Frequency-Domain Approaches To Control-Relevant System Identification

点击翻译按钮获取中文摘要

9504878 Tugnait The goal of the proposed research is to provide estimated system models together with evaluated (quantified) modeling errors given experimental input-output data for a wide variety of systems (electrical, chemical, power, manufacturing and production, robotic, electronic etc.) operating in open or closed loop. Such information (system model and quantified modeling error) is crucial to synthesis of a robust controller which maintains its design objectives under uncertain modeling. As such, successful completion of the proposed research is expected to significantly impact the state-of-the-art in robust control design. Approaches are being considered to transform the given time-domain experimental input-output data into a frequency-domain transform function model description along with the associated noise error characterization. Both second-order spectrum and cross-spectrum of the data record as well as higher-order integrated polyspectrum of the data record is being exploited for such purposes. This allows for far fewer a priori assumptions; rather such modeling is dictated by the acquired data. The case of noisy input measurements is also being considered. A stochastic embedding approach is being followed for quantifying modeling errors. ***