A method, apparatus, and medium for manufacturing execution interaction of an energy storage thermal management unit

Abstract

The application provides a manufacturing execution interaction method of an energy storage thermal management unit, relates to the technical field of industrial manufacturing execution, and realizes efficient cooperation and intelligent decision of the manufacturing process of the energy storage thermal management unit by integrating three function modules of production planning, quality monitoring and scheduling simulation. A unique product serial number is allocated to each unit, and a product-level digital main line is constructed to support full-process tracing. In terms of quality monitoring, visual images and torque-angle time sequence data of key assembly stations are collected in real time, and a lightweight convolutional neural network model is used to generate structured quality judgment results online, which significantly improves the accuracy and timeliness of defect identification. Meanwhile, the candidate scheduling scheme is verified by rolling simulation in the digital twin body synchronized with the physical production line, and the optimal scheme is only executed when the preset scheduling requirements are met, thereby enhancing the production scheduling executability and effectively improving the overall manufacturing efficiency and product quality consistency.

Description

Technical Field

[0001] This application relates to the field of industrial manufacturing execution technology, and in particular to a manufacturing execution interaction method, device and medium for an energy storage thermal management unit. Background Technology

[0002] Against the backdrop of the rapid development of the energy storage industry, energy storage thermal management units, as key equipment to ensure the safety and performance of battery systems, are characterized by multi-model mixed-line production, complex processes, and stringent quality requirements. Existing manufacturing execution systems generally suffer from data silos, making it difficult to effectively connect information across material, process, equipment, and quality inspection stages, resulting in difficulties in tracing the entire product lifecycle. Simultaneously, traditional production scheduling relies heavily on static rules or manual experience, lacking real-time perception and response capabilities to dynamic constraints such as equipment status and material availability, resulting in insufficient flexibility and difficulty in handling disturbances such as order insertions and equipment failures. Furthermore, critical production and quality inspection data are easily tampered with, lacking reliable evidence storage mechanisms, which restricts the reliability of quality control and compliance traceability. Summary of the Invention

[0003] This application provides a manufacturing execution interaction method, device, and medium for energy storage thermal management units to solve one or more technical problems existing in the prior art, and at least provides a beneficial option or creates conditions to achieve real-time connection and reliable storage of data throughout the entire manufacturing process, thereby improving production flexibility, quality traceability, and system collaboration efficiency.

[0004] On the one hand, this application provides a manufacturing execution interaction method for an energy storage thermal management unit, including the following steps: The main interface for manufacturing execution of the energy storage thermal management unit is displayed; wherein, the main interface includes production planning controls, quality monitoring controls, and scheduling simulation controls; In response to the trigger command of the production planning control, the production planning sub-interface is displayed, a unique product serial number is assigned to each unit, and a product-level digital master line including orders, materials, processes, equipment and personnel is constructed based on the product serial number; In response to the trigger command of the quality monitoring control, the quality monitoring sub-interface is displayed, and visual images and torque-angle time series data of the assembly station are collected in real time and input into the lightweight convolutional neural network model to generate structured quality judgment results online; the assembly station includes a sealant coating station, a connector crimping station and a structural fastening screw fastening station. In response to the trigger command of the scheduling simulation control, the scheduling simulation sub-interface is displayed. The scheduling constraints are updated based on the quality judgment result. Combined with the production line disturbance event, the scheduling engine generates candidate scheduling schemes. The candidate scheduling schemes are then simulated in a rolling simulation in a digital twin synchronized with the physical production line. When the simulation result meets the preset scheduling requirements, the optimal scheduling scheme that meets the preset scheduling requirements is sent to the manufacturing execution system for execution.

[0005] Furthermore, the production planning sub-interface includes a work order information entry control, a serial number rule configuration control, a BOM and process binding control, and a digital master line saving control; In the production planning sub-interface, a unique product serial number is assigned to each unit, and a product-level digital master line including orders, materials, processes, equipment, and personnel is constructed based on the product serial number, including the following steps: In response to the trigger command of the work order information entry control, the work order information entry window is displayed, and the customer order number, product model, planned quantity and delivery date are entered to generate basic work order data; In response to the trigger command of the serial number rule configuration control, the serial number rule configuration window is displayed to set the encoding structure of the product serial number, including the prefix field, the year-month-day field and the serial number field, so as to generate a unique product serial number generation rule; In response to the trigger command of the BOM and process binding control, the BOM and process binding window is displayed, the BOM and standard process route corresponding to the product model are associated, and the generated unique product serial number is used as the primary key to automatically construct a product-level digital master line including the basic work order data, material batch information, process list, equipment resource pool and personnel skill matrix; In response to the trigger command of the digital master line saving control, the product-level digital master line is saved, and the production task bound to the unique product serial number is pushed to the manufacturing execution system backend.

[0006] Furthermore, the quality monitoring sub-interface includes a visual acquisition configuration control, a tightening data access control, and a quality judgment display area; In the quality monitoring sub-interface, visual images and torque-angle time-series data of the assembly station are collected in real time and input into a lightweight convolutional neural network model to generate structured quality judgment results online, including the following steps: In response to the trigger command of the vision acquisition configuration control, the vision acquisition configuration window is displayed, and the resolution, frame rate and region of interest of the industrial camera at the assembly station are set to generate vision data acquisition rules. In response to the trigger command of the tightening data access control, a tightening data access window is displayed to configure the protocol type and data field mapping relationship for communication with the tightening device, so as to establish a real-time access channel for torque-angle timing data. After completing the above configuration, the visual images and torque-angle time series data of the assembly station are received in real time in the quality monitoring sub-interface and input into the lightweight convolutional neural network model to generate structured quality judgment results online. The quality judgment results include the sealant coating continuity score, the confidence level of the connector position status, and the screw fastening abnormality type. In response to a refresh event of the quality judgment display area, the quality judgment result and its corresponding original data snapshot are dynamically displayed, including glue line image, connector partial view and torque-angle curve; The quality judgment display area includes result review controls, which allow operators to perform manual review and select to confirm approval, reject, or initiate manual re-inspection. In response to the confirmation operation, the quality judgment result is marked as valid and bound to the corresponding product serial number for updating the scheduling constraints of the scheduling engine.

[0007] Furthermore, the lightweight convolutional neural network model includes a semantic segmentation branch unit, an object detection branch unit, and a temporal classification branch unit; the online generation of structured quality assessment results includes the following steps: The real-time visual image of the sealant coating station is input into the semantic segmentation branch unit, which outputs the glue line pixel mask and calculates the glue width standard deviation and the number of break points per unit length based on the mask to generate a sealant coating continuity score. The real-time visual image of the connector crimping station is input to the target detection branch unit, which outputs the bounding box and center coordinates of the connector and socket, and generates the confidence level of the connector's position status based on their relative pose and a preset alignment threshold. The real-time torque-angle timing data of the structural fastening screw fastening station is input into the timing classification branch unit. The torque rise slope, final torque value and angle increment features are extracted and mapped to screw fastening anomaly types in combination with the process window. The anomaly types include underload, overload, stripping or false fit.

[0008] Furthermore, the scheduling simulation sub-interface includes a schedule generation control, a simulation start control, and a schedule distribution control; In the scheduling simulation sub-interface, scheduling constraints are updated based on the quality judgment results. Combined with production line disturbance events, a candidate scheduling scheme is generated using the scheduling engine. This candidate scheduling scheme is then subjected to rolling simulation in a digital twin synchronized with the physical production line. When the simulation results meet the preset scheduling requirements, the optimal scheduling scheme is sent to the manufacturing execution system for execution. This includes the following steps: In the scheduling simulation sub-interface, production line disturbance events from the manufacturing execution system are automatically received. These production line disturbance events include equipment fault alarms, abnormal material availability signals, or emergency order insertion instructions. In response to the trigger command of the scheduling generation control, the scheduling engine is invoked to generate a candidate scheduling scheme that includes normal production tasks and rework tasks based on the scheduling constraints corresponding to the currently valid quality judgment results and the production line disturbance events. In response to the trigger command of the simulation start control, the candidate scheduling scheme is simulated in a rolling simulation in a digital twin that is synchronized with the physical production line in real time, and the simulation results are dynamically generated and displayed, including the predicted value of capacity utilization and the distribution of work-in-process inventory. In response to the trigger command of the scheduling control, when the simulation results meet the preset scheduling requirements, the optimal scheduling scheme is sent to the manufacturing execution system for execution.

[0009] Furthermore, the invocation of the scheduling engine, based on the scheduling constraints corresponding to the currently valid quality judgment results and the production line disturbance events, generates a candidate scheduling scheme that includes normal production tasks and rework tasks, including the following steps: Based on the anomaly type and severity in the quality judgment results, corresponding process jump rules, rework path assignments or resource priority adjustment instructions are generated as scheduling constraints. Based on the scheduling constraints, identify the set of product serial numbers that need to skip a specific process or be redirected to a rework station; Based on the aforementioned production line disturbance events, update the equipment availability time window, material readiness time, and human resource status; The product serial number set, the updated resource status and the work orders to be executed are input into the scheduling engine, which generates multiple candidate scheduling schemes through reinforcement learning strategy. Each scheme includes the process allocation, workstation assignment and timing arrangement of each product. The scheduling simulation sub-interface lists and displays key indicators for each candidate scheduling scheme, including estimated completion time, bottleneck workstation utilization rate, and rework task ratio.

[0010] Furthermore, the step of performing rolling simulations of the candidate scheduling schemes in a digital twin synchronized in real time with the physical production line, and dynamically generating and displaying the simulation results, includes the following steps: Load a digital twin model that is synchronized with the physical production line status in real time, and inject the candidate scheduling scheme into the model as a driving input; Run multi-cycle rolling simulations to calculate the predicted capacity utilization rate of each workstation and the distribution of work-in-process inventory in the buffer area in real time; In the scheduling simulation sub-interface, the predicted capacity utilization rate and the distribution of work-in-process inventory are dynamically displayed in the form of visual charts, and whether the preset capacity utilization rate threshold or the work-in-process safety limit is highlighted. In response to the simulation completion event, the scheduling distribution control is automatically activated, allowing users to select candidate solutions that meet the scheduling requirements for distribution.

