100] Through the above operations, when the network is interrupted and unable to communicate with the central system, the edge node can independently complete equipment allocation and path planning to maintain normal production.

[0101] In step 6.4, to ensure that the production scheduling instructions under different network conditions can accurately and timely guide production execution, the above processing results need to be integrated and output. The specific process is as follows: First, the dynamic reconstruction results of the production schedule generated when the network is connected in step 6.2 and the edge autonomous decision results generated when the network is disconnected in step 6.3 are summarized and organized to ensure that the processing results in both scenarios are fully incorporated into the output system.

[0102] Secondly, depending on the network connectivity status, the corresponding instruction content is output: in the network connectivity scenario, a full production schedule including a timing adjustment plan is output. This plan covers the latest execution timing, equipment allocation and material flow arrangements of all work orders, and can comprehensively guide the collaborative operation of the production system; in the network interruption scenario, an edge autonomous instruction set including equipment allocation instructions and path avoidance instructions is output. This instruction set focuses on production tasks within the jurisdiction of the edge node, ensuring that the task allocation and execution can be completed independently locally.

[0103] This scenario-based result output method enables the production system to obtain accurate production scheduling guidance under different network conditions, ensuring the stability and continuity of the production process.

[0104] like Figure 2 As shown, an embodiment of the present invention further provides a digital factory flexible scheduling system based on MES, including: The acquisition module is used to capture order change instructions in real time through the integrated interface between MES and ERP and convert them into standardized work order data streams; The construction module is used to integrate real-time equipment sensor data with pre-stored process constraint matrices and dynamic material topology data based on structured data objects to generate a three-layer real-time resource model containing equipment state vectors, process constraint matrices, and material topology data. Hierarchical optimization is performed based on the three-layer real-time resource model. The first layer of optimization uses the equipment state vectors and process constraint matrices as hard constraints to generate an initial feasible solution set. The second layer of optimization performs multi-objective evaluation on the initial feasible solution set to obtain a non-dominated solution set as the production scheduling plan. The detection module is used to activate the equipment-level task diversion mechanism based on the production anomaly detection results during the production scheduling process. Based on the material topology data, the transportation path is abstracted into two-dimensional line segments for spatial intersection detection. The module generates logistics avoidance instructions and process parameter adaptive adjustment instructions to obtain the adjusted tasks and parameters. The output module is used to adjust tasks and parameters, verify operator skill matching, redistribute conflicting tasks, and conduct root cause analysis based on equipment fault codes and historical maintenance logs to obtain an updated task instruction set and resource status database; The processing module is used to fine-tune the production schedule in real time when the network is connected based on the updated task instruction set and resource status database, combined with the edge computing node status, and maintain the optimized production scheduling function based on the local resource model copy when the network is interrupted.

[0105] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A flexible production scheduling method for a digital factory based on MES, characterized in that: The method comprises: Step 1: Capture order change instructions in real time through the integrated interface between MES and ERP and convert them into standardized work order data streams; Step 2: Based on the structured data object, the real-time sensor data of the equipment is integrated with the pre-stored process constraint matrix and dynamic material topology data to generate a three-layer real-time resource model including the equipment state vector, process constraint matrix and material topology data; Step 3: Perform hierarchical optimization based on the three-layer real-time resource model. The first-layer optimization uses the equipment state vector and process constraint matrix as hard constraints to generate an initial feasible solution set. The second-layer optimization performs multi-objective evaluation on the initial feasible solution set to obtain a non-dominated solution set as the production schedule. Step 4: Based on the production anomaly detection results during the production scheduling process, the equipment-level task diversion mechanism is activated. Based on the material topology data, the transportation path is abstracted into two-dimensional line segments for spatial intersection detection. Logistics avoidance instructions and process parameter adaptive adjustment instructions are generated to obtain the adjusted tasks and parameters. Step 5: Based on the adjusted tasks and parameters, verify the operator's skill compatibility, redistribute conflicting tasks, and conduct root cause analysis based on equipment fault codes and historical maintenance logs to obtain an updated task instruction set and resource status database. Step 6: Based on the updated task instruction set and resource status database, combined with the edge computing node status, the production schedule is fine-tuned in real time when the network is connected, and the optimized production schedule function is maintained based on the local resource model copy when the network is interrupted.

