Methods, apparatus, equipment and dielectrics for determining the production process of cathode material systems

Abstract

This application relates to a method, apparatus, equipment, and medium for determining the production process of a cathode material system. The method for determining the production process of a cathode material system includes: determining the parameter information of the cathode material system based on its components; the parameter information includes pre-constructed process information corresponding to the production process of the cathode material system or performance requirement information of the cathode material system; inputting the parameter information into a preset neural network model; and determining the target production process of the cathode material system based on the output of the preset neural network model. The method for determining the production process of a cathode material system provided in this application can rapidly determine the production process of a novel "material-coating" system.

Description

Technical Field

[0001] This application relates to the technical field of lithium-ion batteries, and in particular to a method, apparatus, equipment and medium for determining the production process of a cathode material system. Background Technology

[0002] With the increasing energy density requirements of electric vehicles, high-energy-density ultra-high-nickel ternary cathode materials have become a key cathode system for power batteries. Due to inherent defects such as high residual alkali on the surface, unstable lattice oxygen, and high risk of thermal runaway, ultra-high-nickel ternary cathode materials are currently typically treated with surface coating or bulk doping to improve the stability of the material interface and ensure the smooth industrial application of these materials.

[0003] In related technologies, learning models such as neural networks are usually trained based on historical data to predict the performance indicators such as the interface impedance of the coated cathode material, and the target process used for cathode material interface optimization is determined based on the obtained performance indicator values.

[0004] However, traditional technologies have limitations in determining the optimal process for novel "material-coating" systems. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for determining the production process of a cathode material system, in response to the above-mentioned technical problems.

[0006] In a first aspect, this application provides a method for determining the manufacturing process of a cathode material system, the method comprising:

[0007] Based on the composition of the cathode material system, determine the parameter information of the cathode material system; the parameter information includes the process information corresponding to the pre-constructed production process of the cathode material system or the performance requirement information of the cathode material system.

[0008] The parameter information is input into a preset neural network model, and the target production process of the cathode material system is determined based on the output of the preset neural network model. The preset neural network model is obtained by training an initial neural network model with sample training data, which is simulation data generated based on a dynamic virtual experimental data generator and the cathode material system. The initial neural network model is a neural network model with forward prediction and reverse decoding capabilities.

[0009] In one embodiment, if the parameter information consists of multiple sets of process information, the parameter information is input into a preset neural network model, and based on the output of the preset neural network model, the target production process of the cathode material system is determined, including:

[0010] Input each process information into a preset neural network model to obtain the predicted performance index value corresponding to each process information;

[0011] Based on the predicted values ​​of performance indicators and information on each process, the target production process is determined.

[0012] In one embodiment, the target production process is determined based on predicted performance indicators and process information, including:

[0013] For each piece of process information, based on the predicted performance index of the process information, determine the coating quality index corresponding to the predicted performance index.

[0014] The process information corresponding to the maximum coating quality index is determined as the target production process.

[0015] In one embodiment, if the parameter information is the performance requirement information of the cathode material system; the parameter information is input into a preset neural network model, and based on the output of the preset neural network model, the target production process of the cathode material system is determined, including:

[0016] Input the performance requirement information into a preset neural network model to obtain the process information corresponding to the performance requirement information;

[0017] Based on process information, the target production process is determined.

[0018] In one embodiment, the training process of the preset neural network model includes:

[0019] The initial neural network model is jointly trained using the cycle consistency loss function to obtain the preset neural network model.

[0020] In one embodiment, the method further includes:

[0021] Based on the target production process and the preset construction algorithm, a directed acyclic graph is constructed. The nodes of the directed acyclic graph include preset latent variables, process parameters of the target production process and corresponding actual performance indicators. The directed edges in the directed acyclic graph are used to represent the causal relationship between the corresponding two nodes.

[0022] Based on the comprehensive confidence level corresponding to the edges of the directed acyclic graph, the key process parameters in the target production process are determined.

[0023] In one embodiment, parameter information is input into a preset neural network model, and based on the output of the preset neural network model, the target production process of the cathode material system is determined, including:

[0024] If the model of the first production equipment that generates the cathode material system is different from the model of the second production equipment corresponding to the preset neural network model, the parameters of the preset neural network model are adjusted based on the model of the first production equipment to obtain the adjusted neural network model.

[0025] The parameter information is input into the adjusted neural network model, and the target production process of the cathode material system is determined based on the output of the adjusted neural network model.

[0026] Secondly, this application also provides a device for determining the production process of a cathode material system, the device comprising:

[0027] The parameter information determination module is used to determine the parameter information of the cathode material system based on its composition. The parameter information includes the process information corresponding to the pre-constructed production process of the cathode material system or the performance requirement information of the cathode material system.

[0028] The production process determination module is used to input parameter information into a preset neural network model and determine the target production process of the cathode material system based on the output of the preset neural network model. The preset neural network model is obtained by training an initial neural network model with sample training data, which is simulation data generated based on a dynamic virtual experimental data generator and the cathode material system. The initial neural network model is a neural network model with forward prediction and reverse decoding capabilities.

[0029] Thirdly, this application also provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0030] Based on the composition of the cathode material system, determine the parameter information of the cathode material system; the parameter information includes the process information corresponding to the pre-constructed production process of the cathode material system or the performance requirement information of the cathode material system.

[0031] The parameter information is input into a preset neural network model, and the target production process of the cathode material system is determined based on the output of the preset neural network model. The preset neural network model is obtained by training an initial neural network model with sample training data, which is simulation data generated based on a dynamic virtual experimental data generator and the cathode material system. The initial neural network model is a neural network model with forward prediction and reverse decoding capabilities.

[0032] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0033] Based on the composition of the cathode material system, determine the parameter information of the cathode material system; the parameter information includes the process information corresponding to the pre-constructed production process of the cathode material system or the performance requirement information of the cathode material system.

[0034] The parameter information is input into a preset neural network model, and the target production process of the cathode material system is determined based on the output of the preset neural network model. The preset neural network model is obtained by training an initial neural network model with sample training data, which is simulation data generated based on a dynamic virtual experimental data generator and the cathode material system. The initial neural network model is a neural network model with forward prediction and reverse decoding capabilities.

