A system and method for cell ultrasonic weld state root cause analysis

Abstract

The application discloses an ultrasonic welding state root cause analysis system and method for an electric core, and belongs to the technical field of battery detection. The method comprises the following steps: collecting welding equipment parameters, environmental parameters and implicit parameters in real time, synchronously obtaining a CCD detection result, and forming an original data set; preprocessing the original data set and extracting a feature vector for model training; training multiple basic models respectively by using the feature vector, optimizing model parameters through cross-validation, and establishing an associated model of parameters and NG results; calculating parameter importance scores in each model, determining root cause priorities through weighted voting, and generating a root cause analysis report and a visual chart. The disclosure takes the CCD detection result (OK / NG) as a target label, locates the main root cause and the secondary root cause of the NG electric core through a multi-model voting mechanism, and meets the high consistency requirement of the power battery.

Description

Technical Field

[0001] This application belongs to the field of battery testing technology, and in particular relates to a system and method for root cause analysis of ultrasonic welding status of battery cells. Background Technology

[0002] In the lithium-ion battery cell manufacturing process, ultrasonic welding is the core process for connecting the tabs, and its welding quality directly determines the cell's electrical performance, safety performance, and cycle life. Currently, in the lithium battery production field, the control of welding quality mainly relies on CCD visual inspection technology to screen OK / NG cells. However, when NG cells are produced in batches, process experts need to manually check the root cause parameters, which is inefficient and inaccurate. Existing related technologies mostly focus on welding parameter optimization or correlation analysis between a single parameter and welding quality, and have not yet formed a multi-parameter collaborative analysis root cause localization system.

[0003] Specifically, existing technologies have the following limitations: 1. Root cause analysis is too narrow: Most solutions only focus on core welding parameters such as energy and pressure, ignoring hidden influencing factors such as flow time, ambient humidity, and cell moisture content, resulting in incomplete root cause identification. For example, when the humidity in the workshop rises sharply, an oxide film is easily formed on the surface of the electrode tab, and adjusting the energy parameters alone cannot solve the problem of poor welding. 2. The analysis method relies on human experience: Currently, the root cause investigation of NG cells requires experts to compare the parameter differences between OK and NG cells. Subjective factors have a great influence. For example, under the same solder area, a novice engineer may mistakenly determine "amplitude fluctuation" as the root cause, when it is actually caused by "soldering time deviation". 3. Lack of quantitative analysis model: Existing technology has not established a quantitative correlation model between parameters and NG results, and cannot distinguish between "primary root causes" and "secondary root causes". For example, when there is both insufficient energy and abnormal temperature, it is difficult to determine the contribution of each to the NG results, resulting in insufficient targeted parameter adjustment.

[0004] With the increasing demands for cell consistency in the lithium battery industry (such as controlling the welding defect rate of power battery cells to below 0.1%), traditional manual root cause analysis methods can no longer meet the needs of large-scale, high-speed production.

[0005] Therefore, it is necessary to provide a new system and method for root cause analysis of ultrasonic welding conditions of battery cells to solve the above-mentioned technical problems. Summary of the Invention

[0006] The purpose of this disclosure is to provide a system and method for root cause analysis of ultrasonic welding conditions of battery cells in order to solve the above-mentioned problems.

[0007] This disclosure achieves the above objectives through the following technical solutions: A root cause analysis system for ultrasonic welding conditions of battery cells, comprising: The data acquisition module is used to collect welding equipment parameters, environmental parameters, and implicit parameters in real time, and simultaneously acquire CCD detection results to form a raw dataset; The feature engineering module is used to preprocess the original dataset and extract feature vectors for model training. An integrated model training module is used to train multiple basic models using the feature vectors, optimize model parameters through cross-validation, and establish a correlation model between parameters and NG results. The root cause voting determination module is used to calculate the importance score of parameters in each model and determine the root cause priority through weighted voting; The results output module is used to generate root cause analysis reports and visualization charts.

[0008] As a further optimization of this disclosure, the welding equipment parameters include welding pressure, welding energy, welding amplitude, welding height, welding power, welding frequency, weld area, and welding time; the environmental parameters include workshop humidity and workshop temperature; and the implicit parameters include turnover time and cell moisture content.

