Evaluation of Vehicle Ride Comfort based on Neural Network

The relationship between subjective ride comfort in a vehicle seat and human whole-body vibration can be modeled using frequency weightings and rms(root mean square) averaging as specified in ISO2631. However, recent studies indicate that, there are some flaws in the relationship between subjective response and objective vibration given by the ISO2631. This paper presents an alternative approach based on neural network model. Time-domain vibration acceleration signals are processed as neural network inputs, subjective evaluation results are quantified as outputs, and the weights of neural networks are used as frequency weighting coefficients to evaluate the vehicle ride comfort. The method has been used to evaluate the ride comfort on a number of conditions with good results achieved.