Damage detection based on vibration properties of engineering structures with artificial neural networks

Damage to a structure is defined as changes in geometry and material properties that cause a decrease in structural rigidity that adversely affects the performance of the structure. The reduction in structural stiffness causes changes in vibration parameters such as natural frequencies and mode shapes. Artificial Neural Networks (ANN) have been extensively applied in structural damage detection in recent years, thanks to their excellent pattern recognition capabilities. In this study, an ANN-based damage detection method was used to determine the damage location and size by using the vibration parameters of the selected structures. In order to investigate the applicability of the method firstly, experimental modal analysis was applied on the selected structures and experimental dynamic characteristics were obtained. According to the experimental data, the initial finite element (FE) models of the structures were calibrated and the models that best reflected their current state were tried to be obtained. Then, damage regions were defined on the FE models of the selected structures and modal analyzes for the designed damage scenarios were carried out more quickly with the software developed over the SAP2000 API through the MATLAB program. The frequency and mode shape data were obtained by taking into account the changes in the elasticity modulus in the analyzes. The obtained vibration data were used as the input parameters of the ANN, and the elasticity values of each defined region were used as the output parameters. Effective results were obtained for damage assessment with ANN from the applications performed on steel cantilever, steel frame, dam model, bell tower, stone masonry minaret and inclined minaret. It was seen that using only frequencies as ANN input in symmetrical structures does not give effective results.