Physics-informed GRU model for vehicle road noise prediction: Integrating transfer path analysis and hybrid data

With the continuous release of automobile market potential and the steady growth of automobile ownership, consumers’ concern for automobile NVH problems makes the road noise performance directly affect the sales of automobiles. Therefore, it is of great significance to study the problem of vehicle road noise. In the actual research process, it often faces the challenge of insufficient effective sample size. In this paper, 102 sets of sample data are collected by combining the real vehicle test method and the CAE simulation method. By comparing the road noise prediction model based on the GRU algorithm with the LSTM model and the CNN model, the results show that the GRU model performs roughly similarly to the LSTM model in terms of prediction accuracy (RMSE = 2.18) and robustness (MSE = 7.66%), and the GRU model and the LSTM model are significantly better than the CNN model, but the prediction efficiency of the GRU model is significantly better than the LSTM model. Therefore, the vehicle road noise prediction model based on the GRU algorithm is optimal. This paper provides an efficient method for road noise performance analysis and prediction, which can be applied to the vehicle design and performance improvement process and provide technical support for improving vehicle comfort and market competitiveness.