基于MSIF-CNN的地铁车辆制动系统故障诊断方法  

Method of MSIF⁃CNN⁃based fault diagnosis for subway vehicle braking system

作　　者：陈岩霖 孙庚 汪敏捷 贺鑫来 翟逸男 尹娴 冯艳红 CHEN Yanlin;SUN Geng;WANG Minjie;HE Xinlai;ZHAI Yinan;YIN Xian;FENG Yanhong(College of Information Engineering,Dalian Ocean University,Dalian 116023,China)

机构地区：[1]大连海洋大学信息工程学院,辽宁大连116023

出　　处：《现代电子技术》2024年第24期137-142,共6页Modern Electronics Technique

基　　金：大连海洋大学科研项目:轨道列车智能运维管理平台(2023001)。

摘　　要：研究地铁车辆制动系统的故障诊断对保障交通安全、提高运营效率具有重要意义。针对当前的制动系统故障诊断研究存在过度依赖于专家的知识经验、数据融合效率不高以及现有模型训练参数过多的问题,提出了一种基于多传感器信息融合和改进卷积神经网络的“端到端”制动系统故障诊断方法。该方法不需要专家知识对数据进行特征提取,而是利用一维卷积神经网络(1D-CNN)来处理多传感器信息融合问题,并引入一维全局平均池化层(1D-GAP)改进神经网络结构,以减少模型训练参数。最终利用极端梯度提升模型(XGBoost)作为分类判别器,以提高故障诊断的准确性。实验结果表明,所提方法的准确率、精确率、召回率和F1值分别为95.86%、96.59%、92.68%和93.15%,同时,在地铁车辆制动系统故障诊断方面展现了更优的性能。The research on fault diagnosis of subway vehicle braking systems is of great significance for ensuring traffic safety and operational efficiency.In allusion to the problems in current research on brake system fault diagnosis,such as excessive reliance on expert knowledge and experience,low data fusion efficiency,and excessive training parameters of existing models,a method of"end-to-end"brake system fault diagnosis based on multi-sensor information fusion and improved convolutional neural network(MSIF-CNN)is proposed.This method does not require expert knowledge to extract features from data.It can utilize the one-dimensional convolutional neural network(1D-CNN)to handle multi-sensor information fusion problem and introduce the one-dimensional global average pooling layer(1D-GAP)to improve the neural network structure and reduce model training parameters.The extreme gradient Boost(XGBoost)model is utilized as a classification discriminator to improve the accuracy of fault diagnosis.The experimental results show that the accuracy,precision,recall,and F1 values of the method are 95.86%,96.59%,92.68%,and 93.15%,respectively,demonstrating better performance in the fault diagnosis of subway vehicle braking systems.

关 键 词：多传感器信息融合 卷积神经网络 地铁车辆 制动系统 故障诊断 XGBoost 

分 类 号：TN911.22-34[电子电信—通信与信息系统] TP391.5[电子电信—信息与通信工程]