An approach to construction the neuromorphic classifier for analog fault testing and diagnosis

Testing and diagnosis of analog circuits are very important tasks at the quality assurance of integrated circuits and electronic devices. Faults detection and identification are realized using fault dictionary. The architecture of fault dictionary has an essential influence on time and efficiency of diagnosis at whole. An approach to the construction of fault dictionary as the neuromorphic classifier for analog fault testing and diagnosis is proposed. The approach takes into account the component tolerances, includes the faults clustering as the preprocessing and selection the essential characteristics of CUT´s output responses providing the maximum distinguishability between all fault clusters. The experimental results were obtained for second-order bandpass filter and are presented in the paper for demonstrating the proposed approach.