[0011] Furthermore, the main interface also includes a product standard operating procedure (SOP) control, which displays the product standard operating procedure (SOP) interface in response to a trigger command on the product SOP control. On the product standard operation guidance interface, when a product is detected to have been transferred to the current workstation, the corresponding graphic operation guidance content is automatically loaded based on the product's unique serial number, including process description, wiring diagram, interface definition and debugging steps; In response to the scanning operation of the material or process confirmation code, the operator's identity, operation start and end time and scanning event are recorded on the product standard operation guidance interface and bound to the data main line corresponding to the product serial number; When the current workstation is an assembly workstation, it will automatically jump to the quality monitoring sub-interface and simultaneously display the quality judgment results associated with the product serial number.

[0012] On the other hand, this application provides an electronic device including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned manufacturing execution interaction method for an energy storage thermal management unit.

[0013] On the other hand, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned manufacturing execution interaction method for an energy storage thermal management unit.

[0014] The beneficial effects of this application are: This application provides a manufacturing execution interaction method for energy storage thermal management units. This method integrates three major functional modules: production planning, quality monitoring, and scheduling simulation, to achieve efficient collaboration and intelligent decision-making in the manufacturing process of energy storage thermal management units. This method provides all core controls uniformly in the main interface, facilitating users' quick access to different business scenarios. By assigning a unique product serial number to each unit and constructing a product-level digital masterline covering orders, materials, processes, equipment, and personnel, it establishes a complete data link, supporting end-to-end traceability. In terms of quality monitoring, it collects visual images and torque-angle time-series data in real time from key assembly stations such as sealant coating, connector crimping, and structural fastening screw fastening. It also uses a lightweight convolutional neural network model to generate structured quality judgment results online, significantly improving the accuracy and timeliness of defect identification. Simultaneously, this method integrates quality judgment results as dynamic constraints into the scheduling simulation process. Combined with production line disturbance events, it performs rolling simulation verification of candidate scheduling schemes in a digital twin synchronized with the physical production line. The optimal scheme is only issued and executed when the preset scheduling requirements are met, thereby enhancing the flexibility, robustness, and executability of production scheduling and effectively improving overall manufacturing efficiency and product quality consistency.

[0015] This application also provides devices and media corresponding to the method, which have similar beneficial effects to the method described above, and will not be described in detail here.

[0016] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the description, claims and drawings. Attached Figure Description

[0017] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the technical solutions of the present invention, and do not constitute a limitation on the technical solutions of the present invention.

[0018] Figure 1 This is a flowchart of the manufacturing execution interaction method for the energy storage thermal management unit provided in this application; Figure 2 This is a schematic diagram of the main interface for manufacturing execution interaction of the energy storage thermal management unit provided in this application; Figure 3 This is a schematic diagram of the production planning sub-interface provided in this application; Figure 4 This is a schematic diagram of the quality monitoring sub-interface provided in this application; Figure 5 This is a schematic diagram of the scheduling simulation sub-interface provided in this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0020] The present application will be further described below with reference to the accompanying drawings and specific embodiments. The described embodiments should not be considered as limitations on the present application, and all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present application.

[0021] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0023] As the global energy structure accelerates its transformation towards cleaner and lower-carbon energy sources, electrochemical energy storage systems are widely used in grid peak shaving, new energy infrastructure support, and industrial and commercial backup power. As a core subsystem ensuring the safe operation and stable performance of energy storage batteries, the energy storage thermal management unit undertakes the critical task of precise temperature control of battery modules and preventing thermal runaway. Its manufacturing process is characterized by multi-model mixed-line production, complex assembly processes, high requirements for quality consistency, and tight delivery cycles, posing a significant challenge to the intelligence level of the manufacturing execution system.

[0024] Currently, the traditional Manufacturing Execution Systems (MES) widely used in the industry are mainly based on static Bills of Materials (BOMs) and fixed process routes for production planning and scheduling. Steps such as material kitting checks, process flow, and quality inspection rely on manual input or semi-automatic data collection, leading to information delays and severe data fragmentation. At the production planning level, most systems only support coarse-grained scheduling based on orders or work orders, lacking a comprehensive understanding of constraints such as real-time equipment status, personnel skill matching, and dynamic material arrival. Once disturbances such as order insertions, equipment failures, or material shortages occur, manual intervention is often required to readjust the schedule, resulting in slow response times and a high risk of errors, making it difficult to meet the needs of flexible manufacturing.

[0025] In terms of quality management, existing technologies mostly employ sampling or final inspection methods. Quality control of critical assembly processes, such as sealant application, connector crimping, and screw fastening, still heavily relies on operator experience and offline testing methods. For example, issues such as whether the sealant is applied continuously and evenly, whether connectors are fully crimped, and whether screw fastening torque meets standards are typically discovered through visual inspection or post-production sampling, resulting in high missed detection rates, delayed feedback, and the inability to achieve closed-loop traceability. Even when some production lines introduce sensors to collect torque or image data, the lack of efficient edge intelligence analysis capabilities makes real-time online judgment and automatic interception difficult, leading to defective products flowing into the next process or even leaving the factory, creating safety hazards.

[0026] Furthermore, current manufacturing execution systems (MES) commonly suffer from data silos. Interfaces and data standards are inconsistent between subsystems such as production planning, material management, process execution, equipment monitoring, and quality inspection, making it difficult to effectively integrate data across the entire product lifecycle. When quality issues require tracing, it often necessitates manually querying and piecing together data across multiple systems, which is inefficient and prone to errors. More seriously, critical production and quality inspection records lack tamper-proof mechanisms, making it difficult to provide a credible chain of evidence during compliance audits or liability determination.

[0027] In terms of scheduling optimization, although some advanced enterprises have attempted to introduce digital twin technology for production line simulation, existing solutions are mostly offline static simulations. The simulation model is not synchronized with the physical production line status and cannot reflect dynamic disturbances such as equipment failures, staff absences, or quality anomalies in real time. At the same time, the simulation results are disconnected from the actual execution system, and the optimized scheduling plan still needs to be manually imported, making it difficult to form a closed-loop control of "perception-decision-execution-verification," thus limiting the practicality and adaptability of the scheduling strategy.

[0028] In summary, existing technologies have significant shortcomings in areas such as data connectivity, real-time quality control capabilities, scheduling flexibility, and reliable traceability mechanisms, making it difficult to support the high reliability, high flexibility, and high quality intelligent manufacturing requirements of energy storage thermal management units. Therefore, there is an urgent need for a new manufacturing execution interaction method that can deeply integrate the product digital pipeline, edge intelligent quality inspection, and dynamic scheduling simulation to achieve end-to-end collaborative optimization from planning to execution to feedback, providing key technological support for the digital transformation of the energy storage equipment manufacturing industry.

[0029] To address the aforementioned issues, this application constructs a manufacturing execution interaction method, equipment, and medium for energy storage thermal management units. Through a unified main interface, it integrates three core functional modules: production planning, quality monitoring, and scheduling simulation, achieving visualization, collaboration, and intelligence throughout the entire manufacturing process. This method assigns a unique product serial number to each unit, and uses this as a basis to connect order, material, process, equipment, and personnel information, forming a complete product-level digital thread and breaking down data silos. Vision and torque-angle multimodal sensors are deployed at key assembly stations, combined with a lightweight convolutional neural network model to achieve real-time online quality judgment for processes such as sealant coating, connector crimping, and screw fastening, significantly improving defect identification accuracy and response speed. Simultaneously, structured quality results are integrated as dynamic constraints into the scheduling engine. Combined with production line disturbance events, candidate scheduling schemes are subjected to rolling simulation verification in a digital twin synchronized with the physical production line status. The manufacturing execution system is only driven to implement the current scheme when it meets the preset scheduling requirements, thereby achieving highly flexible, robust, and closed-loop verifiable intelligent scheduling and execution control.

[0030] In the technical solution of this application, the real-time linkage between the digital twin and the physical production line is achieved through a bidirectional data channel between the Manufacturing Execution System (MES) and the edge control layer of the production line. Equipment, sensors, and industrial control systems on the physical production line continuously collect data on operating status, work progress, material flow, personnel operations, and quality inspection, and upload this data to the MES in real time via industrial protocols such as OPC UA, Modbus TCP, or MQTT. The MES synchronously injects this dynamic data into the virtual model of the digital twin, driving its state, behavior, and timing to remain consistent with the physical production line, ensuring that the virtual space accurately maps the current state of the physical entity.

[0031] Digital twins not only passively reflect reality but also possess proactive feedback capabilities. The scheduling simulation module starts from the current real-world state, freezing key variables (such as equipment status, work-in-process location, and material inventory) within the digital twin, and conducts rolling simulations of candidate scheduling schemes. If bottlenecks, conflicts, or timeout risks are detected, the scheme's issuance is blocked; if feasibility is verified, the optimal scheduling instruction is sent to the production line control system via the MES to adjust task sequence, equipment start / stop, or process parameters. After the simulation, the digital twin automatically rolls back to the latest physical state and continues synchronization. This mechanism, relying on high-frequency, low-latency data synchronization and a unified data model, constructs an integrated closed loop of "perception—analysis—decision—execution," making the digital twin a reliable testing ground and execution verifier for scheduling decisions.

[0032] First, the manufacturing execution interaction method of the energy storage thermal management unit provided in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0033] Reference Figures 1 to 5The implementation process of the manufacturing execution interaction method for the energy storage thermal management unit provided in this application embodiment includes, but is not limited to, the following steps.

[0034] Step S110: Display the main interface 100 for manufacturing execution of the energy storage thermal management unit.

[0035] The main interface 100 includes a production planning control 101, a quality monitoring control 102, and a scheduling simulation control 103.