[0009] A method for root cause analysis of ultrasonic welding conditions of battery cells includes the following steps: Real-time acquisition of welding equipment parameters, environmental parameters, and implicit parameters, and simultaneous acquisition of CCD detection results to form a raw dataset; The original dataset is preprocessed, and feature vectors for model training are extracted. Multiple basic models are trained using the aforementioned feature vectors, and the model parameters are optimized through cross-validation to establish a correlation model between the parameters and the NG results; Calculate the importance scores of the parameters in each model, and determine the root cause priority through weighted voting; Generate root cause analysis reports and visualization charts.

[0010] As a further optimization of this disclosure, the original dataset is preprocessed, and feature vectors for model training are extracted, including: The original dataset is cleaned, including missing value handling and outlier filtering; The discrete parameters in the cleaned original dataset are encoded to obtain discrete features; the continuous parameters in the cleaned original dataset are standardized to obtain continuous features; and the time-series parameters in the cleaned original dataset are transformed to obtain time-series features.

[0011] As a further optimization of this disclosure, five basic models are adopted: decision tree model, random forest model, gradient boosting tree model, support vector machine and neural network model.

[0012] As a further optimization of this disclosure, multiple basic models are trained using the aforementioned feature vectors, and the model parameters are optimized through cross-validation to establish a correlation model between the parameters and the NG results, including: The feature vectors are divided into training and test sets according to a preset ratio, and stratified sampling is used to ensure that the proportion of NG cells in the training and test sets is consistent. For each base model, 5-fold cross-validation was used to adjust the hyperparameters, with the root cause localization accuracy on the test set as the optimization objective, to ensure that the accuracy of each model is greater than the preset threshold. The model performance is evaluated using three metrics: accuracy, precision, and recall.

[0013] As a further optimization of this disclosure, the importance score of parameters in each model is calculated, and the root cause priority is determined by weighted voting, including: For each completed base model, calculate the importance score of all input parameters to the NG result; The parameter importance scores for each model are normalized to obtain standardized scores, so that the sum of the scores of all parameters in a single model is 1. The total voting weight for each parameter is calculated based on the standardized score. Sort all parameters in descending order of total voting weight. The parameter with the highest weight is the "primary root cause", and the parameters within the preset weight ranking range are the "secondary root causes". Set a weight threshold, and only include a parameter in the root cause list when the weight exceeds the weight threshold.

[0014] As a further optimization of this disclosure, the root cause analysis report includes the following: Basic information: Analysis batch, cell model, quantity and proportion of NG cells; Root cause list: primary root cause, secondary root cause; Adjustment recommendations: Analyze the parameters for the primary root causes and implement supplementary measures for the secondary root causes.

[0015] An electronic device includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor is used to execute the program stored in the memory to implement the root cause analysis method for the ultrasonic welding status of the battery cell.

[0016] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned root cause analysis method for ultrasonic welding status of battery cells.

[0017] The beneficial effects of this disclosure are as follows: 1. Construction of a multi-dimensional parameter system: For the first time, 12 parameters in three categories, namely "equipment parameters, environmental parameters and implicit parameters", are integrated into the root cause analysis input parameters. Among them, implicit parameters (flow time, moisture content) are collected in real time by sensors, which solves the problem of incomplete parameter coverage in traditional solutions. 2. Ensemble Learning Voting Weight Mechanism: An ensemble architecture is constructed using five models: decision tree, random forest, gradient boosting tree (XGBoost), support vector machine (SVM), and neural network (MLP). The root cause priority is determined by a "weighted voting based on importance score". The vote weight = (parameter importance score in a single model / sum of parameter importance scores in all models) × 100%. The model with the highest weight is the primary root cause, and the models with 2-3 votes are secondary root causes. 3. Dynamic feature engineering method: Differentiated feature processing strategies are designed for different parameter types. For example, one-hot encoding is used for discrete parameters (welding frequency), standardization is used for continuous parameters (pressure, energy), and sliding window statistics are used for time-series parameters (trace moisture content) to improve the model's adaptability to multiple types of parameters. 4. Root Cause Results Visualization Output: Construct a root cause contribution bar chart to intuitively show the degree of influence of each parameter on the NG results. For example, when the main root cause is "insufficient energy", its weight in the bar chart is >50%, which makes it easier for engineers to quickly understand and implement adjustments. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a system structure block diagram of an embodiment of this disclosure; Figure 2 This is a flowchart of a method in an embodiment of this disclosure; Figure 3 This is a block diagram of the device structure in an embodiment of this disclosure. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, 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, 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.

