The nature of nonlinear analog circuit intelligent fault diagnosis is patternrecognition, which the key is feature extraction of the circuit. Diagnosis methodbased on functional series and artificial intelligence has a great practicability, soit is taken seriously. The method based on Wiener functional series and Volterrafunctional series is widely used in actual application. It is difficult to obtainVolterra kernel, and has the matters of test waveform slections. Wiener series is akind of orthogonal expansion, it only needs white gaussian noise as excitationsignal, and the acquisition of Wiener kernel is relatively easy. So the method ofnonlinear analog circuit fault diagnosis based on Wiener kernel is used in thispaper.In order to raise the efficiency of fault diagnosis, this paper studied faultdiagnosis method based on Wiener kernel and neural network, then determineddiagnosis steps. To obtain Wiener kernel, this paper studied discrete Wienerkernel acquisition method and Wiener kernel indirect acquisition method. Aimingat the key point in diagnosis process, this paper studied SA and PSO, thenproposed Improved Particle Swarm Simulated Annealing Hybrid OptimizationAlgorithm and it was validated by simulation. The simulation results show thatits performance is improved significantly. We studied the feature extractionmethod of Wiener kernel based on this algorithm. Simulation results show thatthe dimension of Wiener kernel’s characteristic parameters is reduced, itscharacteristics are compressed, and the accuracy of fault diagnosis is improvedafter extracted.When making theoretical research, this paper designed fault diagnosticinstrument based on DSP and diagnostic applications based on PC. On this basis, formed fault diagnosis system based on Wiener kernel. The fault diagnosismethod studied in this paper is validated by this system, experiments proves itsfeasibility. |