铁路货车空气制动系风压信号多特征提取方法  

Multi-Feature Extraction Method for Railway Air Brake System Wind Pressure Signal of Railway Freight Cars

作　　者：曹孟恺 岳建海[1] 胡准庆[1] CAO Meng-kai;YUE Jian-hai;HU Zhun-qing(School of Mechanical and Electronic Control Engineering,Beijing Jiaotong University,Beijing 100044,China)

机构地区：[1]北京交通大学机械与电子控制工程学院,北京100044

出　　处：《计算机仿真》2024年第12期177-182,233,共7页Computer Simulation

基　　金：3万吨列车制动试验台数据采集系统和制动缸状态诊断显示系统(M21L01550)。

摘　　要：空气制动系统是铁路货车安全运行的重要保障,其复杂的结构使系统故障与风压信号间的关系存在模糊性。基于3万吨列车制动试验台数据采集系统采集到的空气制动系统四通道风压信号数据集,提出了一种风压信号的多特征提取方法。将长时间序列风压信号划分为各制动周期风压信号作为数据样本,对各周期四通道信号分别提取时域、频域、时频域特征以及每两通道间的相关性特征。为防止高维特征造成维度灾难,采用随机森林方法计算各特征的重要性指标,选择重要性前10的特征组成敏感特征集。采用BP神经网络对敏感特征进行训练,通过五折交叉验证的方法在训练集和测试集上的准确率分别为100%和97.619%取得了较好的故障分类效果,验证了特征提取方法的有效性。The railway air brake system is an important guarantee for the safe operation of railway freight cars.Its complex structure makes the relationship between system failure and wind pressure signal ambiguous.Based on the four-channel air pressure signal data set of the railway air brake system collected by the data acquisition system of the 30000-ton train braking test bench,a multi-feature extraction method of air pressure signal is proposed.The longtime series pressure signals were divided into pressure signals of each braking cycle as data samples,and time-domain,frequency-domain,and time-frequency domain features,as well as correlation features between each pair of channels,were extracted for each cycle's four-channel signals.In order to prevent high-dimensional features from causing dimensional disasters,the random forest method was used to calculate the importance index of each feature,and the top 10 features were selected to form a sensitive feature set.The BP neural network was used to train sensitive features,and the accuracy of the 50-fold cross-validation method on the training set and test set was 100%and 97.619%,respectively,and achieving good fault classification results,verifying the effectiveness of the feature extraction method.

关 键 词：特征提取 特征选择 空气制动系统 风压信号 故障诊断 

分 类 号：U260.351[机械工程—车辆工程] TP206.3[交通运输工程—载运工具运用工程]