[0021] like Figure 1 As shown, a root cause analysis system for ultrasonic welding status of battery cells includes: The data acquisition module is used to collect welding equipment parameters, environmental parameters, and implicit parameters in real time, and simultaneously acquire CCD detection results to form a raw dataset; The feature engineering module is used to preprocess the original dataset and extract feature vectors for model training. An integrated model training module is used to train multiple basic models using the feature vectors, optimize model parameters through cross-validation, and establish a correlation model between parameters and NG results. The root cause voting determination module is used to calculate the importance score of parameters in each model and determine the root cause priority through weighted voting; The results output module is used to generate root cause analysis reports and visualization charts.

[0022] like Figure 2 As shown, a method for analyzing the root cause of ultrasonic welding conditions of battery cells applied to the aforementioned battery cell ultrasonic welding condition root cause analysis system includes the following steps: S1. Real-time acquisition of welding equipment parameters, environmental parameters, and implicit parameters, synchronous acquisition of CCD detection results, forming a raw dataset, specifically including: S11. Data Acquisition Scope and Specifications: For a single lithium-ion battery cell welding production line, complete process parameter data for each cell is collected. Each data entry contains 12 input parameters and 1 output label, with the specific specifications as follows:

[0023] S12. Data Synchronization and Storage: The OPC UA protocol is used to realize real-time communication between the device, sensor and data acquisition system. The data sampling frequency is set to 1Hz to ensure that the parameters of each cell correspond to the CCD detection results. The time-series database is used to store the raw data, with a single data storage latency of <100ms, and supports historical data backtracking and batch analysis.

[0024] S2. Preprocess the original dataset and extract feature vectors for model training, specifically including: S21. Data Cleaning: Missing value handling: Missing values ​​are filled using the "parameter type adaptation method". Equipment parameters (such as pressure and energy) are filled using the median of the same batch of data. Environmental parameters (such as humidity and temperature) are filled using the sliding window mean. Latent parameters (such as moisture content) are filled using linear interpolation of adjacent cell data. Outlier filtering: based on A dual filtering mechanism is constructed using criteria and domain knowledge, firstly through... Criterion Filtering Exceeds Discrete points of the range ( The mean of the parameters, (The standard deviation is used as the standard deviation). Then, invalid data (such as outliers with welding time <5ms) are filtered out according to the process standard, and finally the proportion of valid samples is >95%.

[0025] S22. Feature Coding and Standardization: Discrete feature encoding: For discrete parameters such as welding frequency, one-hot encoding is used to convert them into binary feature vectors. For example, frequency = 20kHz is encoded as [1,0,0,0], and frequency = 30kHz is encoded as [0,1,0,0]; Continuous feature standardization: For continuous parameters such as pressure and energy, Z-score standardization is used, with the following formula: ; in, These are the original parameter values. This is the sample mean of the parameter. The standard deviation is the sample standard deviation. After standardization, the mean of the parameters is 0 and the standard deviation is 1. Transformation of temporal characteristics: For time-series parameters of humidity and temperature, calculate the area, such as wet sliding mean = the sum of ambient humidity over the past 120 minutes.

