Genetic Algorithm for Damage Assessment

Abstract

Genetic Algorithms (GAs) are recognized as an alternative class of computational model, which mimic natural evolution to solve problems in a wide domain including machine learning, music generation, genetic synthesis etc. In the present study Genetic Algorithm has been employed to obtain damage assessment of composite structural elements. It is considered that a state of damage can be modeled as reduction in stiffness. The task is to determine the magnitude and location of damage. In a composite plate that is discretized into a set of finite elements, if a j^th element is damaged, the GA based technique will predict the reduction in E_x and E_y and the location j. The fact that the natural frequency decreases with decrease in stiffness is made use of in the method. The natural frequency of any two modes of the damaged plates for the assumed damage parameters is facilitated by the use of Eigen sensitivity analysis. The Eigen value sensitivities are the derivatives of the Eigen values with respect to certain design parameters. If ω_iu is the natural frequency of the i^th mode of the undamaged plate and ω_id is that of the damaged plate, with δω_i as the difference between the two, while δω_k is a similar difference in the k^th mode, R is defined as the ratio of the two. For a random selection of E_x,E_y and j, a ratio R_i is obtained. A proper combination of E_x,E_y and j which makes R_i−R=0 is obtained by Genetic Algorithm.