Application of Multi-frequency Test and Neural Network to Fault Diagnosis in Analog Circuits

摘要

In this paper, the multi-frequency test and neural networks (NNs) are applied to fault diagnosis in analog circuits. The reason is that multi-frequency test can maximize differences between the failure and the normal circuit's response, and NNs can solve complex classification problems. Firstly, using sensitivity analysis, the multi-frequency test vectors of the circuit under test(CUT) are generated. Then, selecting waveform amplitude, fault features of test points in CUT are extracted and fused. Last, NNs are used to classify the features for the faulty components detected and located. The experimental result shows that this approach is effective and practical for fault diagnosis in the analog circuits.