[0026] S3. Train multiple basic models using the aforementioned feature vectors, optimize model parameters through cross-validation, and establish a correlation model between parameters and NG results, specifically including: S31. Basic Model Selection and Parameter Settings: Five basic models with different algorithmic logics were selected to ensure the diversity of the integrated architecture. The parameter settings for each model are as follows: Decision Tree (DT): The CART algorithm is used, with a maximum tree depth of 8 and a minimum number of sample splits of 20. The importance of features is measured by the Gini coefficient. Random Forest (RF): The number of trees is set to 20, the maximum number of features is set to "sqrt(total number of features)", and out-of-bag (OOB) samples are used to evaluate the model's generalization ability; Gradient Boosting Tree (XGBoost): The learning rate is set to 0.1, the number of trees is set to 30, the maximum tree depth is set to 6, and the cross-entropy loss function is used; Support Vector Machine (SVM): The kernel function is RBF (Radial Basis Function), the penalty coefficient C is set to 1.0, and the gamma parameter is set to 0.1; Neural Network (MLP): It adopts a 3-layer fully connected structure. The number of neurons in the input layer is equal to the total number of features. The number of neurons in the hidden layer is 64 and 32 respectively. The output layer has 1 neuron (sigmoid activation function). The optimizer is Adam and the learning rate is set to 0.001.

[0027] S32. Model Training and Validation: Dataset partitioning: The effective samples are divided into a training set (70%) and a test set (30%) in a 7:3 ratio. Stratified sampling is used to ensure that the proportion of NG cells in the training set and the test set is consistent. Cross-validation optimization: For each base model, 5-fold cross-validation is used to adjust hyperparameters (such as the number of trees in a random forest). The optimization target is the root cause localization accuracy on the test set (i.e., the matching rate between the main root cause identified by the model and the actual abnormal process parameters), ensuring that the accuracy of each model is ≥85%. Model performance evaluation: The model performance is evaluated using three metrics: accuracy, precision, and recall. The ensemble model is required to have an accuracy of ≥90% and a recall of ≥88% on the test set (i.e., no critical root causes are missed).

[0028] S4. Calculate the parameter importance score for each model, and determine the root cause priority by adding a vote, specifically including: S41. Parameter Importance Score Calculation: For each trained base model, the importance score of 12 input parameters to the NG result is calculated, and the specific calculation method is as follows: Decision Tree / Random Forest / XGBoost: Calculated by the "impurity reduction" during feature splitting, such as the parameter importance score in a decision tree = the sum of the Gini coefficient reductions of all split nodes for that parameter; Support Vector Machine (SVM): Calculates the contribution of features to the classification hyperplane using the following formula: ; in, Here, j represents the weight coefficient, and n represents the total number of features. Neural Network: The sensitivity of hidden layer neurons to input parameters is calculated, and the sum of the absolute values ​​of the partial derivatives of parameter j is obtained by gradient descent, which is used as the importance score.

[0029] S42. Weighted voting mechanism: Score standardization: The parameter importance scores for each model are normalized so that the sum of the scores of all parameters within a single model is 1. The formula is: ; in, The standardized score of parameter j in model m, This represents the original importance score of parameter j in model m; Weight Calculation: The total voting weight of parameter j is calculated using the following formula: ; in, The weights for model m are set according to the model's accuracy on the test set. For example, the weight of a model with 92% accuracy is 0.22, and the weight of a model with 88% accuracy is 0.18. Ensure that the sum of the weights of all models is 1. Root cause priority determination: Sort the 12 parameters from largest to smallest according to their total voting weight. The parameter with the highest weight is the "primary root cause," and the parameters ranked 2nd and 3rd with the highest weight are the "secondary root causes." A weight threshold is set (e.g., ...). ), and only parameters whose weights exceed the threshold are included in the root cause list (to avoid judging irrelevant parameters).

[0030] S5. Generate root cause analysis reports and visualization charts, specifically including: S51. Automatically generate structured reports, including the following content: Basic information: Analysis batch, cell model, quantity and proportion of NG cells; Root cause list: Primary root causes (including weight percentage and abnormal range, such as "insufficient welding energy, weight 52%, actual value 150 < standard value 180-220J"), Secondary root causes (such as "high workshop humidity, weight 21%; excessive turnaround time, weight 18%)"); Adjustment recommendations: Parameter adjustments for the primary root cause (e.g., "It is recommended to increase the welding energy to 190J"), and auxiliary measures for secondary root causes (e.g., "Turn on the workshop dehumidification system and control the humidity at 40%-50% RH"). Visualization: Displays a "root cause contribution bar chart", allowing engineers to view the root cause trend of historical batches, such as the percentage of times "insufficient welding energy" has become the main root cause in the past 7 days.

