Silicon-carbon negative electrode material large-scale production process precision control system and method

Abstract

This invention relates to the field of precision control technology in production processes, specifically to a precision control system and method for the large-scale production of silicon-carbon anode materials. It includes a data acquisition module, a quality mapping module, a model building module, a parameter optimization module, and a model updating module. The data acquisition module collects production data from each process in the silicon-carbon anode material production process and binds the production data with corresponding batches of products to obtain batch-related production data. Simultaneously, it extracts product quality characterization parameters based on the batch-related production data and classifies these parameters to obtain key quality attributes. By constructing a full-process, multi-dimensional data acquisition and quality feature extraction system, real-time acquisition of process parameters, equipment operating parameters, and environmental parameters for all processes in silicon-carbon anode material production is achieved. Furthermore, a unique batch number enables full-link binding of production data with product batches, forming a standardized batch-related production dataset.

Description

Technical Field

[0001] This invention relates to the field of precision control technology for production processes, and more specifically, to a precision control system and method for the large-scale production process of silicon-carbon anode materials. Background Technology

[0002] With the rapid development of the global new energy vehicle and energy storage industry, the market demand for lithium-ion batteries has experienced explosive growth. Silicon-carbon anode materials, with their theoretical specific capacity far exceeding that of traditional graphite anodes, have become the core material for the next generation of high-energy-density lithium-ion batteries.

[0003] Currently, the core electrochemical performance indicators of silicon-carbon anode materials, such as initial charge-discharge efficiency, cycle capacity, and rate performance, require more than 4 hours to test and involve irreversible damage. This makes it impossible to sample and test in real time on the production line, resulting in a serious lag in quality feedback. Quality problems are often only discovered after the entire batch of products has been produced, causing a significant waste of raw materials, energy, and production time. Secondly, the independent control of each process does not fully consider the strong coupling relationship between each process. For example, changes in sintering temperature directly affect the particle size distribution of the subsequent pulverization process, which in turn affects the compaction density and electrochemical performance of the final product. Existing technologies cannot achieve global collaborative optimization of process parameters throughout the entire process, making it difficult to stably control product quality. Therefore, this paper proposes a precise control system and method for the large-scale production process of silicon-carbon anode materials. Summary of the Invention

[0004] The purpose of this invention is to provide a precise control system and method for the large-scale production process of silicon-carbon anode materials, so as to solve the problems mentioned in the background art.

[0005] To address the aforementioned technical problems, one objective of this invention is to provide a precision control system for the large-scale production process of silicon-carbon anode materials, comprising a data acquisition module, a quality mapping module, a model building module, a parameter optimization module, and a model updating module. The data acquisition module is used to collect production data of each process in the production of silicon-carbon anode materials, and bind the production data with the corresponding batch of products to obtain batch-related production data. At the same time, it extracts product quality characterization parameters based on batch-related production data, classifies the product quality characterization parameters, and obtains key quality attributes. The quality mapping module is used to establish a mapping relationship model between process parameters and key quality attributes based on batch-related production data and key quality attributes, and to establish a soft measurement model based on the mapping relationship model to predict key quality attributes that are difficult to detect in real time during the production process. The model building module is used to perform time-series alignment and correlation processing on batch-related production data. Combined with the mapping relationship model and the soft measurement model, a process digital twin model is constructed to characterize the production process, equipment operating status and material structure evolution relationship. The parameter optimization module is used to set a target quality attribute range based on key quality attributes, and to perform collaborative optimization calculations on the control parameters in the production process based on the process digital twin model and the target quality attribute range to obtain the optimal control parameter combination. Then, the corresponding production equipment is adjusted based on the optimal control parameter combination to perform closed-loop control of the production process, so that the actual operating state of the production process approaches the target quality attribute range. The model update module is used to obtain the actual test results of the batch of products after the production of a single batch is completed, and to perform deviation analysis between the actual test results and the prediction results output by the soft measurement model. Based on the deviation analysis results, the mapping relationship model, the soft measurement model and the process digital twin model are updated.

[0006] As a further improvement to this technical solution, the data acquisition module includes processes such as mixing, granulation, drying, sintering, pulverizing, and sieving in the silicon-carbon anode material production process. The production data for each process includes the corresponding process parameters, equipment operating parameters, and environmental parameters. Each batch of silicon-carbon anode material is assigned a unique batch number. All production data of that batch of products from the mixing process to the screening process are associated with that batch number, forming batch-related production data that uses the batch number as an index and includes production data of each process. Each set of production data can be matched with a specific batch of products.

