Traditional damage detection methods can be costly, error-prone and labor-intensive inspections. Artificial neural networks have been applied broadly over recent years because of their outstanding pattern recognition capability, which is suitable for structural damage detection purposes. In this research, artificial neural networks were developed to predict structural damage resulting from earthquakes. This study investigates structural defects caused by earthquake loading in structural components using an inclusive dataset that covers a variety of structural factors. Different networks were trained to detect patterns engaged with structural faults. These networks included a database of general feature sets containing structural height, different types of material, number of floors, severity of earthquake, damping ratio, and source of cracks for predicting damage index. The results of this study indicate that, an ANN with a configuration of 6-14-7-1 possesses great potential to estimate the damage index, as evidenced by the low error and high correlation values. The performance and efficiency of such network was investigated, demonstrating both improved accuracy and efficiency. The ANN model obtained a correlation coefficient of 0.987 for the training set, 0.969 for the testing set, and 0.975 for the validation set, indicating considerable potential in expressing the non-linear behavior of damage to structures.
