基于Elman神经网络的模拟电路故障诊断研究

doi:10.16180/j.cnki.issn1007-7820.2019.06.006

摘要/Abstract

摘要：

针对模拟电路的软故障,文中提出了一种基于改进Elman神经网络与提高特征向量有效性相结合的诊断方法。该方法对不同情况下的输出信号进行3次小波分析,形成8维的特征向量,再与改进的Elman神经网络结合进行分类与诊断。将改进Elman神经网络应用于非线性模拟电路故障诊断中可提高其诊断率与分类率。文中对其诊断方法进行了实验对比测试,结果表明,该方法提高了诊断性能,其诊断率与分类率分别为92.5%和83%。

关键词: 软故障, Elman神经网络, 特征向量, 小波分析, 故障诊断, 分类率

Abstract:

Aiming at the soft fault of analog circuits, a diagnosis method based on improved Elman neural network combined with the improvement of feature vector effectiveness was proposed. This method performed three wavelet analysis on the output signals under different cases to form an eight-dimensional feature vector, which was combined with the improved Elman neural network for classification and diagnosis. Applying the improved Elman neural network to the fault diagnosis of nonlinear analog circuits could improve its diagnostic rate and classification rate.This paper had carried out experimental comparison tests on the diagnostic methods.The results showed that the method improved the diagnostic performance, of which the diagnostic rate and classification rate were 92.5% and 83%,respectively.

Key words: soft fault, Elman neural network, feature vector, wavelet analysis, fault diagnosis, classification rate

中图分类号:

TP183

孙李阔,王代强. 基于Elman神经网络的模拟电路故障诊断研究[J]. 电子科技, 2019, 32(6): 26-30.

SUN Likuo,WANG Daiqiang. Research on Fault Diagnosis of Analog Circuit Based on Elman Neural Network[J]. Electronic Science and Technology, 2019, 32(6): 26-30.

图/表 6

图1

图2

图3

表1

表2

表3

参考文献 18

[1]	高隽 . 人工神经网络原理及仿真实例[M]. 北京: 机械工业出版社, 2007.
	Gao Wei. Principles and simulation examples of artifi-cial neural networks[M]. Beijing: Mechanical Industry Press, 2007.
[2]	黄丽霞, 肖顺根, 宋萌萌 . 小波包和改进Elman神经网络相融合的异步电动机滚动轴承的故障诊断[J].现代制造工程, 2014(7):127-131.
	Huang Lixia, Xiao Shungen, Song Mengmeng . Fault diagnosis of asynchronous motor rolling bearing based on synergetic wavelet packet and Elman neural network[J].Modern Manufacturing Engineering, 2014(7):127-131.
[3]	Venu P, Madhava Rao, Sundari K Siva . Optimized mu-lti frequency approach to analog fault diagnosis using monte carlo analysis[J].Electrical and Electronic Engineering, 2014(2):25-30.
[4]	戚新波, 刘宏飞, 郑先锋 . 电路的计算机辅助分析——MATLAB与P-Spice技术[M]. 北京: 电子工业出社, 2006.
	Qi Xinbo, Liu Hongfei, Zheng Xianfeng. Computer aided analysis of circuits——MATLAB and P-Spice technology[M]. Beijing: Electronics Industry Press, 2006.
[5]	王浩天, 单甘霖, 段修生 . 基于LM-BP神经网络算法的模拟电路故障诊断[J]. 计算机测量与控制, 2013,21(12):3197-3200.
	Wang Haotian, Shan Ganlin, Duan Xiusheng . Analog circuit fault diagnosis based on LM-BP neural net-work algorithm[J]. Computer Measurement & Controll, 2013,21(12):3197-3200.
[6]	潘强, 孙必伟 . 模拟电路故障诊断中的特征信息提取[J]. 电子科技, 2013,26(8):116-119.
	Pan Qiang, Sun Biwei . Feature information extraction in fault diagnosis of analog circuit[J]. Electronic Science and Technology, 2013,26(8):116-119.
[7]	韩涵, 王厚军, 龙兵 , 等. 基于改进马氏距离的模拟电路故障诊断方法[J]. 控制与决策, 2013,28(11):1713-1717.
	Han Han, Wang Houjun, Long Bing , et al. Method for analog circuit fault diagnosis based on improved Mahalanobis distance[J]. Control and Decision, 2013,28(11):1713-1717.
[8]	黄如科, 梁天生, 余汉华 . 基于小波包变换与BP神经网络的模拟电路故障诊断[J]. 电子科技, 2014,27(8):46-49.
	Huang Ruke, Liang Tiansheng, Yu Hanhua . Analog circuit fault diagnosis based on wavelet packet transform and BP neural network[J]. Electronic Science and Technology, 2014,27(8):46-49.
[9]	喻捷, 严太山, 曾凡勇 . 一种基于神经网络的模拟电路故障诊断方法[J].电子技术, 2014(11):71-73.
	Yu Jie, Yan Taishan, Zeng Fanyong . An analog circuit fault diagnosis method based on neural network[J].Electronics Technology, 2014(11):71-73.
[10]	Mahdieh Jahangiri, Farhad Razaghian . Fault detection in analogue circuits using hybrid evolutionary algori-thm and neural network[J]. Analog Integrated Circuits and Signal Processing, 2014,80(3):551-556. doi: 10.1007/s10470-014-0352-7
[11]	王福涛 . 基于神经网络样本的模拟电路故障诊断研究[J].数字技术与应用, 2015(8):92-97.
	Wang Futao . Research on fault diagnosis of analog circuit based on neural network samples[J].Digital Technology and Application, 2015(8):92-97.
[12]	Han B, Li J, Wu H . Diagnosis method for analog cir-cuit hard fault and soft fault[J]. Telkomnika Indonesian Journal of Electrical Engineering, 2013,11(9):77-84.
[13]	周晶晶, 程惠华, 安明 , 等. 基于神经网络方法的模拟电路故障诊断研究[J]. 现代电子技术, 2015,38(23):47-50.
	Zhou Jingjing, Cheng Huihua, An Ming , et al. Research on fault diagnosis of analog circuits based on neural network method[J]. Modern Electronic Technology, 2015,38(23):47-50.
[14]	Deng Yong, Chai Guodong . Soft fault feature extrac-tion in nonlinear analog circuit fault diagnosis[J]. Circuits, Systems,and Signal Processing, 2016,35(12):4220-4248. doi: 10.1007/s00034-016-0265-z
[15]	Narula A K, Singh A P . Fault diagnosis of mixed-signal analog circuit using artificial neural networks[J]. International Journal of Intelligent Systems Technologies & Applications, 2015,7(7):11-17.
[16]	Li Zhihua . A novel fault diagnostic method based on node-voltage vector ambiguity sets[J].IEEE Transactions on Instrumentation and Measurement, 2014(8):1957-1965.
[17]	Gan X S, Gao W M, Zhe D , et al. Research on WNN soft fault diagnosis for analog circuit based on adaptive UKF algorithm[J]. Applied Soft Computing, 2017,50(C):252-259. doi: 10.1016/j.asoc.2016.11.012
[18]	Wei Y M, Gan X S, Tang X Q .Fault diagnosis based on neural network trained by PSO algorithm with information sharing strategy for analog circuit[J]. Applied Mechanics & Materials, 2014, 687- 691:3320-3323.

