Criterion design for optimizing low-cost approximations of infinite-dimensional systems: towards efficient real-time simulation
Résumé
Linear systems with irrational transfer functions are difficult to simulate in the time domain. We consider causal systems which can be represented with poles and cuts in some left half-plane of the Laplace domain. Whereas standard interpolation methods yield convergent but high-dimensional approximations, it is possible to derive efficient low-cost approximations thanks to optimization procedures. Defining the criterion (choice of weightening measure, regularization and contraints) is however of utmost importance: this must be carefully designed and driven by each application. Two examples will illustrate the methodology, and a third one will present an extension to a compound system with delays.