【作者】 朱佳震；

【导师】 陈铁群； 刘桂雄；

【作者基本信息】 华南理工大学 ， 测试计量技术及仪器， 2010， 硕士

【摘要】 脉冲涡流是近几年迅速发展起来的一种无损检测新技术,其宽频谱激励方式在大面积复杂结构的检测中可获得较多的缺陷信息。本文在分析了国内外脉冲涡流无损检测技术研究现状的基础上,对脉冲涡流无损检测系统的研制、检测信号去噪及特征提取以及缺陷的分类识别等若干关键技术进行了研究,这对脉冲涡流检测技术的发展具有重要的学术价值和实际意义。论文研究内容得到广东省科技攻关计划项目(2006B12401001)资助,论文主要研究成果包括:(1)系统设计。设计了整个脉冲涡流缺陷检测系统,主要包括数据采集与存储系统,并对检测系统中的关键模块:脉冲涡流信号产生模块、信号调理模块及检测探头进行研制。本系统采用DDS技术,在FPGA芯片上设计脉冲涡流激励信号,降低了系统设计成本的同时提高了设计灵活性。(2)针对脉冲涡流检测信号噪声污染较严重的情况,分析比较了不同小波系的优缺点,采用Morlet小波对检测信号进行去噪,并采用信噪比、均方根误差及平滑度指标对去噪效果进行评价。针对传统时域波形特征提取容易受到噪声干扰的问题,采用Daubech -ies小波对去噪后的信号进行3层小波分解,并计算各层小波信号的能量值作为脉冲涡流信号的特征值,相对于传统的信号处理方法,可大大减少运算量。(3)分析蚁群算法和遗传算法的优缺点,结合其在模式识别中的应用,创新性地把蚁群算法和遗传算法应用于脉冲涡流表面缺陷和亚表面缺陷的识别。对比了单独采用蚁群算法和遗传算法进行缺陷分类识别的优点及不足,提出两种算法相结合的蚁群-遗传算法,实验证明,这种算法的分类识别能力优于采用单一的算法。更多还原

【Abstract】 Pulsed Eddy Current(PEC) nondestructive testing(NDT) is a new rapidly developing NDT technologies. Due to its broadband spectrum Excitation, it can get more defects information in large-scale testing of complex structures. Based on the current situation of Pulsed Eddy Current NDT technique both abroad and at home, the progress in the research of Nondestructive Test System, test signal denoising and feature extraction,and recognition of different defects are presented. The research work is funded under the research contract with the Guangdong government (2006B12401001) and is important to PEC inspection not only in theory but also in practice.The main research work is described as follows:(1) System design. The paper built the entire system PEC flaw detection including data collection, storage systems, and the key modules of detection system, such as PEC signal generation module, signal conditioning module and development of detection probes. Based on DDS technology and FPGA chip, the PEC excitation signal module can reduce design costs and improve design flexibility.(2) Addressing on the situation that PEC testing signals are disturbed seriously by environment noise pollution,comparing the advantages and disadvantages of different wavelets, this paper used Morlet wavelet to denoise the detection signals, and used signal to noise ratio, root mean square error and smoothness indicators to evaluate the denoising effect. As Traditional time-domain waveform feature extraction is susceptible to noise interference, the paper chose Daubechies wavelet on layer 3 to decompesite denoised signals, calculated the energy of each level of PEC wavelet signal and set the energy value as characteristic value. Compared with traditional signal compression, the progress can significantly reduce the computation.(3)Based on the analysis of advantages and disadvantages in ant colony algorithm and genetic algorithm, combining with its application to pattern recognition, the paper innovatively presented using ant colony algorithm and genetic algorithm in PEC surface defects and sub-surface defects recognition. Compared separately using ant colony algorithm or genetic algorithm in defects classification and recognition, the paper proposed the combination of the two algorithms--ant colony - the genetic algorithm. Experiments showed that this algorithm is better than both ant colony algorithm and genetic algorithm.更多还原