基于LMD算法的大规模模拟电路软故障诊断仿真

Simulation of large scale Analog Circuit Soft Fault Diagnosis Based on LMD Algorithm

随着电路规模的不断扩展,大规模模拟电路的故障诊断越来越难以实现,尤其是软故障的诊断,由于其存在不可数特性,导致当前缺乏对软故障的有效诊断方法,因此提出了一种基于改进LMD算法的大规模模拟电路软故障诊断方法。根据电路节点电压信号的变化及概率密度,判断电路发生的故障类型,同时利用分量积分推断故障概率。考虑到LMD在非线性信号分析方面的优势,对电路测点信号采取LMD分解,并针对LMD平滑过程中的相位变化引入幂函数插值优化,降低包络分离时的误差。通过信息熵与分形维数,将相关性与门限判断结合,实现电路故障信号与故障类型的对应。仿真过程中,对于大规模模拟电路采用拓扑分解策略划分为非相关子区域,根据测点信号和故障状态的相关性计算,得到故障类型判断均正确;经过与其它方法对比,验证了所提方法可以有效提高大规模模拟电路软故障诊断的精度和效率。

With the continuous expansion of circuit scale, fault diagnosis of large-scale analog circuits is becoming more and more difficult, especially the diagnosis of soft fault, due to its uncountable characteristics, there is a lack of effective diagnosis method for soft fault. Therefore, a soft fault diagnosis method for large-scale analog circuits is proposed based on improved LMD algorithm. According to the change of voltage signal and probability density of circuit node, the fault type of circuit was judged, at the same time, the component integral was used to infer the fault probability. Considering the advantage of LMD in nonlinear signal analysis, LMD was adopted to decompose the circuit test point signal. In order to reduce the error of envelope separation, the power function interpolation optimization was introduced for the phase change in LMD smoothing process. Through information entropy and fractal dimension, the correlation and threshold judgment were combined to realize the correspondence between circuit fault signal and fault type. In the process of simulation experiment, the large-scale analog circuit was divided into uncorrelated sub regions based on topology decomposition strategy. According to the correlation calculation between the signal of the measuring point and the fault state, the judgment of the fault type is correct;compared with other methods, it was verified that the proposed method can effectively improve the accuracy and efficiency of large-scale analog circuit soft fault diagnosis.