采用改进的希尔伯特黄变换的损伤检测特征提取方法
|
摘要
针对损伤检测中存在的难以提取准确描述损伤信息特征的问题,提出了一种基于改进的希尔伯特黄变换的有效特征选择方法。该方法首先对信号进行小波包分解和重构,得到一系列窄带信号,然后对其分别做经验模式分解得到若干个固有模式函数,基于互信息量筛选出真实固有模式函数,并对真实固有模式函数分别做希尔伯特变换得到瞬时属性,最后根据瞬时属性提取相关的时频特征。在损伤检测应用中,采用前向反馈神经网络对提取的特征进行分类,结果表明,利用该方法提取的音频信号特征非常有效,综合分类性能指标F相比于原始的希尔伯特黄变换提高了7.7%。
An effective feature selection method based on an improved Hilbert-Huang transform is proposed to focus on the problem that it is difficult to extract the characteristic features of the defect information in acoustic defect detection systems.Firstly,audio signals are decomposed and reconstructed by the wavelet package to obtain a series of narrow-band signals.Secondly,all of the narrow-band signals are respectively decomposed via the empirical mode decomposition method to obtain several intrinsic mode function components.Then the real intrinsic mode function components are screened out based on mutual information.Instantaneous attributes are obtained from these intrinsic mode function components through Hilbert transformation.Finally,relevant time-frequency features are extracted based on these instantaneous attributes.The extracted features are classified by a back propagation neural network in acoustic defect detection.Results show that the features of audio signals extracted by the proposed method is very effective and the comprehensive classification performance index Fimproves by 7.7% compared with the original Hilbert-Huang transform.
An effective feature selection method based on an improved Hilbert-Huang transform is proposed to focus on the problem that it is difficult to extract the characteristic features of the defect information in acoustic defect detection systems.Firstly,audio signals are decomposed and reconstructed by the wavelet package to obtain a series of narrow-band signals.Secondly,all of the narrow-band signals are respectively decomposed via the empirical mode decomposition method to obtain several intrinsic mode function components.Then the real intrinsic mode function components are screened out based on mutual information.Instantaneous attributes are obtained from these intrinsic mode function components through Hilbert transformation.Finally,relevant time-frequency features are extracted based on these instantaneous attributes.The extracted features are classified by a back propagation neural network in acoustic defect detection.Results show that the features of audio signals extracted by the proposed method is very effective and the comprehensive classification performance index Fimproves by 7.7% compared with the original Hilbert-Huang transform.
引文
[1]刘长福,郝晓军,牛晓光,等.基于BP神经网络的瓷绝缘子振动声学检测结果分类[J].无损检测,2014,36(1):1-4.LIU Changfu,HAO Xiaojun,NIU Xiaoguang,et al.Classification of the insulator inspection data by acoustic vibration based on BP neural network[J].Nondestructive Testing,2014,36(1):1-4.
[2]焦敬品,李勇强,吴斌,等.基于BP神经网络的管道泄漏声信号识别方法研究[J].仪器仪表学报,2016,37(11):2588-2596.JIAO Jingpin,LI Yongqiang,WU Bin,et al.Research on acoustic signal recognition method for pipeline leakage with BP neural network[J].Chinese Journal of Scientific Instrument,2016,37(11):2588-2596.
[3]姜瑞涉,王俊,陆秋君,等.鸡蛋敲击响应特性与蛋壳裂纹检测[J].农业机械学报,2005,36(3):75-78.JIANG Ruishe,WANG Jun,LU Qiujun,et al.Eggshell crack detection by frequency analysis of dynamic resonance[J].Transactions of the Chinese Society for Agricultural Machinery,2005,36(3):75-78.
[4]刘毅,张彩明,赵玉华,等.基于多尺度小波包分析的肺音特征提取与分类[J].计算机学报,2006,29(5):769-777.LIU Yi,ZHANG Caiming,ZHAO Yuhua,et al.The feature extraction and classification of lung sounds based on wavelet packet multiscale analysis[J].Chinese Journal of Computers,2006,29(5):769-777.
[5]王宏宇.基于希尔伯特-黄变换的语音识别特征提取方法研究[D].广州:华南理工大学,2012:6-7.
[6]HUANG N E,SHEN Z,LONG S R.A new view of nonlinear water waves:the Hilbert spectrum[J].Annual Review of Fluid Mechanics,1999,31(1):417-457.
[7]杨光松.损伤力学与复合材料损伤[M].北京:国防工业出版社,1995:6-16.
[8]PENG Z K,TSE P W,CHU F L.A comparison study of improved Hilbert-Huang transform and wavelet transform:application to fault diagnosis for rolling bearing[J].Mechanical Systems&Signal Processing,2005,19(5):974-988.
[9]PENG Z K,TSE P W,CHU F L.An improved Hilbert-Huang transform and its application in vibration signal analysis[J].Journal of Sound&Vibration,2005,286(1/2):187-205.
[10]LIU Zehua,SU Jing,QIN Cheng,et al.Wavelet packet decomposition and grey relational analysis application in fault diagnosis of aero hydraulic pump[C]∥2016IEEE/CSAA International Conference on Aircraft Utility Systems.Piscataway,NJ,USA:IEEE,2016:734-738.
[11]孙延奎.小波分析及其应用[M].北京:机械工业出版社,2005:253-257.
[12]COIFMAN R R,WICKERHAUSER M V.Entropybased algorithms for best basis selection[J].IEEETransactions on Information Theory,1992,38(2):713-718.
[13]BHUIYAN S,KHAN J,MURPHY G.Wavelet packet decomposition for power quality monitoring in smart grid[C]∥52nd Annual Meeting on IEEE Industry Applications Society.Piscataway,NJ,USA:IEEE,2016:1-8.
