Adjoint method is widely used in aerodynamic design because only once solution of flow field is required for adjoint method to obtain the gradients of all design variables. However, the calculation cost of adjoint vector is approximately equal to that of flow computation. In order to accelerate the solution of adjoint vector and improve the adjoint-based optimization efficiency, machine learning for adjoint vector modeling is presented. Deep neural network (DNN) is employed to construct the mapping between the adjoint vector and the local flow variables. DNN can efficiently predict adjoint vector and its generalization is examined by a transonic drag reduction about NACA0012 airfoil. The results indicate that with negligible calculation cost of the adjoint vector, the proposed DNN-based adjoint method can achieve the same optimization results as the traditional adjoint method.
Wind-induced response governs the design of the long-span bridges. The shape of the deck is one of the most important factors that not only affects the mechanical properties but greatly influences the aerodynamic performance of the bridge. An efficient framework is proposed to perform aerodynamic shape optimization of a bridge deck using Computational Fluid Dynamics (CFD) simulations and response surface strategies considering aeroelastic phenomena such as flutter, buffeting and Vortex-induced Vibrations (VIV). A parameterized cross section of the deck has been utilized ensuring required carri…
This paper documents results from a recent computational study aimed at de-sensitizing fan stage aerodynamics---in a modern, high bypass ratio aero-engine---to the effects of rear-seal leakage flows. These flows are the result of seal erosion between a rotor and stator disk in an engine, and deterioration over the life of an engine. The density-matching technique for optimization under uncertainty was applied to this problem. This involved RANS and adjoint flow solves of a full fan stage carried out at two different leakage conditions. Here a detailed analysis of the fan stage aerodynamics is …
In aircraft design, structural optimization and uncertainty quantification concerning transonic aeroelastic issues are computationally impractical, because the iterative process requires great number of aeroelastic analysis. Emerging Reduced-Order Model (ROM) method is convenient for transonic aeroelastic analysis. However, current ROMs cannot be reused during iteration, thus time cost is still way too large. This study proposed an improved ROM suitable for Arbitrary Mode Shapes (ROM-AMS), which is reusable regardless the variation of design variables. By adopting Principal Component Analysis,…
We study the aerodynamic response of a pre-stressed curved aileron. Whilst the fluid flow is standard (high Reynolds air flow undisturbed at infinity), the structure is designed to have a peculiar nonlinear behavior. Specifically, the aileron has only one stable equilibrium when the external forces are vanishing, but it is bistable when relevant aerodynamic loads are applied. Hence, for sufficiently high fluid velocities, another equilibrium branch is possible. We test a prototype of such an aileron in a wind tunnel; a sudden change (snap) of the shell configuration is observed when the fluid …
Efficiently predicting the flowfield and load in aerodynamic shape optimisation remains a highly challenging and relevant task. Deep learning methods have been of particular interest for such problems, due to their success for solving inverse problems in other fields. In the present study, U-net based deep neural network (DNN) models are trained with high-fidelity datasets to infer flow fields, and then employed as surrogate models to carry out the shape optimisation problem, i.e. to find a drag minimal profile with a fixed cross-section area subjected to a two-dimensional steady laminar flow.…
An approximation model based on convolutional neural networks (CNNs) is proposed for flow field predictions. The CNN is used to predict the velocity and pressure field in unseen flow conditions and geometries given the pixelated shape of the object. In particular, we consider Reynolds Averaged Navier-Stokes (RANS) flow solutions over airfoil shapes. The CNN can automatically detect essential features with minimal human supervision and shown to effectively estimate the velocity and pressure field orders of magnitude faster than the RANS solver, making it possible to study the impact of the airf…
National University of Singapore（新加坡国立大学）；University of Michigan（密歇根大学）
