Feed forward neural network approach for diagnosing the faulty functioning in digital circuits

This study explains a technique for modeling multiple faults in digital circuits developed on the basis of Neural Network. This study is performed by collecting information from models expressed in Verilog hardware description language. The method uses various quantities of chosen circuit test data, which is generated by introducing single stuck-at faults into the sequential circuit; the observed test vectors which is generated by using Design for Testability (DFT) are then applied to train models of neural network expressed in MATLAB. The neural network models trained are capable of replicating circuit behavior in the presence of faults. The research is based on a benchmark ISCAS-S27 sequential circuit. This approach generates more accurate models than previously published methods. In the presence of faults, the Artificial Neural Network (ANN) is able to replicate the circuit output