P2 HEV의 LSTM 기반 엔진 클러치 접합/해지 이상치 탐지 알고리즘

Abstract

This paper presents an anomaly detection algorithm for an engine clutch engagement/disengagement process of P2 type hybrid electric vehicles that use long short-term memory(LSTM). We proposed a structure of an LSTM-based model that can predict data at present, and trained the model with normal data. When the difference between the predicted values of the model and the measured values exceeds a certain threshold, the algorithm determines the data as anomalies. We used simulation data in the model training, and developed a threshold that considers the data prediction characteristics of the LSTM-based model. The developed anomaly detection algorithm predicted normal data well, and showed the results of anomaly detection with high accuracy. Since other vehicle data have similar characteristics to the target data of this paper, this algorithm is expected to be applied successfully to other vehicle data.