【作者】 刘波；

【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， 2011， 博士

【摘要】 涡流阵列检测是涡流无损检测一个新兴的分支,既有传统涡流检测方法非接触检测的优点,又融合了阵列检测方法易于实现大面积、快速检测的优势,因而成为目前传感器技术和无损检测技术一个共同的研究热点。本文结合部委级预研项目“××××小型化、集成化、快速化无损检测技术”和“基于阵列式磁场测量方法的××××××微损伤检测技术”,围绕裂纹的定量、可视化涡流阵列检测,基于分时多激励发射接收式涡流阵列传感器,展开了涡流阵列传感器优化设计方法、涡流阵列检测裂纹参数估计方法、涡流阵列传感器不同线圈单元组裂纹参数估计精度不一致性的校准方法、涡流阵列C扫描成像方法等方面的研究。主要内容及创新点如下:(1)针对涡流阵列传感器实施检测时存在不敏感区域的问题,提出了参数——线圈单元组有效检测区域比率,该参数能够量化反映涡流阵列传感器的不敏感区域,并可作为减小传感器不敏感区域优化设计中的参考依据。以电磁场理论为基础,探讨了涡流阵列检测问题的解析计算和数值计算方法。阐述了分时多激励发射接收式涡流阵列检测感应电压的详细计算过程,研究了裂纹长度、深度和位置等参数对感应电压的影响,并给出了数值计算结果的物理解释。借助双线圈单元简化模型,进行了涡流阵列传感器线圈单元间互感干扰的有限元仿真,得到了抑制互感干扰的方法。以线圈单元组(涡流阵列传感器可视为若干个线圈单元组的集成,每个线圈单元组包含数个线圈单元)为简化的研究对象,对检测灵敏度、空间分辨率、线圈单元组有效检测区域比率等参数,随线圈单元排布方式、中心距、外半径、高等的变化规律进行了有限元仿真,得到了可用于指导涡流阵列传感器设计的有意义的结论,验证了在涡流阵列传感器设计中引入参数——线圈单元组有效检测区域比率,对于传感器线圈单元排布方式、中心距、外半径等参数的优化设计具有重要参考价值。(2)基于缺陷重构技术的涡流阵列检测裂纹参数估计方法,存在需要完备准确的先验数据库和反演分析过程实时性较差等问题。分析了检测输出信号波峰、波谷、拐点等的扫描时刻与检测线圈单元中心到达裂纹长度边缘点时刻的对应关系,提出了一种无需受检部件先验知识的裂纹长度估计方法。分析了检测输出信号波峰幅值与裂纹深度的非线性递增映射关系,研究了基于弹性BP神经网络的裂纹深度估计方法,该方法在先验知识较少的受检部件检测中具有优势。针对垂直于扫描方向裂纹位置估计精度较低的问题,提出了基于参数——两波谷幅值偏离比率的垂直于扫描方向裂纹位置估计方法,该方法较基于缺陷重构技术的估计方法实时性更好。进行了上述裂纹长度、深度和位置等参数估计方法的实验验证,分析了结果的精度。实验结果表明,上述裂纹参数估计方法得到的裂纹参数精度较高。(3)针对涡流阵列传感器不同线圈单元组裂纹参数估计精度的不一致性、检测系统精确硬件校准周期较长、完成硬件校准后裂纹参数估计精度仍然较低等问题,提出了基于裂纹参数的涡流阵列检测系统软件校准方法。确立了软件校准模型的基本形式,研究了基于样条变换的偏最小二乘回归、基于核函数变换的偏最小二乘回归和最小二乘支持向量回归机等建模方法,建立了软件校准模型。借助于软件校准模型对初步检测得到的裂纹深度数据进行处理,分析了深度校准结果的精度,得到以下结论:在涡流阵列检测系统硬件校准之后,应用本文提出的软件校准方法,可以有效消除不同线圈单元组裂纹参数估计精度的不一致性,从而显著提高裂纹参数的估计精度;与其它建模方法相比较,基于样条变换的偏最小二乘回归方法所建立的软件校准模型,能够更为准确地估计裂纹深度。(4)在对裂纹参数估计精度有较高要求的部件的无损检测中,针对涡流阵列C扫描图像仅对裂纹参数进行定性反映、存在受污染区域、图像裂纹区域与实际裂纹区域存在位置偏差等不足,提出了定量反映裂纹位置、方向、长度和深度等参数的涡流阵列修正C扫描成像方法。将图像分割、边缘检测、轮廓跟踪、细化等图像处理算法和裂纹参数估计方法相结合,给出了涡流阵列修正C扫描图像生成算法。进行了含不同方向裂纹铝板试件的检测实验,生成了U检测模式和T检测模式的修正C扫描图像。实验结果表明:修正C扫描图像作为C扫描图像的衍生图像,能够较为准确地定量反映裂纹位置、方向、长度、深度等参数;修正C扫描成像方法去除了C扫描图像的受污染区域,消除了图像中与实际中裂纹区域位置的偏差,避免了检测人员视觉误差导致的裂纹区域的错误定位;修正C扫描图像生成算法由于包含较多图像处理算法和裂纹参数估计方法,实时性较C扫描图像生成算法稍差,工程实践中,可在C扫描成像方法完成初检判断有无裂纹后,用修正C扫描成像方法完成二次细化检测,进一步准确估计裂纹参数。(5)详细阐述了涡流阵列检测系统的总体设计方案,研究了DDS信号发生电路、正交锁定放大电路、模拟多路转换器阵列电路等主要硬件模块实现过程中的技术问题。基于本文设计实现的检测系统,分别对飞机输油管轴向和周向裂纹、飞机发动机叶片裂纹进行了检测,估计了裂纹参数,生成了涡流阵列C扫描图像和修正C扫描图像,实现了定量、可视化无损检测,从而验证了涡流阵列检测系统硬件的可靠性和软件算法的有效性。更多还原

【Abstract】 Eddy current array testing (ECAT) is a new branch of eddy current nondestructive testing (NDT) technique developed in recent years. In view that ECAT has the advantages of non-contact detecting which merit conventional eddy current testing (ECT) has and easy to implement large-area and fast testing which merit array testing does in the pattern of electronic scan, it has become currently a common hot subject of research in sensor technonlogy and NDT technonlogy. Supported by ministerial-level pre-research projects‘××××miniaturization, integrated, rapid NDT technonlogy’and‘××××××micro-injury testing technonlogy based on methods of array magnetic field measurement’, this paper focuses on quantitative and visual crack detecting by ECAT. Based on time-sharing multi-excitation transimit-receive (TSMETR) eddy current array sensors, several key technology about optimal design methods of eddy current array sensors, methods of crack parameter estimation, calibration methods of crack parameter estimation precision unconsistency from different coil unit groups, methods of eddy current array C scan imaging are studied in this paper. The brief of these researches and the novel approaches are as follows:(1) To solve the problem that less sensitive zones exist when the workpieces are detected using ECAT, the ratio of effective detection area of coil unit group which can