Prediction algorithm of derailment coefficient in turnout area based on multi-sensor data fusion

For the current vehicle derailment occurred in the switch section, and the existing way of derailing judge to consider a single factor can not fully judge the security issue of wheel-rail contact, an algorithm for predicting derailment coefficients in turnout areas based on multi-sensor data fusion is proposed. The method used the T-snake model to segment the wheel-rail contact image of the No. 9 turnout area acquired by the vehicle camera to obtain the relative lateral displacement of the wheel-rail. The wavelet neural network optimized by genetic algorithm is used to construct the fusion model. The input data is the relative displacement amount, speed amount, acceleration amount, and wheel load reduction ratio to perform data fusion to predict the derailment coefficient. The field test results show the accuracy of the derailment coefficient predicted by this method is high, and the considerations are comprehensive. © 2020, Central South University Press. All rights reserved.