Particle Swarm Neural Network in Application of Soft Fault Diagnosis of Analog Circuit

In order to diagnose single soft fault in analog circuit, the particle swarm neural network was applied to fault diagnosis of analog circuit. The particle swarm neural network training process was divided into two steps. Firstly, BP network weights and threshold values as the position vector of the particle, used PSO algorithm searches for a near-optimal position vector as BP neural network initial weight values and thresholds. Secondly, used the BP algorithm to further optimization based on the initial weights and thresholds, got the optimal network weights and threshold value. The training method can improve the convergence accuracy and learning speed of the network training. The simulation results show that this diagnostic method can effectively achieve the accurate diagnosis of analog circuit soft fault.