基于双高斯衰减模型的超声回波处理方法
|
摘要
信号降噪与特征提取是超声检测数据处理的关键技术.基于超声信号有特定结构而噪声和超声信号的结构无关,本文提出一种旨在解决强噪声背景下超声回波的参数估计和降噪问题的方法.该方法将超声回波的参数估计和降噪问题转换为函数优化问题,首先根据工程经验建立超声信号的双高斯衰减数学模型,然后根据观测回波和建立的超声信号模型确定目标函数,接着选择人工蜂群算法对目标函数进行优化从而得到参数的最优估计值,最后由估计出的参数根据建立的超声信号数学模型重构出无噪的超声估计信号.通过仿真和实验表明本文方法可以准确估计出信噪比大于-10 dB的含噪超声回波中的无噪信号,且效果优于基于自适应阈值的小波降噪方法和经验模态分解方法;此外相比常用的指数模型和高斯模型,本文提出的双高斯衰减超声信号模型与实测超声信号更接近,其均方误差为9.4×10~(-5),波形相似系数为0.98.
Ultrasonic non-destructive testing, which is one of the most important and rapidly developed nondestructive testing technologies, is widely used in industrial production and other areas. Signal de-noising and feature extraction, whose performance directly affects the evaluation of non-destructive testing results, are the key technologies of ultrasonic non-destructive testing data processing, and also the core elements of ultrasonic non-destructive testing. Therefore, the research on them has important academic significance and practical value. In order to solve the problem of parameter estimation and noise reduction of ultrasonic echo in strong noise background, a novel ultrasonic echo processing method is proposed in this paper. The principle of the proposed method in this paper is as follows. The ultrasonic echo, which is generated by modulating the ultrasonic transducer, has a specific structure, but the noise in practical engineering is usually a Gauss random process, therefore the noise is independent of the ultrasonic signal structure. In this paper, the problem of parameter estimation and noise reduction of ultrasonic echo signal are converted into a function optimization problem by establishing the model of ultrasonic signal, determining the objective function, optimizing the objective function, estimating the parameters, and reconstructing the ultrasonic signal. Firstly, a dual gaussian attenuation mathematical model of ultrasonic signal is established based on practical engineering experience.Secondly, the cosine similarity function, an effective measure of data sequence similarity, is selected as an objective function according to the observed echo and the established ultrasonic signal model. Thirdly, the artificial bee colony algorithm is selected to optimize the objective function to obtain the optimal estimation parameters of the ultrasonic signal from the noisy ultrasonic echo. Fourthly, the estimation of de-noising ultrasonic signal is reconstructed by the optimal parameters based on the established ultrasonic signal mathematical model. The processing results of simulated ultrasonic echoes and measured ultrasonic echoes show that the proposed method can accurately estimate the parameters of ultrasonic signal from strong background noise whose signal-to-noise ratio is lowest, as low as-10 dB. In addition, compared with the adaptive threshold based wavelet method and empirical mode decomposition method, the proposed method in this paper shows the good de-noising performance. Furthermore, compared with the commonly used exponential model and Gaussian model in numerical and simulation analysis, the proposed dual gaussian attenuation mathematical model of ultrasonic signal in this paper can well simulate the measured ultrasonic signal, with a mean square error of 9.4 x10-5 and normalized correlation coefficient of 0.98.
