小波分析在渗碳层深度涡流检测信号处理中的应用
|
摘要
金属表面硬化层深度的检测及控制是目前机械工业部门急需解决的一个问题，它属于材料检测的范畴。渗碳层深度是衡量渗碳件质量的主要技术指标之一。涡流检测是一种适用于试件表面和近表面变化的检测方法，利用涡流方法检测渗碳件时，涡流检测信号的变化主要取决于渗碳层组织的物理特性的变化，所以涡流检测方法可以对金属表面渗碳层深度进行检测而不构成任何破坏。但影响涡流检测的因素很多，检测结果不尽如人意，如何提取信号中的反映渗碳层深度的信息非常重要。为此本文采用了小波变换分析信号、提取特征值，并用神经网络对信号进行分类。小波变换是信号时间—尺度(时间—频率)的分析方法，它具有多分辨分析的特点，而且在时频两域都具有表征信号局部特征的能力，是一种窗口面积固定不变但其形状可改变的时频局域化分析方法，即在低频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，这符合工程信号分析的需要。本文介绍了小波变换的理论、基小波的选择和利用小波变换进行信号滤波的方法和滤波处理结果，并提出了一种基于小波理论的新的特征值提取方法。即利用小波包分析方法将信号在低频、高频段作进一步的细分，以各层分解的能量作为信号的频域特征值，以最低频带的极值点作为时域特征值，这样的特征值选取方法较全面的反映了信号的时-频特征，优于传统的傅里叶分析方法。
BP神经网络是一种具有代表性的神经网络模型之一，它适用于信号的分类。本文介绍了BP神经网络的特点、算法和其结构的具体设计方法和设计结果，并将小波包提取的特征值输入到BP网络，对7种不同渗碳层深度的试件进行分类，实验结果表明，小波特征值提取和BP神经网络分类器相结合，可以实现对不同渗碳层深度的分类，效果良好，精度较高，有一定的实用价值。
Nondestructive testing and quality control of metal hardened-depth is an important problem to be solved in machinery industry. It belongs to material property testing. The metal carburized layer depth is one of the major technical parameter to evaluate the quality of carburized components. Eddy current testing is a method suitable for testing changes occurred in the surface or subsurface of specimen. When using this method to test carburized components, the variable of eddy current signal is depend on the change of constituents' physical property in carburized layer. So it can be used to test carburized components nondestructively. However, many factors may affect eddy current testing and it is hard to classify them, and how to extract information which indicating carburized layer depth in signals is very important. So this paper is used Wavelet transform method to analyze signals and extract features in them, then classify them by neutral network. Wavelet transform is a method to analyze signal in the time-
scale (time-frequency) plane. It has a character of multi-resolution, and can be used to characterize detailed features in both time and frequency field. The dimension of the window is constant, but its shape is variable. That is to say, in low frequency band, it has high frequency resolution and low time resolution; in high frequency band, it has low frequency resolution and high time resolution ,which is fit for analyzing requirement of engineering signal . This paper is described Wavelet transform theory , mother wavelet choice , the method to filter signal by Wavelet transform and the result , prospered a way to extract feature originated from Wavelet theory, which using Wavelet packet analyzing method to subdivide signal both in low frequency and high frequency field, and consider energy of every layer as feature in frequency field, and in conjunction with the detailed analyzing character of Wavelet packet in time-frequency plane, consider several minimum or maximum points in the lowest frequency band a
s features in the time field. This is a very good method to charactering feature both in the time field and in the frequency field, and this method is superior to Fourier transform method.
BP neutral network is one of the representative neutral network models, it is suitable for classifying. This paper is introduced BP neutral network character, algorithm , designing principal on its construction and the designed product . Input features extracted by the way described above into a BP neutral network, and using it to classify seven type of different carburized layer depth
specimen .The result is indicated, using Wavelet packet method to extract features and BP neutral network to classifying, is effective and precise to classify different metal carburized layer depth. It is useful and economical.