[0007] As a further improvement to this technical solution, the data acquisition module filters out parameters that can reflect the quality of silicon-carbon anode materials from the obtained batch-related production data. The product quality characterization parameters include direct quality characterization parameters and indirect quality characterization parameters. Direct quality characterization parameters are parameters that are directly related to the quality of the product itself and are obtained directly through detection methods. These include the product's particle size distribution, specific surface area, porosity, initial charge-discharge efficiency, cycle capacity, compaction density, and purity. The indirect quality characterization parameters cannot directly reflect product quality, but are significantly related to product quality and can be indirectly derived from production data. These include parameter fluctuation values ​​in each production process, equipment operation stability parameters, and material loss rate. Correlation analysis was used to analyze all the extracted product quality characterization parameters. Parameters with low correlation to the core quality indicators of silicon-carbon anode materials and weak impact on product quality were removed, while parameters with high correlation and decisive effect on product quality were retained as key quality attributes.

[0008] As a further improvement to this technical solution, in the quality mapping module, the process parameters of each process in the obtained batch-related production data are used as input variables, and the obtained key quality attributes are used as output variables. Multiple batches of batch-related production data and corresponding key quality attribute data are collected as training samples. Machine learning algorithms are used to train the training samples to construct a mapping relationship model between process parameters and key quality attributes. Among them, the key quality attributes that are difficult to detect in real time during the production process are those that take a long time to detect, that can damage the product during the detection process, and those that cannot be sampled and detected in real time during the production process. Based on the mapping relationship model, and combined with the process parameters, equipment operating parameters and key quality attributes that can be collected in real time during the production process, and supplemented with real-time collected data as input, the model structure is optimized and trained to obtain a soft measurement model. The soft measurement model can predict the values ​​of key quality attributes that are difficult to detect in real time based on the real-time data in the production process.

[0009] As a further improvement to this technical solution, in the model building module, the batch-related production data of each batch are sorted according to the time sequence of the production process, the timestamp format of the production data of each process is unified, and the production data collected at different processes and different time points are time-series aligned to ensure that the production data of the same production stage can be matched accordingly. Then, the time-aligned production data is correlated to uncover the inherent correlations between production data of different processes, between production data and key quality attributes, and between production data and equipment operating status, thus obtaining a time-correlated production dataset. Based on the time-series associated production dataset, the mapping relationship between process parameters and key quality attributes of the mapping relationship model and the real-time prediction function of the soft measurement model are integrated into the model building process to build a digital twin scenario corresponding to the actual silicon-carbon anode material production process, thereby obtaining a process digital twin model.

[0010] As a further improvement to this technical solution, the parameter optimization module sets a reasonable target range for each key quality attribute based on the application scenario requirements, industry standards, and product quality requirements of silicon-carbon anode materials, combined with the key quality attribute range of qualified products in historical production data. Among them, the target quality attribute range is the acceptable range of key quality attributes, and the range can be adjusted according to actual production needs and product specifications. The target quality attribute range is used as the optimization objective, and the control parameters of each process in the production process are used as optimization variables. The process digital twin model is used to simulate the production process and corresponding key quality attributes under different combinations of control parameters. The genetic algorithm is used to optimize the control parameters in a coordinated manner. Under the premise of meeting the requirements of the target quality attribute range and taking into account production efficiency, energy consumption and cost, the optimal combination of control parameters is calculated.

[0011] As a further improvement to this technical solution, the parameter optimization module converts the optimal control parameter combination into control commands that can be recognized by each production equipment, and sends them to the production equipment of the corresponding process through the control system. The operating parameters and process execution parameters of the equipment are adjusted in real time, and the production parameters of each process operate according to the optimal control parameter combination. During equipment adjustment, production data, equipment operation data, and key quality attributes that can be detected in real time are collected in real time. The real-time data is fed back to the process digital twin model and soft measurement model to compare the deviation between the actual key quality attributes and the target quality attribute range in real time. Set a deviation range. When the deviation exceeds the deviation range, perform optimization calculations again through the process digital twin model and adjust the control parameters until the key quality attributes under actual operation approach the target quality attribute range.

[0012] As a further improvement to this technical solution, in the model update module, after the production of a single batch of silicon-carbon anode material is completed, a comprehensive test is conducted on the batch of products to obtain the actual test values ​​of all key quality attributes and form an actual test report for the batch of products. Calculate the deviation between the actual measured values ​​of each key quality attribute of the batch of products and the predicted values ​​output by the soft sensor model during the production process of the batch, and analyze the reasons for the deviation. Based on the magnitude of the deviation and the cause of the deviation, the training samples of the mapping relationship model are adjusted, the model parameters are optimized, and the mapping accuracy between process parameters and key quality attributes is improved. At the same time, the input variable weights of the soft measurement model are corrected, the prediction algorithm is optimized, and the prediction deviation is reduced, thereby updating the production process parameters, equipment operating parameters, and material structure evolution laws in the process digital twin model.