Ultrasonic non-destructive testing, which is one of the most important and rapidly developed nondestructive testing technologies, is widely used in industrial production and other areas. Signal de-noising and feature extraction, whose performance directly affects the evaluation of non-destructive testing results, are the key technologies of ultrasonic non-destructive testing data processing, and also the core elements of ultrasonic non-destructive testing. Therefore, the research on them has important academic significance and practical value. In order to solve the problem of parameter estimation and noise reduction of ultrasonic echo in strong noise background, a novel ultrasonic echo processing method is proposed in this paper. The principle of the proposed method in this paper is as follows. The ultrasonic echo, which is generated by modulating the ultrasonic transducer, has a specific structure, but the noise in practical engineering is usually a Gauss random process, therefore the noise is independent of the ultrasonic signal structure. In this paper, the problem of parameter estimation and noise reduction of ultrasonic echo signal are converted into a function optimization problem by establishing the model of ultrasonic signal, determining the objective function, optimizing the objective function, estimating the parameters, and reconstructing the ultrasonic signal. Firstly, a dual gaussian attenuation mathematical model of ultrasonic signal is established based on practical engineering experience.Secondly, the cosine similarity function, an effective measure of data sequence similarity, is selected as an objective function according to the observed echo and the established ultrasonic signal model. Thirdly, the artificial bee colony algorithm is selected to optimize the objective function to obtain the optimal estimation parameters of the ultrasonic signal from the noisy ultrasonic echo. Fourthly, the estimation of de-noising ultrasonic signal is reconstructed by the optimal parameters based on the established ultrasonic signal mathematical model. The processing results of simulated ultrasonic echoes and measured ultrasonic echoes show that the proposed method can accurately estimate the parameters of ultrasonic signal from strong background noise whose signal-to-noise ratio is lowest, as low as-10 dB. In addition, compared with the adaptive threshold based wavelet method and empirical mode decomposition method, the proposed method in this paper shows the good de-noising performance. Furthermore, compared with the commonly used exponential model and Gaussian model in numerical and simulation analysis, the proposed dual gaussian attenuation mathematical model of ultrasonic signal in this paper can well simulate the measured ultrasonic signal, with a mean square error of 9.4 x10-5 and normalized correlation coefficient of 0.98.
引文
[1] Fulin J 2017 Mater. Eval. 75 456
[2] Burkov M V, Eremin A V, Lyubutin P S, Byakov A V, Panin S V 2017 Russ. J. Nondestr. Test. 53 817
[3] Lu Z K, Yang C, Qin D H, Luo Y L 2016 Signal Process. 120607
[4] Wang X K, Guan S Y, Hua L, Wang B, He X M 2019Ultrasonics 91 161
[5] Meng M, Chua Y J, Wouterson E, Ong C P K 2017Neurocomputing 257 128
