Lithium ion battery thermal runaway inducement tracing method and device based on particulate matter morphology

Abstract

The invention discloses a lithium ion battery thermal runaway inducement tracing method and device based on particle morphology, and relates to the technical field of batteries, and the method comprises the steps: receiving an electron microscopic image of micron-scale morphology of thermal runaway eruption particles; performing preprocessing and data enhancement on the electron microscopic image, and processing the standardized image data by using the thermal runaway inducement classification model to obtain an inducement classification result; processing the standardized image data by the thermal runaway cause classification model: extracting multi-scale micron-scale morphology features from the standardized image data based on a preset network; and applying attention mechanism weighting, and carrying out progressive dimension reduction and classification on weighted features through a multi-layer classifier to obtain an inducement classification result. According to the method, scanning electron microscope imaging and deep learning are combined, the problems of low precision and high subjectivity caused by the fact that a traditional method depends on macroscopic features and expert experience are solved, and automatic and high-precision recognition of inducements such as overcharging, heating and needling is achieved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of batteries, in particular to a lithium ion battery thermal runaway cause tracing method and device based on particle morphology. BACKGROUND

[0002] As the core power source of new energy vehicles and large-scale energy storage systems, lithium ion batteries have a continuously expanding application scale and a continuously increasing energy density. However, the batteries are prone to thermal runaway under extreme abuse conditions such as overcharging, overheating and needle piercing, which can lead to serious safety accidents such as smoking, fire and explosion. Therefore, accurate cause tracing of battery thermal runaway accidents is of great significance to accident investigation, responsibility identification, product safety improvement and industry standard improvement.

[0003] At present, the identification of the cause of lithium ion battery thermal runaway mainly relies on macroscopic examination and electrochemical analysis of the battery debris after the accident. This kind of method usually makes inferences based on macroscopic or indirect characteristics such as burn marks, voltage curves and internal resistance changes, and to a great extent depends on the experience of experts. However, the thermal runaway process is violent and complex, and the battery body is often severely damaged, resulting in the absence or distortion of key macroscopic evidence, which makes the analysis results have strong subjectivity and uncertainty, and it is difficult to form an objective and unified identification conclusion.

[0004] The internal materials of the battery during thermal runaway will be violently ejected, forming a large number of micron-sized particles, and the micro-morphology characteristics of particles produced by different causes are significantly different. In order to make up for the defects of macroscopic analysis, the existing technical solutions use a scanning electron microscope to obtain high-resolution morphology images of these micron-sized particles, but the traditional method mainly relies on manual observation of the morphology images for identification, which has inherent defects such as low efficiency, being easily affected by subjective factors and lacking quantitative standards. SUMMARY

[0005] The purpose of the present application is to provide a lithium ion battery thermal runaway cause tracing method and device based on particle morphology, which solves the technical problems of traditional thermal runaway cause identification methods that rely on macroscopic characteristics and expert experience, have strong subjectivity, insufficient precision and low efficiency, by constructing a thermal runaway cause classification model containing EfficientNet feature extraction, channel and spatial attention mechanism and multi-layer classifier, and adopting a data enhancement strategy simulating SEM physical noise and particle deformation, and realizes high-precision identification and tracing of different thermal runaway causes such as overcharging, heating and needle piercing.

[0006] To achieve the above purpose, the present application provides the following solutions: In a first aspect, the present application provides a lithium-ion battery thermal runaway cause tracing method based on particle morphology, comprising: receiving an electron micrograph image of the micron-scale morphology of lithium-ion battery thermal runaway eruption particles; preprocessing the electron micrograph image to obtain preprocessed image data; data augmentation is performed on the preprocessed image data to obtain standardized image data; the standardized image data is processed using a pre-trained thermal runaway cause classification model, and the thermal runaway cause classification model outputs a cause classification result of the lithium-ion battery thermal runaway; wherein the processing of the standardized image data using the pre-trained thermal runaway cause classification model comprises: extracting multi-scale micron-scale morphology features from the standardized image data based on a preset network; applying an attention mechanism weighting to the extracted multi-scale micron-scale morphology features to obtain weighted features; and performing progressive dimension reduction and classification on the weighted features through a multi-layer classifier containing at least three fully connected layers to obtain the cause classification result containing the cause category and the corresponding confidence score.

[0007] Optionally, the lithium-ion battery thermal runaway cause tracing method based on particle morphology further comprises: determining the cause category of the cause classification result to obtain a determination result; the determination of the cause category of the cause classification result to obtain a determination result comprises: if the confidence score of a single cause category exceeds a preset score threshold, determining that the cause category is a thermal runaway cause; if the confidence scores of multiple cause categories are all higher than the preset score threshold, and the difference between the highest score and the second highest score is less than a preset score difference, then determining that it is a mixed cause.