2. A flexible production scheduling method for a digital factory based on MES according to claim 1, characterized in that: Through the integrated interface between MES and ERP, order change instructions are captured in real time and converted into standardized work order data flows, including: Step 1.1: Capture real-time order change instructions through the integrated interface between MES and ERP; Step 1.2: Parse the key business elements in the order change instruction, including order identifier, product specifications, delivery time window, process version, and priority level; In step 1.3, the key business elements are mapped into structured data objects, which include the work order unique identifier, bill of materials fingerprint, process path code, time constraint set and optimization weight factor.

3. A flexible production scheduling method for a digital factory based on MES according to claim 2, characterized in that: Based on structured data objects, the real-time sensor data of the equipment is integrated with the pre-stored process constraint matrix and dynamic material topology data to generate a three-layer real-time resource model containing the equipment state vector, process constraint matrix and material topology data, including: Step 2.1: Receive the output standardized work order data stream and parse the structured data objects therein; Step 2.2: Based on the process path encoding of the structured data object, a pre-stored process constraint matrix is ​​called. The matrix includes the sequence constraints between processes, the equipment-process binding relationship, and the process parameter threshold set. Step 2.3: Based on the BOM fingerprint of the structured data object, dynamic material topology data is constructed. This data defines the physical coordinate set of the material nodes, the geometric parameters of the traversable paths between the nodes, and the real-time occupancy status of each path. Step 2.4: Fusion of the device’s real-time sensor data with the time constraint set of the structured data object, optimization of the weight factors, and generation of the device state vector. In step 2.5, the process constraint matrix, material topology data, and equipment state vector are integrated to form a three-layer real-time resource model.

4. A flexible production scheduling method for a digital factory based on MES according to claim 3, characterized in that: Hierarchical optimization is performed based on a three-layer real-time resource model. The first layer of optimization uses the equipment state vector and process constraint matrix as hard constraints to generate an initial feasible solution set. The second level of optimization performs multi-objective evaluation on the initial feasible solution set and obtains a non-dominated solution set as the production schedule, including: Step 3.1: Extract the equipment state vector and process constraint matrix from the three-layer real-time resource model, and construct a constrained optimization model with process sequence as decision variables, equipment state as boundary conditions, and process constraints as hard rules. Generate an initial feasible solution set by enumerating feasible process permutations and combinations that satisfy all hard constraints. In step 3.2, a multi-objective evaluation is performed on the initial feasible solution set. For each feasible solution, weighted evaluation values ​​of the delivery date achievement rate, equipment overall efficiency, and energy consumption coefficient are calculated in parallel. By comparing the objective function values ​​of each solution pairwise, a solution set is selected in which all objective dimensions are not dominated. In step 3.3, normalize the non-dominated solution set and select the solution with the best weighted evaluation value as the final production schedule.

5. A flexible production scheduling method for a digital factory based on MES according to claim 4, characterized in that: Based on the production anomaly detection results during the production scheduling process, the equipment-level task diversion mechanism is activated. Based on the material topology data, the transportation path is abstracted into two-dimensional line segments for spatial intersection detection. Logistics avoidance instructions and process parameter adaptive adjustment instructions are generated, and the adjusted tasks and parameters are obtained, including: Step 4.1: Receive the production schedule execution data stream in real time and analyze the work order task sequence, equipment allocation plan, and process parameter settings contained therein; Step 4.2: Perform anomaly detection based on the real-time sensor data of the equipment state vector. Compare the actual equipment load rate with the upper load threshold defined in the process constraint matrix. If the actual value continuously exceeds the threshold, it is marked as an equipment overload anomaly. Compare the deviation between the real-time monitored value of the process parameter and the planned value. If the deviation exceeds the preset tolerance range, it is marked as a process deviation anomaly. Step 4.3: Start diversion for abnormal tasks, extract the geometric features of the original transportation path, model the new path as a two-dimensional line segment, and calculate the spatial relationship with the activated path in real time; In step 4.4, dynamic instructions are generated based on the spatial relationship detection results. When a path intersection is detected, the detour path cost is calculated based on the path load-bearing level and the work order priority weight. The non-intersection path with the lowest cost is selected to generate a logistics avoidance instruction. The process parameters are adjusted to within the threshold range through a closed-loop feedback mechanism. When there is no path intersection, the original transportation path is maintained, and adaptive calibration of the process parameters is performed to obtain an updated task instruction containing the adjusted path and parameters.

