Computational issues for two multiple constraint stiffness matrix adjustment methods

Multiple-constraint matrix adjustment methods are used in structural identification in order to correct or update property matrices (i.e., stiffness, mass, or damping), so that mathematical models may be used to accurately predict a system's response to given stimuli. A number of these methods are available to engineers seeking to fine-tune their models. In each case, a constrained optimization problem is formulated and solved for the adjusted property values. This paper illustrates some of the computational issues that accompany the implementation of two of these matrix adjustment methods, the Multiple-Secant Marwil-Toint method (MSMT), and the reformulated Kammer Projector Matrix method. Preconditioning of the auxiliary problem is seen to improve the convergence rate, though not necessarily the result, for the MSMT method. Accurate results may be obtained from the Total Least Squares (TLS) solution of the reformulated approach with few modes, but when noise is present the results improve with the addition of more modes.