An inverse model is proposed based on the extended finite element methods(XFEM)and a modified artificial bee colony(MABC)algorithm to detect and quantify the multiple internal flaws in structures.With this novel model,the location and the size of each flaw are approximated by the level set method.Moreover,a best-and-guided solution updating strategy is used during the global search until convergence is reached.The parameter samples with the different number of flaws are selected greedily by the artificial bee colony.The number of flaws,which is unkown in structures beforehand,is optimized during the iteration.This model avoids the problem of grid reclassification in the inversion process and reduces the amount of calculation for inverse analysis.The results of the numerical examples show that the proposed approach can detect and quantify the multiple internal flaws in structures and the corresponding locations and shapes without any knowledge of the number of flaws beforehand.