6. A flexible production scheduling method for a digital factory based on MES according to claim 5, characterized in that: Based on the adjusted tasks and parameters, the operator skill match is verified, conflicting tasks are redistributed, and root cause analysis is performed based on equipment fault codes and historical maintenance logs. This results in an updated task instruction set and resource status database, including: Step 5.1: Based on the adjusted tasks and parameters, perform operator skill compatibility verification. Extract the skill qualification code set in the task process requirements, retrieve the skill matrix, which stores each operator's qualification certificate number, skill proficiency weight, and current task load, and calculate the skill compatibility. If the compatibility falls below the preset threshold, mark it as a conflicting task. Step 5.2: Generate a reallocation plan for the conflicting tasks. Construct a three-dimensional decision space of task, equipment, and operator. Traverse the set of alternative equipment that meets the process constraints. Calculate the operator skill matching for each alternative equipment. Select the combination with the highest matching degree that meets the equipment capability constraints as the reallocation plan. Step 5.3: Perform root cause analysis based on the equipment fault code and historical maintenance logs. Analyze the pattern category corresponding to the equipment fault code, retrieve the previous occurrence records of the same fault pattern in the historical maintenance logs, calculate the probability distribution of the fault root cause, and select the root cause with the highest probability and associate it with the current task to obtain the root cause analysis results. In step 5.4, based on the reallocation plan and the root cause analysis results, the task instruction set is reconstructed, the reallocation plan is written into the task equipment allocation field, the root cause code is written into the task exception record field, and the resource status database is updated.

7. A flexible production scheduling method for a digital factory based on MES according to claim 6, characterized in that: Based on the updated task instruction set and resource status database, combined with the edge computing node status, the production schedule can be fine-tuned in real time when the network is connected. In the event of a network outage, the optimized production schedule can be maintained based on the local resource model copy, including: Step 6.1: Collect network communication quality data of edge computing nodes in real time, determine network connectivity status by continuously analyzing transmission stability indicators, and output connectivity mark signals; In step 6.2, when the marking signal indicates network connectivity, the system extracts the real-time equipment occupancy status and material flow trajectory based on the updated task instruction set and resource status database. It then identifies the affected work orders that compete with the changed task for equipment resources or have process dependencies. The system then recalculates the process time windows for the affected work orders based on the equipment release time nodes and logistics buffer requirements, dynamically reconstructing the work order execution timing in the production schedule. In step 6.3, when the marking signal indicates a network outage, the three-layer real-time resource model replica stored in the edge node is activated. Based on the equipment state vector and process constraint matrix in the replica, an equipment allocation plan that maximizes processing efficiency is selected. A spatial avoidance path is generated based on the real-time occupancy status of the material topology data. Step 6.4 integrates the processing results of steps 6.2 and 6.

3. In the network connectivity scenario, output a full production schedule including a timing adjustment plan. In the network interruption scenario, output an edge autonomous instruction set including device allocation instructions and path avoidance instructions.

8. A digital factory flexible production scheduling system based on MES, which implements the method according to any one of claims 1 to 7, characterized in that: include: The acquisition module is used to capture order change instructions in real time through the integrated interface between MES and ERP and convert them into standardized work order data streams; A construction module is used to fuse real-time device sensor data with pre-stored process constraint matrices and dynamic material topology data based on structured data objects to generate a three-layer real-time resource model containing device state vectors, process constraint matrices, and material topology data. A hierarchical optimization is performed based on the three-layer real-time resource model. The first layer of optimization uses the device state vectors and process constraint matrices as hard constraints to generate an initial feasible solution set. The second level optimization performs multi-objective evaluation on the initial feasible solution set and obtains the non-dominated solution set as the production schedule; The detection module is used to activate the equipment-level task diversion mechanism based on the production anomaly detection results during the production scheduling process. Based on the material topology data, the transportation path is abstracted into two-dimensional line segments for spatial intersection detection. The module generates logistics avoidance instructions and process parameter adaptive adjustment instructions to obtain the adjusted tasks and parameters. The output module is used to adjust tasks and parameters, verify operator skill matching, redistribute conflicting tasks, and conduct root cause analysis based on equipment fault codes and historical maintenance logs to obtain an updated task instruction set and resource status database; The processing module is used to fine-tune the production schedule in real time when the network is connected based on the updated task instruction set and resource status database, combined with the edge computing node status, and maintain the optimized production scheduling function based on the local resource model copy when the network is interrupted.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images