[0035] Fifthly, this application also provides a computer program product comprising a computer program that, when executed by a processor, performs the following steps:

[0036] Based on the composition of the cathode material system, determine the parameter information of the cathode material system; the parameter information includes the process information corresponding to the pre-constructed production process of the cathode material system or the performance requirement information of the cathode material system.

[0037] The parameter information is input into a preset neural network model, and the target production process of the cathode material system is determined based on the output of the preset neural network model. The preset neural network model is obtained by training an initial neural network model with sample training data, which is simulation data generated based on a dynamic virtual experimental data generator and the cathode material system. The initial neural network model is a neural network model with forward prediction and reverse decoding capabilities.

[0038] The aforementioned method, apparatus, equipment, and medium for determining the production process of the cathode material system determine the parameter information of the cathode material system based on its components. The parameter information includes pre-constructed process information corresponding to the production process of the cathode material system or performance requirement information of the cathode material system. The parameter information is input into a preset neural network model, and the target production process of the cathode material system is determined based on the output of the preset neural network model. The preset neural network model is obtained by training an initial neural network model using sample training data, which is simulated data generated based on a dynamic virtual experimental data generator and the cathode material system. The initial neural network model is a neural network model with both forward prediction and reverse decoding capabilities. This application, using the above method, can determine the optimized process for a novel "material-coating" system, eliminating the need to rely on historical experimental data for cold-start modeling of the new system. It also possesses the ability to reverse-engineer processes based on performance requirements, overcoming the shortcomings of existing technologies such as difficulty in cold-starting and the inability to reverse-engineer only through forward prediction, and contributing to improved efficiency in determining the production process of the cathode material system. Attached Figure Description

[0039] Figure 1 Flowcharts illustrating the method for determining the production process of the cathode material system provided in some embodiments of this application;

[0040] Figure 2 Flowcharts for determining the target production process provided in some embodiments of this application;

[0041] Figure 3 Flowcharts for determining the target production process provided in some embodiments of this application;

[0042] Figure 4 A flowchart for determining a target manufacturing process is provided for some embodiments of this application;

[0043] Figure 5 Flowcharts for determining key process parameters provided in some embodiments of this application;

[0044] Figure 6 Flowcharts for determining the target manufacturing process provided for other embodiments of this application;

[0045] Figure 7 Structural block diagram of the apparatus for determining the production process of cathode material system provided in some embodiments of this application;

[0046] Figure 8 This is an internal structural diagram of a computer device provided in some embodiments of this application. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] The method for determining the production process of the cathode material system provided in this application embodiment can be applied to the production line of the cathode material system. Specifically, it can be used in a terminal connected to each production device on the production line. This terminal can acquire data such as the operating status of each production device and can also input control data into the corresponding production device to ensure that the product performance meets the preset expectations. The terminal can be, but is not limited to, various personal computers, laptops, and tablets.

[0049] When executing the method for determining the production process of the cathode material system provided in this application embodiment, the terminal first determines the parameter information of the cathode material system based on its components. The parameter information includes pre-constructed process information corresponding to the production process of the cathode material system or performance requirement information of the cathode material system. Then, the parameter information is input into a preset neural network model, and the target production process of the cathode material system is determined based on the output of the preset neural network model. The preset neural network model is obtained by training an initial neural network model using sample training data, which is simulated data generated based on a dynamic virtual experimental data generator and the cathode material system. The initial neural network model is a neural network model with both forward prediction and reverse decoding capabilities. Thus, the determination of the optimized process for a novel "material-coating" system can achieve cold-start modeling of the new system without relying on historical experimental data. Simultaneously, it possesses the ability to reverse-engineer the process based on performance requirements, solving the shortcomings of existing technologies such as difficulty in cold-starting and the inability to reverse-engineer only through forward prediction, and helping to improve the efficiency of determining the production process of the cathode material system.

[0050] In one embodiment, such as Figure 1 As shown, this method is illustrated using an example of its application in determining the production process of a cathode material system. In this embodiment, the method includes the following steps:

[0051] Step 102: Determine the parameter information of the cathode material system based on its composition.

[0052] The cathode material system is a composite system consisting of a core cathode active material with a specific chemical composition and a specific surface coating / modification layer. The components are the specific chemical substances that constitute the cathode material system, including the core cathode material and the coating / modifier. For example, taking a high-nickel ternary material@alumina coating, i.e., NCM95@Al2O3, as an example, the corresponding components include NCM95 and Al2O3, where NCM95 is the core cathode material and Al2O3 is the coating.

[0053] The parameter information of the cathode material system includes pre-constructed process information corresponding to the production process of the cathode material system or performance requirement information of the cathode material system. Process information includes process parameters, material characteristic parameters, and system identification parameters, as shown in Table 1. Process parameters may include coating temperature, reaction time, coating layer thickness, coating concentration, stirring rate, drying temperature, calcination atmosphere, slurry pH value, solvent volume ratio, and number of coating steps, as shown in Table 2. Material characteristic parameters may include the specific surface area, average particle size, nickel mass fraction, total impurity content, crystallinity, particle morphology, and tap density of the material, as shown in Table 3. System identification parameters may include cathode type and coating type. In the actual mass production process of the cathode material system, there is at least one set of process information, and each set of process information includes the aforementioned process parameters, material characteristic parameters, and system identification parameters.

[0054] Table 1

[0055]

[0056] Table 2

[0057]

[0058] Table 3

[0059]

[0060] The performance requirements information for the cathode material system refers to the performance requirements proposed for the final product, including multi-dimensional performance indicators. As shown in Table 4, the multi-dimensional performance indicators of the cathode material system can include interfacial binding energy, surface residual alkali content, interfacial impedance, lithium-ion diffusion coefficient, and initial exothermic temperature.

[0061] Table 4

[0062]

[0063] Optionally, when the user's goal is to determine the target production process based on the process information corresponding to the production process of the cathode material system, the composition type of the cathode material system can be obtained first. The composition type includes the cathode material type and the coating type. The composition type can be obtained by user input or by identification equipment on the production line. Then, the composition type, the process parameter information input by the user, and the material characteristic parameter information are used as the parameter information of the cathode material system. Among them, the process parameter information input by the user includes the value range of the process parameter, which can be determined based on historical production process data and experience of material systems similar to the cathode material system. The material characteristic parameter information can be obtained based on the test results of testing equipment that performs physical or chemical tests.