[0013] The second objective of this invention is to provide a method for precise control of the large-scale production process of silicon-carbon anode materials, based on any one of the above-mentioned precise control systems for the large-scale production process of silicon-carbon anode materials, comprising the following steps: S1. Collect production data of each process in the production of silicon-carbon anode materials, and bind the production data with the corresponding batch of products to obtain batch-related production data. At the same time, extract product quality characterization parameters based on batch-related production data, classify the product quality characterization parameters, and obtain key quality attributes. S2. Based on batch-related production data and key quality attributes, establish a mapping relationship model between process parameters and key quality attributes, and based on the mapping relationship model, establish a soft measurement model for predicting key quality attributes that are difficult to detect in real time during the production process. S3. Perform time-series alignment and correlation processing on batch-related production data, and combine the mapping relationship model and soft measurement model to construct a process digital twin model to characterize the production process, equipment operating status and material structure evolution relationship. S4. Based on the key quality attributes, set the target quality attribute range, and based on the process digital twin model and the target quality attribute range, perform collaborative optimization calculations on the control parameters in the production process to obtain the optimal control parameter combination. Then, based on the optimal control parameter combination, adjust the corresponding production equipment to perform closed-loop control of the production process, so that the actual operating state of the production process approaches the target quality attribute range. S5. After a single batch of production is completed, obtain the actual test results of the batch of products, and perform a deviation analysis between the actual test results and the predicted results output by the soft measurement model. Based on the deviation analysis results, update the mapping relationship model, the soft measurement model, and the process digital twin model.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. A precision control system and method for the large-scale production process of silicon-carbon anode materials. By constructing a multi-dimensional data acquisition and quality feature extraction system for the entire process, the system achieves real-time acquisition of process parameters, equipment operating parameters, and environmental parameters for all processes in the silicon-carbon anode production. Furthermore, by using a unique batch number, the system achieves full-chain binding of production data with product batches, forming a standardized batch-related production dataset. Based on this, Pearson correlation analysis is used to accurately identify key quality attributes that play a decisive role in the core quality of the product, clarifying the quantitative correlation between process parameters and product quality. This completely solves the problem of unclear quality influencing factors in traditional methods, providing a reliable data foundation for subsequent modeling analysis and precise control.

[0015] 2. A precision control system and method for the large-scale production process of silicon-carbon anode materials. Based on the XGBoost algorithm, a high-precision mapping model between process parameters and key quality attributes is constructed, ensuring a model determination coefficient of no less than 0.92. Furthermore, real-time measurable input variables are expanded upon this model to optimize a soft-sensor model. This soft-sensor model can predict key electrochemical performance indicators that are difficult to detect in real time, such as initial charge-discharge efficiency and cycle capacity, at 1-second intervals based on real-time data streams during the production process. This fundamentally solves the industry pain point of lagging traditional offline detection, enabling real-time monitoring of quality status during production, early warning of quality anomalies, and effectively reducing the generation of defective products. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall structure of the present invention. Detailed Implementation

[0017] 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, and 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.

[0018] like Figure 1 As shown, one of the objectives of this invention is to provide a precision control system for the large-scale production process of silicon-carbon anode materials, including a data acquisition module, a quality mapping module, a model building module, a parameter optimization module, and a model updating module. The data acquisition module is used to collect production data of each process in the production of silicon-carbon anode materials, and bind the production data with the corresponding batch of products to obtain batch-related production data. At the same time, based on the batch-related production data, product quality characterization parameters are extracted and classified to obtain key quality attributes. The data acquisition module includes data on various processes in the silicon-carbon anode material production process, such as mixing, granulation, drying, sintering, pulverizing, and sieving. Production data for each process includes corresponding process parameters, equipment operating parameters, and environmental parameters. Each batch of silicon-carbon anode material is assigned a unique batch number. All production data of that batch of products from the mixing process to the screening process are associated with that batch number, forming batch-related production data that uses the batch number as an index and includes production data of each process. Each set of production data can be matched with a specific batch of products.

[0019] Each batch of silicon-carbon anode material is assigned a globally unique batch number. All production data of that batch, from the start of the mixing process to the end of the screening process, is linked to the corresponding batch number one by one according to the production time sequence. This forms a batch-linked production dataset with the batch number as the unique index and containing production data of all processes and dimensions, enabling accurate traceability of any set of production data to the corresponding product batch. In the data acquisition module, parameters that can reflect the quality of silicon-carbon anode materials are selected from the batch-related production data. The product quality characterization parameters include direct quality characterization parameters and indirect quality characterization parameters. Direct quality characterization parameters are parameters that are directly related to the quality of the product itself and are obtained directly through detection methods. These include the product's particle size distribution, specific surface area, porosity, initial charge-discharge efficiency, cycle capacity, compaction density, and purity. Indirect quality characterization parameters are parameters that cannot directly reflect product quality but are significantly related to product quality and can be indirectly derived from production data. These include parameter fluctuations in each production process, equipment operating stability parameters, and material loss rates. Correlation analysis was used to analyze all the extracted product quality characterization parameters. Parameters with low correlation to the core quality indicators of silicon-carbon anode materials and weak impact on product quality were removed, while parameters with high correlation and decisive effect on product quality were retained as key quality attributes.