[0008] Optionally, the lithium-ion battery thermal runaway cause tracing method based on particle morphology further comprises: generating a cause classification report of the cause classification result, wherein the cause classification report comprises the determination result, the confidence score, and the historical performance evaluation index of the thermal runaway cause classification model.

[0009] Optionally, the data augmentation of the preprocessed image data comprises at least one of the following ways: simulating the physical noise in the scanning electron microscope imaging process of the preprocessed image data; simulating the elastic deformation characteristics of the particles of the preprocessed image data.

[0010] Optionally, the preset network is an EfficientNet network.

[0011] Optionally, the preprocessing of the electron micrograph image to obtain preprocessed image data comprises: denoising the electron micrograph image; performing size normalization on the denoised image; and performing gray scale standardization on the size-normalized image to obtain the preprocessed image data.

[0012] Optionally, the attention mechanism is a mechanism in series of channel attention and spatial attention.

[0013] In a second aspect, the present application provides a particle morphology-based lithium-ion battery thermal runaway trigger tracing device, comprising: an image receiving module configured to receive an electron micrograph of the micron-scale morphology of lithium-ion battery thermal runaway eruption particles; a preprocessing module configured to preprocess the electron micrograph to obtain preprocessed image data; a data enhancement module configured to perform data enhancement on the preprocessed image data to obtain standardized image data; and a result output module configured to process the standardized image data using a pre-trained thermal runaway trigger classification model, wherein the thermal runaway trigger classification model outputs a trigger classification result of the lithium-ion battery thermal runaway; wherein the result output module comprises: a feature extraction submodule configured to extract multi-scale micron-scale morphology features from the standardized image data based on a preset network; a weighting submodule configured to apply an attention mechanism weighting to the extracted multi-scale micron-scale morphology features to obtain weighted features; and a dimension reduction and classification submodule configured to perform progressive dimension reduction and classification of the weighted features through a multi-layer classifier comprising at least three fully connected layers to obtain the trigger classification result comprising a trigger category and a corresponding confidence score.

[0014] Optionally, the particle morphology-based lithium-ion battery thermal runaway trigger tracing device further comprises a determination module configured to determine the trigger category of the trigger classification result to obtain a determination result.

[0015] Optionally, the particle morphology-based lithium-ion battery thermal runaway trigger tracing device further comprises a report generation module configured to generate a trigger classification report of the trigger classification result, wherein the trigger classification report comprises the determination result, the confidence score, and a historical performance evaluation indicator of the thermal runaway trigger classification model.

[0016] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the steps of the particle morphology-based lithium-ion battery thermal runaway trigger tracing method according to any one of the above.

[0017] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the particle morphology-based lithium-ion battery thermal runaway trigger tracing method according to any one of the above.

[0018] In a fifth aspect, the present application provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of the particle morphology-based lithium-ion battery thermal runaway cause tracing method of any one of the above.

[0019] According to the specific embodiments provided in the present application, the present application discloses the following technical effects: The present application provides a particle morphology-based lithium-ion battery thermal runaway cause tracing method and device. The micron-level morphology of the particle matter ejected by the lithium-ion battery thermal runaway is preprocessed, data is enhanced, and features are extracted. During the battery thermal runaway cause analysis process, the technical problem of high-precision and automated mapping between complex micro-morphology and thermal runaway cause categories is solved through the synergistic cooperation of feature extraction, attention mechanism weighting, and multi-layer classifier classification. The precise and automatic identification from particle images to cause categories is realized. The tracing accuracy is high and far exceeds the level of traditional manual identification methods. After the thermal runaway cause classification model identifies the cause categories, the determination logic based on the confidence score is introduced and a classification report containing quantitative data is generated, which realizes the explicit determination of single causes and the effective prompt of mixed causes, and provides intuitive, reliable, and quantified scientific conclusions for accident investigation and responsibility identification. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0021] Figure 1 An application environment diagram of the particle morphology-based lithium-ion battery thermal runaway cause tracing method in an embodiment of the present application; Figure 2 A flowchart of the particle morphology-based lithium-ion battery thermal runaway cause tracing method provided in an embodiment of the present application; Figure 3 A detailed flowchart of the S14 step in the method; Figure 1 Figure 4 A detailed flowchart of the S12 step in the method Figure 1 Figure 5 A schematic diagram of data enhancement of preprocessed image data in an embodiment of the present application; Figure 6 A schematic diagram of the channel attention mechanism structure in an embodiment of the present application; ​​Figure 7 This is a schematic diagram of the structure of a multi-layer classifier in one embodiment of this application; Figure 8 This is a confusion matrix heatmap for evaluating the performance of a thermal runaway cause classification model in one embodiment of this application; Figure 9 A schematic diagram of the functional modules of a lithium-ion battery thermal runaway cause tracing device based on particulate morphology provided in an embodiment of this application; Figure 10 for Figure 9 A detailed functional module diagram of the result output module; Figure 11 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] The lithium-ion battery thermal runaway cause tracing method based on particulate morphology provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 101 communicates with server 102 via a network. A data storage system can store the data that server 102 needs to process. The data storage system can be set up independently, integrated into server 102, or placed in the cloud or on another server. Terminal 101 can send electron micrographs to server 102. Server 102 receives the electron micrographs, preprocesses them to obtain preprocessed image data, performs data augmentation on the preprocessed image data to obtain standardized image data, and processes the standardized image data using a pre-trained thermal runaway cause classification model. The thermal runaway cause classification model outputs the cause classification result of the lithium-ion battery thermal runaway. Server 102 can feed back the obtained cause classification result to terminal 101. Furthermore, in some embodiments, the cause classification result can also be implemented by server 102 or terminal 101 independently. For example, terminal 101 can directly process the electron micrograph image to be processed, or server 102 can obtain the electron micrograph image from the data storage system.