[0036] In step S110, a centralized and integrated operation entry point is provided to the user. This is achieved by displaying the main interactive interface 100 for the manufacturing execution of the energy storage thermal management unit, which intuitively presents the three most critical functional modules in the manufacturing process as controls. The main interface 100 includes production planning controls 101, quality monitoring controls 102, and scheduling simulation controls 103, each corresponding to a core business dimension of manufacturing execution. This allows operators to quickly switch to the appropriate functional scenario based on actual needs, avoiding frequent switching between multiple independent systems or pages, thereby improving human-computer interaction efficiency and operational continuity. This step serves as the starting point of the entire method, laying the foundation for the orderly triggering and collaborative operation of subsequent sub-functions.

[0037] In step S120, in response to the trigger command of the production planning control 101, the production planning sub-interface 200 is displayed, a unique product serial number is assigned to each unit, and a product-level digital master line including orders, materials, processes, equipment and personnel is constructed based on the product serial number.

[0038] In step S120, a unique identifier and information carrier for the entire product lifecycle data association is established. When the user triggers the production planning control 101, the system displays the production planning sub-interface 200 and assigns a unique product serial number to each energy storage thermal management unit. This serial number serves not only as a digital identity identifier for the physical product but also as a core index for data aggregation, used to construct a product-level digital masterline encompassing elements such as order information, bill of materials, process routes, equipment configuration, and operators. Through this masterline, static and dynamic data that were originally scattered across different business systems can be structured and organized around a single product instance, enabling data traceability, correlation, and queryability throughout the entire process from order acceptance to delivery, providing a complete and consistent data foundation for subsequent quality analysis and scheduling optimization.

[0039] In step S130, in response to the trigger command of the quality monitoring control 102, the quality monitoring sub-interface 300 is displayed, and the visual images and torque-angle time series data of the assembly station are collected in real time and input into the lightweight convolutional neural network model to generate structured quality judgment results online.

[0040] The assembly stations include a sealant application station, a connector crimping station, and a structural fastening screw fastening station.

[0041] In step S130, real-time perception and intelligent judgment of the quality status of key assembly processes are achieved. After the user triggers the quality monitoring control 102, the system displays the quality monitoring sub-interface 300 and simultaneously collects visual images and torque-angle time-series data for the three assembly stages that have a decisive impact on product reliability: the sealant application station, the connector crimping station, and the structural fastening screw fastening station. These multimodal raw data are input in real time into a pre-trained lightweight convolutional neural network model, which outputs structured quality judgment results online, such as whether the sealant application is continuous, the crimping is in place, or the fastening is qualified. This step transforms the traditional quality control method that relies on manual visual inspection or offline sampling into automated, real-time, and quantitative online quality inspection based on edge intelligence, significantly improving the defect detection rate and process control capabilities.

[0042] In step S140, in response to the trigger command of the scheduling simulation control 103, the scheduling simulation sub-interface 400 is displayed. The scheduling constraints are updated based on the quality judgment results. Combined with the production line disturbance event, the scheduling engine generates candidate scheduling schemes. The candidate scheduling schemes are then simulated in a rolling simulation in a digital twin synchronized with the physical production line. When the simulation results meet the preset scheduling requirements, the optimal scheduling scheme that meets the preset scheduling requirements is sent to the manufacturing execution system for execution.

[0043] In step S140, the dynamic optimization and reliable execution of the manufacturing scheduling strategy are realized. When the user triggers the scheduling simulation control 103, the system displays the scheduling simulation sub-interface 400, and uses the quality judgment result generated in step S130 as the key input to dynamically update the constraints required for production scheduling, such as removing work-in-process with quality problems or adjusting their priority. At the same time, the system combines the currently occurring production line disturbance events, such as equipment downtime, personnel absence, or material delays, and uses the scheduling engine to generate multiple candidate scheduling schemes. These schemes are not directly issued, but are subjected to rolling simulation in a digital twin synchronized with the physical production line status in real time to verify their feasibility and timeliness in a real environment. Only when the simulation results meet the preset scheduling requirements, such as the delivery date achievement rate, equipment utilization rate, or work-in-process accumulation threshold, will the system officially issue the optimal scheduling scheme to the manufacturing execution system for execution, thereby ensuring the scientific nature, robustness, and executability of the scheduling decision.

[0044] In some embodiments of this application, reference is made to Figure 3The production planning sub-interface 200 includes a work order information entry control 201, a serial number rule configuration control 202, a BOM and process binding control 203, and a digital master line saving control 204. In step S120, in the production planning sub-interface 200, a unique product serial number is assigned to each unit, and a product-level digital master line including orders, materials, processes, equipment, and personnel is constructed based on the product serial number, including the following steps.

[0045] In step S210, in response to the trigger command of the work order information entry control 201, the work order information entry window is displayed, and the customer order number, product model, planned quantity and delivery date are entered to generate basic work order data.

[0046] In step S210, the business source and planning input basis for the manufacturing task are established. When the user triggers the work order information entry control 201, the system displays the work order information entry window, guiding the user to input key business parameters such as customer order number, product model, planned quantity, and delivery date. This information together constitutes the basic work order data, which not only clarifies the source, target, scale, and time requirements of this production task, but also provides the necessary input basis for subsequent serial number generation, material preparation, process configuration, and production scheduling. This step ensures that manufacturing execution activities always revolve around real customer needs, avoiding planning deviations caused by missing or incorrect information.

[0047] Step S220: In response to the trigger command of the serial number rule configuration control 202, the serial number rule configuration window is displayed, and the encoding structure of the product serial number is set, including the prefix field, the year-month-day field and the serial number field, in order to generate a unique product serial number generation rule.

[0048] In step S220, the logic for generating product identification is defined to ensure that each energy storage thermal management unit has a globally unique and resolvable digital identity throughout its entire lifecycle. Upon responding to the trigger command of the serial number rule configuration control 202, the system displays the serial number rule configuration window, allowing the user to set the coding structure of the product serial number. This includes a prefix field to identify the product category or factory code, a year-month-day field to record the production plan start time, and a serial number field to distinguish different individuals within the same batch. Through this rule, the system can automatically generate unique product serial numbers that conform to the enterprise coding standards and are conflict-free, providing a standardized identification foundation for subsequent data association, traceability queries, and system integration.

[0049] In step S230, in response to the trigger command of the BOM and process binding control 203, the BOM and process binding window is displayed, the BOM and standard process route corresponding to the product model are associated, and the generated unique product serial number is used as the primary key to automatically construct a product-level digital master line including basic work order data, material batch information, process list, equipment resource pool and personnel skill matrix.

[0050] In step S230, the structured aggregation and primary key binding of product-level multi-source manufacturing elements are achieved. After the user triggers the BOM and process binding control 203, the system displays the BOM and process binding window, associating the bill of materials and standard process routes corresponding to the current product model. Based on this, the system uses the unique product serial number generated in step S220 as the primary key, automatically integrates the basic work order data from step S210, and further incorporates dynamic manufacturing elements such as material batch information, process list, equipment resource pool, and personnel skill matrix to construct a complete product-level digital masterline. This masterline takes a single product as its core, unifying and organizing information originally scattered across different dimensions such as planning, procurement, process, equipment, and manpower to form a full-element data view oriented towards individual products.

[0051] In step S240, in response to the trigger command of the digital master line saving control 204, the product-level digital master line is saved, and the production task bound to the unique product serial number is pushed to the manufacturing execution system backend.

[0052] In step S240, the persistent storage and task distribution of the product-level digital masterline are completed, realizing the formal handover from the planning layer to the execution layer. When the user triggers the digital masterline save control 204, the system saves the completed product-level digital masterline to the database, ensuring that it can be called by subsequent modules such as quality monitoring, scheduling simulation, and traceability analysis. At the same time, the system pushes the production task bound to this unique product serial number to the manufacturing execution system backend, enabling the physical production line control system to obtain execution instructions containing complete context information. This step marks the transition of the digital masterline from the configuration phase to the operation phase, laying the data and task foundation for the accurate execution and real-time feedback of the subsequent manufacturing process.

[0053] In some embodiments of this application, reference is made to Figure 4 The quality monitoring sub-interface 300 includes a visual acquisition configuration control 301, a tightening data access control 302, and a quality judgment display area 303. In step S130, in the quality monitoring sub-interface 300, visual images and torque-angle time-series data of the assembly station are acquired in real time, input into a lightweight convolutional neural network model, and structured quality judgment results are generated online, including the following steps.

[0054] In step S310, in response to the trigger command of the vision acquisition configuration control 301, the vision acquisition configuration window is displayed, and the resolution, frame rate and region of interest of the industrial camera at the assembly station are set to generate vision data acquisition rules.

[0055] In step S310, a standardized and configurable image acquisition foundation is established for visual quality inspection. When the user triggers the visual acquisition configuration control 301, the system displays the visual acquisition configuration window, allowing the resolution, frame rate, and region of interest of the industrial camera to be set for key assembly locations such as the sealant coating station, connector crimping station, and structural fastening screw fastening station. These parameters together constitute the visual data acquisition rules, ensuring that the acquired images meet the clarity and detail required for defect identification while avoiding redundant data and processing burdens caused by invalid areas. This step, through the pre-defined and refined acquisition strategy, provides a high-quality, highly relevant input source for subsequent image analysis, which is a prerequisite for achieving stable and reliable visual quality inspection.

[0056] In step S320, in response to the trigger command of the tightening data access control 302, the tightening data access window is displayed, and the protocol type and data field mapping relationship for communication with the tightening device are configured to establish a real-time access channel for torque-angle timing data.

[0057] In step S320, a real-time data path is established between the tightening equipment and the quality monitoring system. Upon responding to the trigger command of the tightening data access control 302, the system displays a tightening data access window, allowing the user to configure the protocol type used for communication with the field tightening equipment, such as Modbus TCP or Profinet, and define the mapping relationship of key data fields such as torque value, angle value, and timestamp in the communication message. Through this configuration, the system can automatically parse the raw signals output by the equipment and construct a continuous, complete, and semantically clear torque-angle time-series data stream. This step ensures accurate capture and timing alignment of process parameters, providing structured and computable data support for subsequent quality judgment based on mechanical behavior.

[0058] Step S330: After completing the above configuration, the visual images and torque-angle timing data of the assembly station are received in real time in the quality monitoring sub-interface 300, and input into the lightweight convolutional neural network model to generate structured quality judgment results online.

[0059] The quality assessment results include the sealant coating continuity score, the confidence level of the connector placement status, and the type of screw tightening abnormality.

[0060] In step S330, the fusion analysis and intelligent judgment of multimodal sensing data are realized. After completing the configuration of the vision and tightening data channels, the system continuously receives real-time visual images and torque-angle time-series data from each assembly station on the quality monitoring sub-interface 300, and synchronously inputs them into a pre-deployed lightweight convolutional neural network model. After training, this model can extract key criteria for different station features, and finally output structured quality judgment results online, including sealant coating continuity score, connector positioning confidence level, and screw tightening anomaly type. This step transforms the raw sensor data into quality conclusions with clear engineering semantics, realizing a leap from "visible" to "accurate judgment," and is the core intelligent link of the entire quality monitoring process.

[0061] In step S340, in response to a refresh event of the quality judgment display area 303, the quality judgment result and its corresponding original data snapshot are dynamically displayed, including the glue line image, the connector partial view and the torque-angle curve.

[0062] The quality judgment display area 303 includes a result review control 304, which supports operators to perform manual review and select to confirm approval, reject, or initiate manual re-inspection.

[0063] In step S340, a human-machine collaborative quality result visualization and interactive review mechanism is provided. After the quality judgment result is generated, the system responds to the refresh event of the quality judgment display area 303, dynamically displaying each judgment result. In addition, the quality judgment display area 303 also displays snapshots of the original data corresponding to each judgment result, including sealant images used to assess the continuity of the adhesive line, magnified views of the connector parts used to determine the insertion status, and torque-angle curves reflecting the characteristics of the fastening process. The display area also integrates a result review control 304. In response to triggering this control, a result review window pops up, allowing the operator to manually review the system output and original evidence, and choose to confirm approval, reject, or initiate a manual re-inspection based on the actual situation. This step retains the efficiency of artificial intelligence while incorporating the operator's judgment, forming an interpretable, interventionable, and traceable quality decision-making closed loop.

[0064] In step S350, in response to the confirmation operation, the quality judgment result is marked as valid and bound to the corresponding product serial number for updating the scheduling constraints of the scheduling engine.

[0065] In step S350, the quality assessment result is formally incorporated into the manufacturing execution data stream and drives downstream scheduling logic. After the operator confirms the successful operation, the system marks the current quality assessment result as valid and binds it to the corresponding product serial number. This binding relationship makes the product's current quality status a reliable fact within the manufacturing execution system, which is then read in real-time by the scheduling engine to update the constraints required for production scheduling, such as allowing qualified products to proceed to the next process or excluding potentially risky work-in-process from participating in high-priority task scheduling. This step achieves a seamless transformation of quality information into production decisions, making quality control a true intrinsic driving force for manufacturing scheduling rather than a reactive remedial measure.

[0066] In some embodiments of this application, the lightweight convolutional neural network model includes a semantic segmentation branch unit, an object detection branch unit, and a temporal classification branch unit. Step S330, generating structured quality assessment results online, includes the following steps.

[0067] Step S410: Input the real-time visual image of the sealant coating station into the semantic segmentation branch unit, output the glue line pixel mask, and calculate the standard deviation of glue width and the number of break points within a unit length based on the mask to generate a sealant coating continuity score.

[0068] In step S410, the sealant coating quality is evaluated at the pixel level. This step inputs real-time visual images from the sealant coating station into the semantic segmentation branch unit of a lightweight convolutional neural network model. This unit outputs a pixel-accurate sealant line mask, clearly identifying the areas in the image that belong to the sealant. Based on this, the system calculates the standard deviation of the sealant width per unit length using this mask to measure coating uniformity and counts the number of sealant line breakpoints to reflect continuity defects. Finally, these two quantitative indicators are combined to generate a sealant coating continuity score, providing an objective and quantifiable basis for judging whether the coating process meets the standards, effectively overcoming the problems of high subjectivity and low accuracy of traditional manual visual inspection.

[0069] Optionally, the sealant coating continuity score is calculated according to the following formula: ,in, The standard deviation of the adhesive width per unit length. The number of break points. and For example, preset weighting coefficients Set to 25. Set to 10. When or hour, Set it directly to 0 to trigger a forced return for repair.

[0070] For example, the system analyzes the acquired visual images and calculates a sealant coating continuity score of 83 points (out of 100). This score is calculated based on a standard deviation of 0.28 mm for the glue width per unit length and a break point of 1, indicating that there is a slight discontinuity in the glue line, but it is still within an acceptable range.

[0071] Step S420: Input the real-time visual image of the connector crimping station to the target detection branch unit, output the bounding box and center coordinates of the connector and socket, and generate the confidence level of the connector's position status based on their relative pose and preset alignment threshold.

[0072] In step S420, intelligent determination of the connector crimping status is achieved through geometric relationship analysis. This step sends the real-time visual image acquired at the connector crimping station to the target detection branch unit, which identifies and outputs the bounding boxes of the connector and its corresponding socket, along with their respective center coordinates. The system further calculates the actual alignment degree based on the relative position and orientation of these two key components and compares it with a preset alignment threshold to generate a numerical confidence level of the connector's positioning status. This confidence level reflects whether the connector has been fully and correctly crimped into the socket, providing a preliminary guarantee for electrical connection reliability and avoiding poor contact or safety hazards caused by loose connections or misalignment.

[0073] Optionally, the confidence level of the connector's in-place status Calculate according to the following rules: if the center offset distance between the connector and the socket is... (Preset alignment threshold, typical value is 1.5 pixels), then ,in This is the scaling factor (e.g., set to 0.063); if ,but The crimping was deemed unqualified.

[0074] For example, at the connector crimping station, the system detects that the center coordinates of the connector are (325.4, 180.7) pixels and the center coordinates of the socket are (326.1, 181.2) pixels. The offset distance between the two is 0.86 pixels, which is less than the preset alignment threshold of 1.5 pixels. Therefore, it is determined to be in place, and the corresponding position confidence is 94.6%.

[0075] Step S430: Input the real-time torque-angle timing data of the structural fastening screw fastening station into the timing classification branch unit, extract the torque rise slope, final torque value and angle increment features, and map them into screw fastening anomaly types in combination with the process window. Anomaly types include underload, overload, stripping or false fit.

[0076] In step S430, the screw fastening quality is accurately classified and diagnosed based on mechanical process characteristics. This step inputs real-time torque-angle timing data collected at the structural fastening screw fastening station into the timing classification branch unit. This unit automatically extracts key process features, including the slope of the torque rise phase, the final torque value, and the angle increment throughout the fastening process. These features collectively characterize the dynamic behavior of screw tightening and are mapped to a predefined process window for pattern matching, thereby determining whether the current fastening process is normal or abnormal. Furthermore, the abnormal type is explicitly categorized as underload, overload, stripping, or false fit. This step achieves in-depth analysis of fastening quality, not only determining "whether it is qualified" but also indicating "why it is unqualified," providing direct evidence for process optimization and fault tracing.

[0077] For example, at the screw fastening station, the system acquires torque-angle timing data: the final torque value is 5.12 N·m, the torque rise slope is 0.84 N·m / deg, and the angle increment is 215 deg, all of which fall within the process window range of [4.8–5.5 N·m], [0.7–1.0 N·m / deg], and [200–230 deg], and are identified as "normal".

[0078] Specifically, the torque range [4.8–5.5 N·m] corresponds to the locking requirements of M4 to M5 steel screws on aluminum alloy housings. This ensures sufficient locking force to prevent loosening during transportation, while avoiding excessive torque that could lead to thread stripping or housing cracking. The torque rise slope [0.7–1.0 N·m / deg] indicates smooth thread engagement without foreign object jamming or misalignment. A slope below 0.7 may indicate a risk of stripping, while a slope above 1.0 may indicate thread misalignment or misalignment. The angle increment [0.7–1.0 N·m / deg] corresponds to the rotation angle of the screw from contact with the housing to complete locking. An angle that is too small may result in false engagement, while an angle that is too large may cause thread stripping.

[0079] In some embodiments of this application, the lightweight convolutional neural network model adopts a shared backbone and multi-head architecture, which balances computational efficiency and multimodal perception capabilities, and is suitable for real-time quality inspection scenarios in industrial settings.

[0080] Specifically, the shared backbone network uses lightweight backbone networks such as MobileNetV3-small or EfficientNet-Lite0. Both are designed for resource-constrained devices and, through depthwise separable convolution or compound scaling strategies, can effectively extract general multi-level feature representations of input data (such as images or sensor signals) while significantly reducing the number of parameters and computational overhead.

[0081] Based on this main branch, the model connects three dedicated task branches in parallel: Semantic segmentation branch unit: Adopting the U-Net structure, its encoder part directly reuses the multi-scale feature map output by the backbone network to avoid redundant calculations; the decoder part consists of several transposed convolutional layers, and fuses the high-resolution features of shallow layers in the backbone network through skip connections, thereby accurately preserving the spatial details of small structures such as glue lines, and finally outputting a pixel-level mask of the same size as the input image, which is used to evaluate the continuity and integrity of sealant coating.

[0082] The object detection branch unit employs the YOLOv5s detection head structure, comprising a classification subnetwork and a regression subnetwork. This branch can quickly locate key components (such as connectors and sockets) in an image during a single forward inference pass, outputting their bounding box coordinates (x, y, w, h), class labels, and confidence scores. YOLOv5s, as a lightweight, single-stage detector, balances speed and accuracy, making it suitable for real-time visual guidance and position status determination on production lines.

[0083] Temporal classification branch unit: For torque-angle time series collected during assembly, a hybrid structure of 1D-CNN and bidirectional LSTM (BiLSTM) is adopted. 1D-CNN slides the convolutional kernel along the time axis to extract local dynamic patterns (such as abrupt changes and plateaus); BiLSTM models long-term dependencies in both forward and backward directions, comprehensively capturing the contextual features of assembly behavior; finally, a fully connected layer maps the features to probability distributions of anomaly types, including typical screw fastening defects such as "underload," "overload," "slippage," and "false fit."

[0084] During the training phase, each of the three branches calculates its own task-related loss (such as Dice loss for segmentation, localization and classification loss for detection, and cross-entropy loss for temporal classification), and then forms a joint loss function for multiple tasks through weighted summation. This joint loss function is then backpropagated to optimize the parameters of the shared backbone and each task head. During the inference phase, each branch can run independently and be invoked as needed, improving system flexibility and response efficiency.

[0085] This multi-task lightweight architecture achieves deep integration of visual (image) and sensor (time series) information, providing end-to-end, high-precision, and low-latency intelligent judgment capabilities for the quality of adhesive lines, the status of component arrival, and assembly anomalies in the manufacturing process of energy storage thermal management units.

[0086] In some embodiments of this application, reference is made to Figure 5The scheduling simulation sub-interface 400 includes a scheduling generation control 402, a simulation start control 404, and a scheduling distribution control 406. In step S140, in the scheduling simulation sub-interface 400, the scheduling constraints are updated based on the quality judgment results. Combined with production line disturbance events, candidate scheduling schemes are generated using the scheduling engine. The candidate scheduling schemes are then subjected to rolling simulation in a digital twin synchronized with the physical production line. When the simulation results meet the preset scheduling requirements, the optimal scheduling scheme is distributed to the manufacturing execution system for execution. This includes the following steps.

[0087] In step S510, the scheduling simulation sub-interface 400 automatically receives production line disturbance events from the manufacturing execution system. These events include equipment fault alarms, abnormal material availability signals, or emergency order insertion instructions.

[0088] In step S510, real-time perception and data access of dynamic disturbance factors in the manufacturing site are achieved. In the production line disturbance event display area 401 of the scheduling simulation sub-interface 400, the system automatically receives and displays production line disturbance events from the manufacturing execution system. These events specifically include equipment fault alarms, abnormal material availability signals, or emergency order insertion instructions. For example, during the manufacturing process of the energy storage thermal management unit, the system received a production line disturbance event from the manufacturing execution system: the automatic tightening equipment (equipment ID: ATU-03) at station 3 of the assembly line reported a "spindle motor overheating" fault at 14:22 on January 27, 2026, and the status changed from "running" to "stopping", with an estimated recovery time of 2 hours; at the same time, the current kitting rate of the liquid cooling plate assembly (material code: LCB-5502) for work order WO-20260127-089 is only 82%, which is lower than the set threshold of 95%; in addition, at 14:30 on the same day, an emergency order insertion instruction KTU-20260127-003 was received, requiring the delivery of 5 STMU-ProX energy storage thermal management units within 48 hours, with the priority marked as "highest".

[0089] By continuously monitoring and capturing such unplanned changes, the system can promptly grasp the key external variables affecting production scheduling feasibility, providing accurate and timely input for subsequent adjustments to the scheduling plan. This step ensures that scheduling decisions are no longer based on static ideal assumptions, but rather on an accurate reflection of the current actual operating status of the physical production line, which is a prerequisite for achieving highly robust scheduling.

[0090] Step S520: In response to the trigger command of the scheduling generation control 402, the scheduling engine is invoked to generate a candidate scheduling scheme that includes normal production tasks and rework tasks based on the scheduling constraints and production line disturbance events corresponding to the currently valid quality judgment results.

[0091] In step S520, a multi-task candidate scheduling scheme that takes into account both quality status and disturbance response is generated and displayed in the candidate scheduling scheme display area 403. When the user triggers the scheduling generation control 402, the system calls the scheduling engine and, based on the scheduling constraints corresponding to the currently valid quality judgment results, marks products with quality problems as needing rework or temporarily suspended, while integrating the production line disturbance events obtained in step S510 to comprehensively optimize the task sequence. The resulting candidate scheduling scheme not only includes regular normal production tasks but also explicitly incorporates rework tasks caused by quality anomalies, ensuring that the scheduling logic truly covers the actual needs of the entire manufacturing process and avoids resource conflicts or delivery delays caused by ignoring the rework flow in traditional production scheduling. For example, candidate scheduling scheme 01 is generated based on the above disturbance events and quality judgment results (such as product serial number STMU202601270045 needing rework). The plan includes: prioritizing the allocation of the 5 units of the urgent order to the normal workstations 1 and 2, skipping the faulty workstation 3, and adjusting the process path to workstation 1 → workstation 2 → quality inspection → packaging; the rework task STMU202601270045 is assigned to the dedicated rework station R1 and will be reassembled from 16:00 to 16:30 on the same day; the other 3 non-urgent products originally scheduled to be worked on at workstation 3 will be postponed to the next day's morning shift and reassigned to the idle time slot of workstation 1; the urgent order is expected to be completed by 13:20 the next day, meeting the 48-hour delivery requirement.

[0092] In step S530, in response to the trigger command of the simulation start control 404, the candidate scheduling scheme is simulated in a rolling simulation in a digital twin that is synchronized with the physical production line in real time, and the simulation results are dynamically generated and displayed, including the predicted value of capacity utilization and the distribution of work-in-process inventory.

[0093] In step S530, the scheduling scheme is used to verify its feasibility and pre-evaluate its performance in a high-fidelity virtual environment. After the user triggers the simulation start control 404, the system loads candidate scheduling schemes in a digital twin that is synchronized with the physical production line in real time and executes a rolling simulation process. The simulation dynamically simulates the execution effect of the scheme over a period of time, generates key performance indicators in real time, and visualizes them in the simulation result display area 405, including the predicted capacity utilization rate and the distribution of work-in-process inventory. For example, the simulation results show that the average utilization rate of the pre-assembly station (W1) is 76% throughout the day, reaching 92% during peak hours, with no overload; the average utilization rate of the assembly station (W2) is 84%, which rises to 96% instantaneously during the processing of rework tasks, but is still below the 98% warning threshold; the maximum inventory of buffer area B1 (W1→W2) is 6 units (occurring at 13:30), which is below the safe limit of 10 units, and the inventory accumulation is stable.

[0094] Through this process, users can intuitively judge whether a solution will lead to bottlenecks, congestion, or idle resources before implementation, thereby completing a quantitative assessment of scheduling quality in a virtual space and effectively avoiding production risks caused by blind deployment.

[0095] In step S540, in response to the trigger command of the scheduling control 406, when the simulation results meet the preset scheduling requirements, the optimal scheduling scheme is sent to the manufacturing execution system for execution.

[0096] In step S540, the validated optimal scheduling scheme is safely and controllably implemented at the physical execution layer. When the user triggers the scheduling deployment control 406, the system first verifies whether the current simulation results meet the preset scheduling requirements, such as capacity utilization not being lower than the threshold and work-in-process inventory not exceeding the limit. Only when all requirements are met is the optimal scheduling scheme from the simulation officially deployed to the manufacturing execution system for execution. This step establishes a closed-loop control mechanism of "simulation verification - compliance confirmation - command deployment," ensuring that only scheduling decisions that have been fully verified by the digital twin and possess high feasibility will drive the actual production line operation, thereby significantly improving the success rate of scheduling execution and the overall stability of the system.

[0097] In some embodiments of this application, step S520 involves invoking the scheduling engine to generate a candidate scheduling scheme that includes normal production tasks and rework tasks based on the scheduling constraints and production line disturbance events corresponding to the currently valid quality judgment results. This includes the following steps.

[0098] Step S610: Based on the anomaly type and severity in the quality judgment result, generate corresponding process jump rules, rework path assignments, or resource priority adjustment instructions as scheduling constraints.

[0099] In step S610, the quality assessment results are transformed into executable scheduling logic rules. Based on the anomaly type and severity contained in the quality assessment results, this step automatically generates corresponding process jump rules, rework path assignments, or resource priority adjustment instructions, formalizing them into constraints that the scheduling engine can recognize and process. For example, when a unit's screw fastening is found to have a stripped thread defect, the system can generate a jump rule to "skip subsequent high-voltage testing processes and redirect to rework station A"; if the sealant continuity score is below a threshold, it may trigger an instruction to "increase the resource priority of the rework station." This step achieves a precise mapping of quality information to scheduling semantics, enabling the scheduling logic to proactively adapt to the product quality status, rather than relying solely on preset static processes.

[0100] Optionally, the system pre-defines a quality-scheduling mapping rule base, for example: If the "screw fastening abnormality type" is "slipped thread" or "false fit", then generate the process jump rule: "Skip the subsequent high-pressure test process and redirect to the rework station R1". If the "Sealant Coating Continuity Score" is less than 80 points, a resource priority adjustment instruction will be generated: "Increase the scheduling priority of rework station R2 to the highest level". If the "confidence level of connector in place" is less than 90%, a rework path assignment is generated: "Assign to manual re-pressing station M3 for secondary crimping".

[0101] The above rules are stored in the database in the form of key-value pairs. After receiving a valid quality judgment result, the scheduling engine queries the corresponding anomaly type / score based on the product serial number and matches it with the rule base to generate specific scheduling constraint instructions.

[0102] Step S620: Based on scheduling constraints, identify the set of product serial numbers that need to skip specific processes or be redirected to rework stations.

[0103] In step S620, the specific product objects affected by the quality anomaly and their processing paths are identified. After obtaining the scheduling constraints generated in step S610, the system identifies a set of product serial numbers that need to skip specific normal processes or be redirected to rework stations. This set, with unique product serial numbers as the granularity, precisely identifies which individuals no longer follow the standard process route due to quality issues and instead enter a differentiated processing flow. This step ensures that the scheduling engine can implement special path management for these products in subsequent scheduling, avoiding the misscheduling of non-conforming products into subsequent critical workstations, thereby ensuring the compliance and efficiency of the overall production flow.

