摘要
针对S700K常见的8种故障模式和正常模式所对应功率曲线,提出一种基于概率神经网络(PNN)与改进的粒子群算法(PSO)相结合的道岔故障诊断方法。首先,在9种功率曲线上分别提取时域、频域特征统计量和时频域小波系数,并用主成分分析法降维每个域的特征量,得到特征向量;其次,以3个改进的PSO-PNN做分类器,并对分类器进行训练和预测;最后,3个分类器的预测结果做三取二表决。仿真结果表明:该方法能有效提高道岔故障诊断的准确率,具有良好的容错性。
Based on the power curves of the eight failure modes and normal mode common to the S700K,a switch failure diagnosis method was proposed based on Probabilistic Neural Network(PNN)combined with improved Particle Swarm Optimization(PSO).First,the time domain and frequency domain feature statistics and time-frequency domain wavelet coefficients were extracted on the nine power curves,and the feature quantity of each domain was reduced by Principal Component Analysis to obtain the feature vectors.Second,three improved PSO-PNN were used as a classifier,and the classifier was trained and predicted.Finally,the prediction results of the three classifiers were evaluated and voted to select two of them.The simulation results show that the method can effectively improve the accuracy of turnout failure diagnosis and has desired failure tolerance.
作者
孔令刚
焦相萌
陈光武
范多旺
KONG Linggang;JIAO Xiangmeng;CHEN Guangwu;FAN Duowang(Automatic Control Institute,Lanzhou Jiaotong University,Lanzhou 730070,China;Gansu Provincial Key Laboratory of Traffic Information Engineering and Control,Lanzhou 730070,China;National Engineering Research Center for Technology and Equipment of Environmental Deposition,Lanzhou Jiaotong University,Lanzhou 730070,China)
出处
《铁道科学与工程学报》
CAS
CSCD
北大核心
2020年第6期1327-1336,共10页
Journal of Railway Science and Engineering
基金
国家自然科学基金资助项目(61863024)
国家科技支撑计划资助项目(2014BAF01B00)
甘肃省科技计划资助项目(18JR3RA116)
甘肃省高等学校科研资助项目(2018C-11).
关键词
道岔故障诊断
S700K转辙机
概率神经网络
粒子群算法
三取二表决
turnout failure diagnosis
S700K switch machine
probabilistic neural network
particle swarm optimization
the vote for selecting two from three
作者简介
通信作者:孔令刚(1978−),男,安徽合肥人,副教授,从事铁路信号设备故障诊断研究,E−mail:konglinggang1978@163.com。
