Research on derailment coefficient prediction model based on deep learning
Mengyao Cai, Yunshui Zheng · 2024
Train derailment is a very complex dynamic process. Aiming at the problem that the current derailment evaluation factors are insufficient to fully study the wheel-rail contact mechanism, a derailment coefficient prediction model based on deep learning is proposed. Firstly, the transverse movement of the wheel relative to the rail is obtained by image segmentation algorithm, and then the derailment coefficient prediction model is constructed by Elman neural network optimized by genetic algorithm, and the input speed, acceleration, wheel load reduction rate and transverse movement are predicted. Field test results show that this model has high prediction accuracy and strong robustness.