An automatic differentiation based nonlinear reduced order modeling technique for unsteady separated flows
This paper discusses the development and implementation of an automatic differentiation based nonlinear reduced order model for unsteady separated flows. The objective of the research presented herein has been to test the capability and limits of the methodology for highly nonlinear unsteady separated flows. In the following, we demonstrate the methodology as applied to unsteady laminar vortex flow shedding aft of a cylinder in crossflow as well as airfoil unsteady transonic flow buffeting. Once constructed, the nonlinear reduced order model provides approximate solutions at a substantial computational cost savings as compared to directly computed flow solutions. However, the limits of the methodology have been found to be rather restrictive for highly nonlinear separated flows, in particular for the airfoil unsteady transonic flow buffet case. For cases where the methodology does work, it is demonstrated that in general the results are more accurate for the nonlinear reduced order model as compared to results for a linear reduced order model. Copyright © 2012 by Jeffrey P. Thomas, Earl H. Dowell, and Kenneth C. Hall.
