高斯模糊核聚类的模拟电路在线故障诊断算法

为提高模拟电路故障在线诊断的运算速度与可靠性,采用高斯模糊核聚类算法对模拟电路故障进行非监督学习。该故障诊断算法的关键是利用已知故障数据类中心点确定故障类。利用模糊核聚类的高效识别树型结构减少训练样本规模、处理模糊类中的野值点,以提高分类器的训练速度和精确度。根据每一类故障数据得到的故障参数均值,设其为故障判断阈值,并赋予类标。在三种不同故障条件下,对Sallen-Key低通滤波器电路进行故障诊断的仿真实验。结果表明：与RBF监督学习方法相比,β-MKFCM方法能够高效地辩识已知故障与未知故障。该研究为电路在线故障诊断提供了参考依据。

Analog circuit faults diagnosis online based on gaussian fuzzy kernel clustering

Aimed at improving the arithmetic speed and the evaluation reliability of the analog circuit fault diagnosis online, this paper proposes unsupervised learning of analog circuit faults diagnosis, a no- vel strategy designed for faults diagnosis, based on Gaussian fuzzy kernel clustering algorithm. The key to the faults diagnosis lies in confirming the fault category by utilizing the known fault data. The strategy consists of decreasing the training sample and eliemating wild value by using the high-performance recog- nition tree structure of fuzzy kernel clustering, improving the training speed and precision of classifier, obtaining the mean value depending on the faults data of each class, and setting this mean values as the thresholds for judging faults and issuing each data point with a class label. The simulation of the fault di- agnosis of the Sallen-Key low pass filters subjected to three different faults reveals that the proposed meth- od, capable of more effectively recognizing the known and unknown faults than does RBF fault diagnosis method, serves as a reference for the circuit faults diagnosis online.