🗞️ Paper “Enhancing battery anomaly detection via trend-reversal spectrograms and multi-modal cross-attention” was published by eTransportation.
Published:
- Phuong-Nam Tran, Trung Thanh Pham, Jae-Jo Lee, Eui-Nam Huh, Choong Seon Hong, “Enhancing battery anomaly detection via trend-reversal spectrograms and multi-modal cross-attention”, eTransportation, vol. 30, 100644, 2026. 📖 Read it for free
This paper detects anomalies in electric vehicle (EV) battery data. We show that the Short-Time Fourier Transform (STFT) captures “trend reversals” in battery signals, and we combine these spectrograms with the raw time series in a multi-modal cross-attention model (STFNet). The model detects anomalies with two branches, one based on reconstruction and one on classification, and fuses their decisions with logic rules. A cross-brand adaptive version transfers knowledge to new EV brands where only normal data are available. On a large real-world EV dataset, it reaches an average F1 score of 0.8667 and an AUROC of 0.9485, outperforming existing methods.