BP神经网络是一种具有代表性的神经网络模型之一，它适用于信号的分类。本文介绍了BP神经网络的特点、算法和其结构的具体设计方法和设计结果，并将小波包提取的特征值输入到BP网络，对7种不同渗碳层深度的试件进行分类，实验结果表明，小波特征值提取和BP神经网络分类器相结合，可以实现对不同渗碳层深度的分类，效果良好，精度较高，有一定的实用价值。
Nondestructive testing and quality control of metal hardened-depth is an important problem to be solved in machinery industry. It belongs to material property testing. The metal carburized layer depth is one of the major technical parameter to evaluate the quality of carburized components. Eddy current testing is a method suitable for testing changes occurred in the surface or subsurface of specimen. When using this method to test carburized components, the variable of eddy current signal is depend on the change of constituents' physical property in carburized layer. So it can be used to test carburized components nondestructively. However, many factors may affect eddy current testing and it is hard to classify them, and how to extract information which indicating carburized layer depth in signals is very important. So this paper is used Wavelet transform method to analyze signals and extract features in them, then classify them by neutral network. Wavelet transform is a method to analyze signal in the time-
scale (time-frequency) plane. It has a character of multi-resolution, and can be used to characterize detailed features in both time and frequency field. The dimension of the window is constant, but its shape is variable. That is to say, in low frequency band, it has high frequency resolution and low time resolution; in high frequency band, it has low frequency resolution and high time resolution ,which is fit for analyzing requirement of engineering signal . This paper is described Wavelet transform theory , mother wavelet choice , the method to filter signal by Wavelet transform and the result , prospered a way to extract feature originated from Wavelet theory, which using Wavelet packet analyzing method to subdivide signal both in low frequency and high frequency field, and consider energy of every layer as feature in frequency field, and in conjunction with the detailed analyzing character of Wavelet packet in time-frequency plane, consider several minimum or maximum points in the lowest frequency band a
s features in the time field. This is a very good method to charactering feature both in the time field and in the frequency field, and this method is superior to Fourier transform method.
BP neutral network is one of the representative neutral network models, it is suitable for classifying. This paper is introduced BP neutral network character, algorithm , designing principal on its construction and the designed product . Input features extracted by the way described above into a BP neutral network, and using it to classify seven type of different carburized layer depth
specimen .The result is indicated, using Wavelet packet method to extract features and BP neutral network to classifying, is effective and precise to classify different metal carburized layer depth. It is useful and economical.
引文
[1] C.J.Renken. The Use of a Personai Computer to Extract Information from Pulsed Eddy Current Tests. Material Evaluation, Mar, 2001
[2] F.Lingvall, T.Stepinski. Automatic Detecting and Classifying Defects During Eddy Current Inspection of Riveted Lap-joints. NDT&E International, Jun, 2000
[3] Sung-Jin Song,Young-Kil Shin. Eddy Current Flaw Characterization in Tubes by Neural Networks and Finite Element Modeling. NDT&E International, Jun, 2000
[4] Hirishi Hoshikawa. Kiyoshi Koyama, Eddy Current Distribution Using parameters Normalized by Standard Penetration Depth. Material Evaluation, Jun, 1999