Machine learning (ML) has been increasingly used to aid aerodynamic shape optimization (ASO), thanks to the availability of aerodynamic data and continued developments in deep learning. We review the applications of ML in ASO to date and provide a perspective on the state-of-the-art and future directions. We first introduce conventional ASO and current challenges. Next, we introduce ML fundamentals and detail ML algorithms that have been successful in ASO. Then, we review ML applications to ASO addressing three aspects: compact geometric design space, fast aerodynamic analysis, and efficient o…
The shape of the trailing edge of a gas turbine nozzle guide vane has a significant effect on the downstream stagnation pressure loss and heat transfer over the surface of the vane. Traditionally, adjoint-based design optimization methods for turbomachinery components have used low-fidelity simulations like Reynolds averaged Navier-Stokes. To reliably capture the complex flow phenomena involved in turbulent flow over a turbine vane, high-fidelity simulations like large eddy simulation (LES) are required. In this paper, an adjoint-based trailing edge shape optimization using LES is performed to…
SuperWing: a comprehensive transonic wing dataset for data-driven aerodynamic design
School of Aerospace Engineering, Tsinghua University, Beijing, China（清华大学北京航空航天工程学院）；School of Computation, Information and Technology, Technical University of Munich, Germany（慕尼黑技术大学计算、信息与技术学院）
The formation and accretion of ice on the leading edge of an airfoil can be detrimental to aerodynamic performance. Furthermore, the geometric shape of leading edge ice profiles can vary significantly depending on a wide range of physical parameters, which can translate into a wide variability in aerodynamic performance. The purpose of this work is to explore the variability in airfoil aerodynamic performance that results from variability in leading edge ice shape profile. First, we demonstrate how to identify a low-dimensional set of parameters that governs ice shape from a database of ice sh…
Offshore wind turbine towers are pre-assembled and temporarily held in close proximity to each other in group arrangements on port quaysides. Accurate estimates of the aerodynamic loads on the individual towers and on the overall group are essential for the safe and economic design of the quayside's supporting structures and foundations. Wind tunnel tests represent the main way to address the problem, but results may lead to overconservative designs due to inevitable mismatches in the Reynolds number. This crucial issue is dealt with here using an original engineering solution based on concent…
This paper reports on the derivation and implementation of a shape optimization procedure for the minimization of hemolysis induction in biomedical devices. Hemolysis is a blood damaging phenomenon that may occur in mechanical blood-processing applications where large velocity gradients are found. An increased level of damaged blood can lead to deterioration of the immune system and quality of life. It is, thus, important to minimize flow-induced hemolysis by improving the design of next-generation biomedical machinery. Emphasis is given to the formulation of a continuous adjoint complement to…
The planar laminar flow resulting from the impingement of two gaseous jets of different density issuing into an open space from aligned steadily fed slot nozzles of semi-width $R$ separated a distance $2H$ is investigated by numerical and analytical methods, with specific consideration given to the high-Reynolds and low-Mach number conditions typically present in counterflow-flame experiments, for which the flow is nearly inviscid and incompressible. It is shown that introduction of a density-weighted vorticity-stream function formulation effectively reduces the problem to one involving two je…
The task of dynamic flow estimation is to construct an approximation of an evolving flow---and particularly, its response to disturbances---using measurements from available sensors. Building from previous work by Darakananda et al.~(Phys Rev Fluids 2018), we further develop an ensemble Kalman filter (EnKF) framework for aerodynamic flows based on an ensemble of randomly-perturbed inviscid vortex models of flow about an infinitely-thin plate. In the forecast step, vortex elements in each ensemble member are advected by the flow and new elements are released from each edge of the plate; the ele…
The paper is concerned with a node-based, gradient-driven, continuous adjoint two-phase flow procedure to optimize the shapes of free-floating vessels and discusses three topics. First, we aim to convey that elements of a Cahn-Hilliard formulation should augment the frequently employed Volume-of-Fluid two-phase flow model to maintain dual consistency. It is seen that such consistency serves as the basis for a robust primal/adjoint coupling in practical applications at huge Reynolds and Froude numbers. The second topic covers different adjoint coupling strategies. A central aspect of the applic…
