Intelligent preparation and coating method of lithium battery electrode slurry

Abstract

The invention discloses an intelligent preparation and coating method for lithium battery electrode slurry in the technical field of information, which comprises the following steps: acquiring particle size distribution, specific surface area and chemical composition attribute parameters of a lithium battery electrode slurry raw material, and establishing a quantitative relation model among the slurry component proportion, rheological property and stability through an autonomous learning algorithm; adjusting the component proportion by adopting a dynamic optimization algorithm to obtain a slurry formula meeting a target attribute; feature extraction and classification are performed on the constructed data set, a decision tree algorithm is adopted to train and generate optimization rules and judgment logic of slurry preparation process parameters, a preparation process knowledge base is formed, and autonomous optimization of the process parameters is realized; aiming at the coating quality problem, slurry preparation process parameters are traced and analyzed, collaborative optimization of a slurry formula, a preparation process and coating parameters is realized through a human-computer interaction interface, and the production efficiency and the yield are improved while the electrode performance is ensured.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for intelligently preparing and coating lithium battery electrode slurry. Background Art

[0002] In the production process of lithium batteries, the preparation and coating of electrode slurry is a key process that directly affects the performance and consistency of the battery. However, the formulation and preparation process of electrode slurry are very complicated, involving the selection, proportioning, mixing and dispersion of multiple raw materials. Among them, the type, morphology, particle size distribution, surface characteristics and other parameters of components such as active materials, conductive agents and binders will affect the rheology, stability and coating quality of the slurry. At the same time, in the preparation process of the slurry, the type of mixing equipment, rotation speed, time, temperature and other process parameters also need to be precisely controlled, otherwise it will lead to problems such as uneven slurry, agglomeration or sedimentation.

[0003] In the slurry coating process, the prepared slurry needs to be evenly coated on the current collector foil to form an electrode layer with uniform thickness and surface density. However, since the rheology and viscosity of the slurry will change with time, temperature, shear rate and other factors, problems such as uneven coating thickness, surface defects, peeling, and cracking will occur during the coating process. In addition, factors such as the model, parameters, and operating status of the coating equipment will also affect the coating quality, such as the shape, gap, pressure, coating speed, tension, and drying conditions of the coating head. Slight changes in these factors may lead to differences and instability in electrode performance.

[0004] Therefore, how to achieve precise control and optimization of electrode slurry preparation and coating process in a complex production environment is a technical problem that needs to be solved urgently. This requires in-depth research on the rheological properties and coating mechanism of the slurry, the establishment of scientific formula design and process optimization methods, and the introduction of advanced technologies such as online monitoring and intelligent control to achieve real-time feedback and closed-loop optimization of the production process, ultimately ensuring high quality and consistency of electrode preparation and improving the performance and yield of lithium batteries. Summary of the invention

[0005] The present invention provides a method for intelligently preparing and coating a lithium battery electrode slurry, the method comprising the following steps:

[0006] S1. Obtain the particle size distribution, specific surface area and chemical composition property parameters of the raw materials of lithium battery electrode slurry, establish a quantitative relationship model between the slurry component ratio and rheology and stability through an autonomous learning algorithm, and use a dynamic optimization algorithm to adjust the component ratio to obtain a slurry formula that meets the target properties;

[0007] S2. On the automated slurry preparation production line, according to the optimized slurry formula, set different mixing equipment types, rotation speeds, time and temperature process parameters, prepare multiple groups of slurry samples, use laser scattering, conductivity, and sedimentation rate measuring instruments to test the slurry uniformity, agglomeration, and sedimentation rate indicators, and construct a data set between the slurry preparation process and quality attributes;

[0008] S3. Extract and classify the constructed data set, use the decision tree algorithm to train and generate the optimization rules and judgment logic of the slurry preparation process parameters, form a preparation process knowledge base, and realize the autonomous optimization of the process parameters;

[0009] S4. According to the target slurry performance requirements, the preparation process knowledge base automatically recommends the corresponding equipment combination and process parameters, and builds a virtual model of the slurry preparation process through digital twin technology to monitor the equipment operation status, material flow and temperature distribution in real time, predict equipment failures and slurry quality fluctuations, and dynamically adjust the production line layout and material distribution to ensure the stability and consistency of slurry preparation;

[0010] S5. After the slurry is prepared, it is transported to the electrode coating process through a pipeline. During the coating process, the sensor collects the slurry temperature, viscosity and shear rate parameters in real time, and combines the slurry rheological model to predict the dynamic changes of the slurry properties during the coating process. The adaptive algorithm based on fuzzy control is used to adjust the coating speed, coating pressure and drying temperature in real time to ensure the uniformity of the coating quality.

[0011] S6. The machine vision system performs online defect detection on the electrode surface after coating, obtains the coating thickness distribution map and defect distribution map, and judges the coating quality according to the preset threshold. If unqualified products are detected, the defect information is fed back to the preparation process knowledge base, and the slurry formula and preparation rules are updated through the online learning algorithm to achieve dynamic optimization of the knowledge base;

[0012] S7. In view of coating quality issues, trace and analyze the slurry preparation process parameters, and achieve coordinated optimization of slurry formula, preparation process and coating parameters through the human-computer interaction interface, so as to improve production efficiency and yield rate while ensuring electrode performance;

[0013] S8. Adopt a networked collaborative manufacturing model to connect the raw material supply, slurry preparation, coating production and battery assembly links, realize end-to-end data flow and logistics integration, and build a digital twin model of the entire electrode manufacturing process, including real-time mapping of equipment operating status, material flow, energy consumption and quality indicators, to achieve real-time monitoring of the process flow, predictive maintenance and intelligent scheduling, to ensure the consistency, stability and efficiency of electrode manufacturing.