[0020] Pearson correlation analysis was used to quantitatively analyze all screened product quality characterization parameters. The correlation coefficients between each quality characterization parameter and the core quality indicators of silicon-carbon anode materials were calculated. A correlation threshold of 0.7 was set, and parameters with absolute correlation coefficients below 0.7 and weak impact on product quality were removed. Parameters with absolute correlation coefficients above 0.7 and decisive impact on product quality were retained as the final key quality attributes. The formula is as follows: ; in, Let x be the Pearson correlation coefficient between the quality characterization parameter x and the core quality indicator y, with a value range of [-1, 1]. Let be the quality characterization parameter value of the i-th sample. Let i be the core quality index value corresponding to the i-th sample. This represents the average of all sample quality characterization parameters. This represents the average of the core quality indicators for all samples. This represents the total number of sample batches used in the analysis.

[0021] The quality mapping module is used to establish a mapping relationship model between process parameters and key quality attributes based on batch-related production data and key quality attributes, and to establish a soft measurement model based on the mapping relationship model to predict key quality attributes that are difficult to detect in real time during the production process. In the quality mapping module, the process parameters of each process in the obtained batch-related production data are used as input variables, and the obtained key quality attributes are used as output variables. Multiple batches of batch-related production data and corresponding key quality attribute data are collected as training samples. Machine learning algorithms are used to train the training samples to build a mapping relationship model between process parameters and key quality attributes. Using the full-process parameters from batch-related production data as input variables and the selected key quality attributes as output variables, data from at least 300 qualified production batches were collected as the original sample set. The original sample set underwent Z-score standardization preprocessing to eliminate dimensional differences between parameters. The preprocessed sample set was then randomly divided into training, validation, and test sets in an 8:1:1 ratio for model training, hyperparameter tuning, and generalization performance testing, respectively.

[0022] The XGBoost algorithm is used as the core modeling algorithm. Using the training set data as input, iterative training is conducted to obtain a nonlinear mapping model between process parameters and key quality attributes. During training, validation set data is used to monitor model overfitting in real time. An early stopping mechanism terminates training when the validation set error does not decrease for 10 consecutive rounds. The generalization performance of the model is validated using test set data, ensuring that the model's coefficient of determination R² ≥ 0.92. Among them, the key quality attributes that are difficult to detect in real time during the production process are those that take a long time to detect, that can damage the product during the detection process, and those that cannot be sampled and detected in real time during the production process. Quality indicators that require more than 4 hours of testing, cause irreversible damage to the product, or cannot be continuously sampled and tested on the production line include the first charge-discharge efficiency, cycle capacity, rate performance, and long-term cycle stability of silicon-carbon anode materials. Based on the mapping relationship model, and combined with the process parameters, equipment operating parameters and key quality attributes that can be collected in real time during the production process, and supplemented with real-time collected data as input, the model structure is optimized and trained to obtain a soft measurement model. The soft measurement model can predict the values ​​of key quality attributes that are difficult to detect in real time based on the real-time data in the production process.

[0023] The optimized model was fine-tuned using data from the latest 50 production batches to obtain the final soft measurement model. This model can output predicted values ​​of difficult-to-measure key quality attributes at 1-second intervals based on real-time data streams during the production process.

[0024] The model building module is used to perform time-series alignment and correlation processing on batch-related production data. Combined with the mapping relationship model and the soft measurement model, it constructs a process digital twin model to characterize the production process, equipment operating status and material structure evolution relationship. In the model building module, the batch-related production data of each batch is sorted according to the time sequence of the production process, the timestamp format of the production data of each process is unified, and the production data collected at different processes and different time points are time-series aligned to ensure that the production data of the same production stage can be matched accordingly. Extract the trigger timestamps for each batch and each process step, including mixing and feeding, granulation start-up, drying feeding, sintering feeding, crushing feeding, and screening feeding, and use these as the baseline time nodes for each process. Standardize the timestamp format of all sensor and equipment data to UTC milliseconds, and divide raw data from different sources into corresponding baseline time intervals according to their respective processes. For data with different sampling frequencies within the same process, use linear interpolation to unify them to a standard sampling frequency of 1Hz, fill in missing time point data, and ensure that process parameters, equipment operating parameters, and environmental parameters at the same production stage correspond and match on the timeline. Then, the time-aligned production data is correlated to uncover the inherent correlations between production data of different processes, between production data and key quality attributes, and between production data and equipment operating status, thus obtaining a time-correlated production dataset. Granger causality analysis was employed to perform correlation processing on the time-series aligned production data. Autoregressive models were constructed between process parameters of different procedures, between process parameters and equipment operating parameters, and between process parameters and key quality attributes. The F-statistic for causality testing was calculated, with a significance level set at 0.05. True causal associations that passed the significance test were retained, while spurious and irrelevant associations were eliminated. The validated associations were integrated with the time-series aligned data to form a time-series correlated production dataset containing time, parameter, and causal association dimensions. Based on the time-series associated production dataset, the mapping relationship between process parameters and key quality attributes of the mapping relationship model and the real-time prediction function of the soft measurement model are integrated into the model building process to build a digital twin scenario corresponding to the actual silicon-carbon anode material production process, thereby obtaining a process digital twin model.

