Regularisation Techniques for Dynamic Model Updating using Input Residual

The updating process to correct finite element models using experimental data is generally an ill conditioned problem. The method here presented is based on the use of the input residual, that presents some important advantages over other proposed techniques. The actions required to limit the solution instability due to ill conditioning are discussed. The importance of using regularisation techniques to minimise the influence of experimental errors is demonstrated. Several procedures are analysed through simulated and experimental tests, finally showing that the most reliable results are obtained using a priori information and the singular values truncation technique based on the minimisation of the output residual.