[0064] When a user's goal is to deduce the target production process from the performance requirements of the cathode material system, the user's expected performance indicators can be obtained first. These expected performance indicators include multi-dimensional expected performance values. Then, the expected performance indicators can be used as the parameter information of the cathode material system.

[0065] Step 104: Input the parameter information into the preset neural network model, and determine the target production process of the cathode material system based on the output of the preset neural network model.

[0066] Among them, the preset neural network model is a pre-trained interpretable dual-channel neural network. The preset neural network model is obtained by training the initial neural network model with sample training data. The sample training data is simulated data generated based on the dynamic virtual experimental data generator and the cathode material system. The initial neural network model is a neural network model with forward prediction and reverse decoding capabilities.

[0067] The sample training data is virtual data generated by the dynamic virtual experimental data generator based on the parameter information of the cathode material system, including virtual process parameters and corresponding performance indicators. The dynamic virtual experimental data generator is a simulator based on physicochemical mechanisms or high-fidelity simulation, capable of replacing some real experiments. It is a module that simulates and calculates the corresponding material performance indicators based on the input process parameters. The target production process of the cathode material system is the optimal set of executable process parameters recommended through model optimization.

[0068] Optionally, when the input parameter information is process information, the forward prediction channel of the preset neural network model is called to output the corresponding prediction performance index, and the target production process with the optimal comprehensive performance index is found through iterative optimization; wherein, the comprehensive performance index can be determined according to the above-mentioned multi-dimensional performance index.

[0069] When the input parameter information is performance requirement information, the reverse decoding channel of the preset neural network model is invoked to output the target production process that meets the performance requirement information.

[0070] The aforementioned method for determining the production process of cathode material systems first acquires the parameter information of the cathode material system. This parameter information can be either the initial process parameters or the final performance requirements, achieving a unified input interface for both working modes. Then, the parameter information is input into a preset neural network model trained by a dynamic virtual experimental data generator, which has both forward prediction and reverse decoding capabilities, directly outputting the target production process. In this way, cold-start modeling of new systems can be achieved without relying on historical experimental data, while also possessing the ability to reverse-engineer processes based on performance requirements. This solves the shortcomings of existing technologies, such as difficulty in cold-starting and the inability to reverse-engineer only through forward prediction, and helps to improve the efficiency of determining the production process of cathode material systems.

[0071] In one embodiment, such as Figure 2 As shown, if the parameter information consists of multiple sets of process information, the parameter information is input into a preset neural network model. Based on the output of the preset neural network model, the target production process of the cathode material system is determined, including:

[0072] Step 202: Input the process information into the preset neural network model to obtain the predicted performance index value corresponding to each process information.

[0073] The performance index prediction value is a quantitative evaluation result of the expected final product performance under the given process, calculated by a pre-set neural network model through its own positive prediction channel for each specific set of input process information. The performance index prediction value can be the specific value corresponding to the five-dimensional performance index mentioned above, namely the specific value corresponding to the interfacial binding energy, surface residual alkali content, interfacial impedance, lithium-ion diffusion coefficient, and initial exothermic temperature.

[0074] Optionally, multiple sets of virtual process information can be acquired. Each set of process information includes process parameters such as coating temperature, reaction time, coating thickness, coating concentration, stirring rate, drying temperature, and slurry pH value. Each set of process information is then input into the forward prediction channel of a preset neural network model. The forward prediction channel of the preset neural network model performs feature extraction and mapping on the input process information and outputs the corresponding predicted values ​​of five-dimensional performance indicators (predicted values ​​corresponding to interface binding energy, surface residual alkali content, interface impedance, lithium ion diffusion coefficient, and initial exothermic temperature). Finally, each set of process information and its corresponding predicted performance indicators are recorded to form a process-performance mapping dataset.

[0075] Step 204: Determine the target production process based on the predicted values ​​of performance indicators and information on each process.

[0076] Optionally, the performance score of each set of process information can be obtained first based on the performance index value of each set of process information; then the target production process can be determined from the process information with higher performance score values.

[0077] In this embodiment, the positive prediction function of the preset neural network model can quickly obtain the predicted values ​​of multi-dimensional performance indicators corresponding to each set of process information, thereby achieving efficient multi-performance synchronous prediction. In addition, the target production process can be screened based on the performance score value corresponding to the predicted performance indicator value, which can ensure the objectivity and accuracy of the determined target production process.

[0078] In one embodiment, such as Figure 3 As shown, based on the predicted values ​​of performance indicators and information on each process, the target production process is determined, including:

[0079] Step 302: For each piece of process information, determine the coating quality index corresponding to the predicted performance index value based on the predicted performance index value of the process information.

[0080] The coating quality index is a quantitative indicator that transforms the predicted values ​​of multi-dimensional performance indicators into a single comprehensive score. The coating quality index characterizes the overall effect of process information; a higher coating quality index indicates a better corresponding process.

[0081] Optionally, the coating quality index can be calculated using the following formula:

[0082] ;

[0083] in, The coating quality index; For the first The weights corresponding to the predicted values ​​of each performance indicator in this embodiment The value of can be 5, which corresponds to the five-dimensional performance index mentioned above; For the first One performance indicator; A standardized scoring function, such as the sigmoid function or a linear normalization function, is used to standardize the score. The original values ​​corresponding to each performance indicator are mapped to the interval [0,1].

[0084] For example, in power-oriented scenarios, such as when batteries are used in electric vehicles, the core requirements are high safety and strong fast charging capability. The key performance indicators in this case are the initial heat release temperature T_onset and the lithium-ion diffusion coefficient D_Li. The weight of the initial heat release temperature T_onset can be 0.3, the weight of the lithium-ion diffusion coefficient D_Li can be 0.25, the weight of the surface residual alkali content C_res can be 0.2, the weight of the interface impedance R_int can be 0.15, and the weight of the interface binding energy E_int can be 0.1.

