<p>This study uses the curvature damage index method based on artificial neural networks to investigate structural faults. The damages are inspected at multiple locations by using a pinned–pinned supported beam and a tubular propped cantilever beam with a rectangular cross-section. Initially, the experimental and numerical data were utilized to observe the mode shapes of undamaged and damaged beam models. The mode shape data was utilized to investigate the curvature damage index for various damage severities. An artificial neural network (ANN) was utilized for training the experimental data to eradicate undesirable peaks caused by data errors in displacement mode shape data. By using the absolute curvature damage index, the numerically obtained modal parameters (displacement mode shape) are highly suitable for calculating damage areas without ANN training. Further, the mode shape curvature was developed by using central difference approximation for each damage case after obtaining the frequency response (FR) data. To display the damages in beam specimens, a modified curvature damage index (MCDI) is created by using trained data. The study has demonstrated that the proposed technique, which utilises ANN-trained FR data instead of directly using untrained FR data, is capable of identifying structural damages with greater accuracy. </p>