CAD-based CFD shape optimisation using discrete adjoint solvers
Abstract
Computational
fluid dynamics is reaching a level of maturity that it can be
used as a predictive tool. Consequently, simulation-driven product design and
optimisation is starting to be deployed for industrial applications. When performing
gradient-based aerodynamic shape optimisation for industrial applications,
adjoint method is preferable as it can compute the design gradient of
a small number of objective functions with respect to a large number of design
variables efficiently. However, for certain industrial cases, the iterative calculation
of steady state nonlinear flow solver based on the Reynolds-averaged
Navier{Stokes equations tends to fail to converge asymptotically. For such
cases, the adjoint solver usually diverges exponentially, due to the inherited
linear instability from the non-converged nonlinear flow. A method for stabilising
both the nonlinear flow and the adjoint solutions via an improved timestepping
method is developed and applied successfully to industrial relevant
test cases. Another challenge in shape optimisation is the shape parametrisation
method. A good parametrisation should represent a rich design space to
be explored and at the same time be flexible to take into account the various
geometric constraints. In addition, it is preferable to be able to transform
from the parametrisation to a format readable by most CAD software, such
as the STEP le. A novel NURBS-based parametrisation method is developed
that uses the control points of the NURBS patches as design variables.
In addition, a test-point approach is used to impose various geometric constraints.
The parametrisation is fully compatible with most CAD software.
The NURBS-based parametrisation is applied to several industrial cases.
Authors
Xu, ShenrenCollections
- Theses [3822]