Neural Battery Cell Pre-Diagnosis for Rapid Defect and Fire Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for diagnosing defects and fires in battery cells are time-consuming and impractical for large-capacity batteries, especially since they require multiple charge and discharge cycles, making it difficult to quickly inspect and ensure the safety of high-capacity battery packs.
Innovation Solution
A neural network-based method and apparatus that collects chemical composition, current, and temperature data during charge and discharge cycles to pre-diagnose defects and fires by training on collected data, predicting deviations from normal behavior, and determining battery health without additional devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current, voltage, and temperature data are collected by charging and discharging the battery to diagnose defects, then diagnosis accuracy is improved, but inspection time becomes excessively long and impractical for large-capacity batteries
Solution Approach 1:
The patent applies preliminary action by pre-training a neural network model using data from multiple charge-discharge cycles before actual inspection. The model learns normal battery behavior patterns in advance, enabling rapid defect detection during actual use without requiring time-consuming full charge-discharge cycles for each inspection
Solution Approach 2:
The patent uses copying by creating a virtual model (neural network) that replicates the complex behavior of batteries under various conditions. This virtual model can simulate and predict battery responses without physically subjecting actual batteries to extensive charge-discharge cycling, thereby reducing inspection time while maintaining diagnostic accuracy
2Reliability
If multiple charge and discharge cycles are performed to ensure battery safety, then reliability of diagnosis is improved, but productivity of battery inspection deteriorates
Solution Approach 1:
The neural network is pre-trained on extensive data from multiple charge-discharge cycles beforehand. This preliminary training phase captures the reliability-critical behavior patterns, allowing the model to perform rapid predictions during actual inspection without repeating time-consuming cycles
Solution Approach 2:
The patent replaces the mechanical process of physically charging and discharging batteries for inspection with an information-processing system (neural network). The model analyzes electrical and thermal data patterns computationally, substituting physical cycling with algorithmic prediction, thereby maintaining diagnostic reliability while dramatically improving inspection speed
3Measurement precision
If a huge warehouse is used to age lithium batteries for one or two months to test them, then defect detection capability is improved, but device complexity and cost increase
Solution Approach 1:
Instead of physically aging batteries in huge warehouses for months, the patent creates a virtual aging process through the neural network model. The model learns from training data that represents long-term aging effects, enabling defect detection without requiring physical storage infrastructure or extended aging periods
Solution Approach 2:
The patent replaces the physical aging process (requiring huge warehouses and months of time) with computational modeling. The neural network simulates the effects of long-term aging by analyzing patterns in electrical and thermal data, eliminating the need for extensive physical infrastructure while maintaining defect detection capability
Data Source
AI summary
The prevent invention relates a method and apparatus based on neural network for pre-diagnosing defect and fire in battery cell. A method based on neural network for pre-diagnosing defect and fire in battery cell in an apparatus for pre-diagnosing defect and fire in battery cell is proposed, the method including collecting, by the battery cell defect and fire pre-diagnosis apparatus, data including at least one of chemical composition, current, voltage, and temperature data measured for each predetermined time interval within each charge and discharge cycle while charging and discharging a plurality of batteries; training, by the battery to cell defect and fire pre-diagnosis apparatus, the neural network by inputting the collected data to the neural network; and predicting, by the battery cell defect and fire pre-diagnosis apparatus, a battery deviated from a main cluster of the neural network as defective.