[0025] Using time-series correlated production datasets as the data foundation, a four-layer process digital twin model is built.

[0026] The first layer is the physical entity mapping layer, which constructs three-dimensional geometric models and physical attribute models of all production equipment such as mixers, granulators, drying kilns, sintering furnaces, crushers, and screening machines, and restores the spatial layout and material flow path of the production workshop.

[0027] The second layer is the data interaction layer, which establishes a two-way real-time data channel between the digital twin model and the actual production PLC and SCADA system, enabling millisecond-level synchronous transmission of production data and issuance of control commands.

[0028] The third layer is the model fusion layer, which embeds the trained process parameter-key quality attribute mapping model and soft measurement model into the digital twin framework, enabling the digital twin model to have the ability to quantitatively analyze process impacts and predict difficult-to-measure quality attributes in real time.

[0029] The fourth layer is the process characterization layer, which dynamically characterizes the material flow status, process parameter change trends, equipment operating health (normal / early warning / fault) during the production process based on real-time data and fusion models, as well as the particle morphology, crystal structure, and pore structure evolution of silicon-carbon anode materials in each process, ultimately obtaining a complete process digital twin model.

[0030] The parameter optimization module is used to set the target quality attribute range based on key quality attributes, and to perform collaborative optimization calculations on the control parameters in the production process based on the process digital twin model and the target quality attribute range to obtain the optimal control parameter combination. Then, the corresponding production equipment is adjusted based on the optimal control parameter combination to perform closed-loop control of the production process, so that the actual operating state of the production process approaches the target quality attribute range. In the parameter optimization module, based on the application scenario requirements, industry standards and product quality requirements of silicon-carbon anode materials, and combined with the key quality attribute range of qualified products in historical production data, a reasonable target range is set for each key quality attribute. Among them, the target quality attribute range is the acceptable range of key quality attributes, and the range can be adjusted according to actual production needs and product specifications. Based on the application scenarios of silicon-carbon anode materials, the national industry standard GB / T38823-2020, and customer-customized quality requirements, the benchmark acceptable range for each key quality attribute is determined. Combining historical data of key quality attributes from qualified batches of products over the past 12 months, the statistical fluctuation range of each parameter is calculated using the 3σ principle. The benchmark range is then corrected, and a reasonable target range is finally set for each key quality attribute. This range can be dynamically updated according to changes in product specifications, adjustments to production processes, or changes in customer needs. The target quality attribute range is used as the optimization objective, and the control parameters of each process in the production process are used as optimization variables. The process digital twin model is used to simulate the production process and corresponding key quality attributes under different combinations of control parameters. The genetic algorithm is used to optimize the control parameters in a coordinated manner. Under the premise of meeting the requirements of the target quality attribute range and taking into account production efficiency, energy consumption and cost, the optimal combination of control parameters is calculated.

[0031] A multi-objective collaborative optimization mathematical model is constructed, with the key quality attributes meeting the target range requirements as the core constraint, and the maximization of production efficiency, the minimization of energy consumption per unit product, and the minimization of production cost per unit product as auxiliary optimization objectives. Adjustable control parameters of the entire process of mixing, granulation, drying, sintering, crushing, and screening are used as optimization variables, and the value range and process constraints of each optimization variable are clarified.

[0032] Using a process digital twin model as a virtual simulation platform, an improved genetic algorithm with elite retention is employed to solve for optimal control parameters. The initial population size is 100, with a crossover probability of 0.8 and a mutation probability of 0.05. Each set of candidate control parameters is input into the digital twin model to simulate the complete production process and output corresponding key quality attributes, production efficiency, energy consumption, and cost data, calculating individual fitness values. In each generation of evolution, the top 10% of individuals with the highest fitness are retained as elites and directly enter the next generation. The remaining individuals generate a new population through selection, crossover, and mutation operations. Iterative evolution continues until the termination condition is met (50 iterations or no significant improvement in fitness value for 10 consecutive generations), outputting the globally optimal combination of control parameters.

