基于神经网络的MCM互连故障模拟器研究

A Study on MCM Interconnect Fault Simulator Based Upon the Neural Network

故障模拟器是基于模拟算法的测试生成系统的重要组成部分,其性能优劣直接影响测试集的各项指标。在基本互连结构的二值Hopfield神经网络模型基础上,提出了一种利用神经网络技术实现MCM互连故障模拟的方法。该方法以能量函数值反映故障状态的原理为基础,将判定测试矢量对故障的检测能力问题转化为能量函数值求解问题,简化了故障模拟的处理过程,提高了模拟效率。实验结果表明,用这一方法实现的验证性互连故障模拟器ICFSim,可以有效地对互连网络中的固定型故障和两两短路型故障进行快速模拟。

The fault simulator is an important part of a simulating based automatic test generating system,and its performance greatly affects the quality of the gained test vector set. Based upon the fundamental models of discrete Hopfield neural network for interconnects, a novel fault simulating method for MCM interconnects is put forward. Since the energy function value could reflect the faults status, the complicated decision toward testability of certain vectors is transferred by the method to calculating the corresponding energy function values of the faulty Hopfield neural network models, to make the fault simulation processing simpler and to acquire a better simulating efficiency. The result of the experiment proves the Hopfield neural network based fault simulating method an effective one.