A Frequency-Domain Model-Order-Deduction Algorithm for Nonlinear Systems

Several model-order deduction algorithms (modas) have been developed to coordinate the synthesis of lumped ((nite-dimensional), linear system models, of accetable order, that accurately characterize the behavior of a system over a frequency range of interest (froi) ! min ; ! max ]. The most recent of these techniques considers the frequency response of the model as the \performance metric" and systematically increases model complexity until the frequency response over a froi has converged to within a user-speciice tolerance. The linear moda algorithm based on frequency response is being extended to support the synthesis of models of nonlinear systems. This technique follows a procedure similar to the linear frequency-domain algorithm, but uses a describing-function approach to develop an amplitude-dependent characterization of the nonlinear system frequency response. The extended algorithm synthesizes models that are also of low order; in addition , they include only those nonlinear eeects that innuence the frequency response signiicantly over the froi and for an amplitude range of interest. This sig-niicantly extends the class of systems to which model-order deduction can be applied.