Island type power battery production data management method and system

Abstract

The invention discloses an island type power battery production data management method and system, and relates to the technical field of battery manufacturing, and the method comprises the steps: employing a distributed architecture based on cloud edge collaboration, and collecting the production data of each process island in real time; performing quality evaluation on the products by using an intelligent prediction model constructed based on machine learning, and verifying and screening the products; monitoring the operation state of each piece of process island equipment; assembling information and timestamps of the materials are recorded through the block chain technology, and information tracing is carried out on the key materials; the production data of each process is integrated through a unified data platform, a virtual mapping model of a production line is established by adopting a digital twinning technology, the state of the corresponding production process on the virtual mapping model is adjusted through the change of real-time data, and the current production takt is adjusted. According to the technical scheme, the quality consistency, the operation efficiency and the intelligent level of power battery production are greatly improved on the whole.

Description

Technical Field

[0001] This invention relates to the field of battery manufacturing technology, and in particular to a method for managing production data of island-type power batteries and a management system for island-type power battery production data. Background Technology

[0002] With the rapid development of the new energy vehicle industry, power battery production faces the dual challenges of improving manufacturing flexibility and production efficiency. Traditional power battery production lines mostly adopt a linear operation mode, and their supporting Manufacturing Execution Systems (MES) typically have significant limitations. Most existing MES systems can only collect partial data from a single production line, production planning relies on manual maintenance and adjustments, and data from different processes are isolated, forming information silos that hinder in-depth data analysis and process optimization. This data fragmentation problem is particularly prominent in island-type production processes, where module and PACK production lines operate independently, leading to barriers to data exchange between systems, making quality verification between upstream and downstream processes difficult, and preventing effective identification and correlation of data from different product systems.

[0003] Furthermore, traditional systems lack real-time monitoring capabilities across the entire production process. Key information such as equipment operating status, material flow data, and quality inspection data cannot be effectively integrated, making it difficult for managers to accurately grasp the actual production line pace and to conduct refined analysis and optimization of the production process. This situation directly restricts the improvement of production flexibility, making it difficult for production lines to quickly respond to product changeovers and process adjustments. When quality anomalies occur, a significant amount of time is often required for manual investigation because the traceability information for key materials is scattered, making accurate end-to-end tracking impossible.

[0004] As power battery production moves towards higher variety and smaller batch sizes, traditional MES systems can no longer meet the demands of modern manufacturing for data-driven and intelligent decision-making. Island-style production processes require each process island to maintain independent operational flexibility while simultaneously achieving high overall collaborative efficiency, placing higher demands on the data integration and analysis capabilities of the MES system. Existing technologies cannot address the challenges of data connectivity, intelligent decision-making, reliable traceability, and real-time optimization throughout the entire island-style power battery production process. Summary of the Invention

[0005] To address the above problems, this invention provides a method for managing production data of island-type power batteries, comprising: Through a unified data platform, a distributed architecture based on cloud-edge collaboration is adopted to collect production data from each process island in real time. Based on the production data of each process island, a machine learning-based intelligent prediction model is used to assess the quality of the products from the previous process island, and the products are verified and screened based on the assessment results. The intelligent prediction model is trained using historical quality data. Based on the collected production data, monitor the operating status of equipment in each process island; Based on the collected production data, the assembly information and timestamps of materials are recorded using blockchain technology to enable information traceability of key materials in the power battery production and assembly process. By integrating production data from various processes through a unified data platform, a virtual mapping model of the production line is established using digital twin technology. The state of the corresponding production process on the virtual mapping model is adjusted based on changes in real-time data, and the current production cycle is regulated.

[0006] In the above technical solution, preferably, the island-type power battery production data management method further includes: The production plan for power batteries is scheduled, the production quantity of power batteries is managed, and the process parameters of power battery products are retrieved from the process parameter database. The production plan, production quantity, and process parameters are respectively sent to the controllers of each process island on the production line, and the required type of power battery products are produced according to the corresponding process parameters, specifically including: Based on production needs, the battery model, order quantity, order priority, and delivery timestamp of the power battery products are determined, and the production plan of the power battery products is obtained through scheduling. Based on the battery model to be produced according to the production plan, the process parameters of the corresponding battery model are automatically retrieved from the process parameter database; The production plan, the production quantity, and the process parameters are serialized in a preset format and asynchronously transmitted to the controller of the corresponding process island via a message queue. The process island controls the actuators to carry out production based on the received battery model and process parameters; During the production process, based on the real-time equipment status and material characteristics collected, the process parameters are dynamically and adaptively optimized using a continuously updated neural network model with an online learning mechanism. This includes adjusting the welding current and welding speed of the welding process parameters, adjusting the charging and discharging parameters of the battery cells, and adjusting the torque and pressure of the wire harness assembly process. The neural network model is trained based on historical production data.

[0007] In the above technical solution, preferably, the step of using a machine learning-based intelligent prediction model to assess the quality of the products from the previous process island, and verifying and screening the products based on the assessment results, includes: Train a machine learning-based intelligent prediction model using historical quality data; Using a pre-verification method, when a product completes the processing of the previous process island and enters the waiting buffer of the next process, all process data of the previous process are retrieved, and the intelligent prediction model is used to conduct an online quality assessment of the product of the previous process island. The product is deemed qualified based on the assessed probability of quality compliance, and unqualified products are removed.

[0008] In the above technical solution, preferably, information traceability is performed on key materials in the power battery production and assembly process, and the specific process includes: Equip the key materials of the power battery with RFID and QR code identification; During the production and assembly process, blockchain technology is used to collect RFID and QR code identification information of each key material being assembled, and a unique traceability label is generated for each battery pack. Based on blockchain technology, the configuration identification and assembly association of the battery cells, housing and wiring harness are recorded throughout the entire life cycle, and the source of materials is traced based on the blockchain record data.

[0009] In the above technical solution, preferably, the island-type power battery production data management method further includes: Based on blockchain technology, the raw information of the manufacturer's incoming materials and battery materials, along with timestamps and supplier digital signatures, are uploaded to the blockchain. The material information is stored, managed, and maintained based on the blockchain information.

[0010] In the above technical solution, preferably, production data from various processes are integrated through a unified data platform, and a virtual mapping model of the production line is established using digital twin technology. The state of the corresponding production process on the virtual mapping model is adjusted based on changes in real-time data. Specifically, this includes: The collected production data is comprehensively analyzed, and a virtual mapping model of the production line is established using digital twin technology; Based on the equipment operation signals and production cycle data of each process island in the production data, the deviation between the production cycle and waiting time of each process and the theoretical cycle time is calculated based on the time series analysis method. The real-time position of the product in each process island is visualized based on the virtual mapping model to realize production line cycle time analysis. Based on the results of the production line cycle time analysis, the equipment utilization rate of each workstation is calculated, the key workstation that has the greatest impact on production efficiency is identified as the production bottleneck workstation, and it is dynamically marked in the virtual mapping model.