[0014] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0015] The present invention discloses a method for intelligent preparation and coating of lithium battery electrode slurry. The method establishes a quantitative relationship model between the slurry component ratio and performance through an autonomous learning algorithm, and uses a dynamic optimization algorithm to adjust the formula. On an automated production line, process parameters are set according to the optimized formula to prepare slurry samples, and a preparation process and quality attribute data set is constructed. The decision tree algorithm is used to train and generate process parameter optimization rules to form a preparation process knowledge base. During the coating process, the slurry rheology model is combined to predict performance changes, and an adaptive algorithm is used to adjust the process parameters in real time. The machine vision system performs online defect detection and feedbacks information to update the knowledge base. The coordinated optimization of slurry formula, preparation process and coating parameters is achieved through human-computer interaction. A full-process digital twin model is constructed to achieve real-time monitoring and intelligent scheduling of the process flow, ensuring the consistency, stability and efficiency of electrode manufacturing. DETAILED DESCRIPTION

[0016] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in combination with the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0017] The intelligent preparation and coating method of lithium battery electrode slurry in this embodiment may specifically include:

[0018] S1. Obtain the particle size distribution, specific surface area and chemical composition property parameters of the raw materials of lithium battery electrode slurry, establish a quantitative relationship model between the slurry component ratio and rheology and stability through an autonomous learning algorithm, and use a dynamic optimization algorithm to adjust the component ratio to obtain a slurry formula that meets the target properties.

[0019] Obtain the particle size distribution, specific surface area and chemical composition attribute parameters of the raw materials of the lithium battery electrode slurry; based on the attribute parameters, use an autonomous learning algorithm to establish a quantitative relationship model between the slurry component ratio and the rheology and stability; based on the quantitative relationship model, use a dynamic optimization algorithm to adjust the slurry component ratio to obtain a candidate formula that meets the target rheology and stability.

[0020] Specifically, the preparation of lithium battery electrode slurry is a complex process that requires precise control of multiple parameters. First, it is crucial to obtain the property parameters of the raw materials. For example, for the positive electrode material lithium cobalt oxide, its particle size distribution range can be measured to be 1-10μm, and the specific surface area is about 0.5-2m 2 / g, the main chemical composition is LiCoO 2Conductive agents such as acetylene black may have a particle size of 20-50 nm and a specific surface area of ​​50-100 m 2 / g. These parameters directly affect the performance of the slurry. Establishing a quantitative relationship model between the proportion of slurry components and performance is a key step. A neural network algorithm can be used, with the input being the proportion percentage of each component and the output being rheological indicators (such as viscosity) and stability indicators (such as sedimentation rate). For example, a three-layer neural network is set with 10 hidden layer neurons, and the model is trained using the back propagation algorithm. If the model prediction error exceeds the preset threshold (such as 5%), the network structure needs to be adjusted or training data needs to be added to remodel the model. After the quantitative relationship model is established, it can be used to optimize the formula. A genetic algorithm is used for dynamic optimization, with the population size set to 100 and the number of iterations set to 1000, and new candidate formulas are generated through crossover and mutation operations. The evaluation function can comprehensively consider multiple indicators such as rheology and stability, such as setting the viscosity target value to 2000mPa·s and the sedimentation rate to less than 1% / h. Finally, the formula with the highest comprehensive score is screened out, such as the mass ratio of positive electrode material, conductive agent, and binder is 92:4:4. The advantage of this method is that it can quickly and systematically optimize the slurry formula, reduce manual trial and error, and improve efficiency. By precisely controlling the properties of the slurry, the electrode coating quality can be improved, thereby improving the capacity and cycle performance of the battery. However, this method also faces challenges, such as the need for a large amount of experimental data to support model training, and the generalization ability of the model may be limited. In the future, it may be considered to introduce technologies such as transfer learning to improve the scope and accuracy of the model.

[0021] S2. On the automated slurry preparation production line, according to the optimized slurry formula, set different mixing equipment types, speed, time and temperature process parameters to prepare multiple groups of slurry samples. Use laser scattering, conductivity, and sedimentation rate measuring instruments to test the slurry uniformity, agglomeration, and sedimentation rate indicators to build a data set between the slurry preparation process and quality attributes.

[0022] S2 also includes: obtaining process data of the slurry sample, the process data of the slurry sample including operating parameters, duration and temperature information of the mixing equipment; recording the formula information of the slurry sample; detecting the particle size distribution data of the slurry sample to determine the degree of agglomeration of the slurry sample; detecting the conductivity of the slurry sample to determine the conductivity of the electrolyte solution in the slurry sample; establishing a temporary database based on the process data and formula information of the slurry sample as well as the conductivity and particle size distribution data; detecting the sedimentation velocity of the slurry sample, and determining the sedimentation ratio relationship of the slurry sample based on the change in sedimentation velocity; based on the sedimentation ratio relationship, merging the associated degree of agglomeration, conductivity data and multiple sets of collected slurry sample process data and formula information to obtain a full slurry database; establishing at least one machine learning model based on the full slurry database ; If the machine learning model is a linear regression model, determine whether there is a linear relationship between the independent variable and the dependent variable in the training set; if there is a linear relationship between the dependent variable and the independent variable in the training set and the determination coefficient is greater than a preset threshold, the linear regression model determines that the result is valid; If the machine learning model is a decision tree model or a random forest model, determine whether there is a nonlinear relationship between the independent variable and the dependent variable in the training set. If the relationship between the dependent variable and the independent variable cannot be described by a straight line and the determination coefficient is less than and the mean square error is greater than a preset threshold, then the decision tree model or random forest model result is invalid; According to the conductive performance and the corresponding operating parameters of the mixing equipment, the slurry sample particle size distribution data and the corresponding formula, the effectively trained machine learning model is iteratively traversed and calculated to obtain the optimal process parameter combination and optimized slurry formula data.

