基于KPCA-SVM的S700K转辙机故障诊断方法  

Fault diagnosis method of S700K switch machine based on KPCA-SVM

作　　者：张友鹏[1] 魏智健 杨妮[1] 张迪 ZHANG Youpeng;WEI Zhijian;YANG Ni;ZHANG Di(School of Automation and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]兰州交通大学自动化与电气工程学院,兰州730070

出　　处：《安全与环境学报》2023年第9期3089-3097,共9页Journal of Safety and Environment

摘　　要：针对S700K转辙机动作功率曲线非线性特征多样化、复杂化的特点,提出了一种基于核主成分分析(Kernel Principal Component Analysis,KPCA)和支持向量机(Support Vector Machine,SVM)的智能故障诊断方法。首先,对S700K转辙机的功率曲线进行分析,研究正常曲线变化规律,总结常见故障类型功率曲线的变化现象和故障原因。然后,从功率曲线中提取10种时域特征值组成初始特征数据集,用KPCA算法将特征数据映射到高维特征空间中对其进行PCA降维,得到故障样本的非线性主成分。最后,将得到的非线性主成分作为多分类SVM的输入样本进行故障模式识别。采用粒子群优化(Particle Swarm Optimization,PSO)算法分别对核函数参数和SVM惩罚因子进行优化,提高模型的诊断精度。仿真结果表明,该模型能够有效提取转辙机故障信号的非线性特征,故障诊断精度达到97%,诊断时间较短,适用于准确性、实时性要求更高的提速道岔。Aiming at the diversified and complex nonlinear characteristics of the S700K switch machine action power curve,an intelligent fault diagnosis method based on kernel principal component analysis(KPCA)and support vector machine(SVM)is proposed in this paper.Firstly,we analyze the normal power curve of the S700K switch machine,study the change law of the normal curve,and summarize the change phenomena and causes of common fault type.Then 10-time domain eigenvalues are extracted from the power curve to form the initial feature data set.The feature data are mapped into the high-dimensional feature space by the KPCA algorithm,the PCA dimensionality reduction is carried out,and the cumulative contribution rate is set to 95%.We extract the nonlinear principal components of the fault sample data,and then sort according to the contribution rate from large to small,and combine them into a new dimensionality reduction matrix.The dimensionality reduction matrix is projected into three-dimensional space for visual display,and the dimensionality reduction visualization effects after dimensionality reduction by different algorithms are compared.Secondly,theobtained nonlinear principal components are used as the input samples of SVM for fault pattern recognition.According to the number of fault types diagnosed,the same number of SVM trainers are trained,and finally,a multi-classification SVM classifier is made up of these trainers.PSO algorithm is used to optimize the kernel function parameters and SVM penalty factors of the SVM classifier to improve the diagnosis accuracy of the model.The simulation results show that the model can effectively extract the nonlinear characteristics of the fault signal of the switch machine and identify the fault pattern of the sample data.The fault diagnosis accuracy is 97%,which is better than other traditional diagnosis methods,and the diagnosis time is shorter than other methods.It is suitable for speed-up turnout with higher fault maintenance accuracy and real-time requirements.

关 键 词：安全工程 S700K转辙机 故障诊断 核主成分分析 粒子群优化算法 支持向量机 

分 类 号：X951[环境科学与工程—安全科学]