【作者】 何赟泽；

【导师】 罗飞路；

【作者基本信息】 国防科学技术大学 ， 仪器科学与技术， 2012， 博士

【摘要】 美国前总统里根曾说过：“没有先进的无损检测技术，美国就不可能享有在众多领域的领先地位”。航空、航天、铁路、核电、新材料等领域的快速发展为无损检测既提供了很好的机遇，同时也提出了极大的挑战。瞬态式电磁检测技术是最具发展前景的一类无损检测技术，具有检测速度快、深度大、灵敏度高、频谱宽、易定量等优势，在金属和复合材料的检测评估中扮演着非常重要的角色。本文从研究脉冲涡流检测技术出发，以涡流脉冲热成像检测技术为重点，对瞬态式电磁无损检测中的缺陷识别与评估方法进行了系统深入的研究，主要研究内容及创新点如下：一、研究了材料属性对脉冲涡流时域和频域响应信号的影响，提出了可表征材料磁导率和电导率的特征值，解决了多种材料和结构中典型缺陷的检测评估问题。分别在时域和频域内研究了材料属性和提离的变化对脉冲涡流时频域响应的影响，结果表明归一化技术可有效消除提离的影响。从时域响应中提取了可表征材料电导率和磁导率（磁场强度）的特征值。使用可表征材料电导率的特征值对航空铝合金材料的应力变化、蜂窝结构复合材料中的分层缺陷和钢结构中的腐蚀缺陷进行了检测和评估。结果表明应力变化与特征值为线性关系，特征值可对蜂窝结构中的缺陷性质进行判断，一年内腐蚀可使用功率函数进行评估。采用可表征材料磁场强度的特征值对碳纤维复合材料中的撞击缺陷进行了评估，表明4J以上的撞击可导致凹坑及损伤，在10J以下损伤面积与撞击能量有单调关系。二、研究了方向性脉冲涡流传感器的时频域响应，提出了可表征纵深位置信息的特征值，抑制了提离和层间隙的负面影响，解决了多层结构中缺陷的分类识别问题。设计了方向性脉冲涡流传感器，引入差分技术提取了可表征纵深位置信息的特征值，分析了多层结构中提离和层间隙对该类型特征值用于缺陷分类的影响。结合频域优化和主成分分析方法，改进了脉冲涡流特征值提取方法，对提离和层间隙进行了抑制，解决了四种缺陷的分类问题。针对目前分类方法多依靠人工的问题，采用改进的支持向量机技术实现了多种缺陷的自动分类识别。并比较了主成分分析和独立成分分析，时域响应和频域响应在该方法中对缺陷分类识别性能的影响。三、研究了涡流持续加热引起的纵向和横向热传递现象，提出了涡流阶跃热成像检测技术，建立了偏离时间和下表面缺陷深度的定量关系，为下表面缺陷和表面平行缺陷的检测评估提供了有效手段。目前的缺陷评估方法主要基于涡流场扰动的形式，无法对平行于线圈的表面缺陷和超出集肤深度的下表面缺陷进行检测。为解决这个问题，提出了涡流阶跃热成像检测技术。针对铁磁性材料，理论推断理想的温度-时间平方根曲线是直线形式。当有缺陷时，该曲线会发生偏离。利用温度-时间平方根曲线对缺陷进行识别。提取了偏离时间作为特征值，采用有限元分析和实验建立了缺陷深度与偏离时间的对应关系。研究了横向热传递现象，建立了表面平行于线圈缺陷的检测方法。四、研究了涡流瞬态加热后导致的热传递现象，提出了涡流瞬态热成像检测技术，采用时域温度曲线及其特征值，在反射模式和穿透模式下建立了缺陷检测和定量评估方法。采用解析分析和三维有限元模型，在反射和穿透两种不同的工作模式下，对内部缺陷和下表面缺陷的定量评估方法进行了研究。得到了峰值时间与缺陷深度的对应关系。为进一步提高定量方法的实用性，提出了基于对数域温度曲线的缺陷评估方法。针对理想的对数域温度曲线为直线形式，采用偏离现象来对缺陷进行识别。提取了对数域偏离时间对下表面缺陷的深度进行定量。针对钢试件中的下表面缺陷，通过实验建立了反射模式下缺陷深度的定量评估方法。五、研究了瞬态式涡流热成像的先进数据处理方法，提出了基于傅立叶变化的涡流脉冲相位热成像检测技术和基于统计分析的图像重构方法，改善了缺陷检测效果，提高了缺陷检测能力。结合脉冲热成像和锁相热成像的优点，提出了涡流脉冲相位热成像检测技术。在差分相位谱上提取了差分过零频率、最小相位、最小相位频率等特征值，对下表面缺陷的深度进行了定量。同时，采用相位谱图检测缺陷，抑制了非均匀加热现象，改善了缺陷检测效果。提出了基于统计分析的图像重构方法，使用主成分分析和独立成分分析方法重构了图像序列，得到了新的二维图像来识别缺陷，有效低抑制了横向热传递导致的模糊效应，提高了微缺陷和深层缺陷的检测能力。针对钢结构中的腐蚀气泡，对提出的方法进行了比较研究，建立了腐蚀气泡的检测方法。六、研究了碳纤维复合材料中的涡流加热效应，提出了碳纤维复合材料典型缺陷检测方法，建立了分层和撞击缺陷的检测评估方法，理解了不同能量撞击的破坏行为。在分析各向同性材料中缺陷识别方法的基础上，针对碳纤维复合材料的非均匀性和各向异性，提出了碳纤维复合材料中缺陷的识别方法。通过实验得知了不同缺陷对瞬态温度信号的影响，碳纤维结构主要体现在早期的加热阶段，撞击缺陷主要体现在晚期的加热阶段，分层缺陷信息主要体现在晚期的冷却阶段。通过该结论对图像重构方法进行了优化，提高了分层缺陷和撞击缺陷的检测效果。最终，获得了分层缺陷深度与特征值的对应关系，理解了不同能量撞击的破坏行为。结果表明，6J以上撞击可以导致损伤，10J以上的撞击可以导致环状损伤。论文不仅对脉冲涡流时频域特征分析、电磁热多物理场耦合、涡流热的三维传递等科学问题进行了探索，而且为航空铝合金应力变化、钢结构腐蚀、蜂窝结构分层、碳纤维复合材料撞击等多种缺陷的检测和评估提供了有效方法和手段。论文的研究成果为瞬态式电磁无损检测技术的发展提供了很好的理论基础和技术指导。更多还原

【Abstract】 With the growing interest to use metal alloy and composite structures in aerospace,marine, traffic and other industrial fields, much attention is devoted to the developmentof non-destructive testing (NDT) techniques. Transient electromagnetic (EM) NDTtechniques have a lot of advantages, such as high speed, great depth, high sensitivity,width spectrum, low cost, and easy to quantification, which are widely investigated inthese fields. Two typical transient EM NDT techniques pulsed eddy current (PEC)testing and eddy current thermography testing (ECPT) are investigated in the thesis. Thebrief of these researches and the novel approaches are listed as follows.