基于小波变换的故障电路特征值提取的研究
|
摘要
随着电子技术的发展，电路日益复杂，集成电路的规模越来越庞大，而且集成电路会越来越多的应用在我们的日常生活中，发挥越来越重要的作用。在集成电路设计中，整个系统的可靠性主要取决于集成电路中元器件的可靠性。元器件的可靠性是由元器件的参数值和容差反映的。电路中元器件参数值的改变超出容差范围，便会形成电路中的软故障，电子电路的功能就不能正常发挥。因此，对元器件参数的可靠性和以及出现的软故障的检测，是集成电路故障诊断中要解决的重点。而要检测软故障，就要针对电路中的软故障信号进行分析，提取能够反映软故障信号的特征值。因此，本文的研究重点就是对电子电路中的软故障信号进行特征提取。
传统的检测故障的方法有很多，但是它们大多是在一种处于特定情况下（大多是类似于开路、短路这种硬故障）的，难以发现在电路中的各个器件中存在类似电路及器件缺陷的软故障。尤其是对于日益复杂的集成电路来说，它们的检测有很大的局限性，不能够很好的解决电路中软故障信号的分析和检测。在电路信号的处理分析方法中，传统的傅立叶变换分析方法也是针对平稳信号经过从时域到频域的变换，从而观察信号的特性。但是它对于电路中存在的这种非平稳的软故障的信号，就会有不小的分析难度，不能够反映出来这种软故障信号的瞬态变化特性。
而本文的研究就针对集成电路中的软故障信号的特征提取，提出了基于小波变换的预处理方法。小波变换技术，是从傅立叶分析和短时傅立叶分析演化过来的，它继承了傅立叶分析的优点，也补充了傅立叶分析在非平稳信号分析中的不足。它在低频部分具有较高的频率分辨率和较低的时间分辨率，在高频部分具有较高的时间分辨率和较低的频率分辨率。同时还有多分辨率分析的特点,在时频两域都具有表征信号局部特征的能力。借助于这个特性，小波变换可以把信号的各个频率层次的特征表达出来，可以把信号分解成为低频的近似信号和高频的细节信号，能够很好的检测
故障信号的局部突变特性，检测瞬态和边沿也具有良好的效果。
本文的研究分为两大部分：电路的容差分析和软故障信号的小波变换算法仿真分析。首先，采用经典的计算机模拟电路仿真软件---PSpice对电路进行仿真分析，利用PSpice的蒙特卡洛分析法分析得到容差之内的正常信号和容差之外的软故障信号，以此作为处理分析的对象。其次，研究小波变换的算法：根据采集信号的特点和分析仿真的要求，适中地选取了能够反映信号各层系数的突变特性相对好的小波基函数，以它作为小波变换的母函数；在选取小波函数的基础上，对多尺度分析进行研究，选取了适中的分解尺度，并结合小波母函数，给出了小波基函数为db2、分解尺度为6的多尺度分析提取软故障信号特征的算法。在本文进行的仿真研究中，将待诊断的软故障信号加载到构造的小波多尺度分析算法中，然后仿真得到故障信号的近似信号分解系数和细节信号分解系数图形。通过与正常信号的分析对比，可以直观地看到电路中与频率特性相关的元器件改变参数所带来的软故障信号的突变特征，成功的提取出来了电路中输出信号的软故障特征，解决了传统的傅立叶分析的难点，而且为进一步诊断电路故障提供了便利的基础。本文的研究中，由于与电路频率特性不相关的元器件的参数值的改变不会产生大的影响，其的故障信号频域特性也就不会在本文的分析图中分别出来。因此，文中的研究方法，作为一种改进的故障信号分析方法，在集成电路的迅猛发展和故障诊断技术的不断提高中会得到更为广泛的探讨，而且会有很大的发展空间。
There are many classical methods to diagnose the fault in the circuits, but they are mostly in the condition such hard fault as the opened circuits and shorted circuits, they are difficult in the diagnosis on the soft fault generated by the changes of the parameters or the deficiencies of the circuits, especially for the more complicated IC. It can not complete the analysis and diagnosis of the soft fault signal. The classical analysis method for the signal is the Fourier analysis, which can get the characters of the signal by transforming the signal from the time scope to the frequency scope. But it can not make good analysis of the soft fault signal in the circuits which is random and unstable or can not reflect and extract the transient characters of the soft fault signal.