[0025] The terminal 101 can be, but is not limited to, various desktop computers, notebook computers, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The Internet of Things device can be a smart speaker, a smart television, a smart air conditioner, a smart vehicle device, etc. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The server 102 can be implemented by a single server or a server cluster composed of multiple servers, and can also be a cloud server.

[0026] In an exemplary embodiment, as shown in Figure 2 , a particle morphology-based lithium-ion battery thermal runaway cause tracing method is provided, which is executed by a computer device, specifically, can be executed by a terminal or a server, or both. In the embodiments of the present application, the method is applied to the server 102 in Figure 1 . As shown in Figure 2 , the particle morphology-based lithium-ion battery thermal runaway cause tracing method includes the following steps: S11, receiving a scanning electron microscope image of the micron-scale morphology of the lithium-ion battery thermal runaway eruption particles; S12, preprocessing the scanning electron microscope image to obtain preprocessed image data; S13, performing data augmentation on the preprocessed image data to obtain standardized image data; S14, processing the standardized image data using a pre-trained thermal runaway cause classification model, and obtaining a cause classification result of the lithium-ion battery thermal runaway from the thermal runaway cause classification model; S15, determining the cause category of the cause classification result to obtain a determination result; S16, generating a cause classification report of the cause classification result; As shown in Figure 3 , the step S14 of processing the standardized image data using the pre-trained thermal runaway cause classification model includes the following steps: S141, extracting multi-scale micron-scale morphology features from the standardized image data based on a preset network; S142, applying an attention mechanism weighting to the extracted multi-scale micron-scale morphology features to obtain weighted features; S143, performing progressive dimension reduction and classification on the weighted features through a multi-layer classifier containing at least three fully connected layers to obtain a cause classification result containing a cause category and a corresponding confidence score.

[0027] The lithium ion battery thermal runaway cause tracing method based on particle morphology in the embodiments of the present application solves the technical problems of strong subjectivity, dependence on expert experience and high analysis cost of traditional macroscopic analysis methods and manual identification by performing a series of preprocessing, data enhancement and feature extraction on the electron micrograph of the micron-level morphology of the lithium ion battery thermal runaway eruption particles, and establishes a standard automated analysis process from image standardization to targeted data enhancement and deep feature extraction, which greatly reduces the analysis cost and dependence on manual experience. In the process of analyzing the battery thermal runaway cause by using the thermal runaway cause classification model, the technical problem of high-precision and automated mapping between complex microscopic morphology and thermal runaway cause categories is solved through the synergistic cooperation of feature extraction, attention mechanism weighting and multi-layer classifier classification, and the accurate and automatic identification from particle images to cause categories is realized, with high tracing accuracy, far exceeding the level of traditional manual identification methods. Among them, the attention mechanism weighting solves the technical problem of complex background of electron micrographs and difficulty in focusing on effective morphology features, and the use of attention mechanism enables the network to adaptively focus on the morphology features most discriminative for thermal runaway cause identification, effectively suppresses the interference of irrelevant background noise, and significantly improves the quality of feature expression and the discriminability of the model. Finally, by introducing the judgment logic based on the confidence score and generating a classification report containing quantitative data, the explicit judgment of a single cause and the effective prompt of mixed causes are realized, providing intuitive, reliable and quantitatively based scientific conclusions for accident investigation and responsibility identification.

