基于BP神经网络的轨道不平顺维修决策建模

Modeling of Track Irregularity Maintenance Policy Based on Back Propagation Neural Network

针对高速铁路轨道在列车荷载运行过程中产生的轨道质量状态劣化情况,本文基于BP神经网络建立了高速铁路的轨道不平顺维修决策模型。模型基于轨道运检过程中采集到的轨道几何特征数据,综合考虑七个几何轨道不平顺几何特征指标涵盖的信息特征,将轨道维修决策抽象为基于多维度特征的二分类问题,然后通过算法定位历史维修点,构造模型训练集,建立BP神经网络模型自主学习,最终达到输入轨道数据,输出铁路维修决策的效果,以达到指导实际铁路维修过程,优化铁路维修决策的目的。经实际运行数据的验证,所建立的神经网络模型在验证集准确率达到80%以上,证明了模型的有效性。

Aiming at the deterioration of the track quality state of the high-speed railway track during the op-eration of the train load, this paper established a decision-making model for the track irregularity maintenance based on the BP neural network. The model is based on the track geometric feature data collected during the track inspection process, considers the information covered by the seven geometric track irregularity geometric index, and abstracts track maintenance decision-making into a binary classification problem based on multi-dimensional features, and then locates historical maintenance points through an algorithm, to construct BP neural network model’s training set, the BP neural network is established to learn autonomously. Finally, it achieves the effect of inputting track data and outputting railway maintenance decision-making, so as to guide the actual railway maintenance process and optimize railway maintenance decision-making. The established neural network model is effective after the verification of the actual operation data, its accuracy rate in the verification set reaches more than 80%.