Research on the DCT vehicle starting process evaluation based on LSTM neural network with attention mechanism

Abstract

Currently, with the advancement of dual-clutch transmission (DCT) control systems and vehicle performance, it is necessary to develop better objective evaluation methods for DCT vehicles. The starting process is a critical element affecting the driving and riding experience of DCT vehicles. Therefore, it is crucial to establish and improve a starting process evaluation model for the objective evaluation to DCT vehicles and optimization to DCT control strategies. This paper proposes a new method to evaluate the DCT vehicle starting process objectively. The method analyzes and models the time-series signals of the driving data using the LSTM neural network and uses the attention mechanism to improve the evaluation performance and enhance the interpretability of the evaluation results. Taking the dynamic performance evaluation as an example, the evaluation results indicate that the proposed model is better than the conventional methods, showing notable efficacy and preponderance.