Lithium-ion batteries (LIBs) are widely used; however, they have safety hazards because of their susceptibility to thermal runaway (TR). Current early warning methods rely on the external monitoring of parameters such as temperature and strain. These methods have an inherent lag, as the signals can only be detected after internal heat and gas accumulation. Internal sensors are difficult to implement due to the harsh environment and high cost, leaving the ultra-early incubation stage of TR poorly addressed. To overcome these limitations, this study introduces acoustic emission (AE) technology for the real-time external detection of internal TR reactions. An experimental platform induced TR through overcharging, integrating multi-source AE and temperature signal acquisition. Continuous AE signals were collected from the onset of overcharging until the valve opened. Time–frequency analysis revealed anomalous waveform features in the early stage of TR; a two-dimensional method enhanced frequency-domain recognition. Combining the processed AE signals with a convolutional neural network achieved high-accuracy phase segmentation. Cross-validation and comparisons with temperature-based methods demonstrate the effectiveness and precision of AE monitoring for ultra-early TR warning. The results highlight the potential of AE-based monitoring as a proactive risk-management strategy, supporting dynamic assessment and safety responses in energy-storage applications.
Liu et al. (Tue,) studied this question.