[14]余春华,张有峰,齐杏林,等.一种基于小波分析的多层粘接结构缺陷信号提取方法[J].无损检测,2012,34(5):13-15.YU Chunhua,ZHANG Youfeng,QI Xinglin,et al.Amethod for extracting defect signal of multi-layer adhesive structure based on wavelet analysis[J].Nondestructive Testing,2012,34(5):13-15.
[15]YANG W C,ZHANG P L,WU D H,et al.A method of false component discriminant of EMD based on Kolmogorov-Smirnov test[C]∥2nd International Conference on Mechanical Engineering,Industrial Electronics and Informatization.Zurich-Durnter,Switzerland:Trans Tech Publications Ltd,2013:2005-2008.
[16]BAO C,HAO H,LI Z X,et al.Time-varying system identification using a newly improved HHT algorithm[J].Computers and Structures,2009,87(23):1611-1623.
[17]PENG H,LONG F,DING C.Feature selection based on mutual information criteria of max-dependency,max-relevance,and min-redundancy[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2005,27(8):1226-1238.
[18]SAINI L M,SONI M K.Artificial neural network based peak load forecasting using LevenbergMarquardt and quasi-Newton methods[C]∥IEEEProceedings:Generation,Transmission and Distribution.Piscataway,NJ,USA:IEEE,2002:578-584.
[2]焦敬品,李勇强,吴斌,等.基于BP神经网络的管道泄漏声信号识别方法研究[J].仪器仪表学报,2016,37(11):2588-2596.JIAO Jingpin,LI Yongqiang,WU Bin,et al.Research on acoustic signal recognition method for pipeline leakage with BP neural network[J].Chinese Journal of Scientific Instrument,2016,37(11):2588-2596.
[3]姜瑞涉,王俊,陆秋君,等.鸡蛋敲击响应特性与蛋壳裂纹检测[J].农业机械学报,2005,36(3):75-78.JIANG Ruishe,WANG Jun,LU Qiujun,et al.Eggshell crack detection by frequency analysis of dynamic resonance[J].Transactions of the Chinese Society for Agricultural Machinery,2005,36(3):75-78.
[4]刘毅,张彩明,赵玉华,等.基于多尺度小波包分析的肺音特征提取与分类[J].计算机学报,2006,29(5):769-777.LIU Yi,ZHANG Caiming,ZHAO Yuhua,et al.The feature extraction and classification of lung sounds based on wavelet packet multiscale analysis[J].Chinese Journal of Computers,2006,29(5):769-777.
[5]王宏宇.基于希尔伯特-黄变换的语音识别特征提取方法研究[D].广州:华南理工大学,2012:6-7.
[6]HUANG N E,SHEN Z,LONG S R.A new view of nonlinear water waves:the Hilbert spectrum[J].Annual Review of Fluid Mechanics,1999,31(1):417-457.
[7]杨光松.损伤力学与复合材料损伤[M].北京:国防工业出版社,1995:6-16.
[8]PENG Z K,TSE P W,CHU F L.A comparison study of improved Hilbert-Huang transform and wavelet transform:application to fault diagnosis for rolling bearing[J].Mechanical Systems&Signal Processing,2005,19(5):974-988.
[9]PENG Z K,TSE P W,CHU F L.An improved Hilbert-Huang transform and its application in vibration signal analysis[J].Journal of Sound&Vibration,2005,286(1/2):187-205.
[10]LIU Zehua,SU Jing,QIN Cheng,et al.Wavelet packet decomposition and grey relational analysis application in fault diagnosis of aero hydraulic pump[C]∥2016IEEE/CSAA International Conference on Aircraft Utility Systems.Piscataway,NJ,USA:IEEE,2016:734-738.
[11]孙延奎.小波分析及其应用[M].北京:机械工业出版社,2005:253-257.
[12]COIFMAN R R,WICKERHAUSER M V.Entropybased algorithms for best basis selection[J].IEEETransactions on Information Theory,1992,38(2):713-718.
[13]BHUIYAN S,KHAN J,MURPHY G.Wavelet packet decomposition for power quality monitoring in smart grid[C]∥52nd Annual Meeting on IEEE Industry Applications Society.Piscataway,NJ,USA:IEEE,2016:1-8.
[14]余春华,张有峰,齐杏林,等.一种基于小波分析的多层粘接结构缺陷信号提取方法[J].无损检测,2012,34(5):13-15.YU Chunhua,ZHANG Youfeng,QI Xinglin,et al.Amethod for extracting defect signal of multi-layer adhesive structure based on wavelet analysis[J].Nondestructive Testing,2012,34(5):13-15.
[15]YANG W C,ZHANG P L,WU D H,et al.A method of false component discriminant of EMD based on Kolmogorov-Smirnov test[C]∥2nd International Conference on Mechanical Engineering,Industrial Electronics and Informatization.Zurich-Durnter,Switzerland:Trans Tech Publications Ltd,2013:2005-2008.
[16]BAO C,HAO H,LI Z X,et al.Time-varying system identification using a newly improved HHT algorithm[J].Computers and Structures,2009,87(23):1611-1623.
[17]PENG H,LONG F,DING C.Feature selection based on mutual information criteria of max-dependency,max-relevance,and min-redundancy[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2005,27(8):1226-1238.
[18]SAINI L M,SONI M K.Artificial neural network based peak load forecasting using LevenbergMarquardt and quasi-Newton methods[C]∥IEEEProceedings:Generation,Transmission and Distribution.Piscataway,NJ,USA:IEEE,2002:578-584.