Burghardt O., Gauger N. R., Gomes P., Palacios R., Economon T.

“Coupled Discrete Adjoints for Multiphysics in SU2”

https://doi.org/10.2514/6.2020-3139

Abstract: Many optimization problems in fluid dynamics are naturally constituted in a larger multiphysics setting, the most prominent ones nowadays being fluid-structure interaction and conjugate heat transfer or combinations of them. In the context of partitioned solution approaches, and from a software perspective, this means that two or more solvers are coupled by exchanging data at common physical boundaries during the simulation. Optimization methods relying on discrete adjoint solutions have to consider such couplings in order to give accurate gradients. This article presents a methodology for how coupled multiphysics adjoints can be computed efficiently in a way that is also independent of the choice or configuration of the underlying physical problem. Based on an implementation in SU2 (an open-source multiphysics simulation and design software) we demonstrate how the same algorithm can be applied for shape sensitivity analysis on a conjugate heat transfer as well as a fluid-structure interaction test case.