基于自适应变异SAPSO-LSSVM的轨道电路故障诊断

Track circuit fault diagnosis based on adaptive mutation SAPSO-LSSVM

针对传统无绝缘轨道电路故障诊断精度不高与诊断结果不稳定的问题,提出一种将模拟退火(Simulated Annealing,SA)算法和粒子群最小二乘支持向量机(PSO-LSSVM)相结合的方法,并用其进行故障诊断.借鉴遗传算法(Genetic Algorithm,GA)中的变异思想,在SA算法中引入简单变异算子,提出动态自适应变异SAPSO算法,以克服传统粒子群算法易陷入局部最优的问题.利用自适应变异SAPSO算法对改进LSSVM核函数的惩罚因子和核函数参数进行优化,避免故障诊断结果不稳定.最后通过仿真将本文算法与近几年比较先进的几种算法进行对比.仿真结果表明:本文算法对无绝缘轨道电路的故障诊断准确率更高、诊断结果更稳定.

Aiming at the problems of low fault diagnosis accuracy and unstable diagnosis results of traditional uninsulated track circuits,a method combining Simulated Annealing(SA)algorithm and Particle Swarm Optimization Least Squares Support Vector Machine(PSO-LSSVM)is proposed for fault diagnosis.This paper draws on the Genetic Algorithm(GA)mutation idea.Firstly,a simple mutation operator is introduced in the SA algorithm,and a dynamic adaptive mutation SAPSO algorithm is proposed to overcome the problem that the traditional particle swarm algorithm is prone to fall into local optimization.Secondly,the adaptive mutation SAPSO algorithm is used to optimize the penalty factor and kernel function parameters of the improved LSSVM kernel function to avoid unstable fault diagnosis results.Finally,the algorithm is compared with several advanced algorithms in recent years by simulation.The simulation results show that the proposed algorithm has higher accuracy and more stable diagnosis results on the fault diagnosis of uninsulated track circuits.