[0023] Specifically, in the production process of lithium battery electrode slurry, collecting process data of multiple groups of slurry samples is the key first step. For example, the operating parameters of the mixing equipment include stirring speed, stirring time, temperature, etc. Assume that the stirring speed of a group of slurry samples in the mixing equipment is 500rpm, the stirring time is 30 minutes, and the temperature is controlled at 25°C. These data are recorded synchronously with the formula information number (such as A001) of the group of slurry samples to ensure the accuracy of subsequent analysis. Next, the slurry sample numbered A001 is tested using a laser scattering instrument to obtain its particle size distribution data. Assume that the test results show that the average particle size of the sample is 10 microns and the particle size distribution range is 5-15 microns. By analyzing these data, the degree of agglomeration of the sample can be evaluated. If the particle size distribution is wide, it means that the agglomeration phenomenon is more serious; otherwise, the degree of agglomeration is low. The conductivity of the A001 slurry sample whose agglomeration degree has been analyzed is tested using a conductivity meter. Assume that the conductivity of the sample is 1.2S / m, which indicates that the electrolyte solution has good conductivity in the slurry. Associate this data with the previous process data and formula information to establish a temporary database for subsequent comprehensive analysis. Use a sedimentation instrument to detect the sedimentation rate of the solid phase particles of the A001 slurry sample. Assume that the test results show that the sedimentation rate of the sample within 1 hour is 0.5mm / h, and the sedimentation ratio is calculated to be 5%. This sedimentation ratio relationship is combined with the degree of agglomeration, conductive performance data, process data and formula information to form a full slurry database. Based on the full slurry database, a machine learning algorithm including a linear regression model, a decision tree model, and a random forest model is established. Assume that when training the linear regression model, the R square threshold is set to 0.8 and the mean square error MSE threshold is set to 0.05. If the training results show that there is a linear trend between the dependent variable Y (such as conductive performance) and the independent variable X (such as stirring speed, particle size distribution, etc.), and the R square is 0.85, which is greater than the set threshold, then the model training result is judged to be valid. On the contrary, if the R square is only 0.6 and the MSE is 0.1, which is greater than the preset threshold, the model is judged to be invalid. For the three established effective machine learning models, iterative traversal calculations of different equipment process combinations are performed. Assuming that in a certain iteration, the process combination scheme with a stirring speed of 600rpm, a stirring time of 40 minutes, and a temperature of 30°C, the conductivity and sedimentation ratio predicted by the model are better than other schemes. After many iterations, it was finally determined that the process parameter combination was optimal, and the slurry formula was optimized accordingly. Through this multi-dimensional data analysis and method application, not only can the various performance indicators of the slurry sample be accurately grasped, but also the production process and formula can be effectively optimized to improve the quality and stability of lithium battery electrode slurry. For example, the optimized slurry formula has a 10% improvement in conductivity while reducing the sedimentation ratio by 15%, significantly improving the comprehensive performance of the slurry.This systematic analysis and optimization process provides solid technical support for lithium battery production and ensures continuous improvement of product quality.

[0024] S3. Perform feature extraction and classification on the constructed data set, use decision tree algorithm to train and generate optimization rules and judgment logic for slurry preparation process parameters, form a preparation process knowledge base, and realize autonomous optimization of process parameters.

[0025] Incomplete information and unreasonable data in the data set are removed to obtain a training set with high slurry quality; the slurry preparation process parameters of the high slurry quality data group in the training set are used to extract multi-dimensional features; if the influencing factor value is higher than the set value, it is set as the target factor, and a rule base is established based on the characteristic value and the target factor, and a judgment table is constructed by judging different feature combinations.

[0026] Specifically, for the constructed data set, all data entries are first extracted as the original data set. Assume that the original data set contains 1000 records, each record includes process parameters such as mixing equipment type, speed, time, temperature, and corresponding slurry uniformity, agglomeration and sedimentation rate. When determining the correspondence between slurry quality and process parameters, the influence of different mixing equipment types on slurry uniformity can be analyzed. For example, the slurry prepared by equipment A at a speed of 1000 rpm, a time of 30 minutes, and a temperature of 50°C has a uniformity of 90%, while the uniformity of equipment B under the same conditions is only 70%. This shows that the type of equipment has a significant impact on slurry quality. In the preprocessing stage, data with incomplete information is removed. For example, if a record lacks temperature information, it is removed. At the same time, unreasonable data is identified and removed. For example, if a record shows a negative speed, it is obviously wrong data. After preprocessing, assume that the remaining 800 high-quality data are used as the training set. When extracting multi-dimensional features, features such as mixing equipment type, speed, time, and temperature can be extracted from the training set. For example, when the speed of equipment A is 1000rpm, the time is 30 minutes, and the temperature is 50℃, the slurry uniformity is 90%, the agglomeration degree is 5%, and the sedimentation rate is 2%. Through analysis, it is found that the slurry uniformity is higher under the combination of high speed and moderate temperature. When judging the quantitative relationship between the quality of slurry and the corresponding characteristic values ​​of different dimensions, it can be observed that high uniformity slurry often appears under high speed and moderate temperature conditions. By calculating the information gain value, it is found that the speed has the highest influence factor on the slurry uniformity. If the influence factor value is higher than 0.5, it is set as the target factor. A rule base is established based on the characteristic value and the target factor, such as "high speed (>900rpm) and moderate temperature (40-60℃) → high uniformity". A judgment table is constructed by judging different feature combinations. For example, when the speed of equipment A is 1000rpm and the temperature is 50℃, it is judged to have high uniformity. When constructing a decision tree model, the judgment condition is set as "speed>900rpm and temperature is 40-60℃". If the condition is met, the next step of the optimization procedure is entered. The optimization direction can be to increase the speed or fine-tune the temperature to generate a new parameter group of different dimensions, such as a speed of 1100rpm and a temperature of 55℃. After obtaining the new parameter group, it is trained through the model with the original parameter group in the data set. Assuming that the slurry uniformity of the new parameter group is increased to 95% during training, which is significantly higher than the original parameter group, the parameter group is determined to be the optimal process parameter group. When building the knowledge base, the optimal process parameter group is stored, such as "equipment A, speed 1100rpm, temperature 55℃→uniformity 95%". By inputting new process parameters to be optimized, such as equipment A at a speed of 1000rpm and a temperature of 50℃, it is compared and optimized according to the optimal process parameters. If it is judged that the new input parameters meet the requirements of the optimization direction, fine-tuning is performed, such as increasing the speed to 1100rpm and adjusting the temperature to 55℃. After determining that the optimization value is obtained after optimization, the uniformity is increased to 95%.The implementation of this method can effectively improve the accuracy and efficiency of the slurry preparation process. Through a data-driven approach, it ensures that each step of optimization has a basis, and ultimately maximizes the slurry quality. Through multi-dimensional feature analysis and information gain calculation, the key factors affecting slurry quality can be accurately identified, so that process parameters can be optimized in a targeted manner to improve production efficiency and product quality. The construction of a knowledge base provides data support for further process optimization, so that new input parameters can quickly obtain optimization suggestions and achieve continuous improvement.