Whereas the research of this paper is to give the preprocessing method based on wavelet transformation in the extraction of the soft fault signal of the IC. The tech of the wavelet transformation, as the development of the Fourier analysis and short time Fourier analysis, inherit the goods of the Fourier analysis and make up for the deficiencies of the Fourier analysis. It has a higher frequency resolution and a lower time resolution in the low frequency part and a lower frequency resolution and a higher time resolution in the high frequency part. In the meantime, it has the character of the multi-resolution and can express the partial signal characters. Because of this character, wavelet transformation can express the signal in all kinds of frequencies. It can decompose the signal into two parts: the approximation signal in the high frequency and the detail signal in the low frequency. It can also express the partial characters of the soft fault signal and diagnose the transient characters and edges of the signal.
The paper is arranged in two parts: analysis of the tolerance for the
circuits and the analysis of the soft trouble detection algorithm. Firstly, emulate the circuits with PSpice, which is a classical computer emulating software for the circuits. Analyze the normal signal and the soft trouble signal through the Mont Carlo method and make them the objects to be processed. Secondly, study the wavelet transform algorithm. Choose the wavelet base function according to the features of the sampled signal and the emulating analysis request. Based on the selection of the wavelet base function, we chose the general decomposing scale and proposed a soft fault signal analysis and extraction method with db2 mother wavelet function and a 6-layer decomposing scale. In this paper, applying the improved multi-resolution wavelet analysis method on the soft fault signal and got the approximate decomposed coefficients and the detailed coefficients of the signal. Compared with the normal signal, we can see the singularity features of the soft fault signal resulted from the changed parameters of the circuits components corresponding to the circuit frequency. The soft fault signal characters in the output of the circuit were extracted successfully, which solve the difficulty of the traditional Fourier analysis. Further more it provides a convenient base for diagnosing the fault in the circuit.
In this paper the proposed method works better for analyzing the soft fault problems in the circuits, which can detect the problem components of the circuits, however, our method cannot detect all the troubled components in the circuits. Because not all the components are related to the frequency character of the circuits. The fault components of the circuits that are unrelated to the frequency of the circuits can be detected by the general fault dictionary method. Our proposed method in this paper can be used in the soft fault diagnose of the IC and will be further developed as the increasing development of the integrated circuits.
传统的检测故障的方法有很多，但是它们大多是在一种处于特定情况下（大多是类似于开路、短路这种硬故障）的，难以发现在电路中的各个器件中存在类似电路及器件缺陷的软故障。尤其是对于日益复杂的集成电路来说，它们的检测有很大的局限性，不能够很好的解决电路中软故障信号的分析和检测。在电路信号的处理分析方法中，传统的傅立叶变换分析方法也是针对平稳信号经过从时域到频域的变换，从而观察信号的特性。但是它对于电路中存在的这种非平稳的软故障的信号，就会有不小的分析难度，不能够反映出来这种软故障信号的瞬态变化特性。
而本文的研究就针对集成电路中的软故障信号的特征提取，提出了基于小波变换的预处理方法。小波变换技术，是从傅立叶分析和短时傅立叶分析演化过来的，它继承了傅立叶分析的优点，也补充了傅立叶分析在非平稳信号分析中的不足。它在低频部分具有较高的频率分辨率和较低的时间分辨率，在高频部分具有较高的时间分辨率和较低的频率分辨率。同时还有多分辨率分析的特点,在时频两域都具有表征信号局部特征的能力。借助于这个特性，小波变换可以把信号的各个频率层次的特征表达出来，可以把信号分解成为低频的近似信号和高频的细节信号，能够很好的检测
故障信号的局部突变特性，检测瞬态和边沿也具有良好的效果。
本文的研究分为两大部分：电路的容差分析和软故障信号的小波变换算法仿真分析。首先，采用经典的计算机模拟电路仿真软件---PSpice对电路进行仿真分析，利用PSpice的蒙特卡洛分析法分析得到容差之内的正常信号和容差之外的软故障信号，以此作为处理分析的对象。其次，研究小波变换的算法：根据采集信号的特点和分析仿真的要求，适中地选取了能够反映信号各层系数的突变特性相对好的小波基函数，以它作为小波变换的母函数；在选取小波函数的基础上，对多尺度分析进行研究，选取了适中的分解尺度，并结合小波母函数，给出了小波基函数为db2、分解尺度为6的多尺度分析提取软故障信号特征的算法。在本文进行的仿真研究中，将待诊断的软故障信号加载到构造的小波多尺度分析算法中，然后仿真得到故障信号的近似信号分解系数和细节信号分解系数图形。通过与正常信号的分析对比，可以直观地看到电路中与频率特性相关的元器件改变参数所带来的软故障信号的突变特征，成功的提取出来了电路中输出信号的软故障特征，解决了传统的傅立叶分析的难点，而且为进一步诊断电路故障提供了便利的基础。本文的研究中，由于与电路频率特性不相关的元器件的参数值的改变不会产生大的影响，其的故障信号频域特性也就不会在本文的分析图中分别出来。因此，文中的研究方法，作为一种改进的故障信号分析方法，在集成电路的迅猛发展和故障诊断技术的不断提高中会得到更为广泛的探讨，而且会有很大的发展空间。
