Data-Based Control of a Free-Free Beam in the Presence of Uncertainty

Abstract

Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing this collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden to Department of Defense, Washington Headquarters Services, Directorate for Information SPONSORING / MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR'S ACRONYM(S)AFRL/VSSV Air Force Research Laboratory** Space Vehicles ABSTRACTLinear control development is typically based on deterministic models that approximate the system under consideration. This approach neglects uncertainty in the system response. System uncertainty can arise from a number of sources including disturbances, noise, unmodeled dynamics, and nonlinearity. This may result in a reduction in performance or even instability in the closed loop system. The goal of this research is to account for measured uncertainty in control design. Our approach is to tune a baseline controller using a cost function that balances performance and robustness given measured system uncertainty. The approach is demonstrated on a free-free beam, with the goal of mitigating the flexural vibration. A lumped mass model is tuned to match the experimentally measured Frequency Response Function (FRF) of an experimental beam. This evaluation model and a reduced order model are used to approximate the beam dynamics. The baseline (LQG) controller is designed around the reduced order model of the beam. This controller is tuned according to the proposed cost function using the FRF and postulated variance. The cost function includes closed loop performance and stability robustness metrics. The resulting baseline and tuned controllers are evaluated on lumped mass models consistent with the measured data and uncertainty. SUBJECT TERMS

Stochastic process models for linear structure behavior

“…These parameters and the form of (3) can be used to transform the measured realizations of the random process into realizations of the random vector : (5) There are M columns in the matrix , each one an m-vector corresponding to a realization . The i(th) row of is a sequence of realizations of the i(th) random variable in the vector , i.e.,…”

“…Then, the underlying random variates are obtained as in (5). Because 12 components are retained in the KLE approximation, each of the vector random variates has dimension .…”