[0028] In specific implementation, the collection of lithium ion battery thermal runaway eruption particles adopts a multi-stage filter grid system. Specifically, the first-stage coarse filter mesh aperture can be 100 microns, the second-stage fine filter mesh aperture can be 50 microns, and the bottom layer uses conductive carbon tape to fix the particles.

[0029] In specific implementation, the collected micron-level particles are imaged at high resolution by a scanning electron microscope (SEM), which is a powerful scientific instrument that uses a focused electron beam to scan point by point on the sample surface to obtain high-resolution, large-depth-of-field microscopic morphology information. In the specific imaging process, the particle sample is first subjected to conductive treatment using sputter gold plating technology, and the gold layer thickness can be controlled to be 5-10 nanometers. Then, the SEM working parameters can be set as follows: acceleration voltage 5-15 kilovolts, working distance 8-12 millimeters, magnification 500-5000 times. Finally, the particle images under different fields of view are collected, at least 50 images are obtained for each sample, and the resolution is not less than 2048x2048 pixels, ensuring the capture of the fine morphology features of the particles.

[0030] In specific implementation, as shown in Figure 4 The step of preprocessing the electron micrograph to obtain preprocessed image data includes: S121, denoising the electron micrograph image; S122, performing size normalization processing on the denoised image; S123, performing gray scale standardization processing on the size normalized image to obtain preprocessed image data.

[0031] The denoising processing adopts a non-local mean filtering algorithm to effectively remove random noise while maintaining edge features; the size normalization processing uniformly adjusts all images to a preset pixel, which can be specifically 224x224 pixels; the gray scale standardization processing can adopt a Z-score standardization method to make the image gray scale value distribution have zero mean and unit variance, and can eliminate the brightness difference under different imaging conditions.

[0032] In specific implementation, as shown in Figure 5 , the data enhancement manner for the preprocessed image data can include geometric transformation, simulation of physical noise, simulation of elastic deformation features, optical adjustment, and online adjustment.

[0033] The geometric transformation includes rotation, flipping, scaling, and shearing of the original electron micrograph image, the rotation can be a random angle of -15° to 15° and a 90° multiple speed angle, the flipping includes horizontal, vertical, and diagonal, and the scaling can be 0.8-1.2 times.

[0034] The simulation of physical noise includes shot noise, charging effect, and scanning distortion. The complete statistical model of the shot noise follows a Poisson distribution for the electron count, and the noise variance formula is , and the signal-to-noise ratio formula is , in which I is the signal current, t is the detection time, e is the elementary charge, is the detector efficiency, is the average number of detected electrons, is the average signal, is the standard deviation of the noise. The electric field distribution model for the charging effect has a potential distribution formula , which calculates the total electrostatic potential generated by the entire sample distribution of electric charges at the sample surface point , in which describes how much charge is accumulated at each point on the sample surface, σ s is for the area irradiated by the electron beam, is the coordinate of the sample charge source point, is the straight-line distance between the observation point and the source point, is the constant in Coulomb's law, is a double integral. The electron beam deflection angle formula is , where, is the electric field component perpendicular to the electron beam, e is the elementary charge, is the electron mass, v is the electron velocity, L is the effective path length of the electron flying in the electric field and being affected by it. For the parametric model of scan distortion, both linear distortion and nonlinear distortion (barrel / pincushion) are included; the matrix equation of linear distortion is , where is the actual pixel coordinate after distortion, is the original ideal pixel coordinate, the 6 parameters (m 11 , m 12 , m 13 , m 21 , m 22 , m 23 ) in the middle 3x3 matrix jointly determine the distortion type; the formula of nonlinear distortion (barrel / pincushion) is , where, is the radial distance of a certain point in the original image to the image center, is the new radial distance after the distortion, 、 、 is the distortion coefficient, is the leading term, describing the basic distortion form, and are high-order correction terms, used to more accurately fit the complex distortion curve.

[0035] By simulating the effects of shot noise, charging effect and scan distortion together, the inherent and multi-physical source imaging defects of SEM images are solved, the generalization ability and robustness of the thermal runaway trigger classification model on real and complex SEM image data are improved, thereby ensuring the high precision and high reliability of the thermal runaway trigger tracing.

[0036] The simulated elastic deformation feature is the image deformation using the moving least squares method, where the weight function is , where is the Euclidean distance of pixel x to , h refers to the smoothing parameter, and the shape function is , where, is a locally weighted coordinate system constructed at point x, is similar to , but only for a certain specific control point.