quantitatively reflect the less sensitive zones is proposed, and this parameter can guide eddy current array sensors’optimal design for reducing the range of less sensitive zones. Based on electromagnetic field theory, ECAT analytical calculation method and numerical method are discussed. Detailed calculation processes of induced voltage of TSMETR ECAT are described. The influence of crack parameters, such as length, depth and position, to induced voltage are researched, and the physical explanation of the results is given. With the help of simplified double coil unit model, finite element simulations of the mutual inductance between eddy current array sensor coil units are implemented, and the elimination methods of mutual inductance are discussed. Using coil unit group (an eddy current array sensor can be seen as integrated by many coil unit groups, and a coil unit group consists of several coil units) as the simplified object of study, the variations of sensitivity, spatial resolution and ratio of effective detection area of coil unit group with diffirent arranged patterns, center distances, outer radius and heights of coil unit are researched by means of finite element simulation method, and several significant conclusions which can guide the design of the eddy current array sensors are drawn. The simulation results verifies that the ratio of effective detection area of coil unit group brings important reference value to the optimal design of arranged pattern, center distance and outer radius of coil unit of eddy current array sensor. (2) The method of crack parameter estimation based on the eddy current array defect reconstruction technique is deficient in complete accurate prior database and time -consuming during the process of inverse analysis. By analysing the mapping between the scan times of crest, trough and inflection point of the signals output by ECAT system and the moments when the center of receiving coil unit reach crack edges in the length direction, this paper presents a new method of crack length estimation without any prior knowledge of the workpieces. A method of crack depth estimation which has the advantage in detecting the workpieces with little prior knowledge based on resilient BP neural network is researched by analyzing the nonlinear incremental mapping between the crest amplititude and crack depth. To solve the problem that it is difficult to improve the eatimation precision of crack position that is perpendicular to the scanning direction, a new method of crack position estimation that is perpendicular to the scanning direction based on the amplitude deviation ratio of two valleys is proposed. This method has better real-time performance compared with the method of crack parameter estimation based on the eddy current array defect reconstruction technique. Experiments are done to verify the methods of crack length, depth and position parameter estimation, and the precision of the experimental results are analysed. The results show the high precision of the above mentioned methods of crack length, depth and position parameter estimation.