[0011] In the above technical solution, preferably, the island-type power battery production data management method further includes: Based on the machine vision inspection data, sensor measurement data, and online test data of each process island in the production data, machine learning algorithms are used for pattern recognition. Based on the recognition results, the quality fluctuation trend of the product and the defect distribution rate of each process island are mapped in real time according to the virtual mapping model to achieve quality analysis. Based on the results of the quality analysis, a clustering algorithm is used to identify similar defect patterns, and the process that caused the defect is determined based on the similar defect identification results. The defect station is then located in the virtual mapping model, and adjacent defect stations are connected to construct a defect tracing path.

[0012] In the above technical solution, preferably, production data from various processes are integrated through a unified data platform, a virtual mapping model of the production line is established using digital twin technology, the state of the corresponding production process on the virtual mapping model is adjusted based on real-time data changes, and the actual production cycle time is adjusted, specifically including: By deploying edge nodes on each process island, real-time data of the corresponding process island is collected and uploaded to the cloud; the real-time data includes the number of products in process, equipment utilization rate, product qualification rate, and the product qualification rate of products waiting to flow into downstream in the buffer zone; Using cloud computing capabilities, control instructions are generated based on the real-time data. The process of generating control instructions is as follows: For the first upstream process island, when the work-in-process quantity of the corresponding downstream process island is lower than the lower limit of work-in-process, and the equipment utilization rate of the corresponding downstream process island is lower than the target utilization rate, and the product qualification rate of the first upstream process island within a preset time window is greater than or equal to the first quality threshold, and the product qualification rate of the buffer of the first upstream process island to flow into the downstream is greater than the first preset qualification rate, then control instructions to shorten the production task issuance interval and shorten the production cycle are generated; when the work-in-process quantity of the corresponding downstream process island exceeds the upper limit of work-in-process, and / or, the equipment utilization rate of the corresponding downstream process island exceeds the utilization rate upper limit, or, the qualification rate of the first upstream process island within a time window is lower than the second quality threshold and the qualification rate of the product to flow into the corresponding downstream process island is less than or equal to the first preset qualification rate, then control instructions to extend the production task issuance interval and extend the production cycle are generated. The state of the corresponding production process on the virtual mapping model is adjusted according to the control instructions.

[0013] In the above technical solution, preferably, the real-time acquisition of production data from each process island specifically includes: Real-time data acquisition is performed on the voltage and internal resistance during the OCV (Open Circuit Voltage) cell testing process, the welding current and temperature during the laser welding process, the torque and pressure during the PACK pre-assembly process, the voltage and current curve data during the charge and discharge test process, and the assembly completion data during the PACK post-assembly process. This data is then stored and retrieved using a time-series database.

[0014] In the above technical solution, preferably, the island-type power battery production data management method further includes: Collect the work-in-process quantity, waiting queue length, equipment runtime, discharge speed and material consumption rate of each process island, and calculate the real-time capacity load rate of each process island in real time. When it is detected that the number of work-in-process products on the first process island exceeds the first threshold and the capacity utilization rate of the downstream second process island is lower than the target range, a production task matching the backlog product model of the first process island is issued to the second process island. When the length of the waiting queue of the third process island in each process island exceeds the queue warning length, and the equipment running time percentage exceeds the intervention limit, the third process island is determined to be in an overload state. When the duration for which the discharge rate of the first process island is higher than the material consumption rate of the second process island is greater than the duration of the higher discharge rate, a control command is generated to reduce the discharge cycle time of the first process island or to reduce the transmission speed of the material flow system.

[0015] This invention also proposes an island-type power battery production data management system, which applies the island-type power battery production data management method disclosed in any of the above technical solutions, including: The production data acquisition module is used to collect production data from each process island in real time through a unified data platform and a distributed architecture based on cloud-edge collaboration. The data evaluation and screening module is used to evaluate the quality of the products from the previous process island based on the production data of each process island using an intelligent prediction model built on machine learning, and to verify and screen the products based on the evaluation results. The intelligent prediction model is trained using historical quality data. The operation status monitoring module is used to monitor the operation status of each process island equipment based on the collected production data; The data traceability and recording module is used to record the assembly information and timestamps of materials based on the collected production data using blockchain technology, enabling information traceability of key materials in the power battery production and assembly process. The data mapping adjustment module is used to integrate production data from various processes through a unified data platform, establish a virtual mapping model of the production line using digital twin technology, adjust the state of the corresponding production process on the virtual mapping model based on real-time data changes, and adjust the current production cycle.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: By constructing a unified data platform based on cloud-edge collaboration, data barriers between various process islands were effectively broken down, enabling real-time collection and integration of production data throughout the entire process, laying the foundation for subsequent analysis. Based on this, an intelligent prediction model based on machine learning was used to conduct online quality assessment and real-time screening of products from the previous process, enabling timely interception of potential defects and significantly reducing resource waste and rework costs caused by defective products. Simultaneously, real-time monitoring of equipment operating status provided data support for predictive maintenance. The introduction of blockchain technology to record key material information created an immutable, full lifecycle traceability system, greatly improving the ability to trace product quality issues and the transparency of the supply chain. Finally, a virtual mapping model of the production line established using digital twin technology enabled real-time interaction and iterative optimization between physical production and virtual space, making dynamic adjustments to production cycle time and process parameters possible, thereby significantly improving the overall quality consistency, operational efficiency, and intelligence level of power battery production. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a method for managing production data of an island-type power battery according to an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The present invention will now be described in further detail with reference to the accompanying drawings: like Figure 1 As shown, a method for managing production data of an island-type power battery according to the present invention includes: Through a unified data platform, a distributed architecture based on cloud-edge collaboration is adopted to collect production data from each process island in real time. Based on the production data of each process island, the quality of the products from the previous process island is evaluated using an intelligent prediction model built on machine learning. The products are then verified and screened based on the evaluation results. The intelligent prediction model is trained using historical quality data. Based on the collected production data, monitor the operating status of equipment in each process island; Based on the collected production data, the assembly information and timestamps of materials are recorded using blockchain technology to enable information traceability of key materials in the power battery production and assembly process. By integrating production data from various processes through a unified data platform, a virtual mapping model of the production line is established using digital twin technology. The status of the corresponding production process on the virtual mapping model is adjusted based on changes in real-time data, and the current production cycle is regulated.

[0020] In this implementation, by constructing a unified data platform based on cloud-edge collaboration, data barriers between various process islands are effectively broken down, enabling real-time collection and integration of production data throughout the entire process, laying the platform foundation for subsequent analysis. Based on this, an intelligent prediction model based on machine learning is used to conduct online quality assessment and real-time screening of products from the previous process, enabling timely interception of potential defects and significantly reducing resource waste and rework costs caused by defective products. Simultaneously, real-time monitoring of equipment operating status provides data support for predictive maintenance. The introduction of blockchain technology to record key material information constructs an immutable, full lifecycle traceability system, greatly improving the ability to trace product quality issues and the transparency of the supply chain. Finally, a virtual mapping model of the production line established using digital twin technology enables real-time interaction and iterative optimization between physical production and virtual space, making dynamic adjustments to production cycle time and process parameters possible, thereby significantly improving the overall quality consistency, operational efficiency, and intelligence level of power battery production.

[0021] Specifically, this method employs a distributed architecture, deploying data acquisition terminals on each process island (OCV cell testing process island, laser welding process island, PACK pre-installation process island, charge / discharge testing island, and PACK post-installation process island). These terminals can connect to a central server (in the cloud, such as AWS IoT or Azure IoT, primarily responsible for data aggregation, model training, and global optimization) via industrial Ethernet (edge ​​nodes). The data acquisition process uses specific protocols to collect real-time equipment data (such as sensor readings and PLC signals) and production data from each process island. Production planning can utilize specific scheduling algorithms to automatically generate plans based on order requirements and distribute process parameters to the controllers on each process island. The material traceability process employs RFID and / or QR code technology to achieve full lifecycle tracking of materials.

[0022] During implementation, the island-type power battery production data management method relies on a unified data platform and a cloud-edge collaborative architecture.

[0023] Based on a unified data platform and a cloud-edge collaborative architecture, edge gateways are deployed on each process island. These edge gateways connect to PLCs, industrial PCs, testing equipment, vision cameras, charge / discharge testers, and other devices within their respective islands via industrial Ethernet or fieldbus, collecting equipment operating signals, process parameters, quality inspection data, and production data in real time. All edge gateways access the unified data platform via message queues or industrial protocols. The unified data platform, deployed in the cloud or data center, provides data access, storage, computing, and visualization capabilities, forming a distributed architecture based on cloud-edge collaboration: the edge side is responsible for real-time data acquisition and preliminary preprocessing within milliseconds to seconds, while the cloud side is responsible for data aggregation, modeling analysis, and cross-island collaborative decision-making.

[0024] Based on this, in the quality assessment and pre-process verification and screening process based on machine learning, historical production data and quality inspection results are maintained in a unified data platform. For each product on each process island, before it enters the next process, the platform calls a pre-trained intelligent prediction model to conduct a quality assessment. The input of the intelligent prediction model includes process parameters of the previous process island (such as welding current, welding speed, cell temperature, pressing pressure, torque curve characteristics), equipment status characteristics, online test results, etc. The model outputs the probability value that the product is qualified.

[0025] The model is trained offline based on historical quality data, with the final product judgment result used as the supervision label during training. During online operation, products below the evaluation probability are marked as high-risk products based on a comparison with a preset threshold. Screening instructions are then sent to the field execution mechanism (sorting device, rejection mechanism) through the industrial control network to pre-screen and isolate unqualified or high-risk products.

[0026] During implementation, equipment operation status monitoring is conducted by a unified data platform, which monitors collected operating signals (start / stop signals, fault signals, alarm codes), deviations in process parameters, and production data according to process islands and equipment dimensions. Edge gateways perform local calculations on the operating status of key equipment and report equipment status summaries (such as running, idle, faulty, and under maintenance) to the cloud in real time. The cloud compares the equipment status with the planned cycle time, identifying abnormal downtime, frequent alarms, and underperformance, providing foundational data for subsequent cycle time analysis and bottleneck identification.

[0027] Furthermore, the unique identifiers, assembly stations, assembly times, assembly parameters, and operator information for each key material such as power battery packs, cells, enclosures, and wiring harnesses are compiled into transaction data and written to a consortium blockchain or enterprise private blockchain via blockchain nodes. The blockchain nodes are deployed on a unified data platform and some edge nodes, employing a consensus mechanism to ensure the immutability of the records. Each chain record is bound with a specific timestamp, enabling traceability of information on key materials throughout the production and assembly process.

[0028] Based on the above data, a digital twin virtual mapping model is established to realize state adjustment. Specifically, the unified data platform constructs the process topology and operating parameters of the production line according to the process islands and workstations based on the collected production data, and uses digital twin technology to establish corresponding virtual process islands, virtual workstations and virtual equipment objects in the virtual environment.

[0029] In the virtual mapping model, each virtual object is bound to actual data such as the operating status, cycle time parameters, and work-in-process quantity of the corresponding physical equipment, and receives real-time data updates to its own status through a unified data platform. Real-time synchronization of physical production line data is achieved. Production cycle time (e.g., robot motion path optimization) and process parameters (e.g., welding speed) are adjusted through simulation. When the production cycle time or equipment status of a certain process island changes, the status of the corresponding object on the virtual mapping model adjusts accordingly, such as changing the color, status marker, and work-in-process count of the virtual equipment. This achieves a one-to-one mapping between the physical and virtual production lines, enabling deep perception and comprehensive analysis of the physical production process. Managers can simulate production cycle time and process parameters without affecting actual production, predicting the effects of adjustments. Control commands can be directly generated and issued to the physical production line to adjust the equipment's production cycle time.

[0030] In the above embodiments, preferably, the island-type power battery production data management method further includes: The production plan for power batteries is scheduled, the production quantity of power batteries is managed, and the process parameters of power battery products are retrieved from the process parameter database. The production plan, production quantity, and process parameters are distributed to the controllers of each process island on the production line. The required types of power battery products are then produced according to the corresponding process parameters, specifically including: Based on production needs, determine the battery model, order quantity, order priority, and delivery time stamp of the power battery products, and schedule the production plan of the power battery products. Based on the battery model to be produced according to the production plan, the corresponding battery model's process parameters are automatically retrieved from the process parameter database; The production plan, production quantity, and process parameters are serialized in a preset format and asynchronously transmitted to the controller of the corresponding process island via a message queue. The process island controls the actuators to carry out production based on the received battery model and process parameters; During the production process, based on the real-time equipment status and material characteristics collected, the process parameters are dynamically and adaptively optimized using a continuously updated neural network model with an online learning mechanism. This includes adjusting the welding current and welding speed of the welding process parameters, adjusting the charging and discharging parameters of the battery cells, and adjusting the torque and pressure of the wire harness assembly process. The neural network model is trained based on historical production data.

[0031] In practice, this method further includes production planning and scheduling as well as online process parameter optimization.

[0032] Specifically, a production planning management module can be configured in the unified data platform. This module receives order information, including battery model, order quantity, order priority, and delivery time. Based on this information, a production plan for battery products is generated using rule-based scheduling or heuristic scheduling algorithms. The plan includes the start and end times of production for each battery model, the production cycle time for each process island, and batch division. The unified data platform maintains a process parameter database, associating the standard process parameters of different power battery models with their model codes. Once the scheduling determines the production plan for a specific model, the system automatically retrieves the corresponding process parameters from the process parameter database, such as welding current, welding speed, cell pre-charge voltage, testing time, torque, and pressure threshold.

