摘要
在综合阐述目前各种模拟集成电路故障诊断方法的基础上,针对中小规模模拟集成电路的软故障诊断问题,提出结合小波分析、聚类分析和神经网络的动态电流测试方法。采用动态电流测试方法的同时,对电流中的稳态部分与瞬态部分进行分析处理,提高故障的覆盖率。首先对待测电路的动态电流信号进行小波分解,将各频段相对能量作为故障特征向量;然后对得到的故障特征向量样本集进行聚类分析,改善样本质量,并将聚类后的样本集输入神经网络进行训练,提高神经网络的学习效率。对实例中小规模模拟电路进行软故障诊断仿真实验,取得了较高的故障识别率,验证了所提方法的有效性。
On the basis of comprehensively stating current various fault diagnosis methods for the analog integrated circuit, this paper presents a kind of dynamic current testing method combining wavelet analysis, clustering analysis and the neural network to solve the problem of soft fault diagnosis of the small-scale analog integrated circuit. At the same time of using the dynamic current testing method, it makes analysis and process for the steady-state part and instantaneous part in the current so as to improve fault coverage rate. It firstly conducts wavelet decomposition for the dynamic current signals of the measured circuit and takes relative energy of each frequency band as the fault characteristic vector. Then by means of clustering analysis for the sample set of fault characteristic vector, it has improved sample quality. By inputting samples after clustering into the neural network for training, it could improve study efficiency of the neural network. By means of simulation experiment for soft fault diagnosis on the small-scale analog circuit, it has obtained higher fault identification rate and verified effectiveness of the proposed method.
作者
彭泽武
黄剑文
冯歆尧
李奇远
李宁
汪樟垚
PENG Zewu;HUANG Jianwen;FENG Xinyao;LI Qiyuan;LI Ning;WANG Zhangyao(Guangdong Power Grid Co., Ltd., Guangzhou, Guangdong 510620, China;College of Electrical and Information Engineering, Hunan University, Changsha 410082, China)
出处
《广东电力》
2019年第2期57-64,共8页
Guangdong Electric Power
基金
广东电网有限责任公司科技项目(000000KK52160001)
湖南省自然科学基金杰出青年基金项目(2017JJ1011)
关键词
模拟电路
电流测试
软故障诊断
小波分析
聚类分析
神经网络
analog circuit
circuit testing
soft fault diagnosis
current testing
wavelet analysis
clustering analysis
neural network