Usability of mathematical models in mechanical decision processes

Mathematical models of mechanical systems are developed to assist decision making for design, dynamic performance, structural interfacing, reliability, etc. Model updating is a costly and time-consuming task, so it is important to evaluate rigorously the quality of a model with respect to the decisions which rely upon it. The measure of model usability is that a model is usable to the extent that decisions based on the model are robust to the associated uncertainties. The analysis of robustness employs convex models of uncertainty, which are particularly suited to situations in which prior information about the uncertainties is severely limited. Several examples of the analysis of model usability for mechanical decisions are considered, including a simple harmonic oscillator, a simple non-linear system, and a multidimensional linear frequency response model for predicting dynamic stability. (C) 1998 Academic Press Limited.