Damage Detection of Beam-like Structures Using an Improved Genetic Algorithm

This paper addresses a proficient strategy for detection of beams damages using the variation of eigenvalues and eigenvectors. We deal with damage detection as an optimization problem in which the difference between of the measured responses and analytical model as the objective function should be minimized. Practically, in damage detection problems the number of equations (measured parameters) is usually less than the unknown variables (damage variables). Hence, this is an undetermined problem in mathematics and has infinite solutions. So, there are many local minimums which the solution may trap on them. In addition, the severity of damage could vary within the damaged elements, and also it is possible the damaged extent doesn’t exactly fit to the pre-generated finite elemental meshing. These facts were the main motivations of this work for non-uniformly modeling of elemental damage distributions by employing the proper shape functions, and also considering nodal positions as design variables. Hence, an improved genetic algorithm is introduced in which three new operators are embedded. In the first and second operators the sensitivity matrices of eigenpairs with respect to damage ratios and nodal positions are established to improve the individuals, respectively. In the last one some elements during optimization process are eliminated from damage candidates to reduce the search space. This strategy is applied to a beam structure, and the numerical results demonstrate high capacity and efficiency of the proposed method for in details structural damage detection.

کلیدواژه ها:

damage detection ، genetic algorithm ، shape function ، sensitivity analysis ، eigenvalue ، eigenvector