Skip to main content
NASA Advanced Supercomputing (NAS) Division World-class modeling, simulation, and supercomputing in support of the NASA mission

The Discrete Adjoint Solver in SU2 and its Application to Multiphysics Problems

AMS Seminar Series

Speaker: Ole Burghardt, University of Kaiserslautern-Landau

Presentation

Seminar Slide Deck (PDF-27.1MB)

Abstract

The discrete adjoint solver in the open-source software SU2 is one of its integral and well-established functionalities. It serves to facilitate, or even enable, the computation of sensitivity information, e.g., of objective function gradients in a design optimization problem. The main technique that is used for generating derivatives of the simulation program code and for solving the discrete adjoint equation is algorithmic differentiation via operator overloading in combination with a fixed-point ansatz. In this talk, we will go through some main aspects of the implementation and we point out how multiple discrete adjoint solvers are combined in the case of multi-disciplinary problems. More precisely, we will discuss how adjoint coupling data is handled and how accurate results can be obtained.

Biography

Photo of Ole Burghardt
Ole Burghardt is a postdoctoral researcher at the University of Kaiserslautern-Landau and focuses on multi-disciplinary optimization, discrete adjoint methods as well as topics in high-performance computing. He is a member of the SU2 developers team and obtained his Ph.D. in 2023.