[0085] For energy storage-focused scenarios, such as when batteries are used for grid energy storage, the core requirements are long-term stability and low cost. In this case, the key performance indicators are surface residual alkali content C_res and interfacial binding energy E_int. The weight of surface residual alkali content C_res can be 0.3, the weight of interfacial binding energy E_int can be 0.25, the weight of interfacial impedance R_int can be 0.2, the weight of lithium-ion diffusion coefficient D_Li can be 0.15, and the weight of initial exothermic temperature T_onset can be 0.1.

[0086] Step 304: Determine the process information corresponding to the maximum coating quality index as the target production process.

[0087] Optionally, all process information can be sorted from high to low according to the coating quality index, and the process information corresponding to the highest coating quality index can be determined as the target production process.

[0088] Furthermore, considering that the process information corresponding to the highest coating quality index may not necessarily meet production requirements, a preliminary screening can be conducted based on the coating quality index of each process. Only processes with a coating quality index ≥ 0.90 can continue to participate in the selection of the target production process. Process information with a coating quality index between [0.80, 0.90) can be fine-tuned; process information with a coating quality index < 0.8 can be directly deleted. Additionally, if the maximum coating quality index corresponding to all process information does not exceed 0.8, the process information needs to be redesigned and determined.

[0089] In this embodiment, the target production process is determined by the coating quality index, which ensures the objectivity and accuracy of the target production process determination.

[0090] In one embodiment, such as Figure 4 As shown, if the parameter information is the performance requirement information of the cathode material system; the parameter information is input into a preset neural network model, and based on the output of the preset neural network model, the target production process of the cathode material system is determined, including:

[0091] Step 402: Input the performance requirement information into the preset neural network model to obtain the process information corresponding to the performance requirement information.

[0092] Optionally, the performance requirement information can be first input into the reverse decoding channel of the preset neural network model, and the target performance index can be extracted through the reverse decoding channel and mapped to the shared feature space; then the corresponding process parameters can be decoded through the reverse decoding head; the process information that meets the performance requirements can be output, including process parameters such as coating temperature, reaction time, coating layer thickness, coating concentration, stirring rate, drying temperature, and slurry pH value.

[0093] Step 404: Determine the target production process based on the process information.

[0094] Optionally, the process information can be first input into the forward prediction channel of a preset neural network model to obtain the verification performance index; then the deviation between the verification performance index and the performance requirement information can be calculated. If the deviation is less than a preset threshold, the process information is determined as the target production process; if the deviation is greater than or equal to the preset threshold, the process information is iteratively corrected and the forward verification process is repeated until the deviation is less than the preset threshold or the maximum number of iterations is reached.

[0095] Alternatively, the process information output in step 402 can be used directly as the target production process. If subsequent experimental verification does not meet the performance requirements, return to step 402 to search again until the experimental verification result meets the performance requirements.

[0096] In this embodiment, the performance requirement information is directly mapped to the corresponding process information through the inverse decoding channel of the preset neural network model, and the initial recommended process can be obtained quickly without the need for external optimization algorithms. Furthermore, the recommended process can be verified through the forward prediction channel, and iterative corrections can be made based on the deviation to form a closed-loop optimization until the performance requirements are met, thereby improving the flexibility of the method for determining the production process of the cathode material system.

[0097] In one embodiment, the training process of the preset neural network model includes: jointly training the initial neural network model using a cycle consistency loss function to obtain the preset neural network model.

[0098] In this embodiment, the cycle consistency loss function is a supervisory signal used to simultaneously train the forward prediction channel and the backward inference channel. It ensures that a logically consistent closed loop is formed between the two parts of a model, thereby improving the overall accuracy of the model. The initial neural network model adopts an architecture of a shared feature extraction layer + bidirectional output head: the input layer receives 19-dimensional parameters (including the aforementioned process parameters, material property parameters, and system identification parameters), and extracts shared features through 3 fully connected layers (128 nodes per layer, Swish activation). These shared features are simultaneously connected to the forward output head and the backward decoding head. The forward output head is a 3-layer fully connected layer (256→128→64, ReLU+Sigmoid activation), which can output 5-dimensional performance indicators (including the aforementioned interface binding energy, residual alkali content, interface impedance, lithium-ion diffusion coefficient, and initial exothermic temperature), realizing forward prediction from process to performance. The backward decoding head is a 4-layer fully connected layer (512→256→128→64, LeakyReLU activation), which outputs 10-dimensional process parameters, realizing backward decoding from performance to process. The two output heads share the same feature space and are jointly trained using cycle consistency loss to ensure they are inverse mappings of each other. The optimizer uses AdamW (learning rate 3e). -4 It can be used in conjunction with cosine annealing scheduling to balance expressive power and computational efficiency.

[0099] Optionally, multiple sets of virtual sample data can be generated first based on the dynamic virtual experimental data generator and the composition information of the cathode material system. The virtual sample data includes virtual process parameters and corresponding virtual performance indicators. Then, the virtual process parameters are input into the forward prediction channel of the initial neural network model to obtain the predicted performance indicators. The forward prediction loss is calculated based on the deviation between the predicted performance indicators and the virtual performance indicators. Next, the virtual performance indicators are input into the reverse decoding channel of the initial neural network model to obtain the predicted process parameters. The reverse decoding loss is calculated based on the deviation between the predicted process parameters and the virtual process parameters. Then, the virtual performance indicators are input into the reverse decoding channel to obtain the reconstructed process parameters, and the first cycle consistency loss is calculated. Then, the predicted process parameters are input into the forward prediction channel to obtain the reconstructed performance indicators. That is, the predicted process obtained by reverse decoding is used again through the performance indicators obtained by the forward prediction channel to calculate the second cycle consistency loss. Then, the forward prediction loss, reverse decoding loss, first cycle consistency loss and second cycle consistency loss are weighted and summed to obtain the total loss. Finally, the parameters of the initial neural network model are updated through the backpropagation algorithm to minimize the total loss. Iterative training is performed until convergence, thereby obtaining the preset neural network model.

