基于粒子群优化的LSSVM在模拟电路故障诊断中的应用

Method of LSSVM optimized by particle swarm and its application in fault diagnosis of analog circuit

提出一种基于粒子群优化的最小二乘支持向量机的模拟电路故障诊断新方法。对模拟电路故障信号采用小波包进行消噪和分解,作为最小二乘支持向量机的输入样本。为了避免参数优化时容易陷入局部最优的缺陷,使用粒子群算法对LSSVM参数进行优化选取。以Sallen-key带通滤波器电路为对象的仿真研究结果表明,提出的基于粒子群优化的最小二乘支持向量机可以对模拟电路有效地进行故障诊断,并且提高了诊断效率。

Based on the LSSVM optimized by particle swarm colony, a new analog circuit diagnosis method is proposed . Using the wavelet packet , this method eliminates the noise and makes the decomposition of fault signals of the analog circuit , then the feature data is organized as input of LSSVM. To avoid the local optimum in parameter optimization , particle swarm optimization method is used in optimizing the selection of parameter of LSSVM. The simulation results of fault diagnosis in Sallen-key band-pass filter shows , the LSSVM fault diagnosis method optimized by particle swarm could diagnose the fault of analog circuit effectively , and it increases the efficiency sharply.