Title: Gradient-Based Shape Optimization using High-Fidelity Simulations with Goal-Oriented Error Control
Abstract: Aerodynamic shape optimization is emerging as an indispensable tool in the design of aerospace vehicles. This presentation focuses on several innovations that have made numerical optimization practical in the context of computational fluid dynamics. We begin with the adjoint method. This method provides optimization gradients and is also used to estimate the level of discretization error in the outputs of interest. The benefits are threefold. First, the computational cost of gradient evaluations is essentially independent of the number of design variables. Second, it offers direct control over discretization error through use of adaptive mesh refinement to improve confidence in the optimized designs and to eliminate the requirement of hand-crafting a sufficiently general grid that is appropriate for all candidate designs. Third, we obtain additional cost savings by using progressive optimization, where the depth of the adaptive mesh refinement is systematically increased as the design improves. In addition, we highlight a component-based geometry approach for flexibility in choosing both the shape parameterization and geometric modelers, along with symbolic definition of objectives and constraints for general problem specification. We present design examples involving real-world aircraft configurations in shock-dominated flows, including sonic boom shaping for NASA’s new X-59 aircraft.
Talk time in other timezones: AEDT 11:00 AM Fri 17 Feb, JST 09:00 AM Fri 17 Feb, CET 01:00 AM Fri 17 Feb, GMT 12:00 AM Fri 17 Feb, UTC 00:00 Fri 17 Feb, EST 7:00 PM Thu 16 Feb, CST 6:00 PM Thu 16 Feb, MST 5:00 PM Thu 16 Feb