一种电路测试的最优神经网络与遗传进化方法

A Neural Network and Genetic Evolution Approach for Digital Circuits Test

针对数字电路中信号线的单故障与多故障 ,本文首先建立电路测试的最优神经网络模型 ,然后设计一种两子群体竞争进化的遗传算法求网络能量函数的最小值点而获得给定故障的测试矢量 ,实验结果表明该方法能降低电路对应的神经网络规模并减少测试生成的计算量。

The optimal neural network model consists of m+n neurons for a logic circuit with m input and n output nodes. A new approach based on optimal neural network model and genetic algorithms for digital circuits test generation is presented, which mainly consists of two modules: the building of optimal neural network for circuit under test, and the finding minimum of the energy function of the network. The approach can be used to generate test patterns for single stuck-at fault and multiple faults, experimental results demonstrate the approach enhances the test generation efficiency based on Hopfied neural networks.