[0027] S4. According to the target slurry performance requirements, the preparation process knowledge base automatically recommends the corresponding equipment combination and process parameters. Through digital twin technology, a virtual model of the slurry preparation process is constructed to monitor the equipment operation status, material flow and temperature distribution in real time, predict equipment failures and slurry quality fluctuations, dynamically adjust the production line layout and material distribution, and ensure the stability and consistency of slurry preparation.

[0028] S4 also includes the following steps: obtaining the slurry performance index requirements and matching similar slurry cases from the index library; recommending equipment groups and initial values ​​of process parameters based on the case results; applying information to create a production plan; synchronously creating a corresponding digital twin model through the production plan, and the model integrates equipment parameters, material properties, and process parameters; obtaining a virtual production space based on the twin model calculation results; obtaining equipment operation data collected by sensors in the actual production line, integrating equipment status information, and comparing it with the virtual model operation information. If the deviation exceeds the preset threshold range, it is judged that the equipment operation is abnormal; collecting space temperature data and material flow speed, updating the virtual model temperature field and material flow distribution, Establish an association relationship based on historical production data. If the fluctuation exceeds the preset threshold, warn of the slurry quality risk. Obtain the warning type for the warning information, and combine it with the knowledge base diagnosis. If the equipment is determined to be abnormal, generate an equipment maintenance plan. If it is determined to be material and environmental factors, generate raw material ratio or environmental parameter adjustment information. According to the fault prediction results, adjust the production plan arrangement, reconstruct it according to the virtual model, obtain a new reconstructed production line layout plan, update the equipment deployment plan, and apply the new plan to adjust the equipment operation plan. According to the quality prediction results, obtain the material ratio plan, generate the raw material distribution task instructions, transmit them to the automatic distribution system through the actuator, and perform real-time material distribution according to the task.

[0029] Specifically, the matching of slurry performance indicators is the key starting point for formulating production plans. For example, for a new type of lithium battery slurry, the viscosity may be required to be between 3000-5000mPa·s and the solid content may be between 60-65%. Through the matching of the indicator library, a similar lithium iron phosphate slurry case may be found, which uses a combination of planetary mixer and three-roll grinder, with an initial stirring speed of 800rpm and a grinding gap of 50μm. This information can be used directly to create a preliminary production plan. The construction of a digital twin model can realize the simulation of a virtual production space. Taking slurry preparation as an example, the geometric parameters, material properties and other information of equipment such as mixing barrels and piping systems can be input into the model. At the same time, process parameters such as stirring speed and temperature are also included in the model. Through methods such as computational fluid dynamics, the flow state and temperature distribution of the slurry in the equipment can be simulated to form a complete virtual production space. In actual production, equipment operation data can be collected through torque sensors and temperature sensors installed on the mixing barrel. For example, if the actual torque of the mixing barrel is 250N·m, and the virtual model predicts 200N·m, the deviation exceeds the set threshold of 20%, it can be judged that the equipment may have abnormalities, such as wear of the mixing blade. Real-time update of temperature field and material flow distribution is crucial to ensure the quality of slurry. For example, the wall temperature of the mixing barrel collected by the infrared thermal imager is 35℃, while historical data shows that the normal temperature should be within the range of 30±2℃. This abnormal fluctuation may indicate abnormal slurry viscosity, which in turn affects the quality of the final product. Timely discovery and early warning of such risks can greatly improve production efficiency and product qualification rate. For different types of early warning information, the system will generate corresponding solutions. If the diagnosis result shows that the equipment is abnormal, such as wear of the mixing blade bearing, the system will generate a detailed maintenance plan including the steps for replacing the bearing, the required tools and the estimated time. If it is determined to be caused by material factors, such as abnormal viscosity caused by changes in raw material batches, a suggestion for adjusting the amount of dispersant may be generated. By reconstructing the virtual model, the production line layout can be optimized. For example, the analysis found that the material transmission distance is too long under the current layout, resulting in increased energy consumption and slurry temperature fluctuations. The reconstructed solution may suggest shortening the distance between the mixing tank and the coating machine by 20% and adding an intermediate storage tank to stabilize the slurry temperature, thereby improving production efficiency and product quality stability. The quality prediction results directly affect the distribution of raw materials. Assuming that the prediction shows that the current slurry is not conductive enough, the system may generate a ratio scheme to increase the conductive agent by 2%. This task instruction is transmitted to the automatic distribution system through the actuator, triggering the corresponding material conveying device to accurately add additional conductive agent to the next batch of raw materials, realizing real-time optimization and adjustment of the production process.

[0030] S5. After the slurry is prepared, it is transported to the electrode coating process through a pipeline. During the coating process, the sensor collects the slurry temperature, viscosity and shear rate parameters in real time. Combined with the slurry rheology model, the dynamic changes of the slurry properties during the coating process are predicted, and the coating speed, coating pressure and drying temperature are adjusted in real time using an adaptive algorithm based on fuzzy control to ensure the uniformity of the coating quality.

[0031] The temperature set T, viscosity set V and shear rate set S of the slurry in the pipeline are collected and obtained, and according to the set T, the set V and the set S, they are transmitted to the pre-established slurry rheology model database to obtain the slurry rheology curve data set K; the data set K is used as the fuzzy control algorithm training data set to obtain the control model C1, and the first coating speed initial value SV is obtained according to the preliminary prediction of the control model C1; the new rheology model C2 is determined according to the initial value SV, the coating temperature initial value and the coating pressure initial value, and the coating quality data set QC_D is collected. If there are data points in the data set QC_D that exceed the preset threshold value TQ, it is judged that there is an unreasonable area in the parameters this time, and the parameters are recalculated to obtain the optimized coating production process parameter set NEW_P.

[0032] Specifically, in the slurry preparation process, temperature, viscosity and shear rate are key parameters that directly affect the coating quality. First, the slurry temperature data is collected through the temperature sensor in the pipeline to obtain a temperature value set T. For example, the sensor may record a series of temperature values ​​such as 30°C, 32°C, and 34°C, which reflect the temperature changes of the slurry at different time points. Next, another sensor obtains the slurry viscosity and shear rate data to form a viscosity set V and a shear rate set S. Assume that the viscosity data is 2000mPa·s, 2200mPa·s, and 2400mPa·s, and the shear rates are 100s-1, 120s-1, and 2400mPa·s, respectively.

