【作者】 白利兵；

【导师】 田书林；

【作者基本信息】 电子科技大学 ， 测试计量技术及仪器， 2013， 博士

【摘要】 无损检测是保障重大工程装备制造质量和运行安全的关键技术，电涡流脉冲热成像无损检测技术综合运用了电涡流与焦耳热现象，具有非接触、单次检测面积大、效率高等优势，是近年来无损检测领域的研究热点。目前，电涡流脉冲热成像中，多通过人工选择特定时刻的热图像实现缺陷检测，对人员要求高，劳动强度大。而材料表面热发射率不均，热传导等的存在，造成缺陷图像模糊，进一步增加了缺陷识别的难度。针对这些问题，本文对金属材料缺陷的电涡流热响应特征、热图像强化及自动识别提取技术进行了研究。首先，构建了电涡流脉冲热成像的涡流分布及热传导模型，分析了不同缺陷与材料表面热发射率的瞬态热响应特征；其次，以瞬态热响应特征为依据，研究了材料表面热发射率不均影响的抑制算法和图像增强算法；最后，根据缺陷的空间和瞬态热响应特征，提出了包含缺陷信息热图像的自动识别提取方法。本文的主要研究内容如下：1）分析了电涡流激励参数，如频率、幅度、时间、线圈材料等对涡流加热效果的影响，设计了电涡流脉冲热成像实验系统，并搭建了实验平台。选取不同电、磁参数的金属材料，制作了表面及深层缺陷的试件，以及可模拟材料表面热发射率变化的试件。2）通过理论分析、仿真和实验验证，探讨了被测材料的表面及深层缺陷的检测机理，深入分析了缺陷与电涡流和热传导的作用过程。在得出不同缺陷涡流和热量空间分布的同时，分别给出了其瞬态温度响应的变化过程和形态特征。另外，还讨论了材料表面热发射率对红外热响应的作用，指出了其对检测结果的干扰机理，为热发射率不均的抑制提供了理论依据。3）根据电、磁、热以及红外热辐射理论，系统的分析了不均匀热发射率条件下缺陷的红外热响应特征。采用频域、时域和统计学分析方法，如傅里叶变换、二次热平衡法、归一化和曲线形态相似性比较等，提取像素级的瞬态红外热响应特征，抑制材料表面热发射率不均的影响，强化突出缺陷信息。给出了每种方法的强化原理及依据，并进行了实验验证。另外，还从自动化程度、缺陷信息的完整性、应用前景等方面对其进行了比较，指出了每种方法的优劣。所提出的热发射率不均抑制算法，在不增加成本和操作复杂度的情况下，提高了电涡流脉冲热成像对复杂环境的适应能力，为其从实验室向现场应用的推广提供了支撑。4）以瞬态红外热响应为观测信号，构建了电涡流脉冲热成像的信号混叠模型。利用不同区域热响应彼此独立的特性，开展了基于主成分分析和独立成分分析的热响应盲分离算法研究，并进行了验证实验，给出了不同主成分（独立成分）的物理含义，实现了无先验信息条件下缺陷信息的强化。分别从信号的分离原理和分离后图像的物理含义等方面对两种分离方法进行了系统的比较分析。论证了两种求解方法在增强缺陷图像对比方面的优势和不足。所提出的图像增强算法，有效克服了横向热传导造成的缺陷图像模糊，提高了电涡流脉冲热成像的缺陷检测能力。5）根据缺陷的瞬态热响应及空间热分布特性，提出了基于混叠向量特征和主成分（独立成分）峰度系数的主成分（独立成分）识别算法，并给出了理论解释。在验证试验中，实现了包含缺陷信息热图像的自动提取，摆脱了对人工筛选热图像的依赖，降低了操作人员的劳动强度，提高了检测效率。更多还原

【Abstract】 Non-destructive testing and evaluation (NDT&E) is important to the structuralhealth monitoring (SHM) and quality control of key structure. Pulsed eddy currentthermography (ECPT), is new emerging NDT&E technique, which combines eddycurrent and thermography. The major advantage of ECPT over other techniques is thepotential in accurate non-contact inspection of a large area with a short time and largestandoff distances.ECPT mainly use a specific manual selected frame to detect the defects. It not onlyrequires experienced operators, but is labor consuming. And worse still, surfaceemissivity variation and heat diffusion causes "blurred effect", which seriously pullsdown the image quality and raises the difficulty of defect detection. To overcome thisissue, this thesis focus on the ECPT features of metallic material, infrared imageenhancement and automatic feature extraction. First, develop an ECPT model to analyzethe transient infrared response behavior in various defects and surface emissivityconditions; following the achievements, reduce the effect of surface emissivity variationto enhance the infrared image; lastly, proposes an automatic image extraction methodfor defect detection.The main contributions are:1) Analyze the influence of different eddy current excitation parameters, such asfrequency, amplitude, time, coil material and so on. Design and develop an ECPTexperiment set-up. Manufacture surface and undersurface defects using materials withdifferent electromagnetic properties. A sample with emissivity variation is also made.2) Through theoretical analyzing, simulation and verified experiment, discuss theinteraction of defect with eddy current and heat diffusion, and point out the mechanismof surface and undersurface defect detection. Not only present eddy current and heatdistribution, but describe the changing behavior of transient response. The effect ofsurface emissivity is also considered in detail, which provides fundamental support forimage enhancement.3) According to electromagnetism, thermology and thermal radiation, the defectfeature under emissivity variation is systematically discussed. Several methods areproposed to extract the pixel level features corresponding to material temperature, such as Fourier transform, two heat balance method, normalization, and similarity oftransient response behavior. Their advantages and disadvantages are also given from thepoint of information integrity, automation and application prospect. All of thesemethods can reduce the effect of emissivity variation, without any further cost andoperation. This benefit pushes ECPT development from a lab technique to in-situapplication.4) Taking the infrared response as an observation, a single channel blind sourceseparation model is developed for ECPT. Since the response of different areas areindependent, both principal component analysis (PCA) and independent componentanalysis (ICA) are employed to enhance image contrast without any training knowledge.Verified experiments are carried out, and the physical meaning of principle componentand independent component are detailly given. Their advantages and disadvantages arealso discussed from both of the mathematical and physical side. Both of these methodscan reduce the "blurred effect", and improve the probability of detection (POD) ofECPT.5) Kurtosis and normalized mixing vector are employed to quantify the spatial andtime features respectively. Both of them are used as indicators for automatic defectdetection. Their efficiency is verified through theoretical analysis and experiment. In theexperiment, defect image is automatically extracted, which shows the potential incutting down labour and raising efficiency.更多还原