(3) To solve the problem that different coil unit groups have different crack parameter estimation precision, the hardware calibration period of the testing system is long and the crack parameter estimation precision are still low after completing the hardware calibration processes, a new software calibration method of ECAT instrument based on the crack parameters is proposed. The basic forms of the software calibration models are established. By researching the partial least-squares (PLS) regression based on the spline transform as well as kernel function and the algorithm based on the least squares support vector machine (LS-SVM), the software calibration models are founded. According to the precision analysis of the crack depth estimated by the software calibration models, the following conclusions can be made, after the hardware calibration process, using the proposed software calibration method can largely reduce the crack parameter estimation precision unconsistency from different coil unit groups so that significantly improving the crack parameter estimation precision. The precision of the crack depth estimated by the software calibration models which are founded by the PLS regression method based on the spline transform is higher than the precision of the crack depth estimated by the software calibration models which are founded by the other modeling approaches.(4) In NDT of components whose crack parameters need to be estimated precisely, eddy current array C scan images have the disadvantages of only qualitatively reflecting the defects, existing contaminated areas, having differences between the positions of crack zones in C scan images and the real positions of crack zones, and et al. Eddy current array modified C scan imaging, which can quantitatively reflect the crack position, direction, length and depth, is proposed. Combining the image processing algorithms, such as image segmentation, edge detection, contour tracking, thinning and et al, with the crack parameter estimation methods, the generation algorithms of the eddy current array modified C scan images are researched. Detect the Aluminum specimen with cracks in different directions, generate the eddy current array modified C scan images in U testing mode and T testing mode. The experiments show the below conclusions. As one of the derivative images of the eddy current array C scan images, eddy current array modified C scan images are adaptive for fast testing some components, whose crack parameters need to be estimated precisely, because of precisely quatitative reflection of crack length, depth, direction and position. For eliminating the contaminated areas of the crack zones in C scan images and modifying the deviations between the positions of crack zones in C scan images and the real positions of crack zones, modified C scan images can effectively avoid the error orientation due to the inspectors’visual errors. However, modified C scan imaging is poor in real-time applications because of many image processing algorithms and methods of crack parameter estimation. Therefore, in practice modified C scan imaging may be used to complete the second-step testing for estimating the crack parameters accurately after C scan imaging has determined whether cracks exist.(5) Overall design scheme of ECAT system is elaborated, and implementation methods of main hardware modules, including DDS signal generation circuit, orthogonal lock-in Amplifier circuit, analog multiplexer array circuit and et al, are researched. Based on the designed ECAT system, airplane oil tubes with circumferential and axial cracks and aircraft engine blades with cracks are detected. The crack parameters are estimated from the detected signals. Eddy current array C scan image and modified C scan image are generated. Quantitative and visual nondestructive detection to the above key components is fulfilled. Therefore, the reliability of the ECAT system’s hardware and the effectiveness of the proposed algorithms are verified.更多还原