Structural evolution indicator for lithium-ion batteries deep discharge via ultrasonic inspection

Deep discharge-induced microstructural evolution significantly affects battery reliability. Conventional methods based on macroscopic electrochemical features, such as incremental capacity analysis and internal resistance, exhibit significant nonlinear hysteresis. This hysteresis occurs because macroscopic indicators only respond after substantial microstructural degradation, as macro-capacity fading lags behind internal structural evolution. Consequently, these methods struggle to capture the real-time evolution of micromechanical properties under dynamic conditions. To address this gap, this study proposes a structural change indicator (SCI) identification method combining ultrasonic inspection with a patching-enhanced scalar LSTM (P-sLSTM). The proposed method employs discrete wavelet transform for time-frequency decomposition of ultrasonic signals. Dynamic features sensitive to the mechanical modulus and interfacial states are extracted. Subsequently, P-sLSTM learns long-term temporal dependencies of feature sequences under normal operating conditions to establish a healthy-state baseline. SCI is identified based on abrupt L2 (Euclidean) reconstruction error. Experimental results show that the SCI exhibits jumps within the 75%–80% depth of discharge, marking the critical boundary of internal structural state changes. Multi-condition tests confirm that the SCI maintains high sensitivity and consistency across different temperatures and discharge rates, providing an indicator with sensitivity and physical meaning for the structural health assessment of lithium-ion batteries.