There are many classical methods to diagnose the fault in the circuits, but they are mostly in the condition such hard fault as the opened circuits and shorted circuits, they are difficult in the diagnosis on the soft fault generated by the changes of the parameters or the deficiencies of the circuits, especially for the more complicated IC. It can not complete the analysis and diagnosis of the soft fault signal. The classical analysis method for the signal is the Fourier analysis, which can get the characters of the signal by transforming the signal from the time scope to the frequency scope. But it can not make good analysis of the soft fault signal in the circuits which is random and unstable or can not reflect and extract the transient characters of the soft fault signal.
Whereas the research of this paper is to give the preprocessing method based on wavelet transformation in the extraction of the soft fault signal of the IC. The tech of the wavelet transformation, as the development of the Fourier analysis and short time Fourier analysis, inherit the goods of the Fourier analysis and make up for the deficiencies of the Fourier analysis. It has a higher frequency resolution and a lower time resolution in the low frequency part and a lower frequency resolution and a higher time resolution in the high frequency part. In the meantime, it has the character of the multi-resolution and can express the partial signal characters. Because of this character, wavelet transformation can express the signal in all kinds of frequencies. It can decompose the signal into two parts: the approximation signal in the high frequency and the detail signal in the low frequency. It can also express the partial characters of the soft fault signal and diagnose the transient characters and edges of the signal.
The paper is arranged in two parts: analysis of the tolerance for the
circuits and the analysis of the soft trouble detection algorithm. Firstly, emulate the circuits with PSpice, which is a classical computer emulating software for the circuits. Analyze the normal signal and the soft trouble signal through the Mont Carlo method and make them the objects to be processed. Secondly, study the wavelet transform algorithm. Choose the wavelet base function according to the features of the sampled signal and the emulating analysis request. Based on the selection of the wavelet base function, we chose the general decomposing scale and proposed a soft fault signal analysis and extraction method with db2 mother wavelet function and a 6-layer decomposing scale. In this paper, applying the improved multi-resolution wavelet analysis method on the soft fault signal and got the approximate decomposed coefficients and the detailed coefficients of the signal. Compared with the normal signal, we can see the singularity features of the soft fault signal resulted from the changed parameters of the circuits components corresponding to the circuit frequency. The soft fault signal characters in the output of the circuit were extracted successfully, which solve the difficulty of the traditional Fourier analysis. Further more it provides a convenient base for diagnosing the fault in the circuit.
In this paper the proposed method works better for analyzing the soft fault problems in the circuits, which can detect the problem components of the circuits, however, our method cannot detect all the troubled components in the circuits. Because not all the components are related to the frequency character of the circuits. The fault components of the circuits that are unrelated to the frequency of the circuits can be detected by the general fault dictionary method. Our proposed method in this paper can be used in the soft fault diagnose of the IC and will be further developed as the increasing development of the integrated circuits.
