Artificial neural network (ANN) application to predict the wheel-rail impact forces due to the short wavelength defects in rail

This paper proposes an application of Artificial Neural Networks (ANN) to predict the wheel-rail impact forces due to the Railway Track Irregularities. The proposed architecture of the ANN is a form of Recurrent Neural Network (RNN), which uses the current sample of the input, a few past samples of the input and also feeds back a couple of previous output samples to the input pattern. This input pattern is then used to predict the new value of the output. The capability of the proposed architecture in predictin the impact forces due to both wheel burns defects and dip weld defects is promising.