The purpose of this paper is to introduce a model refinement method realized by the updating FORTRAN program MORES (Modal Residuals) for time-invariant non-gyroscopic and viscously damped elastomechanical systems, which identifies adjustment parameters of submatrices related to the uncertain substructures of the associated discrete model. The parameters are estimated by minimizing a cost function of modal residuals using a weighted least-squares method to solve an ill-conditioned and overdetermined system of linear equations. The program has been applied to a simulated non-proportionally damped vibrator chain with increasing model order. Even for model parameter errors of 50% the parameter estimate errors are less than O.l% unless additive measurement noise has to be taken into account. The errors of the parameter estimates caused by erroneous measurements have been investigated statistically for elastomechanical models of 8 and 20 degrees-of-freedom by running 50 simulated measurement-data samples with uniformly distributed additive random errors of zero mean. It is shown that the biased parameter estimates caused by disturbed data still yield a satisfactory update of the model. The program execution time and storage request with regard to the increasing size of the problem are discussed briefly.

by ASL