基于IPSO-SVM的地铁车辆牵引控制单元故障诊断

地铁车辆牵引控制单元(TCU)是地铁系统的核心单元之一,准确诊断其故障状态对整个地铁车辆安全运行至关重要.基于数据的故障诊断方法是当前热点方法之一.针对牵引控制单元故障诊断中检测参数多、故障类别多的特点,提出了改进的粒子群优化支持向量机(IPSO-SVM)方法,克服了传统方法存在过拟合、收敛速度慢、易陷入局部最优的缺点.使用UCI机器学习数据库中的5个数据集进行仿真实验,结果表明:IPSO-SVM分类精度高于ICPSO-SVM、PSO-SVM、GA-SVM.进一步将此方法应用于地铁车辆实际数据,同样得到了较好的分类结果,验证了所提方法的有效性.

Fault diagnosis of metro vehicle traction control unit based on IPSO-SVM

Metro vehicle traction control unit is one of the core units of the subway system. Accurate diagnosis of the fault status is very important to the safety running of whole metro vehicle. Data-driven fault diagnosis method is one of current hot methods. Based on the characteristics of multi-parameter and multi-category in traction control unit fault diagnosis, a method of support vector machine (SVM) optimized by improved particle swarm optimization (IPSO) is proposed to overcome the shortcomings of traditional methods, such as overfitting, slow convergence speed and easily being trapped into local optimal solution. Simulation experiments are carried out on five datasets from UCI machine learning repository. The simulation results show that the classification accuracy of IPSO-SVM is higher than that of ICPSO-SVM, PSO-SVM and GA-SVM. Then, this method is applied to metro vehicle actual data, also gets a better classification result, which verifies that IPSO-SVM is an effective method.