Quality improvement of analog circuits fault diagnosis based on ANN using clusterization as preprocessing

The technique of improvement the fault diagnosis quality of analog circuits using artificial neural network is proposed. The technique is based on combining the methods of clustering and classification the output circuits responses taking into consideration component tolerances at data preparation and training the ANN for solving task of fault diagnosis. The decomposition of technique with description of each step is presented. The results of experimental investigation demonstrating high quality of ANN training for effective fault diagnosis of analog circuits with low probability of α- and β-errors are considered.