Continuous Adjoint for Aerodynamic and Aeroacoustic Shape Optimization and Robust Design. Applications to External Flow Problems, Including Wind Turbines

Farhikhteh, Mohammaderfan

dc.contributor.author	Farhikhteh, Mohammaderfan
dc.date.accessioned	2025-09-22T09:11:03Z
dc.date.available	2025-09-22T09:11:03Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/62477
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.30173
dc.relation	Marie Sklodowska Curie Grant Agreement No 860101 (zEPHYR ITN)	el
dc.rights	Default License
dc.subject	Continuous Adjoint Method	en
dc.subject	Shape Optimization	en
dc.subject	Computational Fluid Dynamics	en
dc.subject	Robust Optimization	en
dc.subject	Aerodynamics	en
dc.subject	Aeroacoustics	en
dc.title	Continuous Adjoint for Aerodynamic and Aeroacoustic Shape Optimization and Robust Design. Applications to External Flow Problems, Including Wind Turbines	en
dc.contributor.department	Parallel CFD & Optimization Unit	el
heal.type	doctoralThesis
heal.classification	Mechanical Engineering	en
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2024-12-16
heal.abstract	This PhD thesis presents advancements in aerodynamic and aeroacoustic optimization techniques using the adjoint method, and also focuses on Robust Design Optimization (RDO) under uncertainties related to flow conditions and manufacturing imperfections. The developed methods are applied to horizontal axis Wind Turbine (WT) blades and other aerodynamic bodies, including a car geometry. Objectives of this work are to develop continuous adjoint methods for both the Spalart-Allmaras (SA) and $k-\omega~SST$ turbulence models in rotating reference frames and to assess their application in aerodynamics, to maximize torque, and in aeroacoustics, to minimize generated noise. Additionally, the research includes the developments of non-intrusive Polynomial Chaos Expansion (niPCE) and Karhunen-Loève Expansion (KLE) in order to consider both flow conditions and geometric uncertainties during optimization. The KLE method is developed to generate imperfections onto 3D geometries. The work starts with the development/extension of continuous adjoint equations for the SA turbulence model in a rotating reference frame, by incorporating the linearization of the Coriolis and centrifugal forces. This adjoint method for flows governed by the SA model is assessed by comparing Sensitivity Derivatives (SDs) with those derived from the Finite Differences (FDs) method. Aerodynamic shape optimization is, then, applied to the blades of the MEXICO and NREL Phase VI WTs, aiming for maximum torque. Verification of the flow simulation code against numerical and experimental results precedes the optimization process, which also includes blade parameterization and mesh morphing using volumetric B-Splines. A constraint is imposed in the optimization of the NREL Phase VI blade to maintain the blade volume during optimization. The pressure coefficient and peripheral force distribution along the optimized and baseline blades are investigated to better understand the source of torque maximization. Additionally, optimizations at nominal and off-design conditions, with re--evaluations at other operating points, are carried out. Then, the dissertation elaborates on continuous adjoint equations for the $k-\omega~SST$ turbulence model in a rotating reference frame. This is extended to aeroacoustic shape optimization in 2D and 3D, integrating Proudman's formula into the continuous adjoint equations coupled with the Reynolds--Averaged Navier-Stokes (RANS) equations, for the first time in the corresponding literature. The Proudman's formula computes acoustic emissions of turbulent flows, based on the turbulent kinetic energy and turbulent dissipation rate fields. The objective function is defined as the integral of generated acoustic sources in a user-defined volume around the body, to be minimized. The sensitivity derivatives computed by the developed adjoint method are verified against FDs, on an airfoil case by computing the trailing edge noise. A parametric study is performed to investigate the impact of the integration volume size on the objective function. Aeroacoustic shape optimization is successfully performed for airfoils, the MEXICO WT, and the DrivAer car model, aiming to minimize acoustic sources at the trailing edge, blade, and side mirror, respectively. The broadband noise generated by a trailing edge airfoil, computed by the Proudman's formula and specific spectral properties, is validated against experimental data at the same microphone location. Also, the sound power level of the MEXICO WT blade is validated against numerical and experimental results. Finally, a comparison of broadband noise generation of the optimized and baseline geometries for both the isolated airfoil and the MEXICO WT blade is made. In this PhD thesis, a RDO technique in aerodynamics, under manufacturing imperfections, based on the niPCE and the KLE for 3D geometries, is also developed and applied. The niPCE is used in Uncertainty Quantification (UQ) to model and analyze the impact of input uncertainties on the statistical moments of the performance. By constructing a polynomial representation of the system's response, the niPCE estimates statistical moments of the Quantities of Interest (QoI). Gauss Quadrature (GQ) is used for multi-dimensional numerical integration to compute the PCE coefficients; in specific, the Full Grid and Smolyak Sparse Grid (which is really appreciated in case of many uncertain variables) are employed. The KLE is used to generate stochastic perturbations on the baseline 3D geometry surface in the normal to the wall direction. Initially, using the Full Grid, an RDO with uncertainties related to the flow conditions is performed for the MEXICO WT blade, to maximize the blade torque. Then, using the niPCE and KLE codes, the RDO of an isolated airfoil with manufacturing imperfections, using the Full Grid and Smolyak--based UQ methods for different polynomial orders, is performed. Finally, the KLE is used to generate imperfections on the NREL Phase VI WT blade surface for performing RDO, with the goal of maximizing the blade's torque. For the first three years, this PhD has received funding from the European Unions Horizon 2020 Research and Innovation program, under the Marie Sklodowska Curie Grant Agreement No 860101 (zEPHYR ITN, \textit{" Towards a more efficient exploitation of on-shore and urban wind energy resources "}).	en
heal.sponsor	Marie Sklodowska Curie Grant Agreement No 860101 (zEPHYR ITN)	en
heal.advisorName	Giannakoglou, Kyriakos
heal.committeeMemberName	Voutsinas, Spyridon
heal.committeeMemberName	Riziotis, Vasilios
heal.committeeMemberName	Mathioudakis, Konstantinos
heal.committeeMemberName	Manolesos, Marinos
heal.committeeMemberName	Papadakis, George
heal.committeeMemberName	Nikolos, Ioannis
heal.academicPublisher	Σχολή Μηχανολόγων Μηχανικών	el
heal.academicPublisherID	ntua
heal.numberOfPages	184
heal.fullTextAvailability	false