基于LSTM循环神经网络的转向架部件监测参数趋势预测

Trend Prediction of Monitoring Parameters of Bogie Components Based on LSTM Recurrent Neural Network

故障诊断与预测通过海量数据,采用推断统计、神经网络等研究方法,对监测数据进行分析和预测,从而评估轨道车辆的健康状况,最大限度保证轨道车辆各系统部件健康工作。本文通过研究转向架关键部件的物理特性,结合长短期记忆网络在处理“长期依赖”问题的优势,构建了基于长短期记忆网络的转向架关键参数趋势预测模型。为了验证方法的有效性,本文使用CRH380动车组转向架轴箱轴承和齿轮箱的温度监测数据进行了预测。结果表明,基于长短期记忆网络的转向架关键部件参数趋势预测方法能够有效预测转向架关键参数的变化趋势。

Fault diagnosis and prediction use massive data, inferred statistics, neural network and other research methods to analyze and predict the monitoring data, so as to evaluate the health status of the rail vehicle and ensure the health of each system component of the rail vehicle to the greatest extent. In this paper, by studying the physical characteristics of the key components of the bogie, combined with the advantages of the long and short-term memory network in dealing with the problem of “long-term dependence”, a trend prediction model for the key parameters of the bogie based on the long and short-term memory network is constructed. In order to verify the effectiveness of the method, this paper uses the temperature monitoring data of the CRH380 EMU bogie axle box bearing and gear box to make predictions. The results show that the trend prediction method of bogie key component parameters based on long- and short-term memory network can effectively predict the change trend of bogie key parameters.