An Artificial Neural Network-based Tool for Prediction of Tire-Pavement Interaction Noise

Tire-Pavement Interaction Noise (TPIN) is the dominant noise source for passenger vehicles at speeds above 30 mph. Modelling TPIN is a complex task, due to the multiple noise generation mechanisms involved. A large amount of experimental data was collected (42 tires, and 26 different pavement surfaces). This data was used to develop two Artificial Neural Networks (ANN). Both are configured to predict only positive acoustic sound pressure values. The first ANN uses the tire tread pattern geometry as input to predict tread pattern related noise (TPN). The second ANN uses tread rubber hardness and vehicle speed to predict the noise component not related to the tread pattern (NTPN). TPN is predicted at a fixed vehicle speed (60 mph). Then it is scaled to other speeds using the tire size and a speed scaling law. On the other hand, NTPN is predicted for a fixed tire size (215/60R16) and a reference pavement surface. This is also modified for other user-defined tire sizes and pavement surfaces. Finally, both ANNs were integrated into one TPIN prediction tool using MATLAB. It was validated using experimental data. The overall A-weighted sound pressure level (OASPL) error between measured and predicted total tire noise was 1.1 dBA