Tab tearing defect prediction method, device and equipment and storage medium

Abstract

The invention discloses a tab tearing defect prediction method, device and equipment and a storage medium, and the method comprises the steps: obtaining multi-dimensional data in a battery cell production process, and generating a feature data set based on the multi-dimensional data; training a preset first supervision model based on the feature data set to obtain a second supervision model; training a preset first unsupervised model based on the feature data set to obtain a second unsupervised model; acquiring to-be-predicted data, and inputting the to-be-predicted data into the second supervised model and the second unsupervised model to obtain a first prediction result and a second prediction result; and based on the first prediction result and the second prediction result, obtaining a final prediction result, and realizing prediction of the tab tearing defect. The multi-dimensional feature data in the battery cell production process is predicted based on the supervised model and the unsupervised model, the situation that tab tearing defect prediction depends on a single voltage signal is avoided, and the prediction accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery cell defect prediction, and in particular to a method, device, equipment and storage medium for predicting a tab tear defect. Background Art

[0002] With the rapid development of new energy vehicles, battery technology has become increasingly important. Under the dual effects of the improvement of global environmental awareness and the promotion of policies and regulations, electric vehicles are rapidly replacing traditional fuel vehicles and becoming the mainstream choice for future travel. However, as the core component of electric vehicles, the performance and safety of batteries are directly related to the overall performance of the vehicle and user satisfaction. In the battery structure, the quality and stability of the battery cell as the basic component unit are particularly critical. The battery cell is composed of a positive electrode, a negative electrode, a separator and an electrolyte, and the tab, as a key component connecting the positive and negative electrodes with the external circuit, is made of aluminum or copper. Its thickness, width and shape directly affect the performance of the battery cell. In the production process of the battery cell, due to various factors such as material properties, production process and equipment precision, the problem of tab tearing has become a common challenge. This defect not only threatens product quality, but also may cause safety risks. For example, the tearing of the tab may cause the current path to be unstable, reduce the energy density and power output of the battery, and even cause an internal short circuit, causing the battery to heat up, smoke or even burn, posing a threat to the safety of users.

[0003] The existing method for detecting tab tear defects is mainly to monitor the terminal voltage changes and its derivative indicators during the cycle of the power battery, including the terminal voltage value and the corresponding differential data at different working stages. When the differential value of the terminal voltage to time or power exceeds the preset first threshold, it can be preliminarily determined that the tab has internal structural damage such as tearing. Although this method can detect potential tab tearing problems in real time during production and use, it relies on a single voltage signal as the basis for judgment, is easily affected by other factors, and has low prediction accuracy. Summary of the invention

[0004] The present invention provides a method, device, equipment and storage medium for predicting a tab tear defect, so as to predict a tab tear defect based on multi-dimensional data and improve the prediction accuracy.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a method for predicting a tab tear defect, comprising:

[0006] Acquire multidimensional data in a battery cell production process, and generate a feature data set based on the multidimensional data;

[0007] Training a preset first supervised model based on the feature data set to obtain a second supervised model;

[0008] Training a preset first unsupervised model based on the feature data set to obtain a second unsupervised model;

[0009] Acquire data to be predicted, and input the data to be predicted into the second supervised model and the second unsupervised model respectively to obtain a first prediction result and a second prediction result;

[0010] Based on the first prediction result and the second prediction result, a final prediction result is obtained to achieve the prediction of the tab tear defect.

[0011] The present invention extracts features from multidimensional data in the battery cell production process to avoid the prediction of the tab tear defect from relying on a single voltage signal, thereby improving the prediction accuracy. At the same time, since negative sample data in the battery cell production process is extremely scarce and serious misjudgment of negative samples may cause safety accidents, the first supervised model and the first unsupervised model are trained according to the accuracy of negative samples, so that the multidimensional data is predicted according to the second supervised model and the second unsupervised model, thereby improving the recognition accuracy of the second supervised model and the second unsupervised model for negative samples, and finally, the prediction reliability is improved by integrating the prediction results of the two models.

