论文部分内容阅读
针对模拟电路多频测试问题,提出了一种基于类间马氏距离的测试激励优选方案。通过幅频特性分析,确定频率优选范围。根据参数扫描分析数据,计算不同激励下故障特征样本集的类间马氏距离,并以此为适应度函数,利用离散粒子群算法实现激励优选。相比于传统的欧氏距离,马氏距离不受量纲的影响。仿真结果表明,该方法优选出的测试激励信号降低了故障间的模糊度,提高了故障诊断率。
Aiming at the multi-frequency test problem of analog circuit, a test excitation optimization scheme based on Mahalanobis distance between classes is proposed. Through the analysis of amplitude-frequency characteristics, determine the frequency preferred range. According to the data of parametric sweep analysis, the Mahalanobis distance between different classes of fault feature samples under different excitation is calculated, and as a fitness function, the discrete particle swarm optimization algorithm is used to optimize the excitation. Compared with the traditional Euclidean distance, Mahalanobis distance is not affected by the dimension. The simulation results show that the test excitation signals optimized by this method can reduce the ambiguity among faults and improve the fault diagnosis rate.