Abstract
A control-oriented uncertainty modeling on frequency domain is presented for a class of spectral systems with unknown high-order modal parameters. At any user-specified frequency, the set of all the frequency responses of the feasible systems is characterized on the complex plane in terms of the convex hull of several circle segments, where the system is said to be feasible if partial modal parameters are given and some other conditions are satisfied by the unknown parameters. We emphasize that such a characterization enables us to quantify the least upper bounds of errors for any specified nominal models, and to develop further efficient results using some additional information. It is shown that, if the dc gain information of the system is available, it reduces the size of the feasible set at all frequencies, and a finite-dimensional nominal model gives explicit tight error bounds. The efficiency of the presented scheme is demonstrated by a simple example of ideal flexible beam.