[0031] S52, Dynamic Update and Adaptation Mechanism: Model Iteration Update: When a new type of NG battery cell appears on the production line (such as abnormal solder area due to wear of solder head), the new data is automatically included in the training set. The ensemble model is retrained every 500 valid samples to ensure that the model is adapted to the new scenario. Multi-production line adaptation: For equipment differences between different production lines (such as different welding machine models), a dedicated integrated model is built for each production line. Through transfer learning, the model parameters of the historical production line are used as initial values ​​to shorten the training cycle of the new production line model (from the original 4 hours to 2 hours). Real-time response mechanism: When the CCD detects that the proportion of NG cells exceeds the threshold (e.g., 1%), the root cause analysis process is triggered. The total time from data acquisition to the output of the root cause report is less than 5 minutes, ensuring timely location and resolution of process anomalies.

[0032] Compared to traditional solutions, the disclosed solution: Breakthrough improvement in root cause localization accuracy: By integrating 12 full-dimensional parameters and a five-model integrated voting mechanism, the problem of multi-parameter coupling is overcome, and the accuracy of main root cause localization reaches more than 90%, which is 30 percentage points higher than the traditional manual investigation (accuracy of about 60%). The identification rate of NG battery cell root causes caused by hidden parameters has increased from 70% to 82%, solving the root cause "blind spot" problem. The efficiency of production line anomaly response has been significantly optimized: an adjustment plan with priority can be output within 5 minutes after triggering root cause analysis, which is 24-48 times more efficient than manual investigation by experts (average time 2-4 hours). When the proportion of NG cells exceeds the threshold, losses can be stopped quickly, meeting the high consistency requirements of power batteries. Improved Parameter Adjustment Targeting and Resource Utilization: By quantifying the contribution of root causes and clarifying primary and secondary priorities, resource misallocation is avoided. The hit rate of parameter adjustments has increased from 45% in traditional methods to over 85%, and the proportion of ineffective adjustments to secondary root causes has decreased by 60%. At the same time, material losses caused by trial-and-error parameter changes have been reduced. Enhanced adaptability to complex working conditions and technology reusability: Dynamic feature engineering and model iteration mechanism can adapt to complex working conditions such as tab thickness deviation of +0.03mm and humidity fluctuation of +15%RH. The robustness of the model is improved compared to a single algorithm, and it can be quickly adapted to different welding machine production lines through transfer learning (training cycle shortened by 50%). It can also be extended to welding root cause analysis in different scenarios such as energy storage cells and consumer electronics cells, with a technology reusability rate of 90%.

[0033] See Figure 3 The electronic device provided in the embodiments of this disclosure includes a processor 1110, a communication interface 1120, a memory 1130 and a communication bus 1140, wherein the processor 1110, the communication interface 1120 and the memory 1130 communicate with each other through the communication bus 1140. Memory 1130 is used to store computer programs; When the processor 1110 executes the program stored in the memory 1130, it implements the above-described root cause analysis method for the ultrasonic welding status of the battery cell.

[0034] The aforementioned communication bus 1140 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus 1140 can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, it is represented by only one thick line in the figure, but this does not indicate that there is only one bus or one type of bus.

[0035] The communication interface 1120 is used for communication between the above-mentioned electronic device and other devices.

[0036] The memory 1130 may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory 1130 may also be at least one storage device located remotely from the aforementioned processor 1110.

[0037] Embodiments of this disclosure also provide a computer-readable storage medium. The computer-readable storage medium stores a computer program that, when executed by a processor, implements the root cause analysis method for the ultrasonic welding status of battery cells as described above.

[0038] The embodiments described above are merely examples of several implementations of this disclosure, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent disclosure. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this disclosure, and these modifications and improvements all fall within the protection scope of this disclosure.