[5] G Hands. PMFT-A New Eddy Current Technique for CheckingMaterialHardness.Insight, Vol.40, Jan, 1998
[6] G Chen, A Yamaguchi, K Miya. Singnal Processing for Detection and Characterization of Defects From Composite Eddy Current Singals. Insight, Vol.39, Nov, 1997
[7] Bruno Lebrun,Yves Jayet,Jean-Claude Baboux. Pulsed Eddy Current SingnalAnalysis Application to the Experimen Detection and Characterization of Deep Flaws in Highly ConducterMaterials,NDT&International,Vol.30,No.3,1997
[8] U.patel,D.Rodger. Finite Element Modeling of Pulsed Eddy Current for Nondestructive Testing. IEEE Transaction on Magnetics,Vol.32,N0． 3,Nov,1996
[9] Vetterli ,M,Mathieu.P.Wavelet and filter Banks:Theory and design.IEEE Trans.singnal Proc,Vol.1,No.2,April,1992
[10] S.Mallat,S.Xhong.Chracterization of signal from multiscale edges.IEEE Trans. On Pattern Analysis and Machine Intelligence,Vol.14 ,Sep,1992
[11] S.Mallat,W.L.HWang,Singularity detection and processing with Wavelet. IEEE Trans. On Information Theroy,Vol,38,Feb,1992
[12] MT Shyamsunder,C Rajagopalan, K K Ray,Baldev Raj. A Comparative Study of Conventional and Artificial Neural Network Classifiers for Eddy Current Singal Classification. Insight,Jan, 1995
[13] Bruno Lebrun,Yves Jayet,Iean Claude Baboux. Pulsed Eddy Current Application to the Detection of Deep Cracks.Material Evalution,Nov,1995
[14]G.Mook,A.Tehemov, New Method of Signal Excitation and Processing in Magnetoinductive Testing of Structure and Alloys. NDT&E International,Oct, 1994
[15]Z Liu, K Tsukada. One-dimensional Eddy Current Mulfi-frequency Data Fusion:A Multi-resolution Analysis Approach. Insight,Vol.40,Apr, 1998
[16]Tadeusz Stepinski, Nora Maszi,Conjugate Speetum Filters for Eddy Current Signal Processing. Material Evaluation,Nov, 1995
[17]H.Komatsu. Basie Study on ECT Dated Evaluation Method with Neural Network. Non-Destructive Testing,1992
[18]I.T. Rekanos,T.P.Theodoulidis. Impendenee Inversion in Eddy Current Testing of Layered Planar via Neural Network,NDT&E International,Apr, 1997
[19]Martin Gibbs,Joe Campbell. Pulsed Eddy Current Inspection of Cracks Under Installed Fasteners. Material Evalution,Jan, 1991
[20]Z Liu,K Tsukada. Eddy Current Defect Characterization Using Neural Network. Material Evaluation, Mar, 1990
[21]杨福生．小波变换的工程分析与应用．北京：科学出版社，2001
[22]楼天顺，施阳．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
[23]楼天顺，施阳．基于MATLAB的系统分析与设计—神经网络．西安；西安电子科技大学出版社，1999
[24]吴步宁．小波变换及其在涡流无损检测中的应用．湖北汽车工业学院学报，2001，15(1)：28～31
[25]宋良．神经网络在电磁无损检测中的应用．信息技术，2001，(12)：3～4
[26]吴步宁，梁玉红．涡流阻抗图的傅里叶描述和分析．无损检测，2001，23(9)：367～371
[27]梁玉红，吴步宁．神经网络在涡流检测信号处理中的作用．无损检测，2000，22(11)：545～547
[28]徐科，徐金梧．小波分析在信号滤波中的应用．北京科技大学学报，1997，19(8)：382～385
[29]宋大雷，徐殿国．小波分析在钢丝绳无损检测中的应用．仪器仪表学报，1997，18(4)：151～155
[30]刘建峰，戚飞虎．基于子波变换的局域特征提取方法．红外与毫米波学报，1996，15(4)：123～127
[31]康宜华，武新军，杨叔子．磁性无损检测中的信号处理技术．无损检测，2000，22(6)：
255～259
[32]吴步宁，陈志祥．钢材硬度涡流无损检测技术的研究．无损检测，2000，22(6)：43～45
[33]钟洁，王素菊．涡流检测信号输入神经网络前特征值提取问题的论证．无损检测，1999，21(3)：97-98
[34]辛宁宁，路灿，盛剑霓．小波边缘检测技术在涡流检测中的应用．高压电技术，1999，25(9)；7～10
[35]孙晓云，路灿，袁斌．涡流无损检测中的定量分析．无损检测，2000，21(5)：195～199
[36]雷银照，马信山．人工神经网络在涡流检测中的应用．无损检测，1994，16(2)：31～33
[37]任吉林．涡流检测技术近20年的进展．无损检测，1998，20(5)：121～125
[38]胡光立，谢希文．钢的热处理，西安：西北工业大学出版社，1993
[39]任吉林．电磁无损检测．北京：航空工业出版社，1989
[40]姚天任，孙洪．现代数字信号处理．武汉：华中理工大学出版社，1999．11
[41]殷勤业，杨忠凯．模式识别与神经网络．北京；机械工业出版社，1992．9
[42]魏莉．涡流测试信号处理系统的研究．武汉：武汉理工大学，2001
[2] F.Lingvall, T.Stepinski. Automatic Detecting and Classifying Defects During Eddy Current Inspection of Riveted Lap-joints. NDT&E International, Jun, 2000
[3] Sung-Jin Song,Young-Kil Shin. Eddy Current Flaw Characterization in Tubes by Neural Networks and Finite Element Modeling. NDT&E International, Jun, 2000