[0033] Based on this, the unified data platform serializes the production plan, planned production quantity and corresponding process parameters in a preset JSON or binary format, and asynchronously sends them to the controllers (PLC, industrial PC or host computer) of each process island through a message queue.

[0034] The controllers of each process island analyze the received data and automatically set the formula parameters of the field equipment according to the battery model and process parameters, and control the actuators (welding equipment, handling robots, pressing equipment, torque tools, etc.) to perform production according to the corresponding formula.

[0035] During production, the platform continuously collects real-time equipment status (welding current waveform, temperature curve, torque curve, pressure curve) and material characteristics (cell internal resistance, open circuit voltage, electrode temperature). An online learning neural network model is deployed in the cloud. This model is trained with historical production data and quality results to obtain initial parameters. During online operation, the weights are updated incrementally based on newly collected data to achieve adaptive learning of the relationship between process parameters and quality results.

[0036] When the model determines that there is room for optimization in the current combination of process parameters, it calculates adjustment suggestions. For example, for the laser welding process, it provides small adjustments to the welding current and welding speed; for the cell charging and discharging test process, it provides fine-tuning suggestions for the charging and discharging current and cutoff voltage; and for the wire harness assembly process, it provides adjustment suggestions for the target values ​​of torque and pressure.

[0037] The unified data platform pushes these optimization suggestions to the corresponding process island controllers through message queues. The controllers then dynamically fine-tune the process parameters within a safe range, achieving adaptive optimization of production process parameters.

[0038] In the above embodiments, the step of using a machine learning-based intelligent prediction model to assess the quality of products from the previous process island, and then verifying and screening the products based on the assessment results, includes: Train a machine learning-based intelligent prediction model using historical quality data; By adopting a pre-verification method, when a product completes the processing of the previous process island and enters the waiting buffer of the next process, all process data of the previous process are retrieved, and an intelligent prediction model is used to conduct online quality assessment of the product of the previous process island. The product's qualification is determined based on the probability of quality compliance obtained from the assessment, and unqualified products are screened out.

[0039] In this implementation, during the training of the intelligent prediction model, typical batches are selected from historical data. The process data (process parameters, equipment status, and inspection results) from the previous process island are paired with the final quality judgment result of that batch to form training samples. These samples are used to train a machine learning-based intelligent prediction model. The model can employ a gradient boosting decision tree or a deep neural network structure, learning the mapping relationship between various process parameters and product quality through training. After training, the model is validated and tested to ensure that it achieves the preset accuracy and recall rate on an offline dataset.

[0040] During online production, when a product completes processing at the previous process island and enters the waiting buffer for the next process, the unified data platform retrieves all process data from the previous process based on the product's unique identifier and inputs it into the intelligent prediction model for online evaluation. For example, the model outputs a probability value P that the product is qualified, which is compared with a preset threshold P0. When P ≥ P0, the product is determined to be able to proceed to the next process; when P < P0, the product is determined to be a high-risk product.

[0041] For products deemed high-risk or non-compliant, the platform, through its interface with the on-site control system, sends rejection or re-inspection commands to the corresponding buffer station. This controls the sorting mechanism to separate the product from the normal production cycle and guide it to the re-inspection or scrap channel. Simultaneously, the predicted results, key signals, and handling outcomes for this product are recorded in a unified data platform for subsequent model retraining and quality analysis.

[0042] In the above embodiments, preferably, information traceability is performed on key materials in the power battery production and assembly process, specifically including: Equip key materials in power batteries with RFID and QR code labels; During the production and assembly process, blockchain technology is used to collect RFID and QR code identification information of each key material being assembled, and a unique traceability label is generated for each battery pack. Based on blockchain technology, the configuration identification and assembly relationship of the cells, housings and wiring harnesses of the power battery are recorded throughout their entire life cycle, and the source of materials can be traced based on the blockchain record data.

[0043] In this implementation, key materials are tagged. Specifically, key materials such as battery cells, housings, and wiring harnesses are pre-configured with unique RFID tags and QR codes. The RFID tags are used for automatic long-distance identification, and the QR codes are used for manual or visual reading. Before warehousing or going online, the RFID tags and QR codes are bound to the material's basic information (supplier, batch number, specifications) and stored in a unified data platform.

[0044] During the production and assembly process, each key assembly station is equipped with an RFID reader and a QR code scanner. When materials pass through a station and complete assembly, their RFID and QR code identification information is collected, along with the assembly time, equipment number, operator number, and relevant process parameters. The system generates a unique traceability tag for each battery pack, linking the battery pack to the identification of its constituent cells, casing, and wiring harnesses, forming a hierarchical relationship between the battery pack and key materials.

[0045] The collected assembly records are packaged according to a block data structure and written onto the blockchain via blockchain nodes. The records include: battery pack traceability tags, key material identifiers, assembly workstations, assembly timestamps, process parameter summaries, and operator digital signatures. During battery pack manufacturing, after-sales service, or quality traceability, querying the records on the blockchain allows tracking of the source batch, assembly process, and participating workstations of each key material in the battery pack, achieving full lifecycle traceability.

[0046] In the above embodiments, preferably, the island-type power battery production data management method further includes: Based on blockchain technology, the raw information of the manufacturer's incoming materials and battery materials, along with timestamps and supplier digital signatures, are uploaded to the blockchain. The material information is stored, managed, and maintained based on the blockchain information.

[0047] During implementation, the following on-chain evidence storage and maintenance measures are taken for supplier incoming materials and original material information: Specifically, when shipping materials, suppliers generate electronic data, including material codes, batch numbers, quantities, manufacturing dates, and inspection results, and digitally sign this data using their private keys. After the materials arrive at the factory and undergo warehousing acceptance, the system packages the manufacturer's incoming material data, factory inspection data, original material information, and warehousing timestamp together, and writes it to the blockchain through the enterprise's blockchain node, forming an incoming material storage record.

[0048] Furthermore, a unified data platform is established to create a material storage and management interface, using on-chain hashing and digital signatures to verify the integrity and source credibility of incoming material data. When materials need to be traced in subsequent production, after-sales service, or quality incidents, their original incoming material information can be verified through blockchain records, avoiding difficulties in determining responsibility caused by off-book modifications or data forgery.

[0049] In the above embodiments, preferably, production data from various processes are integrated through a unified data platform, and a virtual mapping model of the production line is established using digital twin technology. The state of the corresponding production process on the virtual mapping model is adjusted based on changes in real-time data. Specifically, this includes: The collected production data is comprehensively analyzed, and a virtual mapping model of the production line is established using digital twin technology; Based on the equipment operation signals and production cycle data of each process island in the production data, the deviation between the production cycle and waiting time of each process and the theoretical cycle time is calculated using time series analysis. The real-time position of the product in each process island is visualized using a virtual mapping model, thus realizing production line cycle time analysis. Based on the results of the production line cycle time analysis, the equipment utilization rate of each workstation is calculated, the key workstation that has the greatest impact on production efficiency is identified as the production bottleneck workstation, and it is dynamically marked in the virtual mapping model.