[0033] 140s-1. These data reflect the rheological properties of the slurry under different process conditions. These data are transferred to the pre-established slurry rheological model database to obtain the slurry rheological curve data set K at specific temperature, viscosity, and shear rate. For example, the database may show that under the conditions of 30°C, 2000mPa·s, and 100s-1, the rheological curve of the slurry exhibits specific nonlinear characteristics. The data set K is used as the training data for the fuzzy control algorithm to obtain the control model C1. The coater makes a preliminary prediction of the coating speed based on C1 to obtain the first coating speed initial value SV. Assuming that SV is 10m / min, but the preset speed range is 8m / min to 12m / min, if SV exceeds this range, it is necessary to adjust the adaptive algorithm, recalculate the coating pressure parameters, and obtain the corrected coating speed initial value VS. For example, after adjustment, VS is 9m / min, which meets the preset range. If VS is still not within the set range, activate the drying temperature correction function of the fuzzy control algorithm to calculate the optimal temperature range. Assuming that the optimal temperature range is 50°C to 60°C, the range is transmitted to the temperature controller of the coating machine to obtain the optimal target coating temperature initial value BT, such as 55°C. Combine VS, BT and the modified coating pressure initial value to determine the new rheological model C2. For example, C2 may show that the rheological properties of the slurry are more stable at a coating speed of 9m / min, a drying temperature of 55°C and a specific pressure. The coating quality data QC under this working condition is collected through a preset sensor to obtain a data set QC_D. Assuming that a data point shows that the coating thickness is 100μm, and the set threshold value TQ is 90μm to 110μm, if the data point exceeds this range, it is judged that the production process parameters are unreasonable and need to be recalculated and optimized. For example, adjust the amount of dispersant or change the stirring time until the data point meets the standard to obtain the optimized coating production process parameter set NEW_P. NEW_P is stored in the process database, and this set is preferred when the same working conditions are required. For example, if the same type of lithium battery slurry is produced again, the system automatically calls NEW_P to ensure the stability and consistency of production parameters. If the parameters of NEW_P are stable during the production process, without obvious data jitter and drift, the coating thickness is recorded as the target variable for later optimization. For example, the coating thickness is stable at 100μm, indicating that the current parameter combination is ideal and can be used as a benchmark for subsequent optimization. The implementation of this method ensures precise control of the coating process and improves product quality and production efficiency. Through real-time data acquisition and model optimization, it can quickly respond to process fluctuations, reduce scrap rates, and improve the flexibility and adaptability of the production line. In the end, a closed-loop intelligent control system was formed to achieve efficient coordination of slurry preparation and coating processes.

[0034] S6. The machine vision system performs online defect detection on the electrode surface after coating, obtains the coating thickness distribution map and defect distribution map, and judges the coating quality according to the preset threshold. If unqualified products are detected, the defect information is fed back to the preparation process knowledge base, and the slurry formula and preparation rules are updated through the online learning algorithm to achieve dynamic optimization of the knowledge base.

[0035] Obtain image information of the electrode surface after coating, the image information is obtained through machine vision technology, and determine the coating thickness distribution map and the defect distribution map; if the coating thickness exceeds a preset coating thickness threshold, or the defect area exceeds a preset defect area threshold, the coating quality of the corresponding electrode is determined to be unqualified; according to the defect information of the unqualified electrode, the defect information includes defect location, defect area and defect type, determine the basis for updating the preparation process knowledge base.

[0036] Specifically, machine vision technology plays a key role in electrode coating quality inspection. Through high-resolution cameras and image processing algorithms, the microscopic morphology of the coating surface can be captured in real time. For example, a line scan camera is used to scan along the electrode width to obtain a coating thickness distribution map. By setting a reasonable grayscale threshold, the coating can be distinguished from the substrate, and then the coating thickness can be calculated. For an electrode with a width of 100mm, a data point can be collected every 0.1mm, 1000 thickness values ​​can be obtained, and a thickness distribution curve can be drawn. Defect detection also relies on image analysis technology. Common defect types include bubbles, cracks, impurities, etc. Taking bubble detection as an example, the circular contour can be identified by the edge detection algorithm, and the area threshold can be combined to determine whether it is a bubble defect. If a bubble with a diameter greater than 0.5mm is detected, it is judged as a defective product. For cracks, morphological processing methods can be used to extract slender structural features. The extraction and feedback of defect information is an important basis for optimizing the preparation process. For example, if it is found that the bubble defects are mainly distributed in the electrode edge area, it may be caused by uneven coating pressure. At this time, it can be improved by adjusting the pressure distribution of the coating roller. For example, if crack defects mostly occur in the drying stage, it may be caused by excessive temperature gradient. You can consider reducing the drying rate or adopting a multi-stage drying process. The application of online learning algorithms makes process optimization a dynamic iterative process. Incremental learning algorithms can continuously absorb new sample information without discarding existing knowledge. For example, for a defect classification model based on a decision tree, new decision nodes can be added incrementally without retraining the entire model. Transfer learning allows knowledge in existing fields to be transferred to new related fields. For example, the coating experience of cylindrical batteries is applied to square batteries to accelerate the process optimization of new products. Data mining technology plays an important role in analyzing the causes of defects. Through association rule mining, the potential relationship between defects and formulations and process parameters can be discovered. For example, it may be found that when the slurry viscosity exceeds 5000mPa·s and the coating speed is higher than 10m / min, the probability of bubble defects increases significantly. This discovery can guide the optimization of process parameters, such as reducing the coating speed or adjusting the slurry formulation to reduce viscosity. The dynamic update of the knowledge base ensures continuous improvement of the production process. For example, when a new type of additive is found to effectively inhibit crack formation, this discovery can be added to the formula database. At the same time, the corresponding preparation rules will also be updated, such as adding 1% of the additive to the slurry formula. This closed-loop optimization mechanism can continuously improve product quality and reduce the defective product rate, thereby improving production efficiency and economic benefits.

