A Comparative Study of Neural Networks and Sensitivity Modal Analysis for Damage Assesment

Nowadays, the advances in neural networks have opened new sights in structural engineering. The robustness and ability in dealing with incomplete and noisy data make neural networks the best alternative for this purpose. Associated to vibration signature analysis, this technique has shown a robust behavior. The assessment of structural damage and identification of the damaged place in a large and complex structure is a hard task. It is well known that from variations in structural natural frequency measured "in situ" and a well-calibrated structural model, it is possible to detect position as well as intensity of damaged states. Some new advances have been made in this area by means of neural networks. A brief review in literature in the application of neural networks in the last decade is outlined. Emphasis is set to the application of neural networks with radial basis functions for damage detection in a civil structure. A numerical example is presented highlighting the main features of neural networks for detection and evaluation of damage. This example shows that the use of neural networks is very promising, indicating great potential in damaged detection tasks for civil structures.