Fast Bi-dimensional Empirical Mode Decomposition And Phase Tracking Method And Their Applications In The Guided-wave-based Nondestructive Testing

Traditional guided-wave-based nondestructive testing uses sensor network to collectwave signals, and then complex1D signal processing methods to implement featureextraction and damage identification. This kind of testing form puts forward highrequirements to position of the sensors, network layout and signal processing methods.Therefore, it cannot satisfy today’s engineering demands any longer.Speckle interferometry is an optical measuring technique, emerging in the1960s. Itholds high precision and sensitivity, simple experimental apparatus, full-field, real-timeand non-contact nature. It is widely applied in the field of measurement of deformation,stress-strain, vibration analysis, nondestructive testing, and so forth. The acquisitiondevice can provide the output of speckle fringe patterns and directly extract fullinformation of the structure. So it is a better choice instead of sensor network forguided-waves-based nondestructive testing. However, speckle fringe patterns are alwayscoupled with serious speckle noise, which reduces the image visibility and spatialresolution of the measuring results.Speckle noises can be classified into three types: random speckle noises, non-uniformbackground and intensity modulation. There are two kinds of methods proposed for thequality enhancement of speckle fringe patterns: signal filtering and phase demodulation.However, there exist some drawbacks in these two approaches, which lead to a limitedapplication range. For most of the signal filtering methods, the universal image processingtechniques are used for the first two adverse variate without signal characteristicsconsidered. Consequently this kind of methods is not so satisfied in some cases. On theother hand, phase demodulation based on signal nature is straightly realized. Although itholds high robustness, this method suffers a lot from noises and background.With the comprehensive consideration of three adverse variate, a three-layer qualityenhancement method combining signal filtering and phase demodulation for specklefringe patterns was proposed in this thesis.In the first layer, a method for signal decomposition and feature classification was proposed. Firstly, fast and adaptive bi-dimensional empirical mode decomposition methodwas introduced, based on which the thesis gave the modifications on disk-shaped structureelements and adaptive neighboring window method. Secondly, a technique based on theenergy estimation of2D intrinsic mode function was proposed for feature classification.The former can decompose a signal into several stationary sub-signals with localnarrow-band from high to low spatial frequencies. The latter can automaticallydiscriminate noises from desired information and then implement feature classification.The first-layer quality enhancement can eliminate high-frequency speckle noises andlow-frequency DC background.In the second layer, a middle-frequency filtering technique based on Hilbert spiraltransform was proposed. Firstly, the thesis proposed five feasible formulas of2D Hilberttransform based on orientation angle and2D signum function, of which Hilbert spiral wasthe best choice. Secondly, the threshold method in amplitude distribution was proposed, inwhich noisy regions with the values lower than threshold were indicated and thendenoised using adaptive averaging filter. The second-layer enhancement can eliminatemid-frequency noises.In the third layer, a modified regularized phase tracking technique was proposed.Firstly, the traditional regularized phase tracker was introduced. And then fourmodification schemes were proposed: Gaussian window was added to cost function.Linear and quadratic polynomial functions were used respectively to approximate differentparts of the local fringe model better, which can also avoid normalization procedure.Gaussian window function with adaptive size was proposed. Levenberg-Marquardt washighly advised to use as an optimization method with property of fast convergence. Themodified approach can process more fringe signals with better demodulation performance.The third-layer enhancement can effectively reduce the influence of intensity modulation.Furthermore, the thesis presented the application of speckle interferometry and itssignal processing method in the field of guided-wave-based nondestructive testing. Thissystem realized nondestructive testing for thin-walled mechanical components and can bedivided into three parts: the generation of the Lamb wave-field, the detection of the Lambwave-field and the processing of speckle fringe patterns. Prism coupling method with narrow-band excitation was adopted to produce Lamb waves in order to simplify thewave-field. Double-pulsed digital speckle interferometry measuring system was designedto capture transient information of wave-field. The temporal resolution can reach a fringepattern in3s. At last, the three-layer quality enhancement method was used to improvethe visibility and spatial resolution of speckle fringe signals so as to realize nondestructivetesting. In addition, synchronizer consisting of computer and synchronizing electronicscontrolled the time sequence of the whole measuring system.