[0012] Furthermore, generating a feature data set based on the multidimensional data includes:

[0013] Perform feature extraction based on the multidimensional data to obtain feature data; the feature data includes labeled data and unlabeled data, and the labeled data includes positive samples and negative samples;

[0014] Draw a first kernel density estimation graph based on the labeled data, and obtain unevenly distributed first feature data based on the first kernel density estimation graph;

[0015] Draw a second kernel density estimation graph based on the labeled data and the unlabeled data, and obtain evenly distributed second feature data based on the second kernel density estimation graph;

[0016] A feature data set is generated based on the first feature data and the second feature data.

[0017] Furthermore, the first supervised model is trained based on the feature data set to obtain a second supervised model, including:

[0018] Determine a first optimization parameter of the first supervised model, and optimize the first optimization parameter according to a grid search algorithm and the accuracy of negative samples to obtain a first parameter;

[0019] The first supervised model is trained according to the first parameter and the feature data set until the accuracy of the negative sample reaches a preset first accuracy threshold, and the training is stopped to obtain a second supervised model.

[0020] Furthermore, the first unsupervised model is trained based on the feature data set to obtain the second unsupervised model, specifically:

[0021] Determine a second optimization parameter of the first unsupervised model, and optimize the second optimization parameter according to a grid search algorithm and the accuracy of negative samples to obtain a second parameter;

[0022] The first unsupervised model is trained according to the second parameter and the feature data set until the accuracy of the negative sample reaches a preset second accuracy threshold, and the training is stopped to obtain a second unsupervised model.

[0023] Further, the obtaining of the final prediction result based on the first prediction result and the second prediction result is specifically:

[0024] The first prediction result and the second prediction result are integrated, and when the first prediction result and the second prediction result are both negative samples, the final prediction result is a negative sample.

[0025] Furthermore, after generating a feature data set based on the multidimensional data, the method further includes:

[0026] Dividing the positive samples and negative samples in the training data according to a preset ratio to obtain a training set and a test set for the positive samples and a training set and a test set for the negative samples;

[0027] Based on the training set and test set of the positive samples and the training set and test set of the negative samples, a final training set and test set are obtained.

[0028] Furthermore, the extracting features based on the multidimensional data to obtain feature data includes:

[0029] Acquire multidimensional data in the core production process, extract key features from the multidimensional data, and obtain production process parameters;

[0030] Null value data and zero value data in the production process parameters are eliminated, and characteristic data are obtained based on the normal distribution of the production process parameters.

[0031] In a second aspect, the present invention provides a tab tear defect prediction device, comprising: a data preprocessing module, a supervised learning module, an unsupervised learning module and a prediction module;

[0032] The data preprocessing module is used to obtain multidimensional data in the battery cell production process and generate a feature data set based on the multidimensional data;

[0033] The supervised learning module is used to train the preset first supervised model based on the feature data set to obtain a second supervised model;

[0034] The unsupervised learning module is used to train the preset first unsupervised model based on the feature data set to obtain a second unsupervised model;

[0035] The prediction module is used to predict the data to be predicted according to the trained linear regression model and the isolation forest model respectively, obtain a first prediction result and a second prediction result, and integrate the first prediction result and the second prediction result to obtain a final prediction result.

[0036] In a third aspect, the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the method for predicting a tab tear defect when executing the computer program.

[0037] In a fourth aspect, the present invention provides a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method for predicting a tab tear defect. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic flow chart of a method for predicting a tab tear defect provided in an embodiment of the present invention;

[0039] Figure 2 A schematic diagram of a feature data set partitioning process provided by an embodiment of the present invention;

[0040] Figure 3 A schematic diagram of a prediction process of data to be predicted provided by an embodiment of the present invention;

[0041] Figure 4 A schematic structural diagram of a tab tear defect prediction device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0043] The terms "first" and "second" and the like in the specification, claims and drawings of this application are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0044] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0045] Example 1

[0046] See also Figure 1 , Figure 1 The flowchart of a method for predicting a tab tear defect provided in an embodiment of the present invention is shown in FIG. The method for predicting a tab tear defect provided in an embodiment of the present invention includes steps 101 to 105, which are as follows:

[0047] Step 101: Acquire multidimensional data in a cell production process, and generate a feature data set based on the multidimensional data;

[0048] In this embodiment, sensors are rationally arranged in various stages of battery cell production (such as coating, winding, assembly, charge and discharge testing, etc.) to obtain multi-dimensional data of the battery cells in real time, wherein the multi-dimensional data includes parameters such as voltage, current, temperature, humidity, charge and discharge rate, pressure, and vibration.