[0037] The elastic deformation is simulated using the moving least squares method, the generated deformation is continuous, smooth and physically reasonable, and the model is forced to focus on more essential microscopic topographic features that remain unchanged under elastic deformation, thereby improving the discriminant ability of the model.

[0038] Optical adjustment mainly refers to the transformation of image pixel values themselves, such as brightness, contrast, and saturation adjustment; online adjustment refers to real-time and random data augmentation transformation on input images in each training batch.

[0039] Specifically, advanced data augmentation techniques such as MixUp, CutMix, and Mosaic can be used to achieve data augmentation.

[0040] In specific implementation, after data augmentation on the preprocessed image data, 30 enhanced samples with different changes can be generated from the original image, prompting the generalization ability of the thermal runaway trigger classification model.

[0041] It can be understood that in step S14, the thermal runaway trigger classification model adopts an extensible framework design, which facilitates the addition of trigger categories. In the embodiments of the present application, the trigger categories can include four types, namely overcharge, heating, needle puncture, and normal.

[0042] It can be understood that in step S14, the trigger classification result output by the thermal runaway trigger classification model includes the trigger category and the corresponding confidence score.

[0043] It can be understood that in step S15, the trigger category of the trigger classification result is determined to obtain a determination result, wherein the determination result includes: If the confidence score of a single trigger category exceeds a preset score threshold, the trigger category is determined to be a thermal runaway trigger.

[0044] If the confidence scores of multiple trigger categories are all higher than the preset score threshold, and the difference between the highest score and the second highest score is less than a preset score difference, then it is determined to be a mixed trigger.

[0045] In specific implementation, the preset score threshold can be 0.8.

[0046] It can be understood that in step S16, the trigger classification report includes the determination result, the confidence score, and the historical performance evaluation index of the thermal runaway trigger classification model, wherein the historical performance evaluation index includes, for example, accuracy, recall rate, F1 score, etc., to prove the reliability of the thermal runaway trigger classification model itself from a macro perspective and provide further credibility support for accident analysis.

[0047] In specific implementation, in step S141, the preset network can be an EfficientNet network, specifically, an EfficientNet-B0 network. The depth, width and resolution of the network can be balanced through a compound scaling strategy. The EfficientNet-B0 network can include 7 groups of MBConv modules, each of which is composed of a 1x1 expansion convolution, a depth separable convolution, a Squeeze-and-Excitation module and a 1x1 projection convolution. In specific implementation, by using the network, through progressive down-sampling, an input 224x224x3 image can be gradually converted into a high-dimensional feature representation of 7x7x1280, realizing effective capture of multi-level features of particulate matter from local texture to overall shape.

[0048] In the above MBConv module, the formula of the expansion convolution is: , and the expansion rate calculation formula is , wherein, means a 1x1 convolution, BN means batch normalization, means an activation function, means the number of output channels of the 1x1 convolution , and the ratio of the number of input channels ; the formula of the depth separable convolution is , wherein, means using a kxk convolution kernel; the formula of the Squeeze-and-Excitation module includes , wherein, means global average pooling of the feature map , the spatial information of each channel is compressed into a numerical value, and the vector z captures the global distribution of each channel, is a fully connected layer for dimension reduction, means introducing nonlinearity, means multiplying the learned weight s with the original feature map channel by channel; the formula of the projection convolution is , wherein, means another 1x1 convolution.

[0049] Further, when the input and output dimensions match, residual connection is also applied in each MBConv module, and the formula of the residual connection is , wherein, is the final output of the MBConv module, means the feature map obtained after all operations (expansion convolution, depth separable convolution, SE module, projection convolution) in the module, xis referred to as the original feature map input into the MBConv module, is referred to as the input feature map x is referred to as the number of channels of the input feature map is referred to as the number of channels of the output feature map is referred to as the number of channels of the output feature map is referred to as the step of the MBConv module.

[0050] Therefore, by taking the EfficientNet-B0 network based on the composite scaling strategy as the backbone, and using the core MBConv module for multi-level feature extraction, the technical problem of efficiently and accurately capturing and fusing the multi-scale morphological features of particulate matter from complex SEM images is solved, providing the most critical feature basis for realizing high-precision thermal runaway cause tracing. Among them, by using 1x1 expansion convolution in the internal components of the MBConv module, the problem of improving feature richness at low computational cost is solved, by using depth separable convolution, the problem of large standard convolution calculation and parameter redundancy is solved, by using the Squeeze-and-Excitation module, the problem of equal importance between feature channels and the inability to adaptively focus on key information is solved, by using 1x1 projection convolution, the problem of connecting the high-dimensional features inside the module with the external network dimension is solved, when the dimensions match, by using the conditional residual connection, the problem of gradient disappearance and network degradation in network training is solved, which ensures the stability of network training and protects the micro-morphological information extracted from SEM images from being destroyed by the deep network.

