电流模电路的小波神经网络测试研究

2014-10-22 12:23郭杰荣何怡刚
关键词:神经网络测试

郭杰荣+何怡刚

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.

关键词电流模;测试;小波分解;神经网络

中图分类号TM315文献标识码A文章编号1000-2537(2014)02-0051-05

与开关电容技术不同,电流模电路采用电流作为信号传输介质,因而呈现较低的电抗特性,具有较小的漂移电感(stray-inductance)[1-3],可以达到较高的速率.电流模电路基本单元如图1所示.但是,将传统的模拟电路测试方法应用在电流模电路方面遇到了困难,电流模电路独特的结构及传输方式需要新的测试方法.可供选择的方法是采用神经网络方法,相关文献[4~9]表明,如果要达到较高的故障识别率,需要在神经网络训练过程中加大训练数据的数量以及神经元数目,这将导致训练过程的复杂化及过长的训练时间.本文在研究电流模电路特殊结构与特性的前提下,提出了一种基于测试节点电压的瞬态测试的多尺度小波分分解及神经网络非线性映射归纳的测试方法,可以在较少训练的前提下获得较高的故障识别率.

4结束语

本文提出了一种针对电流模电路的预先采用多尺度小波分解,再使用神经网络训练测试的方法.即首先将各类故障模型加入电流模电路中,以电流模电路的电流输出响应信号为样本,在正常提取测试信号特性的前提下,采用多尺度小波分解对各类响应数据进行预处理,在保留故障信号特性品质的前提下降低分析的样本数量,将预处理的结果作为神经网络模型训练样本,训练完成后可对各类故障进行识别.本文方法可以适用于电流模式信号传输测试并有效降低训练所需神经网络结构复杂性.

参考文献:

[1]KIM H S, JEON J Y, LEE S W, et al. A compact-sized 9-bit switched-current DAC for AMOLED mobile display drivers[J]. IEEE Trans Circuits Syst Ⅱ: Express Briefs, 2012,58(12):887-891.

[2]SAETHER G E, TOUMAZOU C, TAYLOR G, et al. Concurrent self test of switched current circuits based on the S 2I-technique[J]. IEEE Int Symp Circuits Syst, 1995,12(2):841-844.

[3]戴力,庄奕琪,景鑫,等. 一种新型高性能CMOS电流模式的动态规划电路[J]. 西安交通大学学报, 2012,46(6):29-35.

[4]刘美容,何怡刚,赵新民,等. 基于混沌的模拟电路故障诊断方法[J]. 湖南师范大学自然科学学报, 2012,35(3):34-39.

[5]彭良玉, 禹旺兵. 基于小波分析和克隆选择算法的模拟电路故障诊断[J]. 电工技术学报, 2007,22(6):12-16.

[6]GU H Y, ZHANG F, WANG Z J, et al. Identification method for low-voltage Arc fault based on the loose combination of wavelet transformation and neural network[C]. IEEE Power Engineering and Automation Conference (PEAM), Wuhan, China, 2012.

[7]DAVANIPOOR M, ZEKRI M, SHEIHOLESLAM F. Fuzzy wavelet neural network with an accelerated hybrid learning algorithm[J]. IEEE Trans Fuzzy Sys, 2012,20(3):463-470.

[8]徐崇斌,赵志文,郑慧芳.ELVQ算法实现宽参数偏移的多故障电路诊断[J]. 电子与信息学报, 2011,33(6):1520-1524.

[9]ABOHAGAR A A, MUSTAFA M W. Back propagation neural network aided wavelet transform for high impedance fault detection and faulty phase selection[C]. IEEE International Conference on Power and Energy (PECon), Kota Kinabalu, Malaysia, 2012.

[10]CORDOVA J J, YU W, LI X O. Haar wavelet neural networks for nonlinear system identification[C]. 2012 IEEE International Symposium on Intelligent Control (ISIC), Dubrouk, Croatia, 2012.

[11]ESPINOZA-TREJO D R, CAMPOS-DELGADO D U, BRCENAS E, et al. Robust fault diagnosis scheme for open-circuit faults in voltage source inverters feeding induction motors by using non-linear proportional-integral observers[J]. IET Power Electr, 2012,5(7):1204-1216.

[12]REN S X, GAO L. Application of a wavelet packet transform based radial basis function neural network to analyze overlapping spectra [J]. Congress Image Signal Proc, 2008,5(2):228-232.