[0050] In this implementation, production data from various processes are integrated through a unified data platform, and a virtual mapping model of the production line is established using digital twin technology to achieve cycle time and quality analysis. Specifically, this includes: First, a virtual mapping model is established. Specifically, the collected production data is comprehensively analyzed to extract the process sequence, equipment type, equipment number, process parameters, and material flow relationships for each process island and workstation, forming a topological structure of production line equipment-process-materials. On the digital twin platform, virtual process island objects, virtual workstation objects, and virtual equipment objects are created based on the topological structure, and their positions and layouts in the actual production line are mapped to their geometric positions in virtual space, realizing a virtual replication of the physical production line.

[0051] Based on the collected equipment operation signals and production cycle time data, time series analysis is performed on the deviations of the production cycle, waiting time, and theoretical cycle time for each process island. First, signals are collected for each process island / station and standardized into a time series. The waiting time, processing time, production cycle / actual cycle time, and deviation from the theoretical cycle time for a specific workstation are calculated. At the time series level, aggregation is performed by sliding time window or by shift / hour to analyze the average value, standard deviation, trend line (such as moving average), and outliers (such as deviations from the mean by several times the standard deviation) to identify workstations with abnormal cycle times and large fluctuations. Based on predefined time intervals and the time proportion of operating status signals, the equipment utilization rate of each process island within the statistical period is calculated.

[0052] In the virtual mapping model, the system visualizes the real-time location and flow status of products on each process island using colors, layers, or icons, while overlaying cycle time deviation and equipment utilization information. Based on the cycle time analysis results, workstations with long production cycles, high waiting times, low utilization rates, and the greatest impact on overall capacity are identified and marked as production bottleneck workstations. These workstations are dynamically highlighted in the virtual mapping model, facilitating management identification and optimization measures.

[0053] In the above embodiments, preferably, the island-type power battery production data management method further includes: Based on machine vision inspection data, sensor measurement data, and online test data from each process island in the production data, machine learning algorithms are used for pattern recognition. Based on the recognition results, the quality fluctuation trend of the product and the defect distribution rate of each process island are mapped in real time using a virtual mapping model to achieve quality analysis. Based on the results of the quality analysis, a clustering algorithm is used to identify similar defect patterns. The process that caused the defect is determined based on the similar defect identification results. The defect station is located in the virtual mapping model, and adjacent defect stations are connected to construct a defect tracing path.

[0054] During implementation, machine vision inspection data, sensor measurement data, and online test data from each process island in the production data are used to perform pattern recognition using machine learning algorithms. Specifically, feature extraction is performed on the machine vision inspection data, sensor measurement data, and online test data collected from each process island. Image texture features, shape features, temporal statistical features, and charge / discharge curve features are fused to construct a multi-dimensional quality feature vector to characterize the product quality status. Based on historically labeled quality data, supervised learning models such as convolutional neural networks and gradient boosting trees are used to classify the quality feature vector and output the probability distribution of different quality states and defect types. At the same time, K-Means clustering algorithm or density clustering algorithm is used to cluster unlabeled samples, identify similar defect patterns, and assign defect codes to the defect patterns through manual confirmation. This maps different quality states and defect types to standardized quality feature fields, including quality state, defect type code, quality confidence, and defect severity, which are used to display the product quality fluctuation trend and defect distribution rate of each process island in real time in the digital twin virtual mapping model.

[0055] In the virtual mapping model, the product quality fluctuation trend and the defect distribution rate of each process island are superimposed on the corresponding virtual objects in the form of color gradients, defect heat maps, etc., to realize the visualization of the production line quality status.

[0056] Based on the aforementioned multidimensional quality feature vectors, the defect distribution rate is statistically analyzed according to process islands and time periods to form the quality analysis results. On this basis, historical defect samples with unqualified quality status are selected from the quality analysis results, and their quality feature vectors and process context features are extracted to form a defect feature vector set. This defect feature vector set is then clustered using K-Means clustering or density clustering algorithms. Each cluster is considered a similar defect pattern, and a corresponding defect pattern code is assigned to each cluster. The frequency of these patterns occurring in each process island and their associated process parameters are statistically analyzed. Based on the similar defect identification results, the possible processes and related workstations that caused the defect are determined. The defect workstation is located in the virtual mapping model, and a defect tracing path is constructed by connecting adjacent defect workstations, helping engineers track the complete path of the defect from the source workstation to the discovery workstation.

[0057] In the above embodiments, preferably, production data from various processes are integrated through a unified data platform, a virtual mapping model of the production line is established using digital twin technology, the state of the corresponding production process on the virtual mapping model is adjusted based on real-time data changes, and the actual production cycle time is adjusted, specifically including: By deploying edge nodes on each process island, real-time data of the corresponding process island is collected and uploaded to the cloud; the real-time data includes the number of products in process, equipment utilization rate, product qualification rate, and the product qualification rate of products waiting to flow into downstream in the buffer zone; Using cloud computing capabilities, control instructions are generated based on the real-time data. The process of generating control instructions is as follows: For the first upstream process island, when the work-in-process quantity of the corresponding downstream process island is lower than the lower limit of work-in-process, and the equipment utilization rate of the corresponding downstream process island is lower than the target utilization rate, and the product qualification rate of the first upstream process island within a preset time window is greater than or equal to the first quality threshold, and the product qualification rate of the buffer of the first upstream process island to flow into the downstream is greater than the first preset qualification rate, then control instructions to shorten the production task issuance interval and shorten the production cycle are generated; when the work-in-process quantity of the corresponding downstream process island exceeds the upper limit of work-in-process, and / or, the equipment utilization rate of the corresponding downstream process island exceeds the utilization rate upper limit, or, the qualification rate of the first upstream process island within a time window is lower than the second quality threshold and the qualification rate of the product to flow into the corresponding downstream process island is less than or equal to the first preset qualification rate, then control instructions to extend the production task issuance interval and extend the production cycle are generated. The state of the corresponding production process on the virtual mapping model is adjusted according to the control instructions.

[0058] In this implementation, a cloud-edge collaborative distributed computing architecture is adopted for unified management of cell OCV data, backend MES order data, and equipment sensor data from each process island. Edge nodes are deployed on each process island, equipped with industrial gateways and lightweight rule engines, responsible for extracting features from real-time data and executing controls (such as emergency stops and temperature over-limit alarms).

[0059] The cloud aggregates all data uploaded from edge nodes (such as time-series data, image data, and business data), and processes the data using its storage and computing power.