[0104] Optionally, the system iterates through all product serial numbers currently in work-in-process. If any anomaly is found in the quality assessment result associated with a particular product (e.g., the anomaly type is not "normal," or the continuity score is below a threshold), the product is added to the "Set of Product Serial Numbers to be Returned for Repair"; otherwise, it is added to the "Set of Product Serial Numbers in Normal Circulation." This set is maintained in real-time using a list or set data structure and serves as the input parameter for the scheduling engine.

[0105] Step S630: Based on the production line disturbance event, update the equipment availability time window, material readiness time, and human resource status.

[0106] In step S630, the availability status of manufacturing resources is dynamically refreshed to reflect the actual operating conditions of the production line. The system incorporates production line disturbance events from the manufacturing execution system, such as equipment failure alarms, delayed material arrivals, or temporary staff absences, to update the availability time windows of each piece of equipment, the actual readiness time of materials, and the current status of human resources in real time. These updated resource parameters constitute the basic boundary conditions for scheduling calculations, ensuring that candidate solutions do not assign tasks to downed equipment, workstations without materials, or unattended positions, thereby improving the feasibility and success rate of scheduling solutions.

[0107] In step S640, the product serial number set, the updated resource status, and the work orders to be executed are input into the scheduling engine, which generates multiple candidate scheduling schemes through reinforcement learning strategies. Each scheme includes the process allocation, workstation assignment, and timing arrangement for each product.

[0108] In step S640, the intelligent scheduling engine generates diverse and highly adaptable candidate scheduling schemes. The system inputs the product serial number set determined in step S620, the resource status updated in step S630, and the original work orders to be executed into the scheduling engine, which then performs multi-objective optimization based on its built-in reinforcement learning strategy. This strategy explores different combinations of process allocation, workstation assignment, and timing arrangements while satisfying various constraints through continuous trial and error and feedback learning, ultimately outputting multiple candidate scheduling schemes. Each scheme fully describes the entire process execution plan of all products (including normal products and rework products) on the production line, demonstrating the flexibility and intelligence of scheduling decisions.

[0109] Optionally, the reinforcement learning strategy is implemented using a Deep Q-Network (DQN) or PPO algorithm. Its state space includes: the current work order queue, equipment availability time window, material availability status, work-in-process location, and quality anomaly markers; the action space is a combination of assigning the next process, workstation, and time slot to each product; the reward function is designed as follows: ,in, To ensure on-time delivery rate, To improve equipment utilization, The repair task was delayed; among them , , , where represents the weighting coefficient. The scheduling engine samples the state-action-reward sequence by interacting with the digital twin, iteratively updates the policy network, and finally outputs multiple candidate scheduling schemes with high reward values.

[0110] In step S650, the key indicators of each candidate scheduling scheme are listed in the scheduling simulation sub-interface 400, including the estimated completion time, bottleneck workstation utilization rate and rework task ratio.

[0111] In step S650, operators are provided with an intuitive and comparable view of the scheduling options for evaluation. In the scheduling simulation sub-interface 400, the system displays key performance indicators for each candidate scheduling option in a list format, including estimated completion time, bottleneck workstation utilization, and rework percentage. These indicators quantify the merits of each option from three dimensions: delivery timeliness, resource efficiency, and quality load, helping users quickly understand the combined impact of different scheduling strategies. This step does not directly participate in schedule generation, but by presenting the decision-making basis in a structured manner, it supports the selection of the optimal solution under human-machine collaboration, enhancing the transparency and controllability of the scheduling process.

[0112] In some embodiments of this application, a highly responsive, highly reliable, and highly integrated edge-platform collaborative sensing architecture is constructed to optimize the real-time sensing of production line disturbance events, specifically including the following:

[0113] (1) Integrate the Manufacturing Execution System (MES) with the underlying equipment and control systems. Through unified industrial communication protocols (such as OPCUA, MQTT, or TSN) and standardized data interfaces, ensure that events such as equipment fault alarms, material status changes, and personnel operation triggers can be uploaded to the MES with millisecond-level latency. For critical equipment, an edge intelligent gateway should be deployed to clean, aggregate, and perform preliminary diagnosis of raw signals locally. Only structured disturbance events (such as "Equipment ID_07 shutdown: spindle overheating") should be efficiently pushed to the central system to avoid network congestion and invalid data interference.

[0114] (2) Establish a unified modeling and semanticization mechanism for multi-source heterogeneous disturbance events. Disturbances from different sources—such as equipment fault codes issued by PLCs, material kitting rates below the threshold reported by WMS, and emergency order insertion instructions from ERPs—have different formats and semantics. By predefining a disturbance event ontology model, various raw signals are mapped to standardized event objects with unified fields (such as event type, affected workstation, occurrence time, severity level, and associated product serial number), enabling the scheduling engine to parse and process them unambiguously, thereby improving the consistency and availability of perception.

[0115] (3) Introduce an event-driven stream processing architecture. Utilize Apache Kafka, Flink, or an industrial-grade stream computing engine to subscribe to, filter, correlate, and prioritize perturbation event streams in real time. For example, if the same workstation receives "material shortage" and "equipment standby" events consecutively within a short period, the system can automatically merge them into a "workstation blocking" composite event and increase its scheduling response priority. This stream processing capability enables the system not only to perceive perturbations but also to understand the logical relationships between them, thereby generating more accurate context awareness.

[0116] Fourth, enhance the spatiotemporal correlation capabilities of disturbance events. Bind disturbance events to their specific physical location (e.g., production line workstation coordinates), time window, and affected product batches or serial numbers. This allows the subsequent scheduling engine to accurately assess which tasks are directly affected by the disturbance when generating scheduling plans, avoiding global misjudgments. For example, only adjust the paths of work-in-process tasks flowing through faulty equipment, rather than completely offline and rescheduling them.

[0117] Finally, a closed-loop feedback and confidence verification mechanism for disturbance events is established. When the scheduling plan is adjusted due to a disturbance event, the system should continuously monitor whether the disturbance has been eliminated (e.g., equipment resumes operation, material shortages are replenished), and automatically trigger an event status update. Simultaneously, a confidence score is introduced for disturbance sources with high-frequency false alarms (e.g., frequent false alarms from a sensor), dynamically adjusting their weight in scheduling decisions to prevent noise interference from causing excessive scheduling oscillations.

[0118] In some embodiments of this application, reinforcement learning is introduced into the closed-loop feedback and confidence verification mechanism of disturbance events, which can improve the system's adaptability to dynamic manufacturing environments. Specifically, an intelligent agent with scheduling performance as the optimization objective can be constructed. Its state space includes the current disturbance event type, source device, historical false alarm frequency, impact range of related tasks, and production line operation indicators; the action space includes the handling strategy for disturbance events, such as "immediate response", "delayed confirmation", "downgraded processing", or "ignore".

[0119] This agent continuously interacts with the manufacturing execution environment, observing the actual results of different actions. For example, if production is suspended due to a sensor triggering a "material shortage" alarm, but it is later found that there is actually sufficient material, this response is considered a negative reward. Conversely, if a timely response to a real fault avoids batch defects, a positive reward is obtained. Based on such cumulative reward signals, reinforcement learning algorithms (such as policy gradient methods) can gradually learn the optimal response strategy for different disturbance sources in different contexts and dynamically generate confidence scores.

[0120] This scoring not only reflects the historical reliability of the disturbance source but also incorporates the contextual characteristics of the current operating conditions (such as equipment load rate, material inventory level, and order urgency), thereby enabling intelligent adjustment of the weight of disturbance events. When the confidence level of a sensor is lower than the threshold, the system can automatically mark its alarm as "pending verification," triggering only a warning without directly driving scheduling changes, until it is upgraded to a valid disturbance after cross-verification through multi-source data (such as combining visual recognition or manual confirmation).

[0121] In this way, reinforcement learning enables the disturbance perception mechanism to shift from passive reception to active judgment, effectively suppressing scheduling oscillations caused by noise or false alarms, while retaining a high sensitivity to real anomalies, ultimately achieving intelligent disturbance management.

[0122] In some embodiments of this application, step S530 involves performing rolling simulations of candidate scheduling schemes in a digital twin that is synchronized in real time with the physical production line, dynamically generating and displaying the simulation results, including the following steps.

[0123] Step S710: Load the digital twin model that is synchronized with the physical production line status in real time, and inject the candidate scheduling scheme into the model as the driving input.

[0124] In step S710, a high-fidelity, state-consistent virtual operating environment is provided for the verification of candidate scheduling schemes. This step first loads a digital twin model that is synchronized in real time with the physical production line, ensuring that key elements in the model, such as equipment status, work-in-process location, material inventory, and personnel configuration, are completely aligned with the actual production line. Subsequently, candidate scheduling schemes generated by the scheduling engine are injected into the model as driving input, making it the instruction source for simulation execution. This process ensures that the simulation is not based on idealized or static assumptions, but rather on a dynamic deduction based on the current real production line state, thereby significantly improving the credibility and reference value of the simulation results.

[0125] Step S720: Run multi-cycle rolling simulation to calculate the predicted capacity utilization rate of each workstation and the distribution of work-in-process inventory in the buffer area in real time.

[0126] In step S720, the performance of candidate scheduling schemes under continuous operation is quantitatively evaluated through time-progressive multi-cycle rolling simulation. The system advances the simulation process in a digital twin according to a preset time step (e.g., per minute or per cycle), covering the complete execution process of several future production cycles from the current moment. During this process, the predicted capacity utilization rate of each workstation in each time period is calculated in real time, reflecting the equipment load intensity; simultaneously, the changes in the work-in-process quantity between each buffer area or workstation are tracked, generating work-in-process accumulation distribution data. These dynamic indicators reveal potential bottlenecks, congestion, or resource idle risks of the scheme, providing an objective basis for subsequent decision-making.

[0127] In step S730, the predicted capacity utilization rate and the distribution of work-in-process inventory are dynamically displayed in the form of a visual chart in the scheduling simulation sub-interface 400, and whether the preset capacity utilization rate threshold or the safety limit of work-in-process inventory is exceeded is highlighted.