(1) The PEC features representing material properties are proposed andutilized to solve out the detection problem of the typical defects in aluminium alloy,honeycomb sandwich, steel and carbon fiber reinforcement plastic. PEC response isa complex mix of many factors including conductivity, permeability, lift-off andmaterial thickness variation, which should be all taken into account in PEC testing. Theinfluences of material perperties on PEC responses in time domain and frequencydomain are investigated and normalization technique is used to reduce the lift-off effect.After this, two time-domain features, representing conductivity and permeability(magnetic field intensity) are extracted. These features are utilized to measure stress inaluminum alloy, to detect defect in honeycomb sandwich structure, to evaluatelow-energy impact defect in CFRP material, and to characterize atmospheric corrosionon steel samples in Chapter2. At alst, the methods for evaluating these defects are built.(2) The PEC features representing longitudinal locations are proposed andutilized to solve out the classification problem of the typical defects in multi-layerstructures wildly used in aircrafts. In Chapter3, the directional PEC probe providinguniform eddy current is designed. Then, the PEC feature extraction techniques arestudied in both time and frequency domains. PEC frequency response optimization isinvestigated and used in combination with principal component analysis (PCA) toeliminate the lift-off and interlayer air gap and to classify the defects in multi-layerstructures. Current PEC defect classification methods require highly trained personneland the results are usually influenced by human subjectivity. Therefore, automateddefect classification is desirable in a PEC instrument. The optimized support vectormachine (SVM) is used to build the classifier model and predict the types of defects.PCA and independent component analysis (ICA) are investigated for feature extractionand compared for classification results using SVM. Two-layer Al-Mn alloy specimenswith four kinds of defects are used for classification. The experimental results show thatthe proposed methods have great potential for in-situ defect inspection of multi-layeraircraft structures. (3) Eddy current step heating thermography (ECSHT) testing and relatedquantification methods are proposed in order to solve out the characterizationproblem of subsurface defects and parallel surface defects. Because the conventionalmethods based on eddy current field interruption is invalid to evaluate the subsurfacedefects whichs are beyond the skin depth, Chapter4proposed eddy current step heatingthermography testing based on heat diffusion after long time eddy current excitation. Ifneglecting the skin depth, the ideal temperature-time1/2line for defect-free are is a linearline, while the temperature-time1/2line for defect will separate from that of defect-free.Thus, the temperature-time1/2line can be used to detect the defect. The characteristicfeature separation time is extracted and two features representing separation time aredefined to realize the defect quantification. The conventional methods based on eddycurrent field distribution are difficult to detect the defects which are parallel to theinductive coil. This Chapter also proposed eddy current pulsed thermography (ECPT)based on the lateral heat conduction for detection of these defects. The character bylateral heat conduction is addressed to detect the parallel defects. Due to significanttemperature gradient in the direction of lateral heat conduction, the spatial derivative ofthe thermogram is proposed to improve the defect detectability.