焊管内毛刺超声检测缺陷特征提取与智能识别研究

发布时间：2018-07-09 22:19

本文选题：超声检测 + 内毛刺　；参考：《辽宁科技大学》2017年硕士论文

【摘要】：油气输送过程中要求管道抗挤毁能力强、成本低。在工业生产中,为更好的预防因焊管的质量而带来的潜在安全隐患,延长焊管的使用寿命,对焊缝的质量检测需更加严格。随着技术的发展,超声波探伤技术在无损检测领域应用越来越广。因为传统的时频分析方法时频分辨率不高,对信号所携带的信息很难充分分析利用,检测精度和可靠性也没有明显提高。而超声检测信号的时频局部化特征更能有效的描述其信号特点,对超声信号的分析、识别以及检测精度的提升和可靠性的提高更加有效。在对焊管内毛刺缺陷的工程实际超声检测中,始终不能百分之百的实现某一缺陷的定性分类,需要不断的在这一领域进行探索、研究,实现对缺陷特征量的提取。基于此,本文利用MATLAB软件对焊管内毛刺的超声检测信号进行时频局部化分析,并对缺陷信号进行特征量提取,为今后内毛刺的检测打下坚实的基础。在内毛刺超声检测信号的研究过程当中发现,由于EEMD方法为克服传统的EMD方法中存在的模式混叠问题,在分析处理信号前,需加入大量高斯白噪声,这大大降低了EEMD分析信号的速度,将正交小波包作为EEMD方法的预滤波单元,有效地提高了其时效性。在实际的焊管内毛刺清除过程中,由于刮刀位置的不同和使用时间的长短通常会出现各种类型的毛刺,其超声检测结果也存在较大差异,综合幅值特征和厚度特征可判定有无毛刺及毛刺类型。由于超声探头在液体中的声束指向性差,且存在其他干扰波的影响,采用中心频率分别为2MHz和5MHz的水浸式线聚焦探头对外毛刺刮削干净、但内毛刺未经处理的ERW焊管进行超声信号的样本采集,对焊管内毛刺超声检测信号的缺陷特征量的提取和智能识别研究提供真实、可靠的分析数据。通过观察实验所测得的焊管内毛刺超声回波信号的波形,已知缺陷在波形中对应的采样点数、单个采样点所用时间、超声波在介质中的传播速度以及超声探头的入射角,就可以准确确定内毛刺所在的位置和深度。由于超声回波信号的部分有效信息淹没在了大量噪声当中,采用时频分辨率较高的EEMD方法有效地对信号进行多尺度分解,获得的结果完全可以体现原信号的信息特征。并结合基于Lorenz混沌系统的Volterra级数预测模型预测多尺度IMF信号的系统参数,通过矩阵奇异值的计算,得到系统的最小二乘解,提高了预测精度,且求得的奇异值几乎不受噪声的影响,根据求得的奇异值大小可以有效地判断焊管是否存在毛刺,验证了本文中所使用的EEMD-Volterra方法对内毛刺检测的正确性和有效性。
[Abstract]:In the process of oil and gas transportation, it is required that the pipeline has strong ability to resist squeezing damage and low cost. In industrial production, in order to prevent the potential hidden danger caused by the quality of welded pipe, prolong the service life of welded pipe, and test the quality of weld seam more strictly. With the development of technology, ultrasonic flaw detection technology is more and more widely used in the field of nondestructive testing. Because the time-frequency resolution of the traditional time-frequency analysis method is not high, it is difficult to fully analyze and utilize the information carried by the signal, and the accuracy and reliability of the detection are not obviously improved. The time-frequency localization feature of ultrasonic detection signal can describe the signal characteristics more effectively, and it is more effective for ultrasonic signal analysis, identification, detection accuracy and reliability. In the engineering practice of ultrasonic detection of burr defects in welded pipes, the qualitative classification of certain defects can not be realized 100%. It is necessary to continuously explore and study in this field, to achieve the extraction of defect characteristics. Based on this, this paper makes use of MATLAB software to analyze the ultrasonic detection signal of welded pipe internal burr by time-frequency localization, and extracts the characteristic quantity of defect signal, which lays a solid foundation for the detection of internal burr in the future. During the study of the internal burr ultrasonic signal, it is found that in order to overcome the mode aliasing problem in the traditional Gao Si method, a large amount of white noise should be added before the signal is analyzed and processed. This greatly reduces the speed of the EEMD analysis signal. The orthogonal wavelet packet is used as the pre-filter unit of the EEMD method, and its time-efficiency is improved effectively. In the actual internal burr removal process of welded pipe, because of the different positions of scraper and the length of service time, there are usually various types of burrs, and the ultrasonic testing results are also quite different. Comprehensive amplitude and thickness characteristics can be used to determine whether or not burr and burr type exist. Due to the poor directivity of ultrasonic probe in liquid and the influence of other interference waves, the water immersion wire focusing probe with central frequency of 2 MHz and 5 MHz was used to scratch the external burr. However, the untreated ERW pipe with internal burr is used to collect ultrasonic signal samples, to extract the defect characteristic quantity of ultrasonic detection signal in welded pipe and to provide real and reliable analysis data. By observing the waveform of the burr ultrasonic echo signal in the welded pipe, the sampling points corresponding to the known defects in the waveform, the time used for a single sampling point, the velocity of ultrasonic wave propagation in the medium and the incidence angle of the ultrasonic probe are observed. The position and depth of the internal burr can be determined accurately. Because part of the effective information of ultrasonic echo signal is submerged in a large number of noises, the EEMD method with high time-frequency resolution is used to decompose the signal effectively, and the obtained results can fully reflect the information characteristics of the original signal. Combining with the Volterra series prediction model based on Lorenz chaotic system to predict the system parameters of multi-scale IMF signal, the least square solution of the system is obtained by calculating the singular value of matrix, and the prediction accuracy is improved. The obtained singular value is almost unaffected by noise. According to the obtained singular value, the existence of burr in welded pipe can be effectively determined, and the correctness and validity of EEMD-Volterra method used in this paper for internal burr detection are verified.
【学位授予单位】：辽宁科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TE973.6

【参考文献】