Contactless Fault Diagnosis Technique by Magnetic Field Image Recognition for Lithium Battery

Model-based, signal analysis-based, and data-driven methods are commonly employed for fault detection in lithium-ion batteries. However, these approaches are often limited by their reliance on accurate parameter identification, threshold-setting issues, and the need for large-scale, high-quality data. In recent years, there has been increasing interest in active measurement techniques utilizing sensors for the fault diagnosis of lithium-ion batteries. Among these, magnetic field-sensing presents a contactless solution for fault diagnosis. This article proposes an innovative intelligent magnetic field-sensing technique based on tunnel magnetoresistive (TMR) sensors and magnetic field image recognition for the effective and low-cost fault diagnosis of lithium-ion batteries. First, the electrochemical and magnetic characteristics of lithium-ion batteries are investigated to establish the underlying principles of magnetic field variation and to model the battery. A TMR sensor is employed to measure the magnetic field of lithium-ion batteries under various conditions, and magnetic field distribution patterns and image features are subsequently identified. A contactless fault diagnosis system is then developed using the convolutional neural network (CNN) algorithm and a Raspberry Pi 5. The proposed system is demonstrated to reliably diagnose the healthy, aging, anomalies, and swelling of batteries, offering a promising solution for production quality inspection and recycling of lithium batteries.