引文
[1] 钱志鸿．集成电路故障分析与诊断．长春：吉林科学技术出版社，2003
[2] 戴逸松．微弱信号检测方法及仪器．北京：国防工业出版社，1994
[3] 高伟涛．PSpice 8.0电路设计实例精粹．北京：国防工业出版社，2001
[4] 杨福生．小波变换的工程分析与应用．北京：科学出版社，1999
[5] 彭玉华．小波变换与工程应用．北京：科学出版社，1999
[6] 胡广书．现代信号处理教程．北京：清华大学出版社，2001
[7] 胡广书．数字信号处理—理论、算法与实现．北京：清华大学出版社，1997
[8] 张贤达,保铮．非平稳信号分析与处理．北京：国防工业出版社, 1998
[9] 高艳梅，房蔓楠．Spice/PSpice编程技术．北京：电子工业出版社，2002
[10] 陈艳峰，丘水生，任言真，周小安．MATLAB与PSPICE相结合用于开关功率变换器仿真的方法．电机与控制学报，1999，2
[11] 赵国南，郭裕顺．模拟电路故障诊断．哈尔滨：哈尔滨工业大学出版社，1991
[12] 于开平,邹经湘,杨炳渊．小波函数的性质及其应用研究．哈尔滨工业大学学报，2000,2
[13] 刘贵中, 邸双亮．小波分析及其应用．西安：西安电子科技大学出版社，1995
[14] 秦前清,杨宗凯．实用子波分析．西安：西安电子科技大学出版社，1995
[15] 张静远，张冰，蒋兴舟．基于小波变换的特征提取方法分析．信号处理，2000，6
[16] 胡昌华，张军波 等．基于MATLAB的系统分析与设计—小波分析． 西安：西安电子科技大学出版社，1999
[17] 尹泽明，丁春礼等．精通MATLAB 6．北京：清华大学出版社，2002
[18] 飞思科技产品研发中心．MATLAB 6.5辅助小波分析与应用．北京：电子工业出版社，2003
[19] 董新洲，耿中行等．小波变换用于电力系统故障信号分析初探．中国电机工程学报，1997，11
[20] 王雷．基于小波变换消除信号噪声的研究．现代电子技术，2001，14
[21] 用PSpice分析和设计电路．张凌，华北电力大学学报．2002，6
[22] 田杰，陈杰等．基于小波变换及分形特征的目标检测与识别．北京理工大学学报，2003，2
[23] 雍正正．模拟电路故障诊断:模糊故障字典法．数据采集及处理， 1998，3
[24] 杜鑫，唐大全．模拟电路故障诊断技术的发展．测控技术，2003，7
[25] 谭阳刚．模拟电路故障诊断的新故障字典法．微电子学，2001，4
[26] 王忠鹏，解恒力等．利用PSpice程序对模拟集成电路参数进行设计．辽宁工学院学报，2000，5
[27] 孙云莲，刘敦敏．时域分析与小波变换及其应用．武汉大学学报，2003，4
[28] 肖化铸．小波分析中信号的时频局部化．电子科技大学学报，1995，6
[29] 林京，屈梁生．基于连续小波变换的奇异性检测与故障诊断．振动工程学报，2000，4
[30] 赵学智，林颖等．光滑奇异性信号的小波检测及其应用．信号处理，2000，12
[31] 陈德智，唐磊等．由小波变换的模极大值快速重构信号．电子学报，1998，9
[32] 姜礼平，龚沈光，胡伟文．一种基于小波分解的信号特征提取方法．武汉理工大学学报，2003，6
[33] 代俊光．一种奇异信号测量中的小波分析方法．电子测量与仪器学报，2000，3
[34] 叶昊，王桂增等．小波变换在故障检测中的应用．自动化学报，1997，11
[35] 王真理，李衍达．基于小波的信号部分重构与插值．清华大学学报，1995, 5
[36] S.G. Mallat．Multifrequency Channel Decompositions of Images and Wavelet Models．IEEE. Trans on Acoustics. Speech and Signal Processing, 1989,37：12
[37] I. Daubechies, A.Grossmann．Frames in the Bargmann space of entire functions．Common Pure and ApplMth, 1988,41：151-164
[38] S. Mallat．A Theory of Multiresolution Signal Decomposition: the wavelet represention．IEEE Trans on Pattern Anal Intell, 1989，11：674-614
[39] I. Daubechies．Ten Lectures on Wavelets．SIAM, Philadelphia, PA, l992
[40] M.Gatelani, S.Giraldi．A Measurement System for Fault Detection and Fault Isolation of Analog Circuits．Measurement, 1999,25
[41] Y. Meyer．Wavelets algorithms and applications．Society for Industry and Applied Mathematics, 1993
