ANALOGUE CIRCUIT FAULT DIAGNOSIS METHOD BASED ON GENERALIZED MULTIPLE KERNEL LEARNING-SUPPORT VECTOR MACHINE

摘要

An analogue circuit fault diagnosis method based on a generalized multiple kernel learning-support vector machine (GMKL-SVM), comprising the following steps: (1) collecting a time domain response signal of an analogue circuit, i.e. collecting an output voltage signal of the analogue circuit; (2) performing wavelet transform of the collected voltage signal, calculating the energy of a wavelet coefficient as a characteristic parameter, the set of all the characteristic parameters being sample data; (3) applying, based on the sample data, PSO to optimize a regularization parameter and a compromise parameter of the generalized multiple kernel learning-support vector machine, and constructing a GMKL-SVM-based fault diagnosis model; and (4) using the constructed GMKL-SVM-based fault diagnosis model as a classifier to diagnose a fault of the analogue circuit. The classification property of the GMKL-SVM in the present invention is better than those of other classification algorithms, and the method of applying PSO to optimize the GMKL-SVM parameters is also better than the conventional parameter acquisition methods, being capable of effectively detecting an element fault of the analogue circuit.