[0051] In specific implementation, in step S142, the attention mechanism is a mechanism in which the channel attention and the spatial attention are connected in series. The channel attention obtains a channel descriptor through global average pooling, generates a channel weight through at least two fully connected networks, and the compression ratio can be set to 16. The spatial attention can generate a spatial weight map through 7x7 convolution by fusing features through maximum pooling and average pooling. The two attention mechanisms work together to make the network adaptively focus on the morphological features that are most discriminative for thermal runaway cause identification and suppress the interference of irrelevant background noise.

[0052] The above channel attention mechanism can be as shown in Figure 6 , which is a channel attention mechanism structure diagram. The channel attention mechanism inputs a feature F∈ℝ^(H×W×C), where F is referred to as the input feature map, ℝ is referred to as the real number field, the feature map height H=7, the feature map width W=7, and the feature map channel number C=1280. Then, global average pooling is used to compress global spatial information as the input of the subsequent fully connected layer. Next, the first FC layer is entered for dimension reduction, and the formula is , where is referred to as the weight matrix of the first fully connected layer, and z is referred to as a global descriptor. is a bias vector of the first fully connected layer, is an output vector of the first fully connected layer, and the weight initialization in the dimension reduction of the first FC layer adopts He initialization, and the formula is ; then, ReLU activation is adopted, and the activation function is , wherein is an output vector after the ReLU activation function, is a ReLU function, which sets all negative input values to 0 and keeps positive input values unchanged, thereby enhancing the expression ability of the network; then, the second FC layer is entered for dimension increase, and the formula is , wherein is a weight matrix of the second fully connected layer, is a bias vector of the second fully connected layer, is an original output vector of the second fully connected layer; then, Sigmoid function activation is adopted, and the activation function is , wherein is a channel attention weight vector; then, feature rescaling is performed, and the formula is , wherein is a weighted output feature map, and the dimension is the same as the original feature F , that is, an identity mapping, is respectively a height, a width and a channel index of the feature map; finally, the weighted feature is obtained. It can be understood that, after the input 1280-dimensional feature F is subjected to the channel attention mechanism in Figure 6 , a 1280-dimensional weighted feature is finally output.

[0053] In specific implementation, in step S143, a structure diagram of the multi-layer classifier can be as shown in Figure 7As shown, the multi-layer classifier can map the extracted high-dimensional features to the thermal runaway trigger categories, and the multi-layer classifier adopts a progressive dimension reduction structure, including four fully connected layers, and the calculation rule of the parameter quantity of each layer is parameter quantity=(input dimension x output dimension)+(output dimension). Specifically, 1280-dimensional features are taken as the input of the first fully connected layer (FC1), the output is 512-dimensional, the parameter quantity is 1280x512+512=655872, the activation function is ReLU, and the Dropout rate is 0.3, then the input of the second fully connected layer (FC2) is 512-dimensional, the output is 256-dimensional, the parameter quantity is 512x256+256=131328, the activation function is ReLU, and the Dropout rate is 0.3, then, the input of the third fully connected layer (FC3) is 256-dimensional, the output is 128-dimensional, the parameter quantity is 256x128+128=32896, the activation function is ReLU, and the Dropout rate is 0.2, then, the input of the fourth fully connected layer (FC4) is 128-dimensional, the output is 4-dimensional, the parameter quantity is 128x4+4=516, the fourth fully connected layer is taken as the output layer, and the output layer maps the 128-dimensional features to the final four categories, such as overcharge, heating, needle piercing, and normal, and the output layer activation function is Softmax, which is used to generate the probability distribution of each category, representing the confidence of the trigger category. In addition, each fully connected layer is normalized, which can speed up the training process.

[0054] Therefore, the multi-layer classifier solves the problem of accurately and robustly mapping high-dimensional abstract topographic features to thermal runaway trigger categories by adopting a progressive dimension reduction structure and integrating batch normalization, ReLU activation, and Dropout, provides an objective and quantitative decision basis for accident tracing, and ensures the excellent generalization ability of the model under small sample conditions.

