Computational Methodologies for Vibration-based Damage Assessment of Structures

In the present work, computational methodologies based on artificial neural networks and genetic algorithms (GA) have been developed for identification of structural damage utilizing vibration data. The natural frequencies and mode shapes obtained from the finite element analysis for the first few modes have been considered for this purpose. A multi-stage hybrid methodology combining the modal strain energy criteria with GA has also been proposed, which showed improved damage identification capability as compared to the conventional GA, and proved to be computationally efficient. To demonstrate the efficiency of the proposed hybrid approach, numerical studies have been carried out on the truss structure. The efficacy of mode shape expansion in conjunction with GA is demonstrated for damage identification of reinforced concrete beam based on experimental modal data.