快速二维经验模态分解和相位追踪方法及其在导波无损检测中的应用

发布时间：2018-05-01 17:08

本文选题：导波 + 无损检测　；参考：《上海交通大学》2014年博士论文

【摘要】：传统的导波无损检测是利用传感器网络采集导波信号，再经过复杂的一维信号处理方法提取损伤特征、识别损伤参数。这种检测形式对传感器的位置、网络布置形式及信号处理方法均提出了较高的要求，很难实现工业化检测。散斑干涉术是上世纪六十年代逐渐兴起的一种光学测量技术，它具有高精度、高灵敏度、全场、实时性、结构简单、非接触等特点，广泛应用于变形测量、应力应变、振动分析、无损检测等领域。论文提出以散斑干涉技术作为导波信号的采集装置，输出散斑条纹信号，快速直观地提取结构的全场信息。该技术可以取代现有传感器网络，形成另外一种形式的导波无损检测系统。但是，强烈的散斑噪声导致了较低的可见度和空间分辨率，大大降低了条纹信号的使用效果。散斑噪声可分为三类：随机散斑噪声、非一致背景光、光强调制。传统的散斑条纹信号的质量增强方法可以分为两类：信号滤波法和相位解调法。这两类方法都存在各自的弊端，适用范围受限。信号滤波法，多为通用的图像处理技术，并未考虑条纹信号的特点，只是针对随机散斑噪声和非一致背景光，因此在某些情况下的质量增强效果并不理想；相位解调法，充分考虑了条纹信号的特点，试图在噪声和背景光的干扰下直接进行相位解调，然而即使解调方法的鲁棒性再强，在某些情况下，依旧无法摆脱噪声的干扰。论文综合考虑了多种不利因素的影响，结合信号滤波和相位解调，提出了一种散斑条纹信号的三层质量增强方法：第一，论文提出了信号分解与特征分类方法。首先，引入了快速自适应二维经验模态分解，并提出碟形结构元素和自适应邻域窗的改进；其次，提出了基于二维固有模态函数能量估计的特征分类方法。前者可将散斑条纹信号按空间频率从高到低进行分解，形成若干局部窄带的平稳子信号；后者具有自动识别功能，实现了子信号的特征分类。第一层质量增强可以消除高频散斑噪声和直流背景光。第二，论文提出了基于希尔伯特螺旋变换的第二层质量增强方法。首先，以方向角和二维符号函数为基础，提出了五种可行的二维希尔伯特变换形式，其中以希尔伯特螺旋为最优；其次，提出了幅值分布阈值法，该方法判定低于阈值的区域为噪声，然后以具有自适应窗口的均值滤波器进行去噪。第二层可以消除中频噪声。第三，论文提出了一种修正正则化相位追踪方法。引入了传统的正则化相位追踪方法，并提出了四点修正，其中：能量函数中加入了高斯窗函数，并对局部条纹模型分别作了线性和二次展开，从而解决了归一化处理的难题；提出以迭代次数作为质量谱引导解调过程；提出具有自适应窗口的高斯窗函数；提出使用一种收敛快速的优化方法——Levenberg-Marquardt。修正方法的解调范围更广，解调效果更好。第三层质量增强可以降低光强调制变量的影响。最后，论文研究了散斑干涉测量技术及其信号质量增强方法在导波无损检测中的应用，实现了薄板类机械构件的损伤检测，系统包括三部分：Lamb波波场的激发、Lamb波波场的检测以及散斑条纹信号的质量增强。为简化波场，利用可窄带激发的棱柱耦合块法产生了单模式超声波波场；为捕捉波场的瞬态信息，设计了双脉冲数字散斑干涉测量系统，可每隔3微秒记录1个散斑条纹信号；最后使用三层质量增强方法，提高散斑条纹信号的可见度和空间分辨率，达到无损检测的目的。另外，，计算机和同步电子作为同步装置，控制整个无损检测系统的工作时序。
[Abstract]:Traditional guided wave nondestructive testing uses sensor networks to collect guided wave signals, and then extracts damage characteristics and identification of damage parameters through complex one-dimensional signal processing methods. This detection form puts forward higher requirements for sensor location, network layout and signal processing methods. It is difficult to realize industrial detection.
Speckle interferometry is a kind of optical measurement technology which rose gradually in the 60s of last century. It has the characteristics of high precision, high sensitivity, full field, real time, simple structure, non contact and so on. It is widely used in the fields of deformation measurement, stress strain, vibration analysis, nondestructive testing and so on. This paper proposes the speckle interference technique as the acquisition and assembly of the guided wave signal. It can quickly and intuitively extract the full field information of the structure. This technique can replace the existing sensor network and form another form of guided wave nondestructive testing system. However, the strong speckle noise leads to lower visibility and spatial resolution, and greatly reduces the use effect of the stripe signal.
Speckle noise can be divided into three categories: random speckle noise, non consistent background light and intensity modulation. The quality enhancement methods of traditional speckle stripe signals can be divided into two categories: signal filtering and phase demodulation. These two methods all have their own disadvantages, and the application range is limited. Considering the characteristics of the stripe signal, only for random speckle noise and non coherent background light, the quality enhancement effect is not ideal in some cases. The phase demodulation method takes full consideration of the characteristics of the stripe signal and tries to demodulate the phase directly under the interference of the noise and the background light, but even if the robustness of the demodulation method is strong again In some cases, it is still unable to get rid of noise interference.
