Combining Subset Selection and Parameter Constraints in Model Updating

Abstract

Model updating often produces sets of equations whose solution are ill-conditioned and extra information must be used to produce a well-conditioned estimation problem. One possibility is to change all the parameters, but to introduce extra constraints, for example by taking the minimum norm solution. This paper takes a different approach, by considering only a subset of the parameters to be in error. The critical decision is then the choice of parameters to include in the subset. The methods of subset selection are outlined and extended to the selection of groups of parameters. The incorporation of side constraints is considered and demonstrated using an experimental example.