摘要
当铁路货车通过道岔或高速行驶时,轮轨之间的相互作用会产生瞬间的强烈冲击,货车也会持续受到振动。这种振动和冲击会导致货车关键数据存在波动和噪声干扰,降低了不同传感器采集数据间的一致性,影响对货车运行安全状态的实时感知,对此,研究面向复杂工况的C70E型铁路货车运输安全状态感知方法。在C70E型铁路货车上布设多种传感器,包括风速传感器、载重传感器、振动传感器、速度传感器等,采集其复杂工况下的各种运输安全状态数据。通过深度改进的D-S证据理论算法实施多传感器信息融合,有效解决多传感器数据不一致问题。利用设计的集成卷积神经网络与长短期记忆网络的深度学习架构实现C70E型铁路货车运输安全状态分类。测试结果表明,设计方法能够实现复杂工况下多种运输安全的状态感知,及时发现潜在的安全隐患。随着列车运行次数的变化,设计方法的TPR一直高于0.975,FPR一直低于0.30。
When railway freight cars pass through turnouts or travel at high speeds,the interaction between the wheel and rail will produce a momentary strong impact,and the freight cars will continue to be subjected to vibration.This vibration and impact will lead to fluctuations and noise interference in the key data of freight cars,reduce the consistency between the data collected by different sensors,and affect the real-time perception of the safe state of freight cars.To this end,the C70E railway freight car transportation safety state perception method for complex working conditions is studied.Multiple sensors are installed on the C70E railway freight car,including wind speed sensors,load sensors,vibration sensors,speed sensors,etc.,to collect various transportation safety status data under complex working conditions.Implementing multi-sensor information fusion through a deeply improved D-S evidence theory algorithm effectively solves the problem of inconsistent multi-sensor data.Using a deep learning architecture that integrates convolutional neural networks and long short-term memory networks,the C70E railway freight car transportation safety status classification is implemented.The test results show that the design method can achieve state perception of various transportation safety under complex working conditions,and timely discover potential safety hazards.As the number of train runs changes,the TPR of the design method remains above 0.975 and the FPR remains below 0.30.
作者
李林俊
高峰
李金宝
LI Junlin;GAO Feng;LI Jinbao(Baotou Vehicle Maintenance Branch,Guoneng Railway Equipment Co.,Ltd.,Baotou,Inner Mongolia 014060;Tianjin Hveic Technologies Co.,Ltd.,Tianjin 301799,China)
出处
《自动化与仪器仪表》
2026年第2期135-139,共5页
Automation & Instrumentation
基金
国能铁路装备有限公司铁路安全科技创新项目(GJNY-22-121)。
关键词
复杂工况
C70E型铁路货车
运输安全
卷积神经网络
长短期记忆网络
状态感知
complex working conditions
C70E type railway freight car
transportation safety
convolutional neural network
long short-term memory network
state perception