We discuss exterior and classical interior alternatives for evaluating fluid flow induced forces on bodies. The discussion aims at a reduction of the total shape derivative, achieved through a decoupling of control and objective in the exterior approach. In this case, geometric as well as convective contributions to the shape derivative vanish. Convective contributions depend on primal physics and may disappear, which is not the case for geometric components. The latter can be interpreted as curvatures immanent to industrial applications. The remaining local derivative of the objective functio…
Figuring out the right airfoil is a crucial step in the preliminary stage of any aerial vehicle design, as its shape directly affects the overall aerodynamic characteristics of the aircraft or rotorcraft. Besides being a measure of performance, the aerodynamic coefficients are used to design additional subsystems such as a flight control system, or predict complex dynamic phenomena such as aeroelastic instability. The coefficients in question can either be obtained experimentally through wind tunnel testing or, depending upon the accuracy requirements, by numerically simulating the underlying …
With the aim of providing a first step in the quest for a reduction of the aerodynamic drag on the rear-end of a car, we study the phenomena of separation and reattachment of an incompressible flow focusing on a specific aerodynamic geometry, namely a backward-slanted step at 25 degrees of inclination. The ensuing recirculation bubble provides the basis for an analytical and numerical investigation of streamwise-streak generation, lift-up effect, and turbulent-wake and Kelvin-Helmholtz instabilities. A linear stability analysis is performed, and an optimal control problem with a steady volumic…
DrivAerNet: A Parametric Car Dataset for Data-Driven Aerodynamic Design and Prediction
Massachusetts Institute of Technology（麻省理工学院）；Technical University of Munich（慕尼黑技术大学）；Department of Mechanical Engineering（机械工程系）；Department of Computer Science（计算机科学系）
Wind farm flow control aims to improve wind turbine performance by reducing aerodynamic wake interaction between turbines. Dynamic, physics-based models of wind farm flows have been essential for exploring control strategies such as wake redirection and dynamic induction control. Free vortex methods can provide a computationally efficient way to model wind turbine wake dynamics for control optimisation. We present a control-oriented free-vortex wake model of a 2D and 3D actuator disc to represent wind turbine wakes. The novel derivation of the discrete adjoint equations allows efficient gradie…
Active Control of Turbulent Airfoil Flows Using Adjoint-based Deep Learning
Graduate Student, Department of Aerospace and Mechanical Engineering（航空航天与机械工程系）；Postdoctoral Research Associate, Mathematical Institute（数学学院）；Assistant Professor, Department of Aerospace and Mechanical Engineering（航空航天与机械工程系）
The present work is concerned with the shape optimization of flapping wings in forward flight. The analysis is performed by combining a gradient-based optimizer with the unsteady vortex lattice method (UVLM). We describe the UVLM implementation and provide insights on how to select properly the mesh and time-step sizes to achieve invariant UVLM simulation results under further mesh refinement. Our objective is to identify a set of optimized shapes that maximize the propulsive efficiency, defined as the ratio of the propulsive power over the aerodynamic power, under lift, thrust, and area const…
A range of optimization cases of two-dimensional Stefan problems, solved using a tracking-type cost-functional, is presented. A level set method is used to capture the interface between the liquid and solid phases and an immersed boundary (cut cell) method coupled with an implicit time-advancement scheme is employed to solve the heat equation. A conservative implicit-explicit scheme is then used for solving the level set transport equation. The resulting numerical framework is validated with respect to existing analytical solutions of the forward Stefan problem. An adjoint-based algorithm is t…
This paper analyzes the adjoint solution of the Navier-Stokes equation. We focus on flow across a circular cylinder at three Reynolds numbers, Re_D=20, 100 and 500. The quantity of interest in the adjoint formulation is the drag on the cylinder. We use classical fluid mechanics approaches to analyze the adjoint solution, which is a vector field similar to a flow field. Production and dissipation of kinetic energy of the adjoint field is discussed. We also derive the evolution of circulation of the adjoint field along a closed material contour. These analytical results are used to explain three…
