摘要
提出一种基于多尺度小波分解及神经网络映射归纳的测试电流模电路故障缺陷的方法.针对CMOS器件典型故障建立了测试所需的故障模型,给电路节点加入故障模型进行故障响应测试.对故障信号进行时域采样,采用小波多尺度分解对故障相应信号进行频域多尺度分解,然后将处理数据作为神经网络训练样本,对各类缺陷响应结果进行分类、识别,最后根据可接受偏差范围确定信号为故障或非故障.给出了6类故障的故障覆盖率测试结果.
An efficient defect-oriented parametric test method for current model circuits using Wavelet Neural Networks (WNN) is proposed, in which the nonlinear mapping, Neural Network generalizing and multiple-scale analysis are adopted. Aiming at the typical fault cases of CMOS, the required fault model is established and be joined for fault emulational test. The output was sampled in time domain, wavelet multi-scale decomposition is used for various response data preprocessing and then the Neural network is used to classificat response results of different defects. In the end, signal fault or fault is determined according to the acceptable tolerances. Six kinds of fault coverage are demonstrated.
出处
《湖南师范大学自然科学学报》
CAS
北大核心
2014年第2期51-55,共5页
Journal of Natural Science of Hunan Normal University
基金
国家杰出青年科学基金资助项目(50925727)
国防预研重大基金资助项目(C1120110004)
中国博士后科学基金资助项目(2013M541819)
湖南省科技计划资助项目(2010J4
2011JK2023)
关键词
电流模
测试
小波分解
神经网络
current model
test
wavelet decomposition
neural networks (NN)