Regularisation methods for finite element model updating

This paper addresses the problem of selecting a side constraint and determining the regularisation parameter in model updating. The weight to be attached to the constraint is determined by the regularisation parameter. Methods based on singular value decomposition, cross-validation, and L-curves are considered, and results obtained by applying these methods to a numerical example provide the basis for a comparative study. It is found that the method of cross-validation can be used reliably to truncate the small generalised singular values which contain the measurement noise. The L-curves approach is similarly robust in locating the regularisation parameter, and this is demonstrated in a physical experiment. It is shown that careful selection of the side constraint can lead to updated parameters with physical understanding. (C) 1998 Academic Press Limited.