Analog Circuit Fault Classification and Data Reduction Using PCA-ANFIS Technique Aided by K-means Clustering Approach

The paper work aims to extract effectively the fault feature information of analog integrated circuits and to improve the performance of a fault classification process. Thus, a fault classification method based on principal component analysis (PCA) and adaptive neuro fuzzy inference system classifier (ANFIS) preprocessed by K-means clustering (KMC) is proposed. To effectively extract and select fault features the traditional signal processing based on sampling technique conducts to different signature parameters. A stimulus pulse signal applied to the circuit under test (CUT) allowed us to get a reference output response. Respecting both specific sampling interval and step, the fault free and the faulty output responses are sampled to create amplitude sample features that will serve the fault classification process. The PCA employed for data reduction has lessened the computational complexity and obtaining the optimal features. Thus more than 75% of data volume decreased without loss of original information. The principal components extracted by this reduction data method have been input into ANFIS aided by KMC to obtain the best fault diagnosis results. The experimental results show a score of 100% diagnostic accuracies for the CUTs. Therefore, our approach has achieved best fault classification precision comparing to other research works.