Damage Detection of Rectangular Sections by 2D Curvelet, Contourlet and Wavelet Packet Transform

Abstract

In this research, the application of 2D curvelet, 2D wavelet, and Contourlet transform for damage detection of shell structure are presented. For this study, the equation of discrete curvelet transforms using unequally-spaced fast Fourier and wrapping transforms are employed. Four scenario models with one and more damages with different percentages and locations were considered for investigation of proposed methods. As real data has noise, and the existence of noise in the input data may affect the accuracy of damage detection methods. To consider the natural conditions, white noise has been applied to the data extracted from the FEM. First, the coefficients of transforms related to the noisy signal are calculated, then thresholding the previous coefficients calculated, and finally, the de-noise signal is reconstructed. Based on the presented results, the curvelet transform better identifies the location of damage than the wavelet transforms. Wavelet transform based on the direction of the damage element may be specified in some subbands. But in curvelet and Contourlet, it can better identify different damages in different directions due to the multi-directional transform. The curvelet transform has been modified based on polar space and poses challenges for non-polar (normal) discrete space. Therefore, using the Contourlet is divided into multiple directions because of each of the available general directions. According to the obtained results, the proposed methods are sensitive to the location and percentage of damage and can identify the location of damage with a low percentage of damage. Also, to confirm the results of the proposed methods, the results are evaluated using laboratory results. The results show that the proposed methods can identify the location of damages and with a percentage of damage of up to 3%.