A Vague Decision Method for Analog Circuit Fault Diagnosis Based on Description Sphere

Abstract

This paper proposes a vague decision method for analog circuit fault diagnosis based on description sphere. Firstly, the proposed method uses the wavelet transform as the preprocessor to extract fault features from the output voltages of the circuit under test (CUT). And then, each class sample is trained to produce a minimum description sphere. Finally, the test samples are detected by a defined vague decision rule, which is based on the vague weight distance between the test data and the center of description sphere. The defined decision rule fuses the truth and false membership degrees of the test sample and the weight of the description sphere, and it can effectively deal with the uncertain information. The reliability of the defined decision rule is proved theoretically. This new diagnostic method is first applied to testing two actual circuits, and then it is compared with other two diagnostic methods. The experimental results show that the proposed technique can achieve good performance and reduce the diagnostic time.