[0100] The composition information includes the cathode material type and coating type. Based on the dynamic virtual experimental data generator and the composition information of the cathode material system, multiple sets of virtual sample data are generated, which may include the following steps: First, based on the composition information, the corresponding physical driving empirical formula template is retrieved from a preset equation template library; the material characteristic parameters of the cathode material system are obtained, including specific surface area, nickel mass fraction, and total impurity content; then, surface defect density and coating uniformity are introduced as microstructure correction factors; then, sampling is performed within the feasible domain of the process parameters to generate multiple sets of virtual process parameters; then, the virtual process parameters, material characteristic parameters, and microstructure correction factors are substituted into the matching empirical formula template to calculate the corresponding virtual performance indicators; then, each set of virtual process parameters and its corresponding virtual performance indicators are combined to form a set of virtual sample data; the above process is repeated to generate multiple sets of virtual sample data, constituting a virtual sample dataset. The feasible domain of the process parameters can be determined based on historical production process data and experience of material systems similar to the cathode material system.

[0101] For example, the empirical formula template (typical empirical equation) includes the residual alkali formation rate equation and the interfacial binding energy equation, wherein the residual alkali formation rate equation is expressed in the following form:

[0102] ;

[0103] in, This refers to the residual alkali content on the surface. This is the residual alkali formation rate constant, used to reflect the activity of the material itself. The higher the nickel content, the better. The larger; This refers to the mass fraction of nickel. It is the nickel content sensitivity coefficient; It is the wrapping temperature; It is the temperature conversion constant; It is the reaction time; It is the time offset constant; This is the stirring effect coefficient; This is the defect impact coefficient; Specific surface area; The degree of surface defects can be determined based on the nickel content.

[0104] The interface binding energy equation is expressed as follows:

[0105] ;

[0106] in, This is the initial heat release temperature; The basic initial exothermic temperature is the intrinsic thermal stability of the material without coating. The binding energy influence coefficient; This refers to the interfacial bonding energy, which is also the bonding strength between the coating layer and the cathode material; It is the coefficient of influence of coating thickness; It refers to the thickness of the coating; It is the impurity influence coefficient; It is the total impurity content; It is the uniformity influence coefficient; It refers to the uniformity of the coating, that is, the degree of uniformity of the coating layer.

[0107] Furthermore, when generating virtual sample data based on the preset range of process parameters and the dynamic virtual experimental data generator, the coefficients in the empirical equations can be calibrated using a small amount of real experimental data, such as 12 sets of real experiments, to ensure the reliability of the generated virtual sample data. In addition, during the initial stage of training the initial neural network model, the input training data includes not only virtual sample data but also sample data obtained from real experimental data.

[0108] In this embodiment, training the initial neural network model through joint training ensures that forward prediction and reverse decoding are inverse mappings of each other, thereby solving the problem of bidirectional inconsistency when training separately, and thus achieving model training effects of accurate prediction, reliable reverse decoding, and bidirectional matching.

[0109] In one embodiment, such as Figure 5 As shown, the method also includes:

[0110] Step 502: Construct a directed acyclic graph based on the target production process and the preset construction algorithm.

[0111] The pre-defined construction algorithm learns the causal relationships in the production data corresponding to the target production process and constructs a directed acyclic graph (DAG). The nodes of the DAG include pre-defined latent variables, process parameters of the target production process, and corresponding actual performance indicators. Directed edges in the DAG represent the causal relationships between two corresponding nodes. Pre-defined latent variables are intermediate state variables that are pre-set based on domain knowledge and cannot be directly measured but exist. In this embodiment, the pre-defined latent variables may include slurry dispersibility, coating uniformity, and membrane integrity.

[0112] Optionally, multiple batches of production data generated during the actual production process of the target production process can be obtained first. This production data includes process parameters, latent variable measurements, and corresponding actual performance indicators. Then, the process parameters, latent variable measurements, and corresponding actual performance indicators are used as nodes, and a preset construction algorithm, such as the PC algorithm or gradient-based structure learning algorithm, is used to learn the causal structure of the production data. Then, the causal relationship direction between variables is determined through conditional independence testing or continuous optimization, and an initial directed acyclic graph is constructed. Finally, the initial directed acyclic graph is pruned and optimized to remove edges with a comprehensive confidence level lower than a preset threshold, resulting in the final directed acyclic graph.

[0113] Step 504: Based on the comprehensive confidence level corresponding to the edges of the directed acyclic graph, determine the key process parameters in the target production process.

[0114] The overall confidence level is a quantitative score of the credibility of a causal relationship, ranging from 0 to 1. Key process parameters are root cause process parameters with zero in-degree that significantly affect multiple performance indicators. For example, a key process parameter might be the stirring speed, affecting the surface residual alkali content C_res and the interfacial impedance R_int; another example is the coating temperature, affecting the interfacial binding energy E_int and the lithium-ion diffusion coefficient D_Li.

[0115] Optionally, one can first traverse all nodes in the directed acyclic graph, identifying nodes with an in-degree of zero as candidate root variables and nodes with an out-degree of zero as endpoints, and then identify all causal paths from the nodes corresponding to the candidate root variables to the endpoints. Next, from each candidate root variable, nodes of process parameter type are selected as candidate critical process parameters. Then, the overall confidence level of each candidate critical process parameter is calculated; the overall confidence level is the product of the confidence levels of all directed edges originating from that node, where each directed edge is a directed edge on one of the causal paths from the candidate root variable to the endpoint node (the node with an out-degree of zero). Finally, the candidate critical process parameter with the highest overall confidence level is determined as the critical process parameter in the target production process. For example, assuming that the overall confidence level of a certain identified causal path, "stirring rate → slurry dispersibility → coating uniformity → surface residual alkali content," is the highest, then the stirring rate can be determined as the root variable, i.e., the critical process parameter, and can be prioritized for control.

[0116] Alternatively, in another exemplary embodiment, the comprehensive influence intensity of each candidate critical process parameter can be calculated, and the candidate critical process parameter with the highest comprehensive influence intensity can be determined as the critical process parameter in the target production process. Here, the comprehensive influence intensity is the weighted sum of the confidence level and influence intensity of all directed edges originating from the node. A directed edge is a directed edge on one of the directed paths from the candidate root dependent variable to the endpoint node (a node with an out-degree of zero), and the influence intensity is the product of the influence intensities of each directed edge on that path. The confidence level of a directed edge represents the degree of credibility of the causal relationship (directed edge) and ranges from 0 to 1; a larger value indicates higher credibility. The influence intensity of a directed edge represents the magnitude of the causal relationship and ranges from 0 to 1; a larger value indicates stronger influence.

