PSO优化LS-SVM在模拟电路故障预测中的应用

针对模拟电路故障预测的特点,提出一种基于PSO优化LS-SVM惩罚因子和核参数的模拟电路故障预测方法。利用小波包分解重构构造能量特征向量,通过计算PPMCC和欧氏距离来表征电路中元件的健康程度,定义为健康度,由此推导出电路发生故障时该元件的故障阈值。应用PSO优化的LS-SVM模型来实现模拟电路的故障预测,预测各个时间点的健康度变化轨迹并估计模拟电路的剩余寿命。通过仿真实验得知,该方法简单便捷,能够有效实现模拟电路的故障预测,具有较好的实用性。

Analog circuit fault prediction based on LS-SVM optimized by PSO

A new method of analog circuit fault prediction is put forward based on the LS-SVM optimized by PSO for penalty factor and kernel parameter, in order to adapt the characteristic of the analog circuit fault prognostics. The wavelet packet decomposition can reconstruct feature vector based on energy information. The element fault threshold is derived from calculating Pearson product-moment correlation coefficient(PPMCC) and Euclidean distance to represent circuit components of well-being defined as Health degree. This paper applies PSO optimization principle to the LS-SVM to exactly realize fault forecast of analog circuit; moreover, it can predict the changes of Health degree at each point and estimate residual life of analog circuit. Taking Multisim simulation circuit as an example, the method is not only easy and convenient, but also have batter practicability to realize fault prediction of analog circuit.