[0033] In the parameter optimization module, the optimal combination of control parameters is converted into control commands that can be recognized by each production equipment. These commands are then sent to the production equipment of the corresponding process through the control system. The operating parameters and process execution parameters of the equipment are adjusted in real time, and the production parameters of each process are operated according to the optimal combination of control parameters. During equipment adjustment, production data, equipment operation data, and key quality attributes that can be detected in real time are collected in real time. The real-time data is fed back to the process digital twin model and soft measurement model to compare the deviation between the actual key quality attributes and the target quality attribute range in real time. Set a deviation range. When the deviation exceeds the deviation range, perform optimization calculations again through the process digital twin model and adjust the control parameters until the key quality attributes under actual operation approach the target quality attribute range.

[0034] During equipment adjustment and production operation, production data, equipment operation data, and real-time detectable key quality attributes for each process are collected in real time at a frequency of 1Hz. This real-time data is synchronously fed back to the process digital twin model and soft measurement model to calculate the deviation between the actual key quality attributes and the center value of the target quality attribute interval in real time. The allowable deviation range is set at 10% of the target interval width. When the deviation exceeds the allowable range, a secondary optimization process is immediately triggered. Using the current actual production state as the initial condition, the digital twin simulation and genetic algorithm optimization are rerun, adjusting the control parameter combination and issuing execution until the key quality attributes under the actual operating state approach the center value of the target quality attribute interval.

[0035] The model update module is used to obtain the actual test results of the batch of products after the production of a single batch is completed, and to perform deviation analysis between the actual test results and the prediction results output by the soft measurement model. Based on the deviation analysis results, the mapping relationship model, the soft measurement model and the process digital twin model are updated.

[0036] In the model update module, after a single batch of silicon-carbon anode material is produced, a comprehensive test is conducted on the batch of products to obtain the actual test values ​​of all key quality attributes and generate an actual test report for the batch of products. After each batch of silicon-carbon anode material is produced, sampling inspection is conducted in accordance with the national standard GB / T38823-2020 and customer requirements. The sampling ratio is no less than 0.5% of the total production of that batch, and the sampling points cover different time periods of the production process. The sampled samples undergo comprehensive physicochemical and electrochemical performance testing to obtain actual test values ​​for all key quality attributes, generating a standardized actual test report that includes the testing time, personnel, equipment, methods, and results.

[0037] Calculate the deviation between the actual measured values ​​of each key quality attribute of the batch of products and the predicted values ​​output by the soft sensor model during the production process of the batch, and analyze the reasons for the deviation. Extract the predicted value sequence of all key quality attributes output by the soft measurement model during the production process of this batch, calculate the batch average value of each key quality attribute prediction value, and compare it with the corresponding actual detection value to calculate the absolute deviation and relative deviation respectively. Analyze the average deviation level of the last 10 batches. If the deviation of the current batch exceeds twice the historical average deviation, it is judged as an abnormal deviation, triggering the deep deviation attribution process.

[0038] The SHAP (SHapley Additive ex Planations) value method was used to attribute the deviation. The contribution of each input variable (process parameters, equipment operating parameters, and environmental parameters) to the prediction deviation of the current batch was calculated. The top 5 core influencing factors with the highest contribution were ranked. Combined with production logs, equipment operation records, and raw material testing data, the root causes of the deviation were determined, including raw material batch fluctuations, equipment performance degradation, abnormal environmental temperature and humidity, and sensor drift.

[0039] Based on the magnitude of the deviation and the cause of the deviation, the training samples of the mapping relationship model are adjusted, the model parameters are optimized, and the mapping accuracy between process parameters and key quality attributes is improved. At the same time, the input variable weights of the soft measurement model are corrected, the prediction algorithm is optimized, and the prediction deviation is reduced, thereby updating the production process parameters, equipment operating parameters, and material structure evolution laws in the process digital twin model.

[0040] The second objective of this invention is to provide a method for precise control of the large-scale production process of silicon-carbon anode materials. A precise control system for the large-scale production process of silicon-carbon anode materials, based on any one of the above-mentioned methods, includes the following steps: S1. Collect production data of each process in the production of silicon-carbon anode materials, and bind the production data with the corresponding batch of products to obtain batch-related production data. At the same time, extract product quality characterization parameters based on batch-related production data, classify the product quality characterization parameters, and obtain key quality attributes. S2. Based on batch-related production data and key quality attributes, establish a mapping relationship model between process parameters and key quality attributes, and based on the mapping relationship model, establish a soft measurement model for predicting key quality attributes that are difficult to detect in real time during the production process. S3. Perform time-series alignment and correlation processing on batch-related production data, and combine the mapping relationship model and soft measurement model to construct a process digital twin model to characterize the production process, equipment operating status and material structure evolution relationship. S4. Based on the key quality attributes, set the target quality attribute range, and based on the process digital twin model and the target quality attribute range, perform collaborative optimization calculations on the control parameters in the production process to obtain the optimal control parameter combination. Then, based on the optimal control parameter combination, adjust the corresponding production equipment to perform closed-loop control of the production process, so that the actual operating state of the production process approaches the target quality attribute range. S5. After a single batch of production is completed, obtain the actual test results of the batch of products, and perform a deviation analysis between the actual test results and the predicted results output by the soft measurement model. Based on the deviation analysis results, update the mapping relationship model, the soft measurement model, and the process digital twin model.

