Dynamic XFEM-based detection of multiple flaws using an improved artificial bee colony algorithm

Abstract This paper proposes a novel strategy for detecting multiple flaws based on the dynamic extended finite element method (XFEM) and an improved artificial bee colony (IABC) algorithm. To improve the qualities of the initial iterative candidates, a chaotic sequence is adopted to replace the random sequence in the conventional artificial colony algorithm. Meanwhile, an adaptive parameter dimension search is proposed to prevent the iterative solution from falling into a local optimum trap, and a set of subcandidates is utilized to describe the shape of irregular flaws. The detection process of multiple flaws in the structure is divided into two stages. First, the improved ABC algorithm with preset discrete points is combined with topology variables to obtain the number of and approximate locations of the flaws and output the corresponding narrow subdomains. Second, a set of circular candidates is regenerated in the new subdomains derived from the first stage to capture the target flaws. In addition, XFEM is employed for the forward analysis, and the discontinuities of the flaws are characterized by level set functions. Several numerical experiments are performed to illustrate the performance of the proposed strategy. The effects of sensor layout and measurement noise with different levels are considered to assess the robustness of this strategy. The results show that this strategy is able to quantify the number of and capture the shape of unknown flaws. The narrow distributions of the Monte Carlo simulations indicate that the proposed strategy is robust against noise.