[4] Hirishi Hoshikawa. Kiyoshi Koyama, Eddy Current Distribution Using parameters Normalized by Standard Penetration Depth. Material Evaluation, Jun, 1999
[5] G Hands. PMFT-A New Eddy Current Technique for CheckingMaterialHardness.Insight, Vol.40, Jan, 1998
[6] G Chen, A Yamaguchi, K Miya. Singnal Processing for Detection and Characterization of Defects From Composite Eddy Current Singals. Insight, Vol.39, Nov, 1997
[7] Bruno Lebrun,Yves Jayet,Jean-Claude Baboux. Pulsed Eddy Current SingnalAnalysis Application to the Experimen Detection and Characterization of Deep Flaws in Highly ConducterMaterials,NDT&International,Vol.30,No.3,1997
[8] U.patel,D.Rodger. Finite Element Modeling of Pulsed Eddy Current for Nondestructive Testing. IEEE Transaction on Magnetics,Vol.32,N0． 3,Nov,1996
[9] Vetterli ,M,Mathieu.P.Wavelet and filter Banks:Theory and design.IEEE Trans.singnal Proc,Vol.1,No.2,April,1992
[10] S.Mallat,S.Xhong.Chracterization of signal from multiscale edges.IEEE Trans. On Pattern Analysis and Machine Intelligence,Vol.14 ,Sep,1992
[11] S.Mallat,W.L.HWang,Singularity detection and processing with Wavelet. IEEE Trans. On Information Theroy,Vol,38,Feb,1992
[12] MT Shyamsunder,C Rajagopalan, K K Ray,Baldev Raj. A Comparative Study of Conventional and Artificial Neural Network Classifiers for Eddy Current Singal Classification. Insight,Jan, 1995
[13] Bruno Lebrun,Yves Jayet,Iean Claude Baboux. Pulsed Eddy Current Application to the Detection of Deep Cracks.Material Evalution,Nov,1995
[14]G.Mook,A.Tehemov, New Method of Signal Excitation and Processing in Magnetoinductive Testing of Structure and Alloys. NDT&E International,Oct, 1994
[15]Z Liu, K Tsukada. One-dimensional Eddy Current Mulfi-frequency Data Fusion:A Multi-resolution Analysis Approach. Insight,Vol.40,Apr, 1998
[16]Tadeusz Stepinski, Nora Maszi,Conjugate Speetum Filters for Eddy Current Signal Processing. Material Evaluation,Nov, 1995
[17]H.Komatsu. Basie Study on ECT Dated Evaluation Method with Neural Network. Non-Destructive Testing,1992
[18]I.T. Rekanos,T.P.Theodoulidis. Impendenee Inversion in Eddy Current Testing of Layered Planar via Neural Network,NDT&E International,Apr, 1997
[19]Martin Gibbs,Joe Campbell. Pulsed Eddy Current Inspection of Cracks Under Installed Fasteners. Material Evalution,Jan, 1991
[20]Z Liu,K Tsukada. Eddy Current Defect Characterization Using Neural Network. Material Evaluation, Mar, 1990
[21]杨福生．小波变换的工程分析与应用．北京：科学出版社，2001
[22]楼天顺，施阳．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
[23]楼天顺，施阳．基于MATLAB的系统分析与设计—神经网络．西安；西安电子科技大学出版社，1999
[24]吴步宁．小波变换及其在涡流无损检测中的应用．湖北汽车工业学院学报，2001，15(1)：28～31
[25]宋良．神经网络在电磁无损检测中的应用．信息技术，2001，(12)：3～4
[26]吴步宁，梁玉红．涡流阻抗图的傅里叶描述和分析．无损检测，2001，23(9)：367～371
[27]梁玉红，吴步宁．神经网络在涡流检测信号处理中的作用．无损检测，2000，22(11)：545～547
[28]徐科，徐金梧．小波分析在信号滤波中的应用．北京科技大学学报，1997，19(8)：382～385
[29]宋大雷，徐殿国．小波分析在钢丝绳无损检测中的应用．仪器仪表学报，1997，18(4)：151～155
[30]刘建峰，戚飞虎．基于子波变换的局域特征提取方法．红外与毫米波学报，1996，15(4)：123～127
[31]康宜华，武新军，杨叔子．磁性无损检测中的信号处理技术．无损检测，2000，22(6)：
255～259
[32]吴步宁，陈志祥．钢材硬度涡流无损检测技术的研究．无损检测，2000，22(6)：43～45
[33]钟洁，王素菊．涡流检测信号输入神经网络前特征值提取问题的论证．无损检测，1999，21(3)：97-98
[34]辛宁宁，路灿，盛剑霓．小波边缘检测技术在涡流检测中的应用．高压电技术，1999，25(9)；7～10
[35]孙晓云，路灿，袁斌．涡流无损检测中的定量分析．无损检测，2000，21(5)：195～199
[36]雷银照，马信山．人工神经网络在涡流检测中的应用．无损检测，1994，16(2)：31～33
[37]任吉林．涡流检测技术近20年的进展．无损检测，1998，20(5)：121～125
[38]胡光立，谢希文．钢的热处理，西安：西北工业大学出版社，1993
[39]任吉林．电磁无损检测．北京：航空工业出版社，1989
[40]姚天任，孙洪．现代数字信号处理．武汉：华中理工大学出版社，1999．11
[41]殷勤业，杨忠凯．模式识别与神经网络．北京；机械工业出版社，1992．9
[42]魏莉．涡流测试信号处理系统的研究．武汉：武汉理工大学，2001
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |