Non-destructive parameter identification of structures

Abstract

Non-destructive health monitoring of structures may be achieved by system identification to determine structural parameters based on measured dynamic response and excitation. A hybrid approach of combining genetic algorithms (GA) and a local search (LS) method is developed, which has some advantages over the direct GA (i.e. without LS). The proposed LS algorithm is GA-compatible and hence easy to implement. The deviation size of local search can be adjusted adaptively in both the global search and local search phases. Extensive numerical studies accounting for the effects of measurement noise have shown that the hybrid GA-LS method leads to considerable improvement in the identification results. The hybrid method is applied to plate/shell structures in this study. As an illustration, an aircraft wing model is presented in the numerical simulation study. The accuracy of identified parameters is very good generally. A sensitivity study is carried out by perturbing the identified parameter and computing the corresponding change in fitness value. It is found that, to some extent, the accuracy of identification results is well correlated with the sensitivity in terms of the fitness value used.