A SHapley Additive exPlanations‐Based Data‐Driven Approach for Lithium‐Ion Battery State of Health Estimation Using Ultrasound Technology

Accurate estimation of battery state of health (SoH) is essential for ensuring battery safety and stability. This article presents an interpretable data‐driven framework that leverages ultrasonic technology for rapid SoH estimation. Specifically, an ultrasonic testing platform is developed, with a probe externally placed on the battery surface to transmit and receive signals. From the collected data, three key ultrasonic features (time of flight [ToF], signal maximum [Max], and signal minimum[Min]) are extracted to characterize the battery SoH. Four machine learning models are then employed to estimate the battery's SoH. Finally, the SHapley Additive exPlanations (SHAP) method is applied to these models to identify the most important ultrasonic features for SoH estimation. Experimental results demonstrate that the random forest model achieves the highest accuracy in battery SoH estimation, with an R 2 value of 0.97. The most important ultrasonic feature is time of flight, followed by Min and Max features. This indicates that the proposed method is highly effective and can be practically applied for battery SoH estimation. Furthermore, integrating the interpretable technique SHAP enhances the model's prediction interpretability and offers valuable insights into the relative importance of ultrasonic features, providing a reliable tool for battery health monitoring and diagnostics.