基于神经网络的组合电路测试生成算法

介绍了一种基于神经网络的组合电路测试生成算法。该算法不同于传统的方法是它既不需要回退也不需要故障传播的过程。利用Hopfield神经网络模型将组合电路表示成双向的神经网络，通过故障注入，建立被测电路的约束网络，并构造网络的能量函数，将组合电路的测试矢量对应于神经网络能量函数的最小值点，从而运用遗传算法求解能量函数的最小值点来求得测试矢量。在一些基准电路上的实验结果表明本算法具有较高的故障覆盖率和较短的测试时间。

Neural networks based test generation algorithm for combinational logic circuits

A new test generation algorithm based neural networks which does not need propagation and backtracks is proposed to represent the combinatory circuits as a bidirectional network of neurons using the Hopfield nets. A constraint network of the circuits under test was constructed through a fault injection, and an energy function was also constructed with global minima at test vectors. A genetic algorithm was used for test generation, and the experimental results on some combinational circuits demonstrate the feasibility of this algorithm.