摘要：AbstractIn this paper, we are interested in deriving a non-intrusive numerical approach to construct nonlinear and linear parameter varying reduced order models from data. More specifically, based on data collected from a stable non-linear time-domain simulator or experimental bench, we show how we can infer either a reduced order nonlinear or a (quasi) linear parameter dependent model. The proposed approach is based on a very recent procedure called MII for Mixed Interpolation Inference, involving three steps: pencil method, interpolation and model inference. The complete process is illustrated on a polynomial nonlinear Duffing oscillator use-case showing how a reduced either nonlinear or linear parameter varying model can be obtained from time-domain raw data.