[42] FARZAN AMINIAN,MEHRAN AMINIAN．Fault Diagnosis of Nonlinear Analog Circuits Using Neural Networks with Wavelet and Fourier Transforms as Preprocessors. Journal of Electronic Testing．Theory and Application.2001,17：471-480
[43] FARZAN AMINIAN,MEHRAN AMINIAN．Fault Diagnosis of Analog Circuits Using Bayesian Neural Networks with Wavelet Transform as Preprocessor．Journal of Electronic Testing: Theory and Application.2001,17：29-36
[44] S. G. Mallat．Characterization of signals From Multiscales Edges．NYU, Computer Science Tech.Report, 1991
[45] J.Canny．A Computational Approach to Edge Detection．IEEE Trans. Parts. Anal.Machine Intell. PAMI-8(4), 1986
[46] R. Hummel, R. Moniot．Reconstruction from Zero-Crossing in Scale Space．IEEE Trans. Acoustic Speech Signal Processing, 37(12), 1989
[47] A. Yullie, T.Poggio．Scale Theorems for Zero Crossings．IEEE Trans. Patt. Anal. Machine Intell. PAMI-8(1), 1986
[48] Y.Meyer．Un Contre-example a la Conjecture de Marr et al．Cell de S. Mallat, Preprint, 1991
[49] S.Mallat．Characterization of Signals from Multi-scales Edges．IEEE Trans.Patt. Anal. Machine Intell, 14(7),1992
[50] I.Daubechies．Orthonormal Bases of Commpactly Supported Wavelets．Common. Pure and Appli. Math., XLT, 1988
[2] 戴逸松．微弱信号检测方法及仪器．北京：国防工业出版社，1994
[3] 高伟涛．PSpice 8.0电路设计实例精粹．北京：国防工业出版社，2001
[4] 杨福生．小波变换的工程分析与应用．北京：科学出版社，1999
[5] 彭玉华．小波变换与工程应用．北京：科学出版社，1999
[6] 胡广书．现代信号处理教程．北京：清华大学出版社，2001
[7] 胡广书．数字信号处理—理论、算法与实现．北京：清华大学出版社，1997
[8] 张贤达,保铮．非平稳信号分析与处理．北京：国防工业出版社, 1998
[9] 高艳梅，房蔓楠．Spice/PSpice编程技术．北京：电子工业出版社，2002
[10] 陈艳峰，丘水生，任言真，周小安．MATLAB与PSPICE相结合用于开关功率变换器仿真的方法．电机与控制学报，1999，2
[11] 赵国南，郭裕顺．模拟电路故障诊断．哈尔滨：哈尔滨工业大学出版社，1991
[12] 于开平,邹经湘,杨炳渊．小波函数的性质及其应用研究．哈尔滨工业大学学报，2000,2
[13] 刘贵中, 邸双亮．小波分析及其应用．西安：西安电子科技大学出版社，1995
[14] 秦前清,杨宗凯．实用子波分析．西安：西安电子科技大学出版社，1995
[15] 张静远，张冰，蒋兴舟．基于小波变换的特征提取方法分析．信号处理，2000，6
[16] 胡昌华，张军波 等．基于MATLAB的系统分析与设计—小波分析． 西安：西安电子科技大学出版社，1999
[17] 尹泽明，丁春礼等．精通MATLAB 6．北京：清华大学出版社，2002
[18] 飞思科技产品研发中心．MATLAB 6.5辅助小波分析与应用．北京：电子工业出版社，2003
[19] 董新洲，耿中行等．小波变换用于电力系统故障信号分析初探．中国电机工程学报，1997，11
[20] 王雷．基于小波变换消除信号噪声的研究．现代电子技术，2001，14
[21] 用PSpice分析和设计电路．张凌，华北电力大学学报．2002，6
[22] 田杰，陈杰等．基于小波变换及分形特征的目标检测与识别．北京理工大学学报，2003，2
[23] 雍正正．模拟电路故障诊断:模糊故障字典法．数据采集及处理， 1998，3
[24] 杜鑫，唐大全．模拟电路故障诊断技术的发展．测控技术，2003，7
[25] 谭阳刚．模拟电路故障诊断的新故障字典法．微电子学，2001，4
[26] 王忠鹏，解恒力等．利用PSpice程序对模拟集成电路参数进行设计．辽宁工学院学报，2000，5
[27] 孙云莲，刘敦敏．时域分析与小波变换及其应用．武汉大学学报，2003，4