[0049] In this embodiment, generating a feature data set based on the multidimensional data includes:

[0050] Perform feature extraction based on the multidimensional data to obtain feature data; the feature data includes labeled data and unlabeled data, and the labeled data includes positive samples and negative samples;

[0051] Draw a first kernel density estimation graph based on the labeled data, and obtain unevenly distributed first feature data based on the first kernel density estimation graph;

[0052] Draw a second kernel density estimation graph based on the labeled data and the unlabeled data, and obtain evenly distributed second feature data based on the second kernel density estimation graph;

[0053] A feature data set is generated based on the first feature data and the second feature data.

[0054] In this embodiment, the step of extracting features based on the multidimensional data to obtain feature data includes:

[0055] Acquire multidimensional data in the core production process, extract key features from the multidimensional data, and obtain production process parameters;

[0056] Null value data and zero value data in the production process parameters are eliminated, and characteristic data are obtained based on the normal distribution of the production process parameters.

[0057] In this embodiment, feature extraction is performed on multidimensional data according to preset feature indicators to obtain feature data, namely production process parameters, wherein the production process parameters include the jump value of the distance from the positive electrode to the baseline, hot pressing Hi-pot, K value, etc.

[0058] In this embodiment, it is also necessary to unify the characteristic data rows and uniformly name the production process parameter files to ensure the consistency of the data format and improve the usability of the data.

[0059] In this embodiment, the feature data includes labeled data and unlabeled data, the labeled data includes positive samples and negative samples, and the labeled data in the feature data is preprocessed to ensure the neatness of the labeled data.

[0060] In this embodiment, the labeled data is obtained by the production line workers disassembling the battery cells during the battery cell production process and marking the battery cells after confirming their quality.

[0061] In this embodiment, the feature data is cleaned to remove empty and zero-value data to avoid problems in the model due to missing data during training; and duplicate rows in the feature data are removed to ensure the neatness of the labeled data in the production process parameters.

[0062] In this embodiment, after data cleaning is completed, characteristic data is obtained based on the normal distribution of the production process parameters. Specifically, the mean and standard deviation of the labeled data are calculated, and the labeled data within the range of the mean minus 1 times the standard deviation and the mean plus 1 times the standard deviation are retained.

[0063] In this embodiment, the labeled data includes positive samples and negative samples, wherein the positive samples and negative samples are divided according to the label content. If the label content is an OK label, the sample is a positive sample, and if the label content is NG, the sample is a negative sample.

[0064] In this embodiment, a first kernel density estimation graph is drawn based on the positive samples and negative samples in the labeled data to obtain unevenly distributed first feature data based on the first kernel density estimation graph, and by analyzing the feature distribution of the positive samples and the negative samples, feature samples that can distinguish between OK labels and NG labels are screened out from the feature data, that is, the first feature data, and the first feature data is unevenly distributed feature data.

[0065] In this embodiment, by screening out features that effectively distinguish OK and NG labels, the influence of interfering features is reduced, thereby improving the classification accuracy of the prediction model. At the same time, by focusing on unevenly distributed feature data, the prediction model can be more robust in dealing with noisy data and has a stronger tolerance for a small number of abnormal samples.

[0066] In this embodiment, based on the distribution of labeled data and unlabeled data in the feature data, a second kernel density estimation graph is drawn, and the feature distribution of labeled and unlabeled data is analyzed based on the second kernel density estimation graph, from which second feature data that can evaluate the generalization performance of the model are screened out, and the second feature data is a uniformly distributed feature.

[0067] In this embodiment, by using the evenly distributed second feature data to evaluate the generalization ability of the model, not only can the stability of the model be improved and the interpretability of the features be enhanced, but also unsupervised learning can be supported and the overall data utilization rate can be improved.

[0068] In this embodiment, the first feature data and the second feature data screened twice are integrated as a feature group for model training, and feature data are extracted from the feature data after data cleaning according to the feature group to generate a feature data set.

[0069] In this embodiment, after generating a feature data set based on the multidimensional data, the method further includes:

[0070] Dividing the positive samples and negative samples in the training data according to a preset ratio to obtain a training set and a test set for the positive samples and a training set and a test set for the negative samples;

[0071] Based on the training set and test set of the positive samples and the training set and test set of the negative samples, a final training set and test set are obtained.

