摘要
为保障列车走行部轴承关键部件的可靠监测,确保列车安全稳定运营,文章提出一种高速列车走行部监测系统全冗余架构,其采用"双智能诊断单元+双通道传感器"方案,解决了传统列车走行部监测系统监测盲区问题。基于全冗余架构,根据地面、车载数据特点、平台特点和应用场景的不同,文中给出相应的诊断模型。其中,地面模型以趋势预测为主,基于"深度神经网络+循环神经网络"的模型框架,利用海量历史数据提取时序特征参数,建立早期故障检测模型;车载模型以实时诊断为主,通过双通道波形检测的交互判断,实时对车辆级、列车级双通道传感器波形一致性进行诊断识别,并结合时速为350 km的动车组进行实际应用。应用结果表明,采用全冗余技术可以降低误报传感器故障的风险,提高诊断系统的可靠性,保障列车安全稳定运行。
In order to ensure the reliable monitoring of key components of train running gears and ensure safe and stable operation of a train,a full redundancy architecture of a running gear monitoring system for high-speed train was proposed.The scheme of“dual intelligent diagnosis unit+dual channel sensor”is adopted in the architecture,which solves the problem of monitoring blind area caused by single diagnosis unit or single channel sensor failure.Based on the full redundancy architecture,a corresponding diagnosis model was proposed,according to the characteristics of ground and vehicle data,platform characteristics and application scenarios.The ground model is based on trend prediction;based on the model framework of“deep neural network+recurrent neural network”,the characteristic parameters of time series are extracted from massive historical data,and the early fault detection model is established.The onboard model is mainly based on real time diagnosis,and is easy to implement through the interactive judgment of dual channel waveform detection,the waveform consistency of dual channel sensors at vehicle level and train level can be diagnosed and identified in real time,combined with the practical application of a EMU of 350 km/h.The application results show that the full redundancy technology can reduce the risk of false alarm of sensor fault,improve the reliability of the diagnosis system,and ensure the stable and safe operation of the train.
作者
董威
王云飞
张晓宁
朱慧龙
DONG Wei;WANG Yunfei;ZHANG Xiaoning;ZHU Huilong(CRRC Qingdao Sifang Co.,Ltd.,Qingdao,Shandong 266111,China)
出处
《控制与信息技术》
2020年第6期77-82,共6页
CONTROL AND INFORMATION TECHNOLOGY
基金
国家重点研发计划(2017YFB1201103)。
关键词
列车走行部监测
数据挖掘
全冗余
车地一体化
诊断模型
趋势预测
深度神经网络
循环神经网络
train running gear monitoring
data mining
full redundancy
train-ground integration
diagnosis model
trend prediction
deep neural network
recurrent neural network