Battery Manufacturing Anomaly Detection With ML Cause Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing battery manufacturing processes lack effective methods to systematically manage and analyze abnormal data occurrences, leading to inefficiencies in identifying and addressing causes of abnormalities, which are often time-consuming and costly.
Innovation Solution
A machine-learning-based abnormal point cause analysis model is developed to integrate design, material, and process information, using latent factors selected through decision trees and artificial neural networks to identify and analyze the causes of abnormal manufacturing data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If manual data extraction and analysis methods are used for abnormal battery manufacturing data, then the analysis can be performed with simple tools, but the process becomes time-consuming and costly
Solution Approach 1:
The patent replaces manual mechanical data extraction and analysis processes with an automated machine learning system. The system automatically collects manufacturing data, detects abnormal points, and identifies causes using trained models, eliminating the need for manual intervention while significantly reducing analysis time and costs.
Solution Approach 2:
The patent implements preliminary action by pre-training machine learning models with historical manufacturing data before actual abnormal point detection. The system performs offline training and model preparation in advance, so that when abnormal points occur during manufacturing, the pre-trained models can immediately identify causes without requiring time-consuming manual analysis or real-time model training.
2Measurement precision
If comprehensive material, design, and process information is collected for abnormal point analysis, then the accuracy of cause identification improves, but the system complexity increases
Solution Approach 1:
The patent segments the comprehensive manufacturing information into three distinct modules: material information module, design information module, and process information module. Each module handles specific types of data independently, and the machine learning system integrates results from all three modules to identify abnormal causes. This segmentation reduces system complexity while maintaining comprehensive analysis capability.
Solution Approach 2:
The patent introduces an abnormal point detection model as an intermediary that automatically processes and integrates information from material, design, and process modules. This intermediary model performs the complex integration task, shielding users from the underlying system complexity while delivering accurate cause identification results.
3Measurement precision
If machine learning models are trained with extensive training data, then the abnormal point detection accuracy improves, but the data processing requirements and computational resources increase
Solution Approach 1:
The patent performs model training as a preliminary action before actual manufacturing operations. Extensive training data is processed offline to train the abnormal point detection model and cause analysis model in advance. Once trained, the models are deployed for real-time detection with minimal computational resources required during actual manufacturing, thus achieving high accuracy without continuous high energy consumption.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
The present disclosure relates to a method and system for detecting an abnormal point of battery manufacturing data and analyzing its cause. The method is performed by the system for detecting the abnormal point of the battery manufacturing data and analyzing its cause, and includes collecting training data including material information, design information, process information, and information on whether an abnormal point has occurred, of a battery to be analyzed, selecting latent factors that cause the abnormal point from among material elements, design elements, and process elements on the basis of the training data, forming an artificial neural network structure on the basis of the selected latent factors, training the artificial neural network structure on the basis of the training data, and generating an abnormal point cause analysis model.