[0037] S7. For coating quality issues, trace and analyze the slurry preparation process parameters. Through the human-computer interaction interface, the slurry formula, preparation process and coating parameters are optimized in a coordinated manner to improve production efficiency and yield while ensuring electrode performance.

[0038] S7 also includes: obtaining coating quality data, the coating quality data including coating thickness and uniformity parameters; judging whether the coating quality data meets a preset threshold, and if not, triggering a traceability analysis process; obtaining process parameter data in the slurry preparation process, the process parameter data including raw material ratio, mixing time and temperature; using a data mining algorithm to analyze the correlation between the process parameter data and the coating quality data to obtain key influencing factors; adjusting the slurry formula and preparation process parameters through a human-computer interaction interface according to the key influencing factors; using an expert knowledge base and a machine learning algorithm to optimize the slurry formula and preparation process parameters to obtain an improved slurry preparation plan; obtaining process parameters in the coating process, the process parameters including coating speed degree and pressure; determine the optimal coating parameter range through data analysis; integrate the optimization functions of slurry formula, preparation process and coating parameters in the human-computer interaction interface; adopt a multi-objective optimization algorithm to maximize production efficiency and yield while ensuring electrode performance, and obtain an optimized production plan; obtain performance data of the coated electrode, the performance data including capacity and rate performance; determine whether the performance data meets the preset performance requirements, and if not, return to the slurry formula and preparation process parameter optimization steps; apply the verified production plan to mass production, and monitor the coating quality and electrode performance data in real time; use statistical process control methods to determine whether the production process is under control, and if an abnormality occurs, trigger an early warning, automatically adjust the process parameters or return to the traceability analysis step.

[0039] S8. Adopt a networked collaborative manufacturing model to connect the raw material supply, slurry preparation, coating production and battery assembly links, and achieve end-to-end data flow and logistics integration. Build a digital twin model of the entire electrode manufacturing process, including real-time mapping of equipment operating status, material flow, energy consumption and quality indicators, to achieve real-time monitoring of the process flow, predictive maintenance and intelligent scheduling, and ensure the consistency, stability and efficiency of electrode manufacturing.

[0040] S8 also includes the following steps: obtaining the interface data of the raw material supplier's production information system, collecting the quality inspection information of each batch, and obtaining the preliminary screening results of each batch of raw materials through the pre-established regression model between the quality parameters of each batch and the electrochemical properties of each batch of materials; according to the purity and batch information in the preliminary screening result data of the raw materials, a nonlinear programming algorithm is used in combination with the inventory status for calculation to obtain a production material delivery order that meets the production target, and synchronize it to the raw material stereoscopic warehouse production information system to execute the automatic delivery of raw materials; according to the raw material information feedback from the stereoscopic warehouse production information system, the flow nodes and production rhythm requirements of the production link are generated by the particle swarm algorithm in combination with the real-time data of the digital twin model, and the basis for issuing production instructions and control timing to each production equipment is obtained; the training data in the pre-established process defect information database is used to establish a random forest algorithm classification model, and the electrode samples produced by the coating machine are subjected to quality sampling and automatic image recognition to obtain the types of defects on each sampled sample, and then the sample is automatically identified. Through outlier analysis, a set of process parameter adjustment strategies is obtained; the real-time status of the twin equipment is obtained through various types of sensor networks deployed on the equipment controller and mechanical motion components, and the long short-term memory network classification algorithm is used to identify the change of the status. According to the historical maintenance plan statistics of each status, a data list of active maintenance signals and execution strategies for different statuses is obtained; the process parameter data of each manufacturing process segment is obtained, and according to the final output battery performance parameters of each batch, a reverse iterative learning model based on a deep neural network is constructed to obtain a data list of the optimal process parameter combination dynamically generated and sent to each production unit according to different quality goals; for the difference between the health status and capacity planning of the equipment of each production unit and the actual production capacity requirements, if the difference exceeds the preset capacity utilization threshold, a hybrid algorithm that integrates a preset number of types of optimization algorithms is used to determine the collaborative operation plan between equipment, and according to the pre-established cross-equipment processing task collaboration rules, the identifier of the optimal plan is obtained, and the operation plan is sent to the production execution information system.

[0041] Specifically, the interface data of the raw material supplier's production information system is obtained. For example, the quality inspection information of each batch of materials, including purity, impurity content and other fields, is collected from the system of supplier A. The relationship between these quality parameters and the electrochemical properties of the materials is analyzed through the pre-established regression model. Assuming that the purity of a batch of materials is 99.5% and the impurity content is 0.05%, the model predicts that its electrochemical performance is excellent, and the batch of materials is initially screened to be qualified. According to the preliminary screening results, combined with the inventory status, the nonlinear programming algorithm is used for automatic batching calculation. Assuming that there are three materials of different purities in the current inventory, the algorithm will calculate the optimal batching ratio according to the production target and inventory, and generate a production material delivery order. For example, if 1,000 kg of slurry needs to be produced, the algorithm recommends using 500 kg of materials with a purity of 99.5%, 300 kg of materials with a purity of 99.7%, and 200 kg of materials with a purity of 99.8%, to ensure that the ingredients meet the quality requirements and maximize the use of inventory. The stereoscopic warehouse production information system executes the automatic delivery of raw materials according to the delivery order, and uses the particle swarm algorithm to generate the slurry preparation task plan information table through the real-time data of the digital twin model. Assuming that the digital twin model shows the current equipment status and production rhythm, the algorithm will optimize the task allocation to ensure the efficient operation of each device. For example, equipment A is suitable for high-speed mixing and equipment B is suitable for low-speed stirring. The algorithm will reasonably arrange the working sequence of equipment A and B according to the task requirements. Using the pre-established process defect information database, a random forest algorithm classification model is established to perform quality sampling and automatic image recognition on the electrode samples produced by the coating machine. Assuming that in a batch of electrode samples, image recognition finds that 5% of the samples have scratch defects. Through outlier analysis, the key process parameters that cause scratches are determined, such as too fast coating speed, and then the parameters are adjusted to reduce the defect rate. The real-time status of the twin equipment is obtained through the sensor network, and the long short-term memory network classification algorithm is used to identify state changes. For example, the vibration sensor data of equipment C shows abnormal fluctuations, and the algorithm identifies it as bearing wear. According to the historical maintenance plan, the system automatically sends a maintenance signal to replace the bearing to avoid equipment failure. Obtain the process parameter data of each manufacturing process segment, combine it with the final output battery performance parameters, and build a deep neural network reverse iterative learning model. Assuming that the capacity of a batch of batteries is lower than expected, the model analysis finds that the excessively high viscosity of the slurry is the main reason, and dynamically generates the optimal combination of process parameters, such as reducing the stirring speed and increasing the temperature, to ensure that the performance of subsequent batches of batteries meets the standards. For the health status of the equipment and capacity planning of each production unit, if the difference between the actual capacity and the requirement exceeds the preset threshold, a hybrid algorithm is used to determine the collaborative operation plan between equipment. For example, if the current capacity utilization rate is only 80%, which is lower than the preset 90% threshold, the algorithm will optimize the task allocation between equipment, such as transferring some tasks from busy equipment D to idle equipment E to ensure that the overall capacity is increased to the target level. Through the pre-established collaborative rules for processing tasks across equipment, the identifier of the optimal solution is obtained, and the operation plan is sent to the production execution information system.For example, there is a collaborative rule in the rule base that says "the joint processing efficiency of equipment F and equipment G is the highest." The algorithm generates the optimal solution based on this rule to ensure efficient collaborative operation between equipment. The implementation of these steps and technical topics not only improves production efficiency and product quality, but also achieves optimal allocation of resources and preventive maintenance of equipment, ultimately improving the overall level of intelligence and market competitiveness of the manufacturing business.