Claims

1. A root cause analysis system for ultrasonic welding status of battery cells, characterized in that, include: The data acquisition module is used to collect welding equipment parameters, environmental parameters, and implicit parameters in real time, and simultaneously acquire CCD detection results to form a raw dataset; The feature engineering module is used to preprocess the original dataset and extract feature vectors for model training. An integrated model training module is used to train multiple basic models using the feature vectors, optimize model parameters through cross-validation, and establish a correlation model between parameters and NG results. The root cause voting determination module is used to calculate the importance score of parameters in each model and determine the root cause priority through weighted voting; The results output module is used to generate root cause analysis reports and visualization charts.

2. The root cause analysis system for ultrasonic welding status of battery cells according to claim 1, characterized in that, The welding equipment parameters include welding pressure, welding energy, welding amplitude, welding height, welding power, welding frequency, weld area, and welding time; the environmental parameters include workshop humidity and workshop temperature; and the implicit parameters include turnover time and cell moisture content.

3. A method for root cause analysis of ultrasonic welding conditions of battery cells, characterized in that, Includes the following steps: Real-time acquisition of welding equipment parameters, environmental parameters, and implicit parameters, and simultaneous acquisition of CCD detection results to form a raw dataset; The original dataset is preprocessed, and feature vectors for model training are extracted. Multiple basic models are trained using the aforementioned feature vectors, and the model parameters are optimized through cross-validation to establish a correlation model between the parameters and the NG results; Calculate the importance scores of the parameters in each model, and determine the root cause priority through weighted voting; Generate root cause analysis reports and visualization charts.

4. The method for root cause analysis of ultrasonic welding status of battery cells according to claim 3, characterized in that, The original dataset is preprocessed, and feature vectors for model training are extracted, including: The original dataset is cleaned, including missing value handling and outlier filtering; The discrete parameters in the cleaned original dataset are encoded to obtain discrete features; the continuous parameters in the cleaned original dataset are standardized to obtain continuous features; and the time-series parameters in the cleaned original dataset are transformed to obtain time-series features.

5. The method for root cause analysis of ultrasonic welding status of battery cells according to claim 3, characterized in that, Five basic models are used: decision tree model, random forest model, gradient boosting tree model, support vector machine and neural network model.

6. The method for root cause analysis of ultrasonic welding status of battery cells according to claim 3, characterized in that, Multiple base models are trained using the aforementioned feature vectors, and the model parameters are optimized through cross-validation to establish a correlation model between the parameters and the NG (Not Given) results, including: The feature vectors are divided into training and test sets according to a preset ratio, and stratified sampling is used to ensure that the proportion of NG cells in the training and test sets is consistent. For each base model, 5-fold cross-validation was used to adjust the hyperparameters, with the root cause localization accuracy on the test set as the optimization objective, to ensure that the accuracy of each model is greater than the preset threshold. The model performance is evaluated using three metrics: accuracy, precision, and recall.

7. The method for root cause analysis of ultrasonic welding status of battery cells according to claim 3, characterized in that, Calculate the importance scores of the parameters in each model, and determine the root cause priority through weighted voting, including: For each completed base model, calculate the importance score of all input parameters to the NG result; The parameter importance scores for each model are normalized to obtain standardized scores, so that the sum of the scores of all parameters in a single model is 1. The total voting weight for each parameter is calculated based on the standardized score. Sort all parameters in descending order of total voting weight. The parameter with the highest weight is the "primary root cause", and the parameters within the preset weight ranking range are the "secondary root causes". Set a weight threshold, and only include a parameter in the root cause list when the weight exceeds the weight threshold.

8. The method for root cause analysis of ultrasonic welding status of battery cells according to claim 3, characterized in that, The root cause analysis report includes the following: Basic information: Analysis batch, cell model, quantity and proportion of NG cells; Root cause list: primary root cause, secondary root cause; Adjustment recommendations: Analyze the parameters for the primary root causes and implement supplementary measures for the secondary root causes.

9. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor is configured to execute a program stored in a memory to implement the root cause analysis method for ultrasonic welding status of battery cells as described in any one of claims 3-8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the root cause analysis method for the ultrasonic welding status of battery cells as described in any one of claims 3-8.

Images