On-Track Measurement of Lateral/Vertical Wheel Loads of Running Railway Vehicles Based on the Neural Network

Abstract

A new approach for on-track measurement of the lateral/vertical contact forces is presented in this paper. The proposed method is based on measurement of the strain at two sides of the wheel web. Electric signals generated by the strain gauges are fed into a neural network algorithm in order to predict the lateral/vertical contact forces. Feed-forward technique is used in the neural network algorithm. A sensitivity analysis has been carried out to find the best position for the strain gauges. A dynamic model of a freight wagon is provided and a variety of numerical simulations are performed to obtain the probability distribution of the lateral and vertical contact forces. The obtained probability distribution function is then utilized to generate lateral/vertical contact forces within the practical range. In order to train the neural network system, the generated contact forces are applied to the wheel flange and the strain signals are obtained. More than 100 configurations are fed into the system in order to train it. Reliability, accuracy and sensitivity of the proposed measurement system are then investigated.

Experimental Determination of Wheel and Guiding Forces on the Tracks

“…Application of neural network or wavelet transformation can help to analyze the measured record [6,7].…”