[0072] Please refer to soil 2, Figure 2 A schematic diagram of a feature data set partitioning process provided by an embodiment of the present invention.

[0073] In this embodiment, the feature data set is divided into a training set and a test set. In order to ensure that the training set and the test set have a considerable proportion of positive and negative samples, the positive and negative samples are first distinguished, and the training set and the test set are divided according to the preset ratio, and the training set and the test set of the positive sample and the training set and the test set of the negative sample are obtained, and then the divided training set and the test set are merged to obtain the final training set and the test set. Among them, the division ratio is set to 8:2 by default.

[0074] Step 102: training a preset first supervised model based on the feature data set to obtain a second supervised model;

[0075] In this embodiment, the first supervised model is trained based on the feature data set to obtain the second supervised model, including:

[0076] Determine an optimization parameter of the first supervised model, and optimize the first optimization parameter according to a grid search algorithm and the accuracy of negative samples to obtain a first parameter;

[0077] The first supervised model is trained according to the first parameter and the feature data set until the accuracy of the negative sample reaches a preset first accuracy threshold, and the training is stopped to obtain a second supervised model.

[0078] In this embodiment, a first supervised model is constructed based on a linear regression model, and the optimization parameters of the first supervised model are determined. Specifically, the optimization parameters of the first supervised model are penalty, C, and class_weight, wherein the penalty parameter specifies the type of regularization used by the model. Regularization is one of the techniques used to prevent overfitting, which limits the size of model parameters by adding a penalty term to the loss function. C is the inverse of the regularization strength. C controls the tolerance of the model to misclassification. A larger C value means that the model pays more attention to errors in the training data and tries to reduce these errors, which may lead to increased model complexity and easy overfitting. A smaller C value means that the model is more inclined to use a simple model, even if it means more training errors. The class_weight parameter refers to the weight of the category. When processing an unbalanced data set, the model may tend to predict categories with a larger number of samples. By adjusting class_weight, the relative importance of different categories in the model training process can be changed to help the model better learn the characteristics of the minority class.

[0079] In this embodiment, a first supervised model is constructed based on the initial framework of the linear regression model, and GridSearchCV (cross-validation grid search) is used to perform automatic parameter optimization with the accuracy of negative samples as the target, and the optimal parameters are passed into the first supervised model for training until the accuracy of the negative samples reaches a preset first accuracy threshold, and the training is stopped to obtain the second supervised model.

[0080] In this embodiment, the grid search algorithm can be used to automatically search for the best parameters, reducing the time and manpower required for manual parameter adjustment, improving the model training efficiency, and through the grid search, it is possible to fully explore the combination of specified parameters, find the best performance parameter configuration, and improve the model performance. At the same time, optimizing with the accuracy of negative samples as the target can help the model better identify and distinguish invalid samples, thereby improving the overall recognition accuracy, helping to reduce the false alarm rate, and improving the actual application effect of the model in the production environment.

[0081] Step 103: training a preset first unsupervised model based on the feature data set to obtain a second unsupervised model;

[0082] In this embodiment, the preset first unsupervised model is trained based on the feature data set to obtain the second unsupervised model, specifically:

[0083] Determine a second optimization parameter of the first unsupervised model, and optimize the second optimization parameter according to a grid search algorithm and the accuracy of negative samples to obtain a second parameter;

[0084] The first unsupervised model is trained according to the second parameter and the feature data set until the accuracy of the negative sample reaches a preset second accuracy threshold, and the training is stopped to obtain a second unsupervised model.