[0117] Furthermore, during the construction of the causal graph, updates can be made based on production data generated from the actual production process. Causal graph reconstruction is triggered when any of the following conditions are met: performance prediction deviation >5% for three consecutive batches; adjustment range of major equipment parameters >10%; or change of raw material supplier or coating type. During the update process, an incremental learning approach can be used to make local corrections while retaining high-confidence edges. Moreover, if a new causal path is identified during the update process using the above method, small-batch experiments can be automatically planned for verification. If the experimental verification success rate is ≥67%, the confidence of the path is increased; if the experimental verification success rate is <67%, its weight is reduced. In addition, all causal paths can be stored in structured data format, including causal edges and their attributes (starting variable, ending variable, influence strength, confidence, and last update time), for visualization in the human-computer interaction interface.

[0118] In this embodiment, a directed acyclic graph is automatically constructed from actual production data using a preset construction algorithm. Preset latent variables, process parameters, and performance indicators are used as nodes, and directed edges represent causal relationships. This enables the objective discovery and quantitative expression of causal relationships. Furthermore, based on the comprehensive confidence of the directed edges, the key process parameters that have the greatest impact on performance are determined, which can solve the shortcomings of traditional methods that rely on expert experience and cannot discover unknown relationships.

[0119] In one embodiment, such as Figure 6 As shown, the parameter information is input into a preset neural network model. Based on the output of the preset neural network model, the target production process of the cathode material system is determined, including:

[0120] Step 602: If the model of the first production equipment that generates the cathode material system is different from the model of the second production equipment corresponding to the preset neural network model, then the parameters of the preset neural network model are adjusted based on the model of the first production equipment to obtain the adjusted neural network model.

[0121] The first production equipment is located in the new plant area and is the target deployment equipment, while the second production equipment is located in the original plant area. Because different production equipment exhibit different physical effects on the same process parameters—for example, at the same stirring rate, different stirring equipment results in inconsistent slurry dispersion—directly transplanting the pre-set neural network model adapted to the second production equipment in the original plant area to the first production equipment in the new plant area could easily lead to significant deviations in the prediction results of the pre-set neural network model, thereby affecting the production quality of the cathode material production system. Therefore, it is necessary to adjust the parameters of the pre-set neural network model adapted to the second production equipment in the original plant area.

[0122] Optionally, the equipment model and parameters of the first production equipment can be obtained first, including temperature control accuracy, stirring device type, furnace structure, and atmosphere control method. Then, the global general model corresponding to the second production equipment can be obtained from the federated collaborative cloud platform. Based on the differences between the equipment parameters of the first and second production equipment, the adjustment strategy of the model parameters can be determined. Then, the output layer or specific feature layer of the global general model can be fine-tuned to adapt it to the characteristics of the first production equipment. Finally, a small amount of local production data from the first production equipment can be used to verify the fine-tuned model. If the verification is successful, the adjusted neural network model is obtained.

[0123] Step 604: Input the parameter information into the adjusted neural network model, and determine the target production process of the cathode material system based on the output of the adjusted neural network model.

[0124] Optionally, parameter information can be input into the adjusted neural network model. The parameter information includes process parameters or performance requirement information. When the parameter information is a process parameter, the forward prediction channel of the adjusted model is called to output the corresponding predicted performance index, and the target production process is found through iterative optimization. When the parameter information is a performance requirement information, the reverse decoding channel of the adjusted model is called to output the target production process that meets the performance requirements. Finally, the target production process output by the adjusted model is used as the final production process applicable to the production equipment.

[0125] In this embodiment, by detecting the model differences between the first and second production equipment and determining the adjustment strategy of the model parameters based on the equipment parameters (temperature control accuracy, stirring device type, etc.), the global general model (preset neural network model) can be fine-tuned in a targeted manner. This enables the model to adapt from the original factory area to the new factory area, thereby solving the problems of model failure caused by equipment differences and the need for a large number of experiments for cross-factory migration in traditional methods, and thus realizing a cross-factory knowledge sharing mechanism.

[0126] In an exemplary embodiment, taking NCM95@Al2O3 as an example, the method for determining the production process of the cathode material system of this application will be described in detail.

[0127] 1. Background and Objectives: A cathode material company plans to introduce an Al2O3 wet coating process for its high-nickel NCM95 material, but currently lacks any historical production data. Traditional R&D requires at least 30 sets of Design of Experiments (DOEs), which is expected to take more than 6 weeks. This embodiment verifies the ability of this invention to rapidly construct an effective process solution under zero-sample conditions.

[0128] 2. System input parameters are shown in Table 5:

[0129] Table 5

[0130]

[0131] First, the dynamic virtual experimental data generator is started, and a strongly temperature-dependent empirical equation template is automatically matched according to the type of NCM95@Al2O3. The surface defect density δ_defect=0.12 (obtained from AI analysis of SEM images) and the initial estimate of the coating uniformity σ_uniform=0.65 are introduced. Then, based on the above input parameters and the dynamic virtual experimental data generator, 800 sets of virtual sample data are generated, covering key ranges such as T (680~750℃), t (6~14h), and d (2.0~5.0nm). ±8% noise is added to simulate measurement fluctuations.

[0132] The aforementioned virtual sample data was used to train the initial neural network model. The forward prediction channel output five-dimensional performance metrics, including: interfacial binding energy E_int, surface residual alkali content C_res, interfacial impedance R_int, lithium-ion diffusion coefficient D_Li, and initial exothermic temperature T_onset. The reverse inference channel supports deducing the process from the target and uses the AdamW optimizer with a learning rate of 3e. -4 After 200 training cycles, the algorithm converged. The first batch of 15 small-scale samples were then prepared, and the measured performance is as follows: average C_res = 742 ± 68 ppm, T_onset = 226 ± 3℃, and EIS measured R_int = 0.54 ± 0.07 Ω·cm. 2 Bayesian optimization was used to calibrate the coefficients of the D-VEDG equations, reducing the deviation between virtual and measured data to ≤7%, resulting in a well-trained pre-defined neural network model. This model was then used to predict the input process information and calculate the coating quality index. The coating quality index The process information corresponding to the maximum coating quality index ≥0.90 is determined as the target production process.