[0055] In specific implementation, for the performance evaluation index of the thermal runaway trigger classification model, a 4-class confusion matrix can be used for comprehensive evaluation. Specifically, as shown in Figure 8 the confusion matrix heat map for performance evaluation of the thermal runaway trigger classification model, the total number of samples is 240, the trigger categories can include overcharge, heating, count missing, and normal, the accuracy rates of the overcharge class, the count missing class, and the normal class are all 100%, and the accuracy rate of the heating class is 90%, indicating that the thermal runaway trigger classification model performs excellently in identifying various categories. In addition, the overall accuracy rate of the thermal runaway trigger classification model reaches 92.9%, and the reliability and significance of the performance of the thermal runaway trigger classification model can be verified by using McNemar test and Wilson score interval, wherein the process of Wilson score interval verification is , 0.929 is the observed accuracy, 0.071 is the observed error rate, 240 is the total number of test samples, 1.96 is the standard normal distribution score corresponding to the 95% confidence interval, and the total accuracy of the thermal runaway cause classification model in the embodiment is 92.9% through this validation process. A reliable fluctuation range [89.7%, 96.1%] is provided, which is high in position and narrow in range, strongly proving the stability and reliability of the performance of the thermal runaway cause classification model, indicating that the thermal runaway cause classification model can continuously provide high-precision traceability results in actual deployment, rather than a random phenomenon.

[0056] Exemplarily, as shown in Figure 9 , a lithium ion battery thermal runaway cause traceability device based on particle morphology is provided, which includes an image receiving module 21, a preprocessing module 22, a data enhancement module 23, a result output module 24, a determination module 25, and a report generation module 26.

[0057] The image receiving module 21 is configured to receive an electron micrograph of the micron-scale morphology of the lithium ion battery thermal runaway eruption particles. The preprocessing module 22 is configured to preprocess the electron micrograph to obtain preprocessed image data. The data enhancement module 23 is configured to perform data enhancement on the preprocessed image data to obtain standardized image data. The result output module 24 is configured to process the standardized image data using a pre-trained thermal runaway cause classification model, and the thermal runaway cause classification model outputs the cause classification result of the lithium ion battery thermal runaway.

[0058] The determination module 25 is configured to determine the cause category of the cause classification result to obtain a determination result. The report generation module 26 is configured to generate a cause classification report of the cause classification result, which includes the determination result, the confidence score, and the historical performance evaluation index of the thermal runaway cause classification model.

[0059] In specific implementation, as shown in Figure 10 , the result output module 24 includes a feature extraction submodule 241, a weighting submodule 242, and a dimension reduction and classification submodule 243.

[0060] The feature extraction submodule 241 is configured to extract multi-scale micron-scale morphology features from the standardized image data based on a pre-set network. The weighting submodule 242 is configured to apply an attention mechanism weighting to the extracted multi-scale micron-scale morphology features to obtain weighted features. The dimension reduction and classification submodule 243 is configured to perform progressive dimension reduction and classification of the weighted features through a multi-layer classifier comprising at least three fully connected layers to obtain the cause classification result comprising the cause category and the corresponding confidence score.

[0061] In an exemplary embodiment, a computer device is provided, which can be a server or a terminal. An internal structure diagram of the computer device can be as shown in Figure 11 The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store electron micrographs. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through a network connection. The computer program is executed by the processor to implement a lithium-ion battery thermal runaway cause tracing method based on particle morphology.

[0062] Those skilled in the art can understand that Figure 11 The structure shown in the above

[0063] In an exemplary embodiment, a computer device is provided, which includes a memory and a processor. The memory stores a computer program. The processor executes the computer program to implement the steps in the above method embodiments.

[0064] In an exemplary embodiment, a computer readable storage medium is provided, which stores a computer program. The computer program is executed by a processor to implement the steps in the above method embodiments.

[0065] In an exemplary embodiment, a computer program product is provided, which includes a computer program. The computer program is executed by a processor to implement the steps in the above method embodiments.

[0066] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0067] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related hardware through a computer program, and the computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments of each method. Among them, any reference to memory, database or other medium used in each embodiment provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc.

[0068] The database involved in each embodiment provided by the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in each embodiment provided by the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0069] Each technical feature of the above embodiments can be combined arbitrarily. In order to make the description simple, not all possible combinations of each technical feature in the above embodiments are described, but as long as the combination of these technical features does not exist contradictory, it should be considered as the scope of the present application.

