AI Battery Cell Fault Identification Using Physics-ML Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electric vehicle power storage systems face challenges in accurately detecting and isolating faults within power cells, such as internal short circuits, rising internal resistance, and decreasing capacity, which can lead to vehicle operation disruptions and potential safety issues.
Innovation Solution
A system comprising sensors, a processor, and machine learning models that acquire direct measurement data, apply physics models like recursive least squares and Kalman filters, and generate derived data to identify fault causes by comparing coordinate positions to a fault map, prompting operators to cease vehicle operation when necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physics models and machine learning models are applied to process sensor data for fault detection, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system segments the fault detection process into distinct modules: sensor data acquisition, physics model processing, machine learning model analysis, and fault determination. Each module handles a specific aspect of the detection process, improving overall precision while managing complexity through functional decomposition.
Solution Approach 2:
The patent introduces intermediary processing layers (physics models and machine learning models) between the raw sensor data and the final fault determination. These intermediaries transform and analyze the data to enhance detection accuracy without requiring direct complex interactions between all system components.
2Reliability
If real-time data processing with multiple models is performed, then reliability of fault detection is improved, but loss of time in processing increases
Solution Approach 1:
The system performs preliminary data processing using physics models before applying machine learning models. This staged approach prepares the data in advance, allowing the more complex ML models to work with pre-processed information, thereby maintaining reliability while reducing overall processing time.
Solution Approach 2:
The patent applies partial processing through physics models for all data points, then uses machine learning models selectively for complex pattern recognition. This partial application of different processing levels maintains detection reliability while optimizing processing efficiency.
Data Source
AI summary
A system for monitoring an electric power storage system includes: a processor electrically connected to multiple sensors, each of the sensors being configured to detect at least one parameter of the electric power storage system the processor being configured to acquire a set of direct measurement data of the set of power cells from the plurality of sensors, provide the set of direct measurement data to a physics model within the processor and generate a set of derived measurement data using the physics model, provide the derived measurement data and at least a portion of the direct measurement data to a machine learning model and generate a coordinate position corresponding to a fault condition of the set of power cells, compare the coordinate position to a fault map and identifying a probable fault cause based on the comparison, and alter an operation of the vehicle based on the comparison.