[0060] In the cloud, the virtual mapping model continuously receives real-time data streams from edge nodes, driving the virtual production lines and virtual products within the model to remain synchronized with the physical world. The system executes the following closed-loop control logic within a preset time window (e.g., 5 minutes): First, the conditions for triggering accelerated production instructions are as follows: For the first upstream process island (e.g., laser welding island), the system monitors the work-in-process quantity and equipment utilization rate of its downstream process island (e.g., PACK assembly island) in real time. If the following conditions are met simultaneously: (1) the downstream work-in-process quantity is lower than the set lower limit threshold L (e.g., L=5 pieces); (2) the downstream equipment utilization rate is lower than the target utilization rate UT (e.g., UT=85%); and (3) in the most recent time window, the product qualification rate of the upstream process is greater than or equal to the first quality threshold Q1 (e.g., Q1=98.5%), and the product qualification rate of the products to be flowed into the downstream in its output buffer is greater than 95% as determined by the quality prediction model (e.g., the first preset qualification rate is set to 95% by default), then the decision engine in the digital twin model will generate control instructions for "shortening the interval between production task issuance" and "shortening the production cycle". The state of the corresponding production process on the virtual mapping model is adjusted according to the control instructions, specifically: shortening the time interval between production task issuances of the first upstream process island and shortening the operation cycle of the robot in the first upstream process island. For example, the interval between production tasks of the laser welding island is reduced by 10 seconds, and the operation cycle of the welding robot is adjusted from 60 seconds to 55 seconds.

[0061] Second, the conditions for triggering the deceleration / pause production instruction: For the first upstream process island (e.g., laser welding island), the system monitors the work-in-process quantity and equipment utilization rate of its downstream process island (e.g., PACK assembly island) in real time. If the system detects that: (1) the work-in-process quantity of the downstream process island exceeds the upper limit threshold H (e.g., H=30 pieces) and / or the equipment utilization rate exceeds the upper limit UH (e.g., UH=95%), it indicates that congestion has occurred downstream; or (2) the product qualification rate of the upstream process island within the time window is lower than the second quality threshold Q2 (e.g., Q2=95%), and the proportion of products marked as "quality risk status" by the quality prediction model in its buffer zone exceeds the preset risk ratio R (e.g., R=10%), it indicates that there is a risk of batch quality risk flowing downstream. Once either condition is met, the decision engine will generate control instructions to "extend the production task issuance interval" and "extend the production cycle time" to alleviate downstream pressure or prevent the continued production of defective products. Based on these control instructions, the engine will adjust the state of the corresponding production process on the virtual mapping model. Specifically, this involves extending the time interval for issuing production tasks to the first upstream process island and extending the robot's work cycle within that island. For example, increasing the production task issuance interval of the laser welding island by 10 seconds would adjust the welding robot's work cycle from 60 seconds to 65 seconds.

[0062] The optimized control commands generated in the cloud are instantly sent to the edge controllers and production scheduling system of the corresponding process islands. The edge controllers are responsible for converting abstract cycle time commands (such as "shorten to 55 seconds") into specific programs or parameters that can be executed by the equipment. At the same time, the execution action and the new data generated are collected again and fed back to the digital twin model, forming a continuous "perception-decision-execution-feedback" optimization closed loop, thereby achieving adaptive and self-optimized production cycle time.

[0063] In the above embodiments, preferably, real-time collection of production data from each process island specifically includes: The system collects real-time data on voltage and internal resistance during OCV cell testing, welding current and temperature during laser welding, torque and pressure during PACK pre-assembly, voltage and current curves during charge and discharge testing, and assembly completion data during PACK post-assembly. This data is then stored and retrieved using a time-series database.

[0064] Specifically, the real-time collection of production data from each process island is implemented as follows: (1) OCV cell testing data acquisition On the OCV cell testing process island, the testing equipment is connected to the edge gateway via industrial Ethernet to collect the voltage and internal resistance data of each cell in real time, and to mark the cell number and testing timestamp.

[0065] (2) Laser welding process data acquisition In the laser welding process island, process parameters such as welding current, welding speed, and welding temperature are collected through the welding controller interface. The parameters of each weld point are associated with the corresponding cell or electrode marker to form a welding process data record.

[0066] (3) PACK pre-installation process data acquisition In the PACK pre-assembly process island, torque and pressure sensors are installed at each tightening and pressing station. The control system collects the torque value and pressure curve of each fastener and pressing action.

[0067] (4) Data acquisition for charge and discharge testing process In the charge and discharge test process island, the test equipment records the charging voltage curve, discharging voltage curve and corresponding current curve of each battery pack, and forms time series data according to the sampling period.

[0068] (5) Pack assembly process data acquisition In the PACK post-assembly process island, the system collects assembly completion data such as assembly completion signals, appearance inspection results, and waterproof test results, providing a basis for finished product judgment.

[0069] (6) Time series database storage and query The unified data platform is configured with a time series database, which stores the above types of data according to the dimensions of "process island - workstation - product identifier - timestamp". It supports quick queries by time interval, process island, product batch and other conditions, providing a data foundation for subsequent model training, cycle time analysis and quality traceability.

[0070] In the above embodiments, preferably, the unified data platform adopts a microservice architecture based on an API gateway to interact with production line data in the module segment and PACK segment, integrates data from all processes in the entire production line, and supports flexible system expansion and function upgrades based on the unified data platform.

[0071] During implementation, a unified data bus is established, a data serialization format is adopted, and a data exchange protocol between module segments and PACK segments is defined. Data association between preceding and following processes is achieved through product serial numbers, and the system automatically verifies data integrity and consistency. When data anomalies occur, an early warning mechanism is triggered to prevent defective products from flowing into the next process. Data recognition employs an intelligent matching algorithm, supporting automatic conversion and mapping of data in different formats.

[0072] In the above embodiments, preferably, the island-type power battery production data management method further includes: Collect the work-in-process quantity, waiting queue length, equipment runtime, discharge speed and material consumption rate of each process island, and calculate the real-time capacity load rate of each process island in real time. When it is detected that the number of work-in-process products on the first process island exceeds the first threshold and the capacity utilization rate of the downstream second process island is lower than the target range, a production task matching the backlog product model of the first process island is issued to the second process island. When the length of the waiting queue of the third process island in each process island exceeds the queue warning length, and the equipment running time percentage exceeds the intervention limit, the third process island is determined to be in an overload state. When the duration for which the discharge rate of the first process island is higher than the material consumption rate of the second process island is greater than the duration of the higher discharge rate, a control command is generated to reduce the discharge cycle time of the first process island or to reduce the transmission speed of the material flow system.

[0073] During implementation, to achieve balanced production capacity across upstream and downstream process islands, a unified data platform comprehensively collects production data from all processes on each process island, including the quantity of work-in-process (WIP), queue length, equipment runtime, discharge speed, and material consumption rate. Simultaneously, the real-time capacity load rate is calculated by dividing the actual output of each process island within a time window by the preset maximum output. The WIP quantity is counted using scanners (such as RFID or visual recognition) to track the number of unfinished products located in the processing area, buffer area, and conveyor lines of the process island. The queue length is obtained in real-time from the scheduling queue, showing the number of tasks waiting for processing on the process island in the material buffer. Equipment runtime is read from the equipment's PLC, showing its net running time in the previous statistical period. The discharge speed is the rate at which the output products from one process island are delivered to its downstream process island. The material consumption rate is the rate at which materials are consumed during production on the process island. Reducing the transmission speed of the material flow system involves reducing the speed of automated guided vehicles (AGVs) or the frequency of conveyor belts connecting upstream and downstream process islands.