[0085] In this embodiment, a first unsupervised model is constructed based on the isolation forest algorithm, and a second optimization parameter of the first unsupervised model is determined, wherein the second optimization parameter includes n_estimatiors, max_samples, max_features, random_state, and contamination, wherein the n_estimators parameter indicates the number of decision trees to be constructed. This parameter controls the overall complexity of the isolation forest and its ability to detect anomalies. The max_samples parameter defines the number of samples used when constructing each tree. It can be an integer or a floating point number less than 1. If it is an integer, it indicates the specific number of samples; if it is a floating point number, it indicates the proportion of the number of samples to the total number of samples. The max_features parameter controls the maximum number of features considered when constructing each tree. It can be an integer or a floating point number less than 1. If it is an integer, it indicates the specific number of features; if it is a floating point number, it indicates the proportion of the number of features to the total number of features. The random_state parameter is used to control the seed of the random number generator. In the isolation forest, since random sampling is required when constructing each tree, setting random_state can make the experimental results repeatable. The contamination parameter is used to estimate the proportion of anomalies in the dataset. Isolation Forest uses this proportion to determine the threshold for the anomaly score. It can be a floating point number between 0 and 0.5, or the string "auto", which means that the model will automatically adjust the threshold based on the data.

[0086] In this embodiment, the isolation forest is an outlier detection algorithm, and the prediction result outputs -1 and 1, where -1 indicates that the outlier corresponds to a negative sample.

[0087] In this embodiment, a first unsupervised model is constructed based on the isolation forest algorithm, and GridSearchCV (cross-validation grid search) is used to perform automatic parameter optimization with the accuracy of negative samples as the target, and the optimal parameters are passed into the first unsupervised model for training until the accuracy of the negative samples reaches a preset first accuracy threshold, and the training is stopped to obtain a second unsupervised model.

[0088] In this embodiment, the grid search algorithm can be used to automatically search for the best parameters, reducing the time and manpower required for manual parameter adjustment, improving the model training efficiency, and through the grid search, it is possible to fully explore the combination of specified parameters, find the best performance parameter configuration, and improve the model performance. At the same time, optimizing with the accuracy of negative samples as the target can help the model better identify and distinguish invalid samples, thereby improving the overall recognition accuracy, helping to reduce the false alarm rate, and improving the actual application effect of the model in the production environment.

[0089] Step 104: Obtain data to be predicted, and input the data to be predicted into the second supervised model and the second unsupervised model respectively to obtain a first prediction result and a second prediction result;

[0090] Please refer to Figure 3 , Figure 3 A schematic diagram of a prediction process for data to be predicted provided by an embodiment of the present invention.

[0091] In this embodiment, data to be predicted, i.e., unlabeled data, is obtained, and preprocessed. After the data to be predicted is subjected to the same outlier processing and feature screening, the data to be predicted is input into a second supervised model and a second unsupervised model, and a first prediction result and a second prediction result are obtained, wherein the first prediction result includes positive samples and negative samples, and the second prediction result includes positive samples and negative samples.

[0092] Step 105: Based on the first prediction result and the second prediction result, a final prediction result is obtained to predict the tab tear defect.

[0093] In this embodiment, the obtaining of the final prediction result based on the first prediction result and the second prediction result is specifically as follows:

[0094] The first prediction result and the second prediction result are integrated, and when the first prediction result and the second prediction result are both negative samples, the final prediction result is a negative sample.

[0095] In this embodiment, the first prediction result and the second prediction result are integrated, and the second supervised model and the second unsupervised model simultaneously predict the battery cell as a negative sample, then the final prediction result is a negative sample, and the others are positive samples.

[0096] In this embodiment, a comprehensive analysis of the first prediction result and the second prediction result can obtain a more stable and accurate final prediction result, reducing the risk of misjudgment in the verification process; achieving accurate prediction of the tab tear defect is of great significance for monitoring and prevention in the actual production process, and can identify potential problems at an earlier stage, reducing resource waste and safety hazards.

[0097] In this embodiment, the Django framework is used to encapsulate the API service. The interface program obtains the predicted data by sending a POST request when calling the interface, and predicts the tab tearing defect based on the encapsulated second supervised model, the second unsupervised model and the integrated prediction results.

[0098] In this embodiment, the battery cell production data is predicted in real time through a POST request, and the prediction results of the tab tearing are obtained online, overcoming the limitation of traditional detection methods that rely on a single voltage signal as a basis for judgment, and effectively improving the accuracy of the prediction.

[0099] In this embodiment, the present invention extracts features from multidimensional data in the battery cell production process to avoid the prediction of the tab tear defect from relying on a single voltage signal, thereby improving the prediction accuracy; at the same time, since negative sample data in the battery cell generation process is extremely scarce and serious misjudgment of negative samples may cause safety accidents, the first supervised model and the first unsupervised model are trained according to the accuracy of negative samples, so that the multidimensional data is predicted according to the second supervised model and the second unsupervised model, thereby improving the recognition accuracy of the second supervised model and the second unsupervised model for negative samples, and finally, the prediction reliability is improved by integrating the prediction results of the two models.