[0041] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A precision control system for the large-scale production process of silicon-carbon anode materials, characterized in that: It includes a data acquisition module, a quality mapping module, a model building module, a parameter optimization module, and a model update module; The data acquisition module is used to collect production data of each process in the production of silicon-carbon anode materials, and bind the production data with the corresponding batch of products to obtain batch-related production data. At the same time, it extracts product quality characterization parameters based on batch-related production data, classifies the product quality characterization parameters, and obtains key quality attributes. The quality mapping module is used to establish a mapping relationship model between process parameters and key quality attributes based on batch-related production data and key quality attributes, and to establish a soft measurement model based on the mapping relationship model to predict key quality attributes that are difficult to detect in real time during the production process. The model building module is used to perform time-series alignment and correlation processing on batch-related production data. Combined with the mapping relationship model and the soft measurement model, a process digital twin model is constructed to characterize the production process, equipment operating status and material structure evolution relationship. The parameter optimization module is used to set a target quality attribute range based on key quality attributes, and to perform collaborative optimization calculations on the control parameters in the production process based on the process digital twin model and the target quality attribute range to obtain the optimal control parameter combination. Then, the corresponding production equipment is adjusted based on the optimal control parameter combination to perform closed-loop control of the production process, so that the actual operating state of the production process approaches the target quality attribute range. The model update module is used to obtain the actual test results of the batch of products after the production of a single batch is completed, and to perform deviation analysis between the actual test results and the prediction results output by the soft measurement model. Based on the deviation analysis results, the mapping relationship model, the soft measurement model and the process digital twin model are updated.

2. The precision control system for the large-scale production process of silicon-carbon anode materials according to claim 1, characterized in that: The data acquisition module includes processes such as mixing, granulation, drying, sintering, pulverizing, and sieving in the silicon-carbon anode material production process. Production data for each process includes corresponding process parameters, equipment operating parameters, and environmental parameters. Each batch of silicon-carbon anode material is assigned a unique batch number. All production data of that batch of products from the mixing process to the screening process are associated with that batch number, forming batch-related production data that uses the batch number as an index and includes production data of each process. Each set of production data can be matched with a specific batch of products.

3. The precision control system for the large-scale production process of silicon-carbon anode materials according to claim 2, characterized in that: In the data acquisition module, parameters that can reflect the quality of silicon-carbon anode materials are selected from the obtained batch-related production data. The product quality characterization parameters include direct quality characterization parameters and indirect quality characterization parameters. Direct quality characterization parameters are parameters that are directly related to the quality of the product itself and are obtained directly through detection methods. These include the product's particle size distribution, specific surface area, porosity, initial charge-discharge efficiency, cycle capacity, compaction density, and purity. The indirect quality characterization parameters cannot directly reflect product quality, but are significantly related to product quality and can be indirectly derived from production data. These include parameter fluctuation values ​​in each production process, equipment operation stability parameters, and material loss rate. Correlation analysis was used to analyze all the extracted product quality characterization parameters. Parameters with low correlation to the core quality indicators of silicon-carbon anode materials and weak impact on product quality were removed, while parameters with high correlation and decisive effect on product quality were retained as key quality attributes.

4. The precision control system for the large-scale production process of silicon-carbon anode materials according to claim 1, characterized in that: In the quality mapping module, the process parameters of each process in the obtained batch-related production data are used as input variables, and the obtained key quality attributes are used as output variables. Multiple batches of batch-related production data and corresponding key quality attribute data are collected as training samples. Machine learning algorithms are used to train the training samples to build a mapping relationship model between process parameters and key quality attributes. Among them, the key quality attributes that are difficult to detect in real time during the production process are those that take a long time to detect, that can damage the product during the detection process, and those that cannot be sampled and detected in real time during the production process. Based on the mapping relationship model, and combined with the process parameters, equipment operating parameters and key quality attributes that can be collected in real time during the production process, and supplemented with real-time collected data as input, the model structure is optimized and trained to obtain a soft measurement model. The soft measurement model can predict the values ​​of key quality attributes that are difficult to detect in real time based on the real-time data in the production process.

