In this paper, we propose anomaly detection for engine clutch engagement/disengagement using LSTM to evaluate the HEV driving mode performance of TMED parallel hybrid electric vehicle. The LSTM-based regression model is trained using vehicle simulatio...
In this paper, we propose anomaly detection for engine clutch engagement/disengagement using LSTM to evaluate the HEV driving mode performance of TMED parallel hybrid electric vehicle. The LSTM-based regression model is trained using vehicle simulation data that can be obtained through the HEV P2 reference application provided by Mathworks and approximated to a mathematical model tor physical properties related to the engine clutch. In other words, the LSTM-based regression model can conduct time series prediction like an estimator. Therefore outliers are detected by inputting the defect-injected signal into the trained model and calculating the residuals from the predicted and observed signal.