[0042] The above embodiments are only used to illustrate the technical solution of the present invention but not to limit it. The present invention is described in detail with reference to the preferred embodiments. It should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, and should be included in the scope of the claims of the present invention.

Claims

1. Intelligent preparation and coating method of lithium battery electrode slurry, characterized in that: The method comprises the following steps: S1. Obtain the particle size distribution, specific surface area and chemical composition property parameters of the raw materials of lithium battery electrode slurry, establish a quantitative relationship model between the slurry component ratio and rheology and stability through an autonomous learning algorithm, and use a dynamic optimization algorithm to adjust the component ratio to obtain a slurry formula that meets the target properties; S2. On the automated slurry preparation production line, according to the optimized slurry formula, set different mixing equipment types, rotation speeds, time and temperature process parameters, prepare multiple groups of slurry samples, use laser scattering, conductivity, and sedimentation rate measuring instruments to test the slurry uniformity, agglomeration, and sedimentation rate indicators, and construct a data set between the slurry preparation process and quality attributes; S3. Extract and classify the constructed data set, use the decision tree algorithm to train and generate the optimization rules and judgment logic of the slurry preparation process parameters, form a preparation process knowledge base, and realize the autonomous optimization of the process parameters; S4. According to the target slurry performance requirements, the preparation process knowledge base automatically recommends the corresponding equipment combination and process parameters, and builds a virtual model of the slurry preparation process through digital twin technology to monitor the equipment operation status, material flow and temperature distribution in real time, predict equipment failures and slurry quality fluctuations, and dynamically adjust the production line layout and material distribution to ensure the stability and consistency of slurry preparation; S5. After the slurry is prepared, it is transported to the electrode coating process through a pipeline. During the coating process, the sensor collects the slurry temperature, viscosity and shear rate parameters in real time, and combines the slurry rheological model to predict the dynamic changes of the slurry properties during the coating process. The adaptive algorithm based on fuzzy control is used to adjust the coating speed, coating pressure and drying temperature in real time to ensure the uniformity of the coating quality. S6. The machine vision system performs online defect detection on the electrode surface after coating, obtains the coating thickness distribution map and defect distribution map, and judges the coating quality according to the preset threshold. If unqualified products are detected, the defect information is fed back to the preparation process knowledge base, and the slurry formula and preparation rules are updated through the online learning algorithm to achieve dynamic optimization of the knowledge base; S7. In view of coating quality issues, trace and analyze the slurry preparation process parameters, and achieve coordinated optimization of slurry formula, preparation process and coating parameters through the human-computer interaction interface, so as to improve production efficiency and yield rate while ensuring electrode performance; S8. Adopt a networked collaborative manufacturing model to connect the raw material supply, slurry preparation, coating production and battery assembly links, realize end-to-end data flow and logistics integration, and build a digital twin model of the entire electrode manufacturing process, including real-time mapping of equipment operating status, material flow, energy consumption and quality indicators, to achieve real-time monitoring of the process flow, predictive maintenance and intelligent scheduling, to ensure the consistency, stability and efficiency of electrode manufacturing.

2. The method for intelligent preparation and coating of lithium battery electrode slurry according to claim 1, characterized in that: The S2 also includes: Acquiring process data of the slurry sample, wherein the process data of the slurry sample includes operating parameters, duration, and temperature information of the mixing equipment; Recording the formula information of the slurry sample; Detecting the particle size distribution data of the slurry sample to determine the degree of agglomeration of the slurry sample; Detecting the conductivity of the slurry sample to determine the conductivity of the electrolyte solution in the slurry sample; Establishing a temporary database based on process data and formulation information as well as conductivity and particle size distribution data of the slurry sample; Detecting the sedimentation velocity of the slurry sample, and determining the sedimentation ratio relationship of the slurry sample according to the change of the sedimentation velocity; According to the sedimentation ratio relationship, the associated agglomeration degree, the conductive performance data and the collected multiple groups of slurry sample process data and formula information are combined to obtain a full slurry database; Establishing at least one machine learning model based on the full slurry database; If the machine learning model is a linear regression model, determine whether there is a linear relationship between the independent variable and the dependent variable in the training set; If there is a linear relationship between the dependent variable and the independent variable in the training set and the determination coefficient is greater than a preset threshold, the linear regression model determines that the result is valid; If the machine learning model is a decision tree model or a random forest model, determine whether there is a nonlinear relationship between the independent variable and the dependent variable in the training set; if the relationship between the dependent variable and the independent variable cannot be described by a straight line and the determination coefficient is less than and the mean square error is greater than a pre-established threshold, then determine that the decision tree model or random forest model result is invalid; According to the conductive properties and corresponding operating parameters of the mixing equipment, the particle size distribution data of the slurry sample particles and the corresponding formula, the effectively trained machine learning model is iteratively traversed and calculated to obtain the optimal process parameter combination and optimized slurry formula data.

