Improvement and Application of a Real-timing Analog Circuit Fault Diagnosis Method

The paper first makes a thorough research on the method for analog circuit fault diagnosis based on kurtosis and negentropy, and then theoretically analyses it´s advantage and disadvantage, which is followed by introducing the idea of centroid to overcome the method´s shortcoming, making the improved method can extract the signal´s feature more efficiency. Finally, it applies the improved method to an actual circuit. The simulation result shows that the improved method not only improves the fault diagnosis ratio of neural network but also can be used in the real-timing condition such as online fault diagnosis.