Anomaly Detection-Based Early Warning Framework for Lithium-Ion Batteries under Dynamic Impact Conditions in Electric Mobility

Kyutae Park; Hyungi Kim; Xianglin Huang; Heung Soo Kim

doi:10.7734/COSEIK.2026.39.5.327

All Issue

2026 Vol.39, Issue 5 Preview Page Next Page

Research Paper

Anomaly Detection-Based Early Warning Framework for Lithium-Ion Batteries under Dynamic Impact Conditions in Electric Mobility 동적 충격 조건에서 전기차용 리튬이온 배터리 이상 탐지 기반 조기 경고 프레임워크

31 October 2026. pp. 327-333

PDF XML

Abstract

Dynamic impacts during electric vehicle operation can induce electrode deformation and active material particle cracking in lithium-ion batteries, potentially leading to internal short circuits and thermal runaway. This necessitates the development of an early warning system capable of rapidly identifying damage immediately after impact. In this study, acoustic emission (AE) signals were collected from pouch-type lithium-ion batteries under impact energies ranging from 0.1 J to 2.0 J through drop-weight impact experiments. Scanning electron microscopy (SEM) analysis was conducted to establish a physics-based classification criterion, defining impact energies of 0.2 J and below as normal and 0.3 J and above as anomalous. Seven physically meaningful features spanning energy, decay, frequency, and statistical domains were extracted from the AE waveforms, and an unsupervised autoencoder was employed for real-time anomaly detection based on the reconstruction error. The proposed model achieved a Precision, Recall, and F1-score of 1.0 and an ROC-AUC of 1.0, demonstrating a clear distinction between normal and anomalous samples based on the reconstruction error and confirming that changes in the elastic wave characteristic according to the impact energy levels are effectively quantified. Notably, the model achieved a robust detection performance trained exclusively on normal patterns without any anomaly data, demonstrating its potential as a practical early warning framework applicable to real-world driving environments where diverse impact conditions coexist.

Keywords

lithium-ion battery

acoustic emission

dynamic impact

autoencoder

anomaly detection

전기차 운행 중 발생하는 동적 충격은 리튬이온 배터리 내부의 전극 변형 및 활물질 입자 균열을 유발하여 내부 단락과 열폭주로 이어질 수 있으며, 이를 충격 직후 신속하게 판별하는 조기 경고 체계의 구축이 요구된다. 본 연구에서는 낙하 충격 실험을 통해 0.1J에서 2.0J의 다양한 충격 에너지 조건에서 파우치형 리튬이온 배터리의 음향 방출(Acoustic Emission, AE) 신호를 수집하고, 주사 전자 현미경(SEM) 분석을 통해 0.2J 이하를 정상, 0.3J 이상을 이상으로 정의하는 물리적 판단 기준을 제안하였다. AE 파형으로부터 에너지, 감쇠, 주파수, 통계 영역에 걸친 7개의 물리적 의미 기반 특징을 추출하고, 비지도학습 기반 오토인코더(AutoEncoder)를 활용하여 복원 오차 기반의 실시간 이상 탐지를 수행하였다. 제안한 모델은 Precision, Recall, F1-score 1.000, ROC-AUC 1.0000을 달성하였으며, 정상과 이상 샘플 간 복원 오차에 따른 우수한 분리 성능을 보이며 충격 에너지 수준에 따른 탄성파 특성 변화가 효과적으로 정량화되고 있음을 확인하였다. 특히 이상 데이터 없이 정상 패턴만으로 학습하여 강건한 탐지 성능을 달성하였다는 점에서 다양한 충격 상황이 혼재하는 실제 주행 환경에서도 적용 가능한 실용적 조기 경고 프레임워크로서의 가능성을 제시한다.

키워드

리튬이온 배터리

음향 방출

동적 충격

오토인코더

이상 탐지

References

Chen, Y., Kang, Y., Zhao, Y., Wang, L., Liu, J., Li, Y., Liang, Z., He, X., Li, X., Tavajohi, N., Li, B. (2021) A Review of Lithium-Ion Battery Safety Concerns: The Issues, Strategies, and Testing Standards, J. Energy Chem., 59, pp.83~99.

10.1016/j.jechem.2020.10.017

Deng, Y., Li, J., Huang, X. (2026) Deep Learning-Based Real-Time Damage Assessment of Lithium-Ion Batteries under Dynamic Impact, J. Power Sources, 662, 238737.

10.1016/j.jpowsour.2025.238737

Hinton, G.E., Salakhutdinov, R.R. (2006) Reducing the Dimensionality of Data with Neural Networks, Science, 313(5786), pp.504~507.

10.1126/science.1127647

Kalnaus, S., Wang, Y., Turner, J.A. (2017) Mechanical Behavior and Failure Mechanisms of Li-Ion Battery Separators, J. Power Sources, 348, pp.255~263.

10.1016/j.jpowsour.2017.03.003

Li, B., Zhou, Y., Zhou, X., Huang, Z., Zhang, X. (2026) Data-Driven Machine Learning Prediction of Impact Failure in Cylindrical Lithium-Ion Batteries, Energies, 19(6), 1435.

10.3390/en19061435

Liu, B., Jia, Y., Yuan, C., Wang, L., Gao, X., Yin, S., Xu, J. (2020) Safety Issues and Mechanisms of Lithium-Ion Battery Cell upon Mechanical Abusive Loading: A Review, Energy Storage Mater., 24, pp.85~112.

10.1016/j.ensm.2019.06.036

Pang, G., Shen, C., Cao, L., Hengel, A.V.D. (2021) Deep Learning for Anomaly Detection: A Review, ACM Comput. Surv., 54(2), pp.38:1~38:38.

10.1145/3439950

Vincent, P., Larochelle, H. (2010) Stacked Denoising Autoencoders: Learning Useful Representations in a Deep Network with a Local Denoising Criterion, J. Mach. Learn. Res., 11, pp.3371~3408.

Zhang, X., Zhang, T., Liu, N., Yin, X., Wu, X., Han, H., Wang, Q., Zhang, Y. (2023) Dynamic Crushing Behaviors and Failure of Cylindrical Lithium-Ion Batteries Subjected to Impact Loading, Eng. Fail. Anal., 154, 107653.

10.1016/j.engfailanal.2023.107653

Zhou, C., Paffenroth, R.C. (2017) Anomaly Detection with Robust Deep Autoencoders, Proc. 23rd ACM SIGKDD Int. Conf. Knowl. Discov. Data Min. (KDD '17), pp.665~674.

10.1145/3097983.3098052

Information

Publisher :Computational Structural Engineering Institute of Korea
Publisher(Ko) :한국전산구조공학회
Journal Title :Journal of the Computational Structural Engineering Institute of Korea
Journal Title(Ko) :한국전산구조공학회 논문집
Volume : 39
No :5
Pages :327-333
Received Date : 2026-06-01
Revised Date : 2026-06-21
Accepted Date : 2026-06-22
DOI :https://doi.org/10.7734/COSEIK.2026.39.5.327

Journal of the Computational Structural Engineering Institute of Korea ISSN:1229-3059(Print) 2287-2302(Online) 한국전산구조공학회 논문집

All Issue