Neural Network Method of Test Pattern Generation and Some Problems in Design

A test pattern generation system implemented with neural network method is introduced in this paper.The system consist of circuit description,preprocessing,the build of neural network,test pattern generation and test informations module.A lot of strategies are discussed to improve the performances of the method,such as the selection of constrained network that accords with the circuit under test,evaluating the values of neurons,circuit partition,the selection of optimization algorithms.Some problems in the test pattern generation method based on neural network is presented and discussed.The experimental results indicate the test system is effective.