(4) Defect quantification methods of eddy current pulsed thermography(ECPT) in time domain and logarithm domain are proposed to solve out thequantification problem of subsurface defects. Chapter5proposed the defectcharacterization methods of eddy current pulsed thermography (ECPT) based on heatdiffusion after eddy current pulsed excitation. The proposed methods are investigatedunder transmission mode and reflection mode through1D analytical analysis,3Dnumerical studies, and experimental studies. Time-based feature from ECPT transientresponse is an effective way to predict the defect depth in steel. Transmission mode issuitable for wall thinning defect quantification, while reflection mode is more suitablefor inner defect quantification. The relationship between peak time and residualthickness is linear when thickness ratio y<0.5. Under reflection mode, the logarithmicanalysis of ECPT is proposed. The ideal temperature curve in logarithm domain is linearand can be used to detect defects. Two features representing separation time inlogarithm domain are defined. At last, the mild steel specimen providing subsurfacedefects are tested and the quantification method is built.(5) The advanced signal processing methods for transient eddy currentthermography are proposed to improve the detection effectiveness anddetectability for deep defects and micro defects. Firstly, Chapter6proposed an eddycurrent pulsed phase thermography (ECPPT) technique combing eddy current excitation,infrared imaging and phase analysis. The experimental results show that this proposedmethod can eliminate non-uniform heating and improve defect detectability. Severalfeatures like blind frequency, min phase, and frequency to min phase are extracted from differential phase spectra and the preliminary linear relationships are built to measurethese subsurface defects’ depth. Chapter6also presented the PCA/ICA based imagereconstruction approach for eddy current pulsed thermography to avoid the severelateral heat diffusion (Blur effect). In the proposed image reconstruction approach,several hundred frames of raw data representing the time history are processed andsome principal components and independent components are selected to improve thedetectability for micro defect and deep defects. By analysis of induction heating andheat diffusion, the early stage thermal images are proposed to improve the proposedmethod.(6) Transient eddy current thermography is investigated for CFRP testingand defect evaluation through the analytical analysis and experimental studies. Thedetection mechanism for carbon fiber structure, delamination, impact and thickness areanalyzed and compared under reflection mode and transmission mode. At last,delaminations with different depth from0.5to3mm, thickness variation from1to3.5mm, and impacts with different energy from4J to12J are characterized and tested usingeddy current thermography. The results show that carbon fiber structure and impactleading to lower conductivity can be detected directly in the heating phase. Thedelamination can be detected using the later phase in transient temperature response.Impact shows the different hot spot shapes at the thermograms. The impact behaviourfor real damages are drawn. The hot area by impacts with10J and12J is like circleshape; the hot area by impact with6J and8J is concentrated. Two detection modes arecompared. Reflection mode is more suitable for in-situ inspection, because there is nodirect access to both sides for many practical components. However, the transmissionmode is more suitable for manufacturing and testing, because the coil doesn’t affect thecamera view to object under this mode.At last, the conclusion and further work are outlined in last chapter.更多还原