[0100] Please refer to Figure 4 , Figure 4 A schematic diagram of a structure of a tab tear defect prediction device provided by an embodiment of the present invention, comprising: a data preprocessing module 401, a supervised learning module 402, an unsupervised learning module 403 and a prediction module 404;

[0101] The data preprocessing module 401 is used to obtain multidimensional data in the battery cell production process and generate a feature data set based on the multidimensional data;

[0102] The supervised learning module 402 is used to train the preset first supervised model based on the feature data set to obtain a second supervised model;

[0103] The unsupervised learning module 403 is used to train the preset first unsupervised model based on the feature data set to obtain a second unsupervised model;

[0104] The prediction module 404 is used to predict the data to be predicted according to the trained linear regression model and the isolation forest model respectively, obtain a first prediction result and a second prediction result, and integrate the first prediction result and the second prediction result to obtain a final prediction result.

[0105] In this embodiment, the data preprocessing module includes:

[0106] Perform feature extraction based on the multidimensional data to obtain feature data; the feature data includes labeled data and unlabeled data, and the labeled data includes positive samples and negative samples;

[0107] Draw a first kernel density estimation graph based on the labeled data, and obtain unevenly distributed first feature data based on the first kernel density estimation graph;

[0108] Draw a second kernel density estimation graph based on the labeled data and the unlabeled data, and obtain evenly distributed second feature data based on the second kernel density estimation graph;

[0109] A feature data set is generated based on the first feature data and the second feature data.

[0110] In this embodiment, there is a supervised learning module, including:

[0111] Determine a first optimization parameter of the first supervised model, and optimize the first optimization parameter according to a grid search algorithm and the accuracy of negative samples to obtain a first parameter;

[0112] The first supervised model is trained according to the first parameter and the feature data set until the accuracy of the negative sample reaches a preset first accuracy threshold, and the training is stopped to obtain a second supervised model.

[0113] In this embodiment, the unsupervised learning module includes:

[0114] Determine a second optimization parameter of the first unsupervised model, and optimize the second optimization parameter according to a grid search algorithm and the accuracy of negative samples to obtain a second parameter;

[0115] The first unsupervised model is trained according to the second parameter and the feature data set until the accuracy of the negative sample reaches a preset second accuracy threshold, and the training is stopped to obtain a second unsupervised model.

[0116] In this embodiment, the prediction module includes:

[0117] The first prediction result and the second prediction result are integrated, and when the first prediction result and the second prediction result are both negative samples, the final prediction result is a negative sample.

[0118] In this embodiment, the data preprocessing module includes:

[0119] Dividing the positive samples and negative samples in the training data according to a preset ratio to obtain a training set and a test set for the positive samples and a training set and a test set for the negative samples;

[0120] Based on the training set and test set of the positive samples and the training set and test set of the negative samples, a final training set and test set are obtained.

[0121] In this embodiment, the data preprocessing module includes:

[0122] Acquire multidimensional data in the core production process, extract key features from the multidimensional data, and obtain production process parameters;

[0123] Null value data and zero value data in the production process parameters are eliminated, and characteristic data are obtained based on the normal distribution of the production process parameters.

[0124] The present invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the method for predicting a tab tear defect when executing the computer program.

[0125] The present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method for predicting a tab tear defect.

[0126] Exemplarily, the computer program may be divided into one or more modules, one or more modules are stored in a memory and executed by a processor to implement the present invention. One or more modules may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in a terminal device.

[0127] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The terminal device may include, but is not limited to, a processor, a memory, and a display. Those skilled in the art will appreciate that the above components are merely examples of terminal devices and do not constitute a limitation on the terminal device. The terminal device may include more or fewer components than the components, or may combine certain components, or different components. For example, a multi-device access platform processing device may also include input and output devices, network access devices, buses, and the like.