[0074] The system monitors the production and work-in-process quantities of each process island in real time to determine if there are bottlenecks in a particular process island or insufficient material supply to downstream process islands. Specifically, the system sets dynamically adjustable target ranges, warning upper limits, warning lower limits, intervention upper limits, and intervention lower limits for various indicators and composite load rates of each process island. For example, when the work-in-process quantity of a process island (defined as the first process island) consistently exceeds the intervention upper limit (e.g., set to 90% of the island's buffer capacity), and the capacity load rate of its downstream second process island is below the target range lower limit (e.g., 65%), the system determines that the first process island is "congested," while the second process island is in a "starved" state, indicating a capacity mismatch. As another example, when the queue length of a process island exceeds the queue warning length (e.g., the island's 2-hour processing capacity), and the equipment runtime percentage (i.e., utilization rate) has exceeded the intervention upper limit (e.g., 92%), the system determines that the process island is in an "overloaded" state.

[0075] When a significant deviation is detected between a process island and the preset ideal production output—for example, if the first process island is congested and the second process island is underutilized—production tasks that perfectly match the backlog of products on the first process island are prioritized for distribution to the second process island. This is achieved by modifying the work order priority in the Manufacturing Execution System (MES), allowing the second process island to "pull" the backlog of work-in-process inventory, rather than processing new tasks in the original order. Simultaneously, the material feeding rate upstream of the first process island can be temporarily reduced. In other words, the material feeding sequence of production orders is dynamically adjusted.

[0076] If the output rate of the upstream process island consistently exceeds the consumption rate of the downstream process island, and the duration exceeds a preset duration (which is manually set), meaning the upstream work-in-process (WIP) is decreasing slowly while the downstream WIP is increasing rapidly, the system can generate control commands to reduce the speed of the automated guided vehicles or the frequency of the conveyor belts connecting the upstream and downstream process islands (e.g., from 1.0 m / s to 0.7 m / s) to physically limit the rate at which WIP flows downstream. Additionally, commands can be sent to the controller of the congested upstream process island to extend its production cycle time (e.g., extend the cycle by 5%), reducing its output speed at the source until the system returns to balance.

[0077] The system periodically (e.g., every 30 minutes) scans the real-time capacity load rate of all process islands. If the process islands with the highest load rate (system bottleneck) and the lowest load rate (capacity surplus) are identified, a digital twin model is used for simulation to assess the feasibility of transferring or sharing some non-core tasks (such as pre-assembly and inspection) from the bottleneck workstation to the surplus workstation. After virtual verification, some process paths are reconstructed on the actual production line to achieve capacity balancing.

[0078] This invention also proposes an island-type power battery production data management system, which applies the island-type power battery production data management method disclosed in any of the above embodiments, and includes the following functional modules: The production data acquisition module is deployed on edge nodes and a unified data platform. It is used to collect production data from each process island in real time through the unified data platform and adopts a distributed architecture based on cloud-edge collaboration. This data includes OCV detection data, laser welding data, PACK pre- and post-assembly data, charge and discharge test data, and equipment operating status signals, and is then uniformly aggregated into the data platform.

[0079] The data evaluation and screening module, deployed on a cloud-based data platform, is used to evaluate the quality of products from the previous process island based on production data from each process island, employing an intelligent prediction model built on machine learning. It outputs the probability of quality compliance and verifies and screens products based on the evaluation results. The intelligent prediction model is trained using historical quality data. This module sends screening instructions to the corresponding process island's actuators via a control interface, achieving pre-process verification and screening.

[0080] The operation status monitoring module is used to monitor the operation status of each process island equipment based on the collected production data, identify the equipment's running, idle, and fault states, and display the operation status of each process island equipment in real time on the system interface, providing a monitoring view for operation and maintenance personnel.

[0081] The data traceability and recording module integrates a blockchain client to record the assembly information and timestamps of materials based on the collected production data using blockchain technology. It packages and writes key material assembly information, timestamps, supplier digital signatures, and other data into the blockchain network, realizing on-chain storage and full lifecycle traceability of material information. It also provides an on-chain data query interface to trace information of key materials in the power battery production and assembly process.

[0082] The data mapping adjustment module, integrating a digital twin engine, consolidates production data from various processes through a unified data platform. Using digital twin technology, it establishes a virtual mapping model of the production line, mapping the real-time status of each process island and workstation onto the virtual model. Based on changes in real-time data, it adjusts the status of the corresponding production process on the virtual mapping model. According to cycle time analysis and quality analysis results, it identifies bottleneck workstations and defect tracing paths, and adjusts the current production cycle time. This module further integrates with a cloud-edge collaboration mechanism, automatically generating cycle time adjustment suggestions or process sequence adjustment strategies based on status changes in the virtual model, and distributing them to the process island controller via a message queue, achieving a closed loop from data analysis to production control.

[0083] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for managing production data of island-type power batteries, characterized in that, include: Through a unified data platform, a distributed architecture based on cloud-edge collaboration is adopted to collect production data from each process island in real time. Based on the production data of each process island, a machine learning-based intelligent prediction model is used to assess the quality of the products from the previous process island, and the products are verified and screened based on the assessment results. The intelligent prediction model is trained using historical quality data. Based on the collected production data, monitor the operating status of equipment in each process island; Based on the collected production data, the assembly information and timestamps of materials are recorded using blockchain technology to enable information traceability of key materials in the power battery production and assembly process. By integrating production data from various processes through a unified data platform, a virtual mapping model of the production line is established using digital twin technology. The state of the corresponding production process on the virtual mapping model is adjusted based on changes in real-time data, and the current production cycle is regulated.

2. The island-type power battery production data management method according to claim 1, characterized in that, Also includes: The production plan for power batteries is scheduled, the production quantity of power batteries is managed, and the process parameters of power battery products are retrieved from the process parameter database. The production plan, production quantity, and process parameters are respectively sent to the controllers of each process island on the production line, and the required type of power battery products are produced according to the corresponding process parameters, specifically including: Based on production needs, the battery model, order quantity, order priority, and delivery timestamp of the power battery products are determined, and the production plan of the power battery products is obtained through scheduling. Based on the battery model to be produced according to the production plan, the process parameters of the corresponding battery model are automatically retrieved from the process parameter database; The production plan, the production quantity, and the process parameters are serialized in a preset format and asynchronously transmitted to the controller of the corresponding process island via a message queue. The process island controls the actuators to carry out production based on the received battery model and process parameters; During the production process, based on the real-time equipment status and material characteristics collected, the process parameters are dynamically and adaptively optimized using a continuously updated neural network model with an online learning mechanism. This includes adjusting the welding current and welding speed of the welding process parameters, adjusting the charging and discharging parameters of the battery cells, and adjusting the torque and pressure of the wire harness assembly process. The neural network model is trained based on historical production data.