[6] Sun L F, Wang T T, Xu M F, Li X, Pu H 2017 Chin. J. Sci.Instrum.38 2879(in Chinese)[孙灵芳,王彤彤,徐曼菲,李霞,朴亨2017仪器仪表学报38 2879]
[7] Wang D W, Wang Z B 2018 Acta Phys. Sin. 67 210501(in Chinese)[王大为,王召巴2018物理学报67 210501]
[8] Qi A L, Zhang G M, Dong M, Ma H W, Harvey D M 2018Ultrasonics 88 1
[9] Wu J, Zhu J G, Yang L H, Shen M T, Xue B, Liu Z X 2014Measurement 47 433
[10] Fang Z H, Hu L, Qin L H, Mao K, Chen W Y, Fu X 2017Flow Meas. Instrum. 55 1
[11] Demirli R, Saniie J 2001 IEEE Trans. Ultrason. Ferr. 48 787
[12] Rathee N, Ganotra D 2018 Signal Image Video P. 12 1141
[13] Kirkpatrick S, Gelatt C D, Vecchi M P 1983 Science 220 671
[14] Tamizharasan T, Barnabas J K, Pakkirisamy V 2012 P. I.Mech. Eng. B:J. Eng. 226 1159
[15] Li S Y, Du Z H, Wu M Y, Zhu J, Li S L 2001 Acta Phys.Sin. 50 1260(in Chinese)[李树有,都志辉,吴梦月,朱静,李三立2001物理学报50 1260]
[16] Hasanoglu M S, Dolen M 2018 Eng. Optimiz. 50 2013
[17] Zhan Z H, Zhang J, Li Y, Chung H S H 2009 IEEE Trans.Cybern. 39 1362
[18] Li Y B, Zhang B L, Liu Z X, Zhang Z Y 2014 Acta Phys. Sin.63 160504(in Chinese)[李一博,张博林,刘自鑫,张振宇2014物理学报63 160504]
[19] Karaboga D, Ozturk C 2011 Appl. Soft Comput. 11 652
[20] Kiran M S, Findik O 2015 Appl. Soft Comput. 26 454
[21] Li G M, Hu Z H 2016 Acta Phys. Sin. 65 230501(in Chinese)[李广明,胡志辉2016物理学报65 230501]
[22] Zhu J J, Li X L 2017 Healthcare. Technol. Lett. 4 134
[23] Tang J, Gao L, Peng L, Zhou Q 2007 High Voltage Eng. 1266(in Chinese)[唐炬,高丽,彭莉,周倩2007高电压技术1266]
[2] Burkov M V, Eremin A V, Lyubutin P S, Byakov A V, Panin S V 2017 Russ. J. Nondestr. Test. 53 817
[3] Lu Z K, Yang C, Qin D H, Luo Y L 2016 Signal Process. 120607
[4] Wang X K, Guan S Y, Hua L, Wang B, He X M 2019Ultrasonics 91 161
[5] Meng M, Chua Y J, Wouterson E, Ong C P K 2017Neurocomputing 257 128
[6] Sun L F, Wang T T, Xu M F, Li X, Pu H 2017 Chin. J. Sci.Instrum.38 2879(in Chinese)[孙灵芳,王彤彤,徐曼菲,李霞,朴亨2017仪器仪表学报38 2879]
[7] Wang D W, Wang Z B 2018 Acta Phys. Sin. 67 210501(in Chinese)[王大为,王召巴2018物理学报67 210501]
[8] Qi A L, Zhang G M, Dong M, Ma H W, Harvey D M 2018Ultrasonics 88 1
[9] Wu J, Zhu J G, Yang L H, Shen M T, Xue B, Liu Z X 2014Measurement 47 433
[10] Fang Z H, Hu L, Qin L H, Mao K, Chen W Y, Fu X 2017Flow Meas. Instrum. 55 1
[11] Demirli R, Saniie J 2001 IEEE Trans. Ultrason. Ferr. 48 787
[12] Rathee N, Ganotra D 2018 Signal Image Video P. 12 1141
[13] Kirkpatrick S, Gelatt C D, Vecchi M P 1983 Science 220 671
[14] Tamizharasan T, Barnabas J K, Pakkirisamy V 2012 P. I.Mech. Eng. B:J. Eng. 226 1159
[15] Li S Y, Du Z H, Wu M Y, Zhu J, Li S L 2001 Acta Phys.Sin. 50 1260(in Chinese)[李树有,都志辉,吴梦月,朱静,李三立2001物理学报50 1260]
[16] Hasanoglu M S, Dolen M 2018 Eng. Optimiz. 50 2013
[17] Zhan Z H, Zhang J, Li Y, Chung H S H 2009 IEEE Trans.Cybern. 39 1362
[18] Li Y B, Zhang B L, Liu Z X, Zhang Z Y 2014 Acta Phys. Sin.63 160504(in Chinese)[李一博,张博林,刘自鑫,张振宇2014物理学报63 160504]
[19] Karaboga D, Ozturk C 2011 Appl. Soft Comput. 11 652
[20] Kiran M S, Findik O 2015 Appl. Soft Comput. 26 454
[21] Li G M, Hu Z H 2016 Acta Phys. Sin. 65 230501(in Chinese)[李广明,胡志辉2016物理学报65 230501]
[22] Zhu J J, Li X L 2017 Healthcare. Technol. Lett. 4 134
[23] Tang J, Gao L, Peng L, Zhou Q 2007 High Voltage Eng. 1266(in Chinese)[唐炬,高丽,彭莉,周倩2007高电压技术1266]