[0128] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0129] The memory can be used to store computer programs and / or modules. The processor realizes various functions of the multi-device access platform processing device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, a text conversion function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, text message data, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0130] Wherein, if the module based on the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0131] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting tab tearing defects, characterized in that: include: Acquire multidimensional data in a battery cell production process, and generate a feature data set based on the multidimensional data; Based on the feature data set, a preset first supervised model is trained to obtain a second supervised model; the first supervised model is constructed based on a supervised algorithm; Based on the feature data set, a preset first unsupervised model is trained to obtain a second unsupervised model; the first unsupervised model is constructed based on an unsupervised algorithm; Acquire data to be predicted, and input the data to be predicted into the second supervised model and the second unsupervised model respectively to obtain a first prediction result and a second prediction result; Based on the first prediction result and the second prediction result, a final prediction result is obtained to achieve the prediction of the tab tear defect.

2. A method for predicting a tab tear defect according to claim 1, characterized in that: The generating a feature data set based on the multidimensional data includes: Perform feature extraction based on the multidimensional data to obtain feature data; the feature data includes labeled data and unlabeled data, and the labeled data includes positive samples and negative samples; Draw a first kernel density estimation graph based on the labeled data, and obtain unevenly distributed first feature data based on the first kernel density estimation graph; Draw a second kernel density estimation graph based on the labeled data and the unlabeled data, and obtain evenly distributed second feature data based on the second kernel density estimation graph; A feature data set is generated based on the first feature data and the second feature data.

3. A method for predicting a tab tear defect according to claim 2, characterized in that: The step of training a preset first supervision model based on the feature data set to obtain a second supervision model includes: Determine a first optimization parameter of the first supervised model, and optimize the first optimization parameter according to a grid search algorithm and the accuracy of the negative sample to obtain a first parameter; The first supervised model is trained according to the first parameter and the feature data set until the accuracy of the negative sample reaches a preset first accuracy threshold, and the training is stopped to obtain a second supervised model.

4. A method for predicting a tab tear defect according to claim 2, characterized in that: The step of training a preset first unsupervised model based on the feature data set to obtain a second unsupervised model includes: Determine a second optimization parameter of the first unsupervised model, and optimize the second optimization parameter according to a grid search algorithm and the accuracy of the negative sample to obtain a second parameter; The first unsupervised model is trained according to the second parameter and the feature data set until the accuracy of the negative sample reaches a preset second accuracy threshold, and the training is stopped to obtain a second unsupervised model.

5. The method for predicting a tab tear defect according to claim 1, characterized in that: The obtaining a final prediction result based on the first prediction result and the second prediction result includes: The first prediction result and the second prediction result are integrated, and when the first prediction result and the second prediction result are both negative samples, the final prediction result is a negative sample.

6. A method for predicting a tab tear defect according to claim 2, characterized in that: After generating a feature data set based on the multidimensional data, the method further includes: Dividing the positive samples and negative samples in the training data according to a preset ratio to obtain a training set and a test set of the positive samples and a training set and a test set of the negative samples; Based on the training set and test set of the positive samples and the training set and test set of the negative samples, a final training set and test set are obtained.

7. A method for predicting a tab tear defect according to any one of claims 1 to 6, characterized in that: The step of extracting features based on the multidimensional data to obtain feature data includes: Acquire multidimensional data in the core production process, extract key features from the multidimensional data, and obtain production process parameters; Null value data and zero value data in the production process parameters are eliminated, and characteristic data are obtained based on the normal distribution of the production process parameters.

8. A device for predicting tab tearing defects, characterized in that: include: Data preprocessing module, supervised learning module, unsupervised learning module and prediction module; The data preprocessing module is used to obtain multidimensional data in the battery cell production process and generate a feature data set based on the multidimensional data; The supervised learning module is used to train a preset first supervised model based on the feature data set to obtain a second supervised model; the first supervised model is constructed based on machine learning; The unsupervised learning module is used to train a preset first unsupervised model based on the feature data set to obtain a second unsupervised model; the first unsupervised model is constructed based on machine learning; The prediction module is used to predict the data to be predicted according to the trained linear regression model and the isolation forest model respectively, obtain a first prediction result and a second prediction result, and integrate the first prediction result and the second prediction result to obtain a final prediction result.

9. A terminal device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, a method for predicting a tab tear defect as described in claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a tab tear defect prediction method as described in claims 1 to 7.

Images