[0133] When the performance requirements are determined to be C_res < 800, T_onset > 225℃, and D_Li > 1.0 × 10⁻¹⁴ cm² / s, the recommended process obtained through the reverse inference channel of the preset neural network model is: T = 725℃, t = 11.5h, d = 3.8nm, C_coat = 1.2wt%, R_stir = 290r / min. After verification through the forward prediction channel, the predicted performance indicators are: E_int = 2.12eV, C_res = 730ppm, and R_int = 0.53Ω·cm. 2 , D_Li=1.12×10-14cm2 / s, T_onset=228℃.

[0134] Final conclusions: Development cycle was only 14 days; no large-scale DOE experiments were required; cold start modeling and continuous optimization were successfully achieved; It meets the standards and has passed safety certification.

[0135] In another exemplary embodiment, a detailed description is given using NCA@organic polymer as the cathode material system:

[0136] 1. Background and Objectives: A battery manufacturer uses organic polymers to coat NCA materials to improve cycle life, but has repeatedly encountered the risk of thermal runaway. The objective is to increase T_onset to above 230°C while controlling C_res < 700 ppm.

[0137] 2. System input parameters are shown in Table 6:

[0138] Table 6

[0139] `

[0140] 3. Model Operation and Causal Discovery: The preset neural network model was fine-tuned using existing local data. Although C_res was well controlled, T_onset fluctuated greatly. A directed acyclic graph, i.e. a causal graph, was constructed, and a new causal path was found: drying temperature → membrane integrity → exothermic initiation temperature. The confidence level of this key causal path was 0.89 and the influence strength was 0.78.

[0141] 4. Verify the experimental design to validate the newly discovered causal path: Trigger the causal verification mechanism to automatically generate three sets of verification experiments: Group A: T_dry=160℃, Group B: T_dry=180℃, Group C: T_dry=200℃, with other parameters kept consistent, and test T_onset.

[0142] 5. The experimental results are shown in Table 7:

[0143] Table 7

[0144]

[0145] According to Table 7, Group B is the best, being intact and uncarbonized.

[0146] 6. Update the model and recommendation process: The causal module confirms the validity of the newly discovered causal path, with the confidence level rising to 0.95; Reverse decoding recommendation: T_dry=180℃, with other parameters fine-tuned;

[0147] The new predicted performance is: T_onset = 235℃, C_res = 680ppm. =0.91.

[0148] 7. Final conclusions: The thermal stability was successfully improved to an industry-leading level; the practical guiding value of causal reasoning was verified; the transformation from "experience-based guessing" to "mechanism-driven" was achieved; and the development cycle was only 12 days.

[0149] In another exemplary embodiment, taking the scenario of NCM90@Li3PO4 as the cathode material system and multi-factory collaborative migration as an example, the specific content is as follows:

[0150] 1. Background and Objectives: Plant A has established a mature NCM90@Li3PO4 coating process, and now it is necessary to replicate this process in Plant B. The two plants have different equipment models (Plant B uses a new type of closed-loop calcining furnace), making direct transfer impossible.

[0151] 2. The differences in local parameters are shown in Table 8:

[0152] Table 8

[0153]

[0154] 3. Collaborative Process: Plant B logs into the Fed-Swarm cloud platform; uploads local equipment parameters and incoming material information; the system automatically retrieves similar scene models (92% matching accuracy); downloads the global model Global-XBiPath-v3 trained by Plant A; performs minor fine-tuning based on local sensor data (only 5 sets of data required); and outputs the adapted recommended process.

[0155] 4. Recommendations and actual test results are shown in Table 9:

[0156] Table 9

[0157]

[0158] The coating quality index was calculated. =0.95.

[0159] 5. Final conclusion: No DOE experiment required, the first chip met the requirements; saved approximately 14 days of debugging time; verified the feasibility of cross-plant knowledge transfer; It is higher than the original level of Plant A, reflecting the advantages of equipment upgrade.

[0160] In addition, this application provides three comparative examples, as detailed below:

[0161] Comparative Example 1: Traditional DOE experimental method (control baseline), method description: L9(3 4 An orthogonal experimental design was used, with three levels for each of the four variables T, t, d, and C_coat, resulting in nine experimental groups. Each group was repeated three times, for a total of 27 batches. The process range is shown in Table 10.

[0162] Table 10

[0163]

[0164] The results include: the optimal combination of process parameters is: T=720℃, t=12h, d=4.0nm, C_coat=1.5wt%, C_res=780ppm, T_onset=220℃. =0.82; took 42 days and was costly; no higher potential areas were found (due to sparse sampling).

[0165] Comparative Example 2: Pure Black-Box AI Prediction (No Inverse or Causal Analysis), Method Description: Using a Deep Neural Network (DNN) for forward prediction only; input parameters → output performance, no inversion or interpretation supported. Problem Exposure: Engineers requested "reducing impedance," but the system could not provide specific adjustment suggestions; it recommended increasing pH_slurry (the pH value of the coating slurry) to 11.5, leading to severe gelation of the slurry; coating failed, resulting in the loss of a batch of raw materials. =0.79, although the prediction accuracy is acceptable, it leads to decision-making errors; the lack of causal understanding results in non-transferability.

[0166] Comparative Example 3: Optimization ignoring causal relationships. Method description: Optimization based solely on statistical correlation revealed "high pH → low residual alkali", leading to blindly increasing the pH value.

[0167] The implementation process included: gradually increasing pH_slurry from 9.2 to 11.5; initially, C_res decreased to 690 ppm, seemingly a success; however, severe gelation occurred in the subsequent coating stage; electrode peel strength decreased by 40%; the battery experienced severe gas swelling, and ARC testing showed T_onset=195℃, lower than the original level.

[0168] The results include: =0.65, judged as "poor"; due to ignoring the real mechanism of "pH too high → slurry instability → interface deterioration", significant losses were caused.

[0169] The summary comparison table is shown in Table 11:

[0170] Table 11

[0171]

[0172] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0173] Based on the same inventive concept, this application also provides a cathode material system production process determination apparatus for implementing the above-described method for determining the cathode material system production process. The solution provided by this apparatus is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the cathode material system production process determination apparatus provided below can be found in the limitations of the cathode material system production process determination method described above, and will not be repeated here.

