Prediction Study on Lateral Acceleration of Railway Coach Based on RBF Neural Networks

The lateral acceleration of railway coach is chaotic time series with certain law when it passes the curve. The passing curve unbalanced acceleration can be predicted with the law and the experiment data in certain time. And the predicted values can be used as the input reference signals of the active control system of the vehicle. The center and normalizing parameters of the active function of the hidden unit of the RBF neural networks’ hidden layer is computed, and the online self-adaptive algorithm to adjust the RBF network weight vector and to predict the acceleration in multi-step. The data detected is simulated and the conclusion is that the RBF neural network can predict the lateral acceleration of the train in the multi-step and what the lag result from all kinds of reason may be compensated, and it can replace the gyroscope and justify the curve direction.