[0128] In step S730, the complex simulation results are transformed into intuitive and easily interpretable visual information to help users quickly identify the merits of different solutions. In the scheduling simulation sub-interface 400, the system dynamically presents the predicted capacity utilization rate and work-in-process inventory distribution for each workstation in the form of line graphs, heatmaps, or bar charts. It also highlights situations exceeding preset thresholds—for example, when the utilization rate of a workstation consistently exceeds 95%, or when the quantity of work-in-process in a buffer zone exceeds the safety limit, the relevant chart areas are marked with flashing red or thick borders. This visual enhancement mechanism allows operators to clearly judge whether the solution meets the requirements for production stability and smoothness, reducing the risk of misjudgment.

[0129] In step S740, in response to the simulation completion event, the scheduling distribution control 406 is automatically activated, allowing the user to select a candidate scheme that meets the scheduling requirements for distribution.

[0130] In step S740, a controllable connection mechanism between simulation verification and actual execution is established. When the system detects a rolling simulation completion event, it automatically activates the scheduling issuance control 406, indicating that the current candidate solution has completed the virtual verification process and meets the prerequisites for entering the execution stage. At this time, the user can select the optimal solution from the multiple displayed solutions that meets the preset scheduling requirements (such as not exceeding the threshold, meeting the delivery date, etc.) and trigger the issuance operation through this control. This step ensures that only solutions that have undergone sufficient simulation verification are allowed to enter the physical production line for execution, effectively preventing unverified scheduling instructions from directly interfering with actual production, thereby realizing the closed-loop control logic of "verification before execution".

[0131] In some embodiments of this application, reference is made to Figure 1 The main interface 100 also includes a product standard operation guidance control 104; in response to the trigger command of the product standard operation guidance control 104, the product standard operation guidance interface is displayed and the following operations are performed.

[0132] Step S810: On the product standard operation guidance interface, when the product is detected to have been transferred to the current workstation, the corresponding graphic operation guidance content is automatically loaded based on the product's unique serial number, including process description, wiring diagram, interface definition and debugging steps.

[0133] In step S810, the precise and automated delivery of work instructions is achieved, ensuring that operators obtain process information perfectly matching the current product at the correct time. When a product flows to a certain workstation, the system automatically retrieves and loads the corresponding graphic and textual work instructions from the background database by identifying the product's unique serial number. These instructions include detailed process descriptions, electrical wiring diagrams, interface definition specifications, and debugging operation steps. This process eliminates the need for manual searching or selection of machine model documents, avoiding operational deviations caused by misuse of general or outdated work instructions, and significantly improving the level of work standardization and the first-time execution accuracy.

[0134] In step S820, in response to the scanning operation of the material or process confirmation code, the operator's identity, operation start and end time and scanning event are recorded on the product standard operation instruction interface, and this operation information is bound to the data main line corresponding to the product serial number.

[0135] In step S820, a traceable and auditable work execution record chain is constructed. In the product's standard operating procedure interface, when an operator scans a material barcode or process confirmation code, the system automatically captures the scanning event and simultaneously records the operator's identity information, the start and end times of the work, and the specific material or process code scanned. This structured work information is then bound to the data thread corresponding to the product serial number, becoming part of the product's entire lifecycle data. This step not only solidifies the association of the four elements—person, material, time, and sequence—but also provides authentic and tamper-proof operational evidence for subsequent quality backtracking, time analysis, and responsibility determination.

[0136] Step S830: When the current workstation is an assembly workstation, automatically jump to the quality monitoring sub-interface 300 and synchronously display the quality judgment result associated with the product serial number.

[0137] In step S830, seamless linkage between quality status and work process is achieved, strengthening the closed-loop quality management of the assembly process. When the system determines that the current workstation belongs to an assembly workstation, it automatically jumps from the product standard work instruction interface to the quality monitoring sub-interface 300, and simultaneously displays the quality judgment results associated with the product serial number, such as sealant coating score, connector position reliability, or screw fastening anomaly type. This jump allows operators or quality inspectors to instantly view the quality assessment conclusions of the product at this workstation or previous key points while performing or reviewing assembly tasks, facilitating timely intervention, rework, or release decisions. This deeply embeds quality information into the work process, improving the self-inspection and corrective capabilities of the manufacturing process.

[0138] In some embodiments of this application, the method further includes: automatically performing hash calculations on the structured quality judgment results, scheduling decision logs, and operation records, and then writing them into the blockchain to form a manufacturing evidence chain. The main interface 100 also includes a blockchain evidence viewing control 105. Specifically, in response to a trigger command on the blockchain evidence viewing control 105, an evidence verification interface is displayed, any product serial number is entered, the corresponding evidence hash value on the chain is retrieved, and compared with the hash value recalculated from local data, outputting a consistency verification result.

[0139] The above steps aim to construct an immutable and independently verifiable chain of credible evidence covering the entire manufacturing process. The system automatically performs hash calculations on key manufacturing data such as structured quality assessment results, scheduling decision logs, and operation records to generate a unique digital fingerprint, which is then written into the blockchain distributed ledger. Because blockchain possesses decentralized, irreversible, and tamper-proof characteristics, once written, it cannot be unilaterally modified or deleted, thus ensuring the long-term preservation of the integrity and authenticity of core decisions and execution actions during the manufacturing process. This provides a highly reliable data foundation for quality traceability, compliance auditing, and liability determination.

[0140] In response to the trigger command of the blockchain evidence viewing control 105, the system displays the evidence verification interface, allowing users to enter any product serial number to initiate a verification request. The system then retrieves the historical evidence hash value associated with that serial number from the blockchain, and simultaneously recalculates the current hash value based on the locally stored original quality judgment results, scheduling logs, or operation records, comparing the two bit by bit. If they match, the system outputs the verification conclusion that "the data has not been tampered with"; if they do not match, it indicates that the data is abnormal. This mechanism allows any authorized party to independently verify the authenticity of manufacturing data without relying on the trust of a centralized database, significantly improving the transparency, credibility, and non-repudiation capability of the entire manufacturing execution system.

[0141] In some embodiments of this application, the main interface 100 also includes a digital masterline traceability control 106. Specifically, in response to a trigger command on the digital masterline traceability control 106, a product digital masterline traceability sub-interface is displayed. After entering the product serial number, the full lifecycle data of the product is aggregated from the product-level digital masterline, and the on-chain evidence hash value is retrieved for integrity verification. After successful verification, the full lifecycle data, including quality judgment results, scheduling instructions, and operation execution logs, is displayed.

[0142] The above steps provide users with a comprehensive view of the entire lifecycle data for a single product, ensuring the authenticity and completeness of the displayed information. When the user triggers the digital lead traceability control 106, the system displays the product digital lead traceability sub-interface, guiding the input of a specific product serial number. The system then aggregates all associated data for that product since the work order was created from the constructed product-level digital lead based on this serial number, covering multi-dimensional information such as production plans, material usage, process execution, quality judgment results, scheduling instructions, and operation execution logs.

[0143] Building upon this, the system further retrieves the corresponding notarized hash value previously written to the blockchain and compares it with the recalculated hash value of the currently aggregated data to perform integrity verification. Only after successful verification and confirmation that the data has not been tampered with is the system fully displayed to the user throughout the product's entire lifecycle. This mechanism not only enables data connectivity and one-click traceability across business processes but also endows the traceability results with non-repudiable trustworthiness through blockchain technology, effectively supporting demanding application scenarios such as quality analysis, fault diagnosis, and compliance auditing.

[0144] Secondly, this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned manufacturing execution interaction method for an energy storage thermal management unit.

[0145] Furthermore, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned manufacturing execution interaction method for an energy storage thermal management unit.

[0146] In summary, the manufacturing execution interaction method, equipment, and medium for energy storage thermal management units provided in this application have the following technical effects.

[0147] This method constructs a product-level digital masterline centered on a unique product serial number, achieving structured association and full lifecycle connectivity of multi-source manufacturing elements such as orders, materials, processes, equipment, and personnel. This effectively solves the problems of severe data silos and difficult traceability in traditional manufacturing execution systems. By deploying real-time acquisition of visual images and torque-angle time-series data at key assembly stations and combining this with online analysis using a lightweight convolutional neural network model, it can accurately and automatically determine quality characteristics such as the continuity of sealant coating, the position of connectors, and the types of screw tightening anomalies, significantly improving the real-time nature, objectivity, and interpretability of quality control. Furthermore, this method integrates quality judgment results as dynamic constraints into the scheduling decision-making process. Combined with production line disturbances such as equipment failures, material anomalies, or emergency order insertions, it performs rolling simulation verification of candidate scheduling schemes in a digital twin synchronized with the physical production line in real time. Execution is only initiated when the simulation results meet the preset scheduling requirements, thereby greatly enhancing the flexibility, robustness, and executability of production scheduling. Furthermore, by hashing quality assessment results, scheduling decision logs, and operation records and writing them to the blockchain, an immutable manufacturing evidence chain is constructed. This supports one-click traceability and integrity verification based on product serial numbers, providing highly reliable data assurance for quality auditing, responsibility definition, and compliance management. Overall, this method achieves deep collaboration between planning, execution, quality inspection, and scheduling, driving the manufacturing of energy storage thermal management units towards a more reliable, flexible, transparent, and intelligent direction.

[0148] It should be noted that in all specific embodiments of this application, all data processing activities related to user identity or personal characteristics, such as user information, user behavior data, historical data, and location information, will be conducted in accordance with the principles of legality, legitimacy, and necessity. All data collection, use, storage, and processing will be subject to compliance with applicable national and regional laws, regulations, and industry standards, and informed consent from users will be obtained in a clear and explicit manner before processing. For the processing of sensitive personal information, separate consent from users will be obtained through prominent means such as pop-up prompts and independent confirmation pages. If any processing conflicts with laws and regulations, the laws and regulations will prevail, and necessary data processing will only be carried out within the scope permitted by laws and regulations, ensuring that all data-based applications, analyses, and technical implementations are conducted within the scope permitted by laws and regulations.

