On the Use of Machine Learning for the Development of an Acoustic-based Detection System for Early-Stage Thermal Runaway

Published

March 15, 2026

Author(s)

Wai Cheong Tam, Md. Ismail Siddiqi Emon, Jian Chen, Hongqiang Fang, Jun Deng, Anthony Putorti

Abstract

The paper presents the development of a multi-class classification model for the detection of early-stage thermal runaway events for button-top single-cell lithium-ion batteries. A signal gate mechanism is introduced to extract relevant acoustic data. A data alignment technique is applied to enhance the model training. A multi-layer two-dimensional convolutional neural network is utilized to learn the important features to differentiate non-thermal runaway events and thermal runaway events. Results show that the proposed model can detect the thermal runaway events with an overall accuracy, precision, and recall of 96.7 %, 78.8 %, and 91.8 %, respectively. Sensitivity studies are conducted and the results indicate that the data alignment and data augmentation techniques help to enhance the model performance significantly. The finding from this paper hopes to contribute to the development of a practical and accurate early-stage thermal runaway detection model that can provide early warning to users to avoid battery fires.

Citation

Process Safety and Environmental Protection

Pub Type

Journals

Download Paper

Local Download

Citation

Tam, W. , Emon, M. , Chen, J. , FANG, H. , Deng, J. and Putorti, A. (2026), On the Use of Machine Learning for the Development of an Acoustic-based Detection System for Early-Stage Thermal Runaway, Process Safety and Environmental Protection, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=960166 (Accessed March 16, 2026)