ADJOINT METHODS FOR OPTIMISATION, MESH ADAPTATION AND UNCERTAINTY QUANTIFICATION ECCOMAS 2016
Gradient-based methods are essential for efficient CFD optimisation is method, an essential ingredient is the adjoint method which allows to compute these gradients at near constant cost, independent of the number of design variables. Adjoint methods appear in two forms: continuous and discrete. This Minisymposium emphasises:
a) the robustness and versatility of the adjoint solvers
b) its application to industrial and unsteady flows
c) the efficient but flexible and automatic parametrisation of arbitrary shapes
d) the imposition of design constraints in either a CAD-based or CAD-free way
e) the integration of shape and topology optimization. Industrial applications will be presented
The adjoint method is recognised as the most efficient method to compute gradients for numerical optimisation, mesh adaptation, or uncertainty quantification. The adjoint approach has been adopted by major European industries, research institutions and commercial vendors, a range of open-source adjoint solvers is also available, all generating a wide range of applications.
The focus of this minisymposium was to review the recent progress in computing and apply adjoint sensitivities in CFD, CSM and related disciplines. Contributions included:
* Improving robustness and versatility of adjoint solvers,
* Adjoints for unsteady models,
* Adjoints for complex physics and multi-disciplinary systems,
* Integration into the workflow with parametrisation, optimisation and return to CAD,
* Adjoint methods in uncertainty quantification
* Error analysis and adjoint-driven mesh adaptation
* Applications of adjoint design in industrial cases.