3. The island-type power battery production data management method according to claim 1, characterized in that, The process of using a machine learning-based intelligent prediction model to assess the quality of products from the previous process island, and then verifying and screening the products based on the assessment results, includes: Train a machine learning-based intelligent prediction model using historical quality data; Using a pre-verification method, when a product completes the processing of the previous process island and enters the waiting buffer of the next process, all process data of the previous process are retrieved, and the intelligent prediction model is used to conduct an online quality assessment of the product of the previous process island. The product is deemed qualified based on the assessed probability of quality compliance, and unqualified products are removed.

4. The island-type power battery production data management method according to claim 1, characterized in that, Information traceability is implemented for key materials in the production and assembly process of power batteries. The specific process includes: Equip the key materials of the power battery with RFID and QR code identification; During the production and assembly process, blockchain technology is used to collect RFID and QR code identification information of each key material being assembled, and a unique traceability label is generated for each battery pack. Based on blockchain technology, the configuration identification and assembly association of the battery cells, housing and wiring harness are recorded throughout the entire life cycle, and the source of materials is traced based on the blockchain record data.

5. The island-type power battery production data management method according to claim 4, characterized in that, Also includes: Based on blockchain technology, the raw information of the manufacturer's incoming materials and battery materials, along with timestamps and supplier digital signatures, are uploaded to the blockchain. The material information is stored, managed, and maintained based on the blockchain information.

6. The island-type power battery production data management method according to claim 1, characterized in that, By integrating production data from various processes through a unified data platform, and establishing a virtual mapping model of the production line using digital twin technology, the state of the corresponding production process on the virtual mapping model is adjusted based on real-time data changes. Specifically, this includes: The collected production data is comprehensively analyzed, and a virtual mapping model of the production line is established using digital twin technology; Based on the equipment operation signals and production cycle data of each process island in the production data, the deviation between the production cycle and waiting time of each process and the theoretical cycle time is calculated based on the time series analysis method. The real-time position of the product in each process island is visualized based on the virtual mapping model to realize production line cycle time analysis. Based on the results of the production line cycle time analysis, the equipment utilization rate of each workstation is calculated, the key workstation that has the greatest impact on production efficiency is identified as the production bottleneck workstation, and it is dynamically marked in the virtual mapping model.

7. The island-type power battery production data management method according to claim 6, characterized in that, Also includes: Based on the machine vision inspection data, sensor measurement data, and online test data of each process island in the production data, machine learning algorithms are used for pattern recognition. Based on the recognition results, the quality fluctuation trend of the product and the defect distribution rate of each process island are mapped in real time according to the virtual mapping model to achieve quality analysis. Based on the results of the quality analysis, a clustering algorithm is used to identify similar defect patterns, and the process that caused the defect is determined based on the similar defect identification results. The defect station is then located in the virtual mapping model, and adjacent defect stations are connected to construct a defect tracing path.

8. The island-type power battery production data management method according to claim 6, characterized in that, By integrating production data from various processes through a unified data platform, and establishing a virtual mapping model of the production line using digital twin technology, the state of the corresponding production process on the virtual mapping model is adjusted based on real-time data changes, and the actual production cycle time is also adjusted. Specifically, this includes: By deploying edge nodes on each process island, real-time data of the corresponding process island is collected and uploaded to the cloud; the real-time data includes the number of products in process, equipment utilization rate, product qualification rate, and the product qualification rate of products waiting to flow into downstream in the buffer zone; Using cloud computing capabilities, control instructions are generated based on the real-time data. The process of generating control instructions is as follows: For the first upstream process island, when the work-in-process quantity of the corresponding downstream process island is lower than the lower limit of work-in-process, and the equipment utilization rate of the corresponding downstream process island is lower than the target utilization rate, and the product qualification rate of the first upstream process island within a preset time window is greater than or equal to the first quality threshold, and the product qualification rate of the buffer of the first upstream process island to flow into the downstream is greater than the first preset qualification rate, then control instructions to shorten the production task issuance interval and shorten the production cycle are generated; when the work-in-process quantity of the corresponding downstream process island exceeds the upper limit of work-in-process, and / or, the equipment utilization rate of the corresponding downstream process island exceeds the utilization rate upper limit, or, the qualification rate of the first upstream process island within a time window is lower than the second quality threshold and the qualification rate of the product to flow into the corresponding downstream process island is less than or equal to the first preset qualification rate, then control instructions to extend the production task issuance interval and extend the production cycle are generated. The state of the corresponding production process on the virtual mapping model is adjusted according to the control instructions.

9. The island-type power battery production data management method according to claim 1, characterized in that, The real-time acquisition of production data from each process island specifically includes: The system collects real-time data on voltage and internal resistance during OCV cell testing, welding current and temperature during laser welding, torque and pressure during PACK pre-assembly, voltage and current curves during charge and discharge testing, and assembly completion data during PACK post-assembly. This data is then stored and retrieved using a time-series database.

10. The island-type power battery production data management method according to claim 1, characterized in that, Also includes: Collect the work-in-process quantity, waiting queue length, equipment runtime, discharge speed and material consumption rate of each process island, and calculate the real-time capacity load rate of each process island in real time. When it is detected that the number of work-in-process products on the first process island exceeds the first threshold and the capacity utilization rate of the downstream second process island is lower than the target range, a production task matching the backlog product model of the first process island is issued to the second process island. When the length of the waiting queue of the third process island in each process island exceeds the queue warning length, and the equipment running time percentage exceeds the intervention limit, the third process island is determined to be in an overload state. When the duration for which the discharge rate of the first process island is higher than the material consumption rate of the second process island is greater than the duration of the higher discharge rate, a control command is generated to reduce the discharge cycle time of the first process island or to reduce the transmission speed of the material flow system.

11. A data management system for island-type power battery production, characterized in that, The method for managing production data of island-type power batteries as described in any one of claims 1 to 10 includes: The production data acquisition module is used to collect production data from each process island in real time through a unified data platform and a distributed architecture based on cloud-edge collaboration. The data evaluation and screening module is used to evaluate the quality of the products from the previous process island based on the production data of each process island using an intelligent prediction model built on machine learning, and to verify and screen the products based on the evaluation results. The intelligent prediction model is trained using historical quality data. The operation status monitoring module is used to monitor the operation status of each process island equipment based on the collected production data; The data traceability and recording module is used to record the assembly information and timestamps of materials based on the collected production data using blockchain technology, enabling information traceability of key materials in the power battery production and assembly process. The data mapping adjustment module is used to integrate production data from various processes through a unified data platform, establish a virtual mapping model of the production line using digital twin technology, adjust the state of the corresponding production process on the virtual mapping model based on real-time data changes, and adjust the current production cycle.

Images