3. The method for intelligent preparation and coating of lithium battery electrode slurry according to claim 1, characterized in that: The S4 also includes the following steps: Obtain slurry performance index requirements and match similar slurry cases from the index library; Based on the case results, recommend equipment groups and initial values ​​of process parameters; Apply information to create production plans; Through the production plan, a corresponding digital twin model is simultaneously created, wherein the model integrates equipment parameters, material properties and process parameters; According to the twin model calculation results, a virtual production space is obtained; Obtain equipment operation data collected by sensors in the actual production line, integrate equipment status information, and compare it with the virtual model operation information. If the deviation exceeds the preset threshold range, it is determined that the equipment operation is abnormal; Collect spatial temperature data and material flow speed, update the virtual model temperature field and material flow distribution, establish a correlation relationship based on historical production data, and warn of slurry quality risks if the fluctuation exceeds a preset threshold; For the warning information, obtain the warning type, combine with the knowledge base diagnosis, if it is determined that the equipment is abnormal, generate an equipment maintenance plan, if it is determined to be material and environmental factors, generate raw material ratio or environmental parameter adjustment information; According to the fault prediction results, the production plan is adjusted, and the virtual model is reconstructed to obtain a new plan for the reconstructed production line layout, the equipment deployment plan is updated, and the equipment operation plan is adjusted by applying the new plan; According to the quality prediction results, the material ratio scheme is obtained, and the raw material distribution task instruction is generated and transmitted to the automatic distribution system through the execution mechanism, and the materials are distributed in real time according to the task.

4. The method for intelligent preparation and coating of lithium battery electrode slurry according to claim 1, characterized in that: The S5 includes: Acquire a slurry temperature set T, a viscosity set V, and a shear rate set S in the pipeline, and transmit the set T, the set V, and the set S to a pre-established slurry rheology model database to obtain a slurry rheology curve data set K; The data set K is used as a training data set for the fuzzy control algorithm to obtain a control model C1, and a first coating speed initial value SV is obtained based on a preliminary prediction of the control model C1; A new rheological model C2 is determined according to the initial value SV, the initial value of the coating temperature and the initial value of the coating pressure, and the coating quality data set QC_D is collected. If there are data points in the data set QC_D that exceed the preset threshold value TQ, it is determined that there is an unreasonable area in the parameters this time, and the parameters are recalculated to obtain the optimized coating production process parameter set NEW_P.

5. The method for intelligent preparation and coating of lithium battery electrode slurry according to any one of claims 1 to 4, characterized in that: The S7 comprises: Acquiring coating quality data, wherein the coating quality data includes coating thickness and uniformity parameters; Determine whether the coating quality data meets a preset threshold, and if not, trigger a traceability analysis process; Acquiring process parameter data during slurry preparation, wherein the process parameter data includes raw material ratio, mixing time and temperature; Using a data mining algorithm to analyze the correlation between the process parameter data and the coating quality data to obtain key influencing factors; According to the key influencing factors, the slurry formula and preparation process parameters are adjusted through the human-computer interaction interface; An expert knowledge base and a machine learning algorithm are used to optimize the slurry formula and preparation process parameters to obtain an improved slurry preparation scheme; Acquiring process parameters during the coating process, wherein the process parameters include coating speed and pressure; Determine the optimal coating parameter range through data analysis; Integrate the optimization functions of slurry formula, preparation process and coating parameters in the human-computer interaction interface; Adopting multi-objective optimization algorithm, we can maximize production efficiency and yield rate while ensuring electrode performance, and obtain the optimized production plan; Obtaining performance data of the coated electrode, the performance data including capacity and rate performance; Determine whether the performance data meets the preset performance requirements. If not, return to the slurry formulation and preparation process parameter optimization step; Apply the verified production plan to mass production and monitor the coating quality and electrode performance data in real time; Statistical process control methods are used to determine whether the production process is under control. If an abnormality occurs, an early warning is triggered, and the process parameters are automatically adjusted or the traceability analysis step is returned.

6. The method for intelligent preparation and coating of lithium battery electrode slurry according to any one of claims 1 to 4, characterized in that: The S8 also includes the following steps: Obtain the interface data of the raw material supplier's production information system, collect the quality inspection information of each batch, and obtain the preliminary screening results of each batch of raw materials through the pre-established regression model between the quality parameters of each batch and the electrochemical properties of each batch of materials; According to the purity and batch information in the preliminary screening result data of the raw materials, a nonlinear programming algorithm is used in combination with the inventory status to calculate, obtain a production material delivery order that meets the production target, and synchronize it to the raw material warehouse production information system to execute automatic delivery of raw materials; According to the raw material information feedback from the stereoscopic warehouse production information system, the production process flow nodes and production rhythm requirements, combined with the real-time data of the digital twin model, the particle swarm algorithm is used to generate the slurry preparation task plan information table, and the basis for issuing production instructions and control timing to each production equipment is obtained; The random forest algorithm classification model is established using the training data in the pre-established process defect information database. The quality sampling and automatic image recognition of the electrode samples produced by the coating machine are carried out to obtain the defect types on each sample. Through outlier analysis, a set of process parameter adjustment strategies is obtained. The real-time status of the twin equipment is obtained through various types of sensor networks deployed on the equipment controller and mechanical motion components. The long short-term memory network classification algorithm is used to identify the change of the status. According to the statistics of the historical maintenance plan of each status, a data list of active maintenance signals and execution strategies for different statuses is obtained. Obtain process parameter data for each manufacturing process segment, build a reverse iterative learning model based on a deep neural network based on the final output battery performance parameters of each batch, and obtain a data list of the optimal process parameter combination dynamically generated and sent to each production unit according to different quality goals; For the difference between the health status and capacity planning of the equipment of each production unit and the actual production capacity requirements, if the difference exceeds the preset capacity utilization threshold, a hybrid algorithm that integrates a preset number of types of optimization algorithms is used to determine the collaborative operation plan between equipment. According to the pre-established cross-equipment processing task collaboration rules, the identifier of the optimal plan is obtained, and the operation plan is sent to the production execution information system.