故障元器件	正常值	故障状态	故障值/单位	识别编码	诊断编码
R₆	6.2 kΩ	R₆↑	7 kΩ	0001	0001
		R₆↓	4.8 kΩ	0011
R₁	6.2 kΩ	R₁↑	7.2 kΩ	0100	0100
		R₁↓	5 kΩ	0101
C₂	5 nF	C₂↑	6 nF	0110	0110
		C₂↓	4 nF	0111
R₉	1 kΩ	R₉↑	1.5 kΩ	1000	1000
		R₉↓	0.6 kΩ	1001

故障状态	诊断输出结果				数据整理	分类结果	诊断结果
R₆↓	0.036 2	0.005 3	0.974 1	0.993 4	0011	0011	0001
	0.009 5	0.023 1	0.999 9	0.944 0	0011	0011
R₁↑	0.039 3	0.994 7	0.012 2	0.014 5	0100	0100	0100
	0.015 2	0.979 8	0.004 4	0.000 0	0100	0100
	0.080 0	0.968 6	0.001 1	0.002 7	0100	0100
C₂↓	0.000 0	0.624 1	1.000 0	1.000 0	0111	0111	0110
	0.000 0	0.978 3	0.875 5	1.000 0	0111	0111
	0.034 2	0.998 4	0.963 0	1.000 0	0111	0111
R₉↓	0.938 4	0.053 3	0.000 7	0.698 4	1001	1001	1000
	0.986 8	0.008 2	0.001 6	0.933 7	1001	1001
	0.580 4	0.009 7	0.000 1	0.649 8	1001	1001

	未改进Elman 和特征向量		未改进Elman 和改进特征向量		改进Elman 和改进特征向量
故障状态	分类率	诊断率	分类率	诊断率	分类率	诊断率
R₆↑ R₆↓ R₁↑ R₁↓ C₂↑ C₂↓ R₉↑ R₉↓	77% 65% 68% 60% 77.5% 70% 68% 60%	77.5%	82% 77.5% 77% 73% 82% 80% 77% 70%	85%	87% 83% 85% 79% 87% 83% 82.5% 75%	92.5%