[28] 肖化铸．小波分析中信号的时频局部化．电子科技大学学报，1995，6
[29] 林京，屈梁生．基于连续小波变换的奇异性检测与故障诊断．振动工程学报，2000，4
[30] 赵学智，林颖等．光滑奇异性信号的小波检测及其应用．信号处理，2000，12
[31] 陈德智，唐磊等．由小波变换的模极大值快速重构信号．电子学报，1998，9
[32] 姜礼平，龚沈光，胡伟文．一种基于小波分解的信号特征提取方法．武汉理工大学学报，2003，6
[33] 代俊光．一种奇异信号测量中的小波分析方法．电子测量与仪器学报，2000，3
[34] 叶昊，王桂增等．小波变换在故障检测中的应用．自动化学报，1997，11
[35] 王真理，李衍达．基于小波的信号部分重构与插值．清华大学学报，1995, 5
[36] S.G. Mallat．Multifrequency Channel Decompositions of Images and Wavelet Models．IEEE. Trans on Acoustics. Speech and Signal Processing, 1989,37：12
[37] I. Daubechies, A.Grossmann．Frames in the Bargmann space of entire functions．Common Pure and ApplMth, 1988,41：151-164
[38] S. Mallat．A Theory of Multiresolution Signal Decomposition: the wavelet represention．IEEE Trans on Pattern Anal Intell, 1989，11：674-614
[39] I. Daubechies．Ten Lectures on Wavelets．SIAM, Philadelphia, PA, l992
[40] M.Gatelani, S.Giraldi．A Measurement System for Fault Detection and Fault Isolation of Analog Circuits．Measurement, 1999,25
[41] Y. Meyer．Wavelets algorithms and applications．Society for Industry and Applied Mathematics, 1993
[42] FARZAN AMINIAN,MEHRAN AMINIAN．Fault Diagnosis of Nonlinear Analog Circuits Using Neural Networks with Wavelet and Fourier Transforms as Preprocessors. Journal of Electronic Testing．Theory and Application.2001,17：471-480
[43] FARZAN AMINIAN,MEHRAN AMINIAN．Fault Diagnosis of Analog Circuits Using Bayesian Neural Networks with Wavelet Transform as Preprocessor．Journal of Electronic Testing: Theory and Application.2001,17：29-36
[44] S. G. Mallat．Characterization of signals From Multiscales Edges．NYU, Computer Science Tech.Report, 1991
[45] J.Canny．A Computational Approach to Edge Detection．IEEE Trans. Parts. Anal.Machine Intell. PAMI-8(4), 1986
[46] R. Hummel, R. Moniot．Reconstruction from Zero-Crossing in Scale Space．IEEE Trans. Acoustic Speech Signal Processing, 37(12), 1989
[47] A. Yullie, T.Poggio．Scale Theorems for Zero Crossings．IEEE Trans. Patt. Anal. Machine Intell. PAMI-8(1), 1986
[48] Y.Meyer．Un Contre-example a la Conjecture de Marr et al．Cell de S. Mallat, Preprint, 1991
[49] S.Mallat．Characterization of Signals from Multi-scales Edges．IEEE Trans.Patt. Anal. Machine Intell, 14(7),1992
[50] I.Daubechies．Orthonormal Bases of Commpactly Supported Wavelets．Common. Pure and Appli. Math., XLT, 1988