Consideration of uncertainty in damage detection using interval analysis wavelet without baseline data

Abstract

The effectiveness of vibration-based damage detection (VBDD) method has been demonstrated by researchers to provide reliable results. However, the existence of uncertainties in measurement and modelling data hinders the accuracy of results obtained from VBDD. Researches have yielded favourable results by integrating probabilistic method. Despite these successes, the probabilistic method faces the problem of obtaining an unbiased probabilistic distribution of uncertainties. Furthermore, the probabilistic method involves long and complex computations. In dealing with these problems, the nonprobabilistic method that requires no assumptions of the uncertainties distribution was proposed. It involves estimating only the upper and lower bounds of the uncertain parameter. However, the success of the nonprobabilistic method is shortened by its reliance on baseline (undamaged) data that is often not available for existing structures. In this study, a nonprobabilistic interval analysis wavelet (NIAW) method to consider uncertainties in damage identification without using baseline healthy data is proposed. The proposed method is demonstrated by using a plate structure and applying the symmetrical properties of the plate structure. The wavelet coefficient of the plate mode shape is divided along the line of symmetry to obtain wavelet coefficients WL and WR, and the bounds (upper and lower) of WL and WR are estimated. The PoDE and wavelet coefficient increment factor (WCIF) are estimated to obtain damage identity by using the bounds of WL and WR. The product of PoDE and WCIF provides the value of DMI which indicates the level of damage severity. This method is demonstrated using numerical models of a steel plate. The results show that the proposed method accurately identifies damage when noise-contaminated mode shape data is applied.