[0174] In one embodiment, such as Figure 7 As shown, a device for determining the production process of a cathode material system is provided, comprising: a parameter information determination module 702 and a production process determination module 704, wherein:

[0175] The parameter information determination module 702 is used to determine the parameter information of the cathode material system based on the composition of the cathode material system; the parameter information includes the process information corresponding to the pre-constructed production process of the cathode material system or the performance requirement information of the cathode material system.

[0176] The production process determination module 704 is used to input parameter information into a preset neural network model and determine the target production process of the cathode material system based on the output of the preset neural network model. The preset neural network model is obtained by training an initial neural network model with sample training data. The sample training data is simulated data generated based on a dynamic virtual experimental data generator and the cathode material system. The initial neural network model is a neural network model with forward prediction and reverse decoding capabilities.

[0177] In one embodiment, the production process determination module 704 includes a performance index prediction value determination submodule and a target production process determination submodule. The performance index prediction value determination submodule is used to input each process information into a preset neural network model to obtain the performance index prediction value corresponding to each process information. The target production process determination submodule is used to determine the target production process based on the performance index prediction value and each process information.

[0178] In one embodiment, the target production process determination submodule is used to determine the coating quality index corresponding to the predicted performance index value for each process information, based on the predicted performance index value of the process information; and to determine the process information corresponding to the maximum coating quality index as the target production process.

[0179] In one embodiment, the production process determination module 704 is used to input performance requirement information into a preset neural network model to obtain process information corresponding to the performance requirement information; and to determine the target production process based on the process information.

[0180] In one embodiment, the apparatus further includes a model training module, which is used to jointly train an initial neural network model using a cycle consistency loss function to obtain a preset neural network model.

[0181] In one embodiment, the apparatus further includes a key process parameter determination module, which is used to construct a directed acyclic graph based on the target production process and a preset construction algorithm; the nodes of the directed acyclic graph include preset latent variables, process parameters of the target production process and corresponding actual performance indicators, and the directed edges in the directed acyclic graph are used to represent the causal relationship between the corresponding two nodes; the key process parameters in the target production process are determined based on the comprehensive confidence level corresponding to the edges of the directed acyclic graph.

[0182] In one embodiment, the production process determination module 704 is used to adjust the parameters of the preset neural network model based on the model of the first production equipment to obtain the adjusted neural network model if the model of the first production equipment that generates the cathode material system is different from the model of the second production equipment corresponding to the preset neural network model; input the parameter information into the adjusted neural network model; and determine the target production process of the cathode material system based on the output of the adjusted neural network model.

[0183] Each module in the aforementioned cathode material system production process determination device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0184] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 8 As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a method for determining a cathode material system production process. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0185] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0186] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0187] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0188] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0189] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0190] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0191] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0192] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for determining the production process of a cathode material system, characterized in that, The method includes: Based on the composition of the cathode material system, the parameter information of the cathode material system is determined; the parameter information includes the process information corresponding to the pre-constructed production process of the cathode material system or the performance requirement information of the cathode material system. The parameter information is input into a preset neural network model, and the target production process of the cathode material system is determined based on the output of the preset neural network model. The preset neural network model is obtained by training an initial neural network model with sample training data, and the sample training data is simulated data generated based on a dynamic virtual experimental data generator and the cathode material system. The initial neural network model is a neural network model with forward prediction and reverse decoding capabilities.

2. The method according to claim 1, characterized in that, If the parameter information consists of multiple sets of process information; the step of inputting the parameter information into a preset neural network model and determining the target production process of the cathode material system based on the output of the preset neural network model includes: The process information is input into a preset neural network model to obtain the predicted performance index value corresponding to each process information. Based on the predicted values ​​of the performance indicators and the process information, the target production process is determined.

3. The method according to claim 2, characterized in that, The step of determining the target production process based on the predicted performance index values ​​and the process information includes: For each piece of process information, based on the predicted performance index of the process information, a coating quality index corresponding to the predicted performance index is determined; The process information corresponding to the maximum coating quality index is determined as the target production process.

4. The method according to claim 1, characterized in that, If the parameter information is performance requirement information for the cathode material system; the step of inputting the parameter information into a preset neural network model and determining the target production process of the cathode material system based on the output of the preset neural network model includes: The performance requirement information is input into the preset neural network model to obtain the process information corresponding to the performance requirement information; Based on the process information, the target production process is determined.

5. The method according to any one of claims 1-4, characterized in that, The training process of the preset neural network model includes: The initial neural network model is jointly trained using the cycle consistency loss function to obtain the preset neural network model.

6. The method according to any one of claims 1-4, characterized in that, The method further includes: Based on the target production process and the preset construction algorithm, a directed acyclic graph is constructed; the nodes of the directed acyclic graph include preset latent variables, the process parameters of the target production process and the corresponding actual performance indicators, and the directed edges in the directed acyclic graph are used to represent the causal relationship between the corresponding two nodes. Based on the comprehensive confidence level corresponding to the edges of the directed acyclic graph, the key process parameters in the target production process are determined.

7. The method according to claim 1, characterized in that, The step of inputting the parameter information into a preset neural network model and determining the target production process of the cathode material system based on the output of the preset neural network model includes: If the model of the first production equipment that generates the cathode material system is different from the model of the second production equipment corresponding to the preset neural network model, the parameters of the preset neural network model are adjusted based on the model of the first production equipment to obtain the adjusted neural network model. The parameter information is input into the adjusted neural network model, and the target production process of the cathode material system is determined based on the output of the adjusted neural network model.

8. A device for determining the production process of a cathode material system, characterized in that, The device includes: The parameter information determination module is used to determine the parameter information of the cathode material system based on the composition of the cathode material system; the parameter information includes pre-constructed process information corresponding to the production process of the cathode material system or performance requirement information of the cathode material system; The production process determination module is used to input the parameter information into a preset neural network model and determine the target production process of the cathode material system based on the output of the preset neural network model; wherein, the preset neural network model is obtained by training an initial neural network model with sample training data, and the sample training data is simulated data generated based on a dynamic virtual experimental data generator and the cathode material system; the initial neural network model is a neural network model with forward prediction and reverse decoding capabilities.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images