地铁列车空调系统的BPNN故障诊断模型研究  

BPNN fault diagnosis model of air-conditioning system for subway train

作　　者：杨闯 陈焕新[1] 程亨达 杨宇 梁宇星 Yang Chuang;Chen Huanxin;Cheng Hengda;Yang Yu;Liang Yuxing(School of Energy and Power Engineering,Huazhong University of Science and Technology;Guangzhou Dinghan Railway Vehicles Equipment Co.,Ltd.)

机构地区：[1]华中科技大学能源与动力工程学院 [2]广州鼎汉轨道交通车辆装备有限公司

出　　处：《制冷与空调》2023年第8期71-77,共7页Refrigeration and Air-Conditioning

基　　金：国家自然科学基金项目(51876070):大数据构架下模式辨识和集成学习的制冷空调故障诊断方法研究

摘　　要：地铁列车空调系统对调节车厢温湿度极为重要,其发生故障会降低车厢舒适度并增大能耗。构建制冷剂欠充、新风脏堵与冷凝器脏堵的单故障及并发故障模拟试验,并基于随机森林对采集数据进行特征选择,有效降低数据集的维度,最后基于BPNN构建单故障诊断模型和并发故障诊断模型。结果表明,构建的单故障诊断模型和并发故障诊断模型能够有效识别故障类型及故障等级,精度分别达到99.92%和98.63%。The air-conditioning system of subway train is extremely important for regulating the temperature and humidity of carriages,and its fault will reduce the comfort of carriages and increase energy consumption.The simulation experiments on single fault and concurrent fault of refrigerant undercharge,fresh air dirty plugging and condenser dirty plugging are constructed.The feature selection of the collected data based on random forest is carried out to effectively reduce the dimension of the dataset.Finally,the single fault and concurrent fault diagnosis models are constructed based on BPNN.The results show that the constructed single fault diagnosis and concurrent fault diagnosis models can effectively identify fault types and fault levels,and the accuracy reaches 99.92%and 98.63%,respectively.

关 键 词：地铁列车空调系统 故障诊断 BP神经网络 单故障 并发故障 

分 类 号：TB6[一般工业技术—制冷工程]