5. The precision control system for the large-scale production process of silicon-carbon anode materials according to claim 1, characterized in that: In the model building module, the batch-related production data of each batch is sorted according to the time sequence of the production process, the timestamp format of the production data of each process is unified, and the production data collected at different processes and different time points are time-series aligned to ensure that the production data of the same production stage can be matched accordingly. Then, the time-aligned production data is correlated to uncover the inherent correlations between production data of different processes, between production data and key quality attributes, and between production data and equipment operating status, thus obtaining a time-correlated production dataset. Based on the time-series associated production dataset, the mapping relationship between process parameters and key quality attributes of the mapping relationship model and the real-time prediction function of the soft measurement model are integrated into the model building process to build a digital twin scenario corresponding to the actual silicon-carbon anode material production process, thereby obtaining a process digital twin model.

6. The precision control system for the large-scale production process of silicon-carbon anode materials according to claim 1, characterized in that: In the parameter optimization module, based on the application scenario requirements, industry standards and product quality requirements of silicon-carbon anode materials, and combined with the key quality attribute range of qualified products in historical production data, a reasonable target range is set for each key quality attribute. Among them, the target quality attribute range is the acceptable range of key quality attributes, and the range can be adjusted according to actual production needs and product specifications. The target quality attribute range is used as the optimization objective, and the control parameters of each process in the production process are used as optimization variables. The process digital twin model is used to simulate the production process and corresponding key quality attributes under different combinations of control parameters. The genetic algorithm is used to optimize the control parameters in a coordinated manner. Under the premise of meeting the requirements of the target quality attribute range and taking into account production efficiency, energy consumption and cost, the optimal combination of control parameters is calculated.

7. The precision control system for the large-scale production process of silicon-carbon anode materials according to claim 1, characterized in that: In the parameter optimization module, the optimal combination of control parameters is converted into control commands that can be recognized by each production equipment. These commands are then sent to the production equipment of the corresponding process through the control system. The operating parameters and process execution parameters of the equipment are adjusted in real time, and the production parameters of each process are operated according to the optimal combination of control parameters. During equipment adjustment, production data, equipment operation data, and key quality attributes that can be detected in real time are collected in real time. The real-time data is fed back to the process digital twin model and soft measurement model to compare the deviation between the actual key quality attributes and the target quality attribute range in real time. Set a deviation range. When the deviation exceeds the deviation range, perform optimization calculations again through the process digital twin model and adjust the control parameters until the key quality attributes under actual operation approach the target quality attribute range.

8. The precision control system for the large-scale production process of silicon-carbon anode materials according to claim 1, characterized in that: In the model update module, after a single batch of silicon-carbon anode material is produced, a comprehensive test is conducted on the batch of products to obtain the actual test values ​​of all key quality attributes and form an actual test report for the batch of products. Calculate the deviation between the actual measured values ​​of each key quality attribute of the batch of products and the predicted values ​​output by the soft sensor model during the production process of the batch, and analyze the reasons for the deviation. Based on the magnitude of the deviation and the cause of the deviation, the training samples of the mapping relationship model are adjusted, the model parameters are optimized, and the mapping accuracy between process parameters and key quality attributes is improved. At the same time, the input variable weights of the soft measurement model are corrected, the prediction algorithm is optimized, and the prediction deviation is reduced, thereby updating the production process parameters, equipment operating parameters, and material structure evolution laws in the process digital twin model.

9. A method for precise control of the large-scale production process of silicon-carbon anode materials, based on the precise control system for the large-scale production process of silicon-carbon anode materials as described in any one of claims 1-8, characterized in that: Includes the following steps: S1. Collect production data of each process in the production of silicon-carbon anode materials, and bind the production data with the corresponding batch of products to obtain batch-related production data. At the same time, extract product quality characterization parameters based on batch-related production data, classify the product quality characterization parameters, and obtain key quality attributes. S2. Based on batch-related production data and key quality attributes, establish a mapping relationship model between process parameters and key quality attributes, and based on the mapping relationship model, establish a soft measurement model for predicting key quality attributes that are difficult to detect in real time during the production process. S3. Perform time-series alignment and correlation processing on batch-related production data, and combine the mapping relationship model and soft measurement model to construct a process digital twin model to characterize the production process, equipment operating status and material structure evolution relationship. S4. Based on the key quality attributes, set the target quality attribute range, and based on the process digital twin model and the target quality attribute range, perform collaborative optimization calculations on the control parameters in the production process to obtain the optimal control parameter combination. Then, based on the optimal control parameter combination, adjust the corresponding production equipment to perform closed-loop control of the production process, so that the actual operating state of the production process approaches the target quality attribute range. S5. After a single batch of production is completed, obtain the actual test results of the batch of products, and perform a deviation analysis between the actual test results and the predicted results output by the soft measurement model. Based on the deviation analysis results, update the mapping relationship model, the soft measurement model, and the process digital twin model.

Images