MS 1009 - 1: ADJOINT METHODS FOR STEADY & UNSTEADY OPTIMIZATION
MS Organizers: Kyriakos C. Giannakoglou, Jens Dominik Mueller
Chair: Jens Dominik Mueller
7708 KEYNOTE: CAD-FREE ADJOINT SHAPE OPTIMISATION IN MARITIME TWO-PHASE FLOWS - Thomas Rung, Jörn Kröger
4881 TWO-STEPS SHAPE OPTIMIZATION ALGORITHM IMPROVING HYDRODYNAMICS STABILITY - Takashi Nakazawa
9306 AERODYNAMIC OPTIMISATION USING ADJOINT METHODS AND PARAMETRIC CAD MODELS - Philip Hewitt, Simão Marques, Trevor Robinson, Dheeraj Agarwal
This paper presents a CAD-based optimization framework using adjoint functions for aerodynamic design. In this work, the SU2 code is used to obtain high-fidelity flow solutions and surface sensitivities using adjoint methods. This work proposes methodologies to exploit CAD models created using standard commercial modelling software like CATIA V5 in the optimization workflow. A formulation to obtain geometric sensitivities is introduced, enabling the calculation of gradients with respect to these CAD variables. The performance and robustness of the optimization framework is assessed using a range of inviscid and viscous problems. The results show the CAD parameterisation can be efficiently used in obtaining reliable optimums, while operating directly on feature based CAD systems.
10552 A CONTINUOUS ADJOINT APPROACH FOR VEHICLE INTERIOR NOISE REDUCTION - Christos Kapellos, Michael Hartmann
In this paper the continuous adjoint method is developed for a vibroacoustic model that predicts the interior noise of a vehicle induced by the airflow. The model simulates the front side window vibration, excited by the acoustic and hydrodynamic pressure load, and the resulting sound wave propagation into the cabin. Targeting interior noise reduction, the continuous adjoint formulation is used to derive the adjoint to the state equations, namely the bending wave and the wave equations, whilst taking into consideration their coupling. The developed method is applied to a generic vehicle test body for the minimisation of the interior noise.
10065 OPTIMISATION OF A U-BEND USING A CAD-BASED ADJOINT METHOD WITH DIFFERENTIATED CAD KERNEL - Salvatore Auriemma, Mladen Banovic, Orest Mykhaskiv, Herve Legrand, Jens-Dominik Mueller, Tom Verstraete, Andrea Walther
In order to optimise the shape of a three-dimensional CAD-based model using the computationally efficient adjoint methods, the calculation of shape sensitivities, the derivatives of the surface position with respect to the design parameters, is required. This sensitivity is usually not available with CAD systems, but can be obtained by applying the Finite Difference method to CAD-system. Finite-Differences or part-analytic differentiation have been proposed to obtain sensitivities, but have their drawbacks. If source code is available, automatic differentiation can provide accurate derivatives without incurring topology changes or requiring hand-differentiation. This paper proposes the differentiation of the open-source CAD kernel - OpenCascade Technology (OCCT) with AD software tool ADOL-C (Automatic Differentiation by Overloading in C++). As a case study we consider the optimisation of pressure loss in a U-bend pipe. The geometry of the U-bend is parametrised in OCCT with a number of cross-sections lofted along a guiding path line. The corresponding geometric derivatives are used in CFD optimisation loops with the resulting shape outperforming the initial design.