[13]IOKIBE K, AMANO T, OKAMOTO K, et al. Improvement of linear equivalent circuit model to identify simultaneous switching noise current in cryptographic integrated circuits[C]. IEEE International Symposium on Electromagnetic Compatibility (EMC), Denver, America, 2013.

[14]YAN Y Y, LEE F C, MATTAVELLI P. Analysis and design of average current mode control using a describing-function-based equivalent circuit model [J]. IEEE Trans Power Electr, 2013,28(10):4732-4741.

[15]KHANG H V, ARKKIO A. Eddy-current loss modeling for a form-wound induction motor using circuit model[J]. IEEE Trans Magnet, 2012,48(2):1059-1062.

[16]ARUMUGAM P, HAMITI T, GERADA C. Modeling of different winding configurations for fault-tolerant permanent magnet machines to restrain interturn short-circuit current [J]. IEEE Trans Energy Conver, 2012,27(2):351-361.

[17]郭杰荣,李长生,刘长青. 基于0.18 μm CMOS的电流模单元最优化设计[J]. 湖南文理学院学报:自然科学版, 2012,24(1):39-41,45.

(编辑陈笑梅)

[12]REN S X, GAO L. Application of a wavelet packet transform based radial basis function neural network to analyze overlapping spectra [J]. Congress Image Signal Proc, 2008,5(2):228-232.

[13]IOKIBE K, AMANO T, OKAMOTO K, et al. Improvement of linear equivalent circuit model to identify simultaneous switching noise current in cryptographic integrated circuits[C]. IEEE International Symposium on Electromagnetic Compatibility (EMC), Denver, America, 2013.

[14]YAN Y Y, LEE F C, MATTAVELLI P. Analysis and design of average current mode control using a describing-function-based equivalent circuit model [J]. IEEE Trans Power Electr, 2013,28(10):4732-4741.

[15]KHANG H V, ARKKIO A. Eddy-current loss modeling for a form-wound induction motor using circuit model[J]. IEEE Trans Magnet, 2012,48(2):1059-1062.

[16]ARUMUGAM P, HAMITI T, GERADA C. Modeling of different winding configurations for fault-tolerant permanent magnet machines to restrain interturn short-circuit current [J]. IEEE Trans Energy Conver, 2012,27(2):351-361.

[17]郭杰荣,李长生,刘长青. 基于0.18 μm CMOS的电流模单元最优化设计[J]. 湖南文理学院学报:自然科学版, 2012,24(1):39-41,45.

(编辑陈笑梅)

[12]REN S X, GAO L. Application of a wavelet packet transform based radial basis function neural network to analyze overlapping spectra [J]. Congress Image Signal Proc, 2008,5(2):228-232.

[13]IOKIBE K, AMANO T, OKAMOTO K, et al. Improvement of linear equivalent circuit model to identify simultaneous switching noise current in cryptographic integrated circuits[C]. IEEE International Symposium on Electromagnetic Compatibility (EMC), Denver, America, 2013.

[14]YAN Y Y, LEE F C, MATTAVELLI P. Analysis and design of average current mode control using a describing-function-based equivalent circuit model [J]. IEEE Trans Power Electr, 2013,28(10):4732-4741.

[15]KHANG H V, ARKKIO A. Eddy-current loss modeling for a form-wound induction motor using circuit model[J]. IEEE Trans Magnet, 2012,48(2):1059-1062.

[16]ARUMUGAM P, HAMITI T, GERADA C. Modeling of different winding configurations for fault-tolerant permanent magnet machines to restrain interturn short-circuit current [J]. IEEE Trans Energy Conver, 2012,27(2):351-361.

[17]郭杰荣,李长生,刘长青. 基于0.18 μm CMOS的电流模单元最优化设计[J]. 湖南文理学院学报:自然科学版, 2012,24(1):39-41,45.

(编辑陈笑梅)

猜你喜欢
神经网络测试
基于人工智能LSTM循环神经网络的学习成绩预测
基于图像处理与卷积神经网络的零件识别
基于改进VGG-16神经网络的图像分类方法
基于自适应神经网络的电网稳定性预测
心理测试
一年级上册期末小测试
逃离体能测试
三次样条和二次删除相辅助的WASD神经网络与日本人口预测
基于BP神经网络PID控制的无刷直流电动机调速系统设计
心理小测试