Neural networks on the dynamic models updating

The correlation.: between theoretical and experimental models, which takes into account the eigenfrequencies and eigenmodes obtaneid from an eigenvalue problem and experimental modal analysis, permits to match the theorectical and experimental results. Updating aims to modify the structural theoretical model matrices, such as mass and stiffness, to reproduce closely as possible the measured response from the data. A new methodology for updating, based on neural networks, is proposed. The neural network will represent the inverse of the parameters sensitivity matrix. The input data for the neural network (changes on natural frequencies and modes) will correspond to the pertubation of measured parameters and the output neurons will represent changes on discrete masses or physical boundary conditions, associated to the pertubation paramenter. A spatial frame is used for evaluation of the neural network approach.