The use of modal analysis and neural networks is proposed to detect damages in parts and structures. Comparing the natural frequencies of a test structure with the ones of a non-damaged structure, a neural net, using its capabilities in recognize standards, can deliver the location and extension of the probable damages.
Two examples are presented. At the first one, a notch is located on a free-free beam. The second example deals with a truss-like tower where the damage is a loss of stiffness in some structural members.
The results obtained point as very promising the use of neural network as computational tool in nondestructive testing.