[0070] The principles and implementations of the present application are described in detail with specific examples in this paper, and the above examples are only used to help understand the method of the present application and its core idea; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation and application range will be changed. Therefore, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method for tracing the causes of thermal runaway in lithium-ion batteries based on particulate morphology, characterized in that, include: Electron micrographs of the micron-scale morphology of particulate matter ejected during thermal runaway from lithium-ion batteries were obtained. The electron microscope image is preprocessed to obtain preprocessed image data; The preprocessed image data is augmented to obtain standardized image data; The standardized image data is processed using a pre-trained thermal runaway cause classification model, and the thermal runaway cause classification model outputs the cause classification result of the thermal runaway of the lithium-ion battery. The step of processing the standardized image data using a pre-trained thermal runaway cause classification model includes: Multi-scale micron-level morphological features are extracted from the standardized image data based on a preset network. The extracted multi-scale micron-level morphology features are weighted by an attention mechanism to obtain weighted features; The weighted features are progressively reduced and classified using a multi-layer classifier containing at least three fully connected layers to obtain the cause classification result, which includes the cause category and the corresponding confidence score.

2. The method for tracing the causes of thermal runaway in lithium-ion batteries based on particulate morphology according to claim 1, characterized in that, The method for tracing the causes of thermal runaway in lithium-ion batteries based on particulate morphology further includes: determining the cause category of the cause classification result to obtain the determination result; The determination of the cause category based on the cause classification results, to obtain the determination result, includes: If the confidence score of a single cause category exceeds the preset score threshold, then that cause category is determined to be a cause of thermal runaway. If multiple cause categories have confidence scores higher than the preset score threshold, and the difference between the highest and second-highest scores is less than a preset score difference, then it is determined to be a mixed cause.

3. The method for tracing the causes of thermal runaway in lithium-ion batteries based on particulate morphology according to claim 2, characterized in that, The lithium-ion battery thermal runaway cause tracing method based on particulate morphology further includes: generating a cause classification report of the cause classification results, wherein the cause classification report includes the judgment result, the confidence score, and the historical performance evaluation index of the thermal runaway cause classification model.

4. The method for tracing the causes of thermal runaway in lithium-ion batteries based on particulate morphology according to claim 1, characterized in that, The data augmentation of the preprocessed image data includes at least one of the following methods: The preprocessed image data is used to simulate physical noise during the scanning electron microscope imaging process; The preprocessed image data is used to simulate the elastic deformation characteristics of particulate matter.

5. The method for tracing the causes of thermal runaway in lithium-ion batteries based on particulate morphology according to claim 1, characterized in that, The preset network is the EfficientNet network.

6. The method for tracing the causes of thermal runaway in lithium-ion batteries based on particulate morphology according to claim 1, characterized in that, The preprocessing of the electron microscope image to obtain preprocessed image data includes: The electron micrograph is then denoised. The size of the denoised image is normalized. The image after size normalization is subjected to grayscale normalization to obtain the preprocessed image data.

7. The method for tracing the causes of thermal runaway in lithium-ion batteries based on particulate morphology according to claim 1, characterized in that, The attention mechanism is a mechanism that combines channel attention and spatial attention in a series.

8. A device for tracing the causes of thermal runaway in lithium-ion batteries based on particulate morphology, characterized in that, include: The image receiving module is configured to receive electron micrographs of the micron-scale morphology of particulate matter ejected from a lithium-ion battery during thermal runaway. The preprocessing module is configured to perform preprocessing on the electron microscope image to obtain preprocessed image data; The data augmentation module is configured to perform data augmentation on the preprocessed image data to obtain standardized image data; The result output module is configured to process the standardized image data using a pre-trained thermal runaway cause classification model, and the thermal runaway cause classification model outputs the cause classification result of the lithium-ion battery thermal runaway. The result output module includes: The feature extraction submodule is configured to extract multi-scale micron-level morphological features from the standardized image data based on a preset network; The weighting submodule is configured to perform an attention mechanism weighting on the extracted multi-scale micron-level topographic features to obtain weighted features; The dimensionality reduction and classification submodule is configured to perform progressive dimensionality reduction and classification of the weighted features through a multi-layer classifier containing at least three fully connected layers, to obtain the cause classification result containing the cause category and corresponding confidence score.

9. The lithium-ion battery thermal runaway cause tracing device based on particulate morphology according to claim 8, characterized in that, The lithium-ion battery thermal runaway cause tracing device based on particulate morphology further includes: a determination module configured to determine the cause category of the cause classification result and obtain a determination result.

10. The lithium-ion battery thermal runaway cause tracing device based on particulate morphology according to claim 9, characterized in that, The lithium-ion battery thermal runaway cause tracing device based on particulate morphology further includes: a report generation module configured to generate a cause classification report of the cause classification result, wherein the cause classification report includes the judgment result, the confidence score, and the historical performance evaluation index of the thermal runaway cause classification model.

Images