Considering the influence of many unfavorable factors, combining the signal filtering and phase demodulation, a three layer quality enhancement method of speckle fringe signal is proposed.
Firstly, the method of signal decomposition and feature classification is proposed. First, the fast adaptive two-dimensional empirical mode decomposition is introduced, and the improvement of the dish structure element and adaptive neighborhood window is proposed. Secondly, a feature classification method based on the energy estimation of the two-dimensional intrinsic modal function is proposed. The former can make the speckle stripe signal in space frequency. From high to low to decompose to form a number of local narrow band stationary signal; the latter has automatic recognition function, and realizes the feature classification of the subsignal. The first layer quality enhancement can eliminate the high frequency speckle noise and the DC background light.
Second, the paper proposes a second layer quality enhancement method based on Hilbert spiral transformation. First, based on the direction angle and the two-dimensional symbolic function, five feasible two-dimensional Hilbert transform forms are proposed, in which the Hilbert spiral is the best. Secondly, the amplitude distribution threshold method is proposed, which determines the area below the threshold value. The domain is noise and then denoised by means of adaptive window averaging filter. The second level can eliminate intermediate frequency noise.
Third, a modified regularized phase tracing method is proposed. The traditional regularization phase tracing method is introduced, and the four point correction is proposed, in which the Gauss window function is added to the energy function, and the linear and two expansion of the local fringe model are made respectively, and the problem of normalization is solved. The number of times is used as the quality spectrum to guide the demodulation process; the Gauss window function with adaptive window is proposed, and a fast convergence method is proposed - the Levenberg-Marquardt. correction method has a wider demodulation range and better demodulation effect. The third layer quality enhancement can reduce the influence of the intensity modulation variable.
Finally, the application of speckle interferometry and signal quality enhancement in guided wave nondestructive testing is studied. The damage detection of thin plate mechanical components is realized. The system includes three parts: the excitation of the Lamb wave field, the detection of the Lamb wave field and the quality enhancement of the speckle stripe signal. A single mode ultrasonic wave field is produced by the prism coupled block method. In order to capture the transient information of the wave field, a dual pulse digital speckle interferometry system is designed, which can record 1 speckle stripe signals every 3 microseconds. Finally, the visibility and spatial resolution of the speckle stripe signal are improved by using the three layer quality enhancement method, and the nondestructive testing is achieved. In addition, computers and synchronous electronic devices are used as synchronous devices to control the working sequence of the whole NDT system.

【学位授予单位】：上海交通大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN911.7

【参考文献】