[0149] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0150] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of ordinary skill of an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary skill. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0151] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several programs to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0152] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequential list of executable programs for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, a program execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can retrieve and execute a program from or in conjunction with such a program execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit a program for use by or in conjunction with a program execution system, apparatus, or device.

[0153] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Additionally, computer-readable media can even be paper or other suitable media on which programs can be printed, for example, by optically scanning the paper or other media, then editing, interpreting, or, if necessary, processing it in a suitable manner to obtain the program electronically, and then storing it in computer memory.

[0154] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable program execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0155] In the foregoing description of this specification, the reference to terms such as "one embodiment / implementation," "another embodiment / implementation," or "certain embodiments / implementations," etc., indicates that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in an embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0156] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0157] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of the present invention.

Claims

1. A manufacturing execution interaction method for an energy storage thermal management unit, characterized in that, Includes the following steps: The main interface for manufacturing execution of the energy storage thermal management unit is displayed; wherein, the main interface includes production planning controls, quality monitoring controls, and scheduling simulation controls; In response to the trigger command of the production planning control, the production planning sub-interface is displayed, a unique product serial number is assigned to each unit, and a product-level digital master line including orders, materials, processes, equipment and personnel is constructed based on the product serial number; In response to the trigger command of the quality monitoring control, the quality monitoring sub-interface is displayed, and visual images and torque-angle time series data of the assembly station are collected in real time and input into the lightweight convolutional neural network model to generate structured quality judgment results online; the assembly station includes a sealant coating station, a connector crimping station and a structural fastening screw fastening station. In response to the trigger command of the scheduling simulation control, the scheduling simulation sub-interface is displayed. The scheduling constraints are updated based on the quality judgment result. Combined with the production line disturbance event, the scheduling engine generates candidate scheduling schemes. The candidate scheduling schemes are then simulated in a rolling simulation in a digital twin synchronized with the physical production line. When the simulation result meets the preset scheduling requirements, the optimal scheduling scheme that meets the preset scheduling requirements is sent to the manufacturing execution system for execution.

2. The manufacturing execution interaction method for the energy storage thermal management unit according to claim 1, characterized in that, The production planning sub-interface includes a work order information entry control, a serial number rule configuration control, a BOM and process binding control, and a digital mainline saving control. In the production planning sub-interface, a unique product serial number is assigned to each unit, and a product-level digital master line including orders, materials, processes, equipment, and personnel is constructed based on the product serial number, including the following steps: In response to the trigger command of the work order information entry control, the work order information entry window is displayed, and the customer order number, product model, planned quantity and delivery date are entered to generate basic work order data; In response to the trigger command of the serial number rule configuration control, the serial number rule configuration window is displayed to set the encoding structure of the product serial number, including the prefix field, the year-month-day field and the serial number field, so as to generate a unique product serial number generation rule; In response to the trigger command of the BOM and process binding control, the BOM and process binding window is displayed, the BOM and standard process route corresponding to the product model are associated, and the generated unique product serial number is used as the primary key to automatically construct a product-level digital master line including the basic work order data, material batch information, process list, equipment resource pool and personnel skill matrix; In response to the trigger command of the digital master line saving control, the product-level digital master line is saved, and the production task bound to the unique product serial number is pushed to the manufacturing execution system backend.

3. The manufacturing execution interaction method for the energy storage thermal management unit according to claim 1, characterized in that, The quality monitoring sub-interface includes a visual acquisition configuration control, a tightening data access control, and a quality judgment display area. In the quality monitoring sub-interface, visual images and torque-angle time-series data of the assembly station are collected in real time and input into a lightweight convolutional neural network model to generate structured quality judgment results online, including the following steps: In response to the trigger command of the vision acquisition configuration control, the vision acquisition configuration window is displayed, and the resolution, frame rate and region of interest of the industrial camera at the assembly station are set to generate vision data acquisition rules. In response to the trigger command of the tightening data access control, a tightening data access window is displayed to configure the protocol type and data field mapping relationship for communication with the tightening device, so as to establish a real-time access channel for torque-angle timing data. After completing the above configuration, the visual images and torque-angle time series data of the assembly station are received in real time in the quality monitoring sub-interface and input into the lightweight convolutional neural network model to generate structured quality judgment results online. The quality judgment results include the sealant coating continuity score, the confidence level of the connector position status, and the screw fastening abnormality type. In response to a refresh event of the quality judgment display area, the quality judgment result and its corresponding original data snapshot are dynamically displayed, including glue line image, connector partial view and torque-angle curve; The quality judgment display area includes result review controls, which allow operators to perform manual review and select to confirm approval, reject, or initiate manual re-inspection. In response to the confirmation operation, the quality judgment result is marked as valid and bound to the corresponding product serial number for updating the scheduling constraints of the scheduling engine.

4. The manufacturing execution interaction method for the energy storage thermal management unit according to claim 3, characterized in that, The lightweight convolutional neural network model includes a semantic segmentation branch unit, an object detection branch unit, and a temporal classification branch unit; the online generation of structured quality assessment results includes the following steps: The real-time visual image of the sealant coating station is input into the semantic segmentation branch unit, which outputs the glue line pixel mask and calculates the glue width standard deviation and the number of break points per unit length based on the mask to generate a sealant coating continuity score. The real-time visual image of the connector crimping station is input to the target detection branch unit, which outputs the bounding box and center coordinates of the connector and socket, and generates the confidence level of the connector's position status based on their relative pose and a preset alignment threshold. The real-time torque-angle timing data of the structural fastening screw fastening station is input into the timing classification branch unit. The torque rise slope, final torque value and angle increment features are extracted and mapped to screw fastening anomaly types in combination with the process window. The anomaly types include underload, overload, stripping or false fit.

5. The manufacturing execution interaction method for the energy storage thermal management unit according to claim 1, characterized in that, The scheduling simulation sub-interface includes a schedule generation control, a simulation start control, and a schedule distribution control; In the scheduling simulation sub-interface, scheduling constraints are updated based on the quality judgment results. Combined with production line disturbance events, a candidate scheduling scheme is generated using the scheduling engine. This candidate scheduling scheme is then subjected to rolling simulation in a digital twin synchronized with the physical production line. When the simulation results meet the preset scheduling requirements, the optimal scheduling scheme is sent to the manufacturing execution system for execution. This includes the following steps: In the scheduling simulation sub-interface, production line disturbance events from the manufacturing execution system are automatically received. These production line disturbance events include equipment fault alarms, abnormal material availability signals, or emergency order insertion instructions. In response to the trigger command of the scheduling generation control, the scheduling engine is invoked to generate a candidate scheduling scheme that includes normal production tasks and rework tasks based on the scheduling constraints corresponding to the currently valid quality judgment results and the production line disturbance events. In response to the trigger command of the simulation start control, the candidate scheduling scheme is simulated in a rolling simulation in a digital twin that is synchronized with the physical production line in real time, and the simulation results are dynamically generated and displayed, including the predicted value of capacity utilization and the distribution of work-in-process inventory. In response to the trigger command of the scheduling control, when the simulation results meet the preset scheduling requirements, the optimal scheduling scheme is sent to the manufacturing execution system for execution.

6. The manufacturing execution interaction method for the energy storage thermal management unit according to claim 5, characterized in that, The invocation of the scheduling engine, based on the scheduling constraints corresponding to the currently valid quality judgment results and the production line disturbance events, generates a candidate scheduling scheme that includes normal production tasks and rework tasks, including the following steps: Based on the anomaly type and severity in the quality judgment results, corresponding process jump rules, rework path assignments or resource priority adjustment instructions are generated as scheduling constraints. Based on the scheduling constraints, identify the set of product serial numbers that need to skip a specific process or be redirected to a rework station; Based on the aforementioned production line disturbance events, update the equipment availability time window, material readiness time, and human resource status; The product serial number set, the updated resource status and the work orders to be executed are input into the scheduling engine, which generates multiple candidate scheduling schemes through reinforcement learning strategy. Each scheme includes the process allocation, workstation assignment and timing arrangement of each product. The scheduling simulation sub-interface lists and displays key indicators for each candidate scheduling scheme, including estimated completion time, bottleneck workstation utilization rate, and rework task ratio.

7. The manufacturing execution interaction method for the energy storage thermal management unit according to claim 6, characterized in that, The process of performing rolling simulations of the candidate scheduling schemes in a digital twin synchronized in real time with the physical production line, and dynamically generating and displaying the simulation results, includes the following steps: Load a digital twin model that is synchronized with the physical production line status in real time, and inject the candidate scheduling scheme into the model as a driving input; Run multi-cycle rolling simulations to calculate the predicted capacity utilization rate of each workstation and the distribution of work-in-process inventory in the buffer area in real time; In the scheduling simulation sub-interface, the predicted capacity utilization rate and the distribution of work-in-process inventory are dynamically displayed in the form of visual charts, and whether the preset capacity utilization rate threshold or the work-in-process safety limit is highlighted. In response to the simulation completion event, the scheduling distribution control is automatically activated, allowing users to select candidate solutions that meet the scheduling requirements for distribution.

8. The manufacturing execution interaction method for the energy storage thermal management unit according to claim 1, characterized in that, The main interface also includes a product standard operation guidance control, which displays the product standard operation guidance interface in response to a trigger command on the product standard operation guidance control; On the product standard operation guidance interface, when a product is detected to have been transferred to the current workstation, the corresponding graphic operation guidance content is automatically loaded based on the product's unique serial number, including process description, wiring diagram, interface definition and debugging steps; In response to the scanning operation of the material or process confirmation code, the operator's identity, operation start and end time and scanning event are recorded on the product standard operation guidance interface and bound to the data main line corresponding to the product serial number; When the current workstation is an assembly workstation, it will automatically jump to the quality monitoring sub-interface and simultaneously display the quality judgment results associated with the product serial number.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the manufacturing execution interaction method of the energy storage thermal management unit according to any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the manufacturing execution interaction method for the energy storage thermal management unit as described in any one of claims 1 to 8.

Images