Pavement roughness identification research in time domain based on neural network

A new simulation study method based on general regression neural network (GRNN) is proposed for identifying the pavement roughness in the time domain. First, a seven degree-of-freedoms vehicle vibration model is estbalished for the vehicle's riding comfort analysis. The vertical acceleration and pitching angular acceleration of vehicle body centroid are calculated by simulation. The nonlinear mapping relations between the two above accelerations and pavement roughness in time domain are built by GRNN, and then the pavement roughness is identified by training the networks. Finally, the vertical acceleration and pitching angular acceleration of the vehicle body centriod are acquired by ADAMS/View virtual experiment simulation and the result are used to identify pavement roughness. In the end, the availability for identifying the pavement roughness by GRNN is confirmed.