|
|
Adjoints for Unsteady Problems
Adjoints for time-dependent problems use reverse-in-time integration
- Need to store entire time-history of the primal solution
- Classical techniques include check-pointing
|
Research topics
- Information compression techniques - use a ROM to recompute quantities required by the backward adjoint solver. The ROM is calibrated during the primal computation.
- Identify region of solution dependency: for convection dominated problem use the method of characteristics
- Monte Carlo based method: recover reverse-in-time integration only in expectation
|
_______________________________________
- Choi, S., Alonso, J.J. “Adjoint-based Optimization for Unsteady Rotorcraft Aerodynamics”, in preparation.
- Wang, Q. et al. “A Monte Carlo Method for Solving Unsteady Adjoint Equations”, submitted to JCP
|
| |
|
