Search

A multi-fidelity (MF) active learning method is presented for design optimization problems characterized by noisy evaluations of the performance metrics. Namely, a generalized MF surrogate model is used for design-space exploration, exploiting an arbitrary number of hierarchical fidelity levels, i.e., performance evaluations coming from different models, solvers, or discretizations, characterized by different accuracy. The method is intended to accurately predict the design performance while reducing the computational effort required by simulation-driven design (SDD) to achieve the global optimum. The overall MF prediction is evaluated as a low-fidelity trained surrogate corrected with the surrogates of the errors between consecutive fidelity levels. Surrogates are based on stochastic radial basis functions (SRBF) with least squares regression and in-the-loop optimization of hyperparameters to deal with noisy training data. The method adaptively queries new training data, selecting both the design points and the required fidelity level via an active learning approach. This is based on the lower confidence bounding method, which combines the performance prediction and the associated uncertainty to select the most promising design regions. The fidelity levels are selected considering the benefit-cost ratio associated with their use in the training. The method’s performance is assessed and discussed using four analytical tests and three SDD problems based on computational fluid dynamics simulations, namely the shape optimization of a NACA hydrofoil, the DTMB 5415 destroyer, and a roll-on/roll-off passenger ferry. Fidelity levels are provided by both adaptive grid refinement and multi-grid resolution approaches. Under the assumption of a limited budget for function evaluations, the proposed MF method shows better performance in comparison with the model trained by high-fidelity evaluations only.

The performance of surrogate-based optimization is dependent on the surrogate training set, certainly for realistic optimizations where the high cost of computing the training set data imposes small training set sizes. This is especially true for multi-fidelity surrogate models, where different training sets exist for each fidelity. Adaptive sampling methods have been developed to improve the fitting capabilities of surrogate models, adding training points only where necessary or most useful to the optimization process (i.e., providing the highest knowledge gain) and avoiding the need for an a priori design of experiments. Nevertheless, the efficiency of the adaptive sampling is highly affected by its initialization. The paper presents and discusses a novel initialization strategy with a limited training set for adaptive sampling. The proposed strategy aims to reduce the computational cost of evaluating the initial training set. Furthermore, it allows the surrogate model to adapt more freely to the data. In this work, the proposed approach is applied to single- and multi-fidelity stochastic radial basis functions for an analytical test problem and the shape optimization of a NACA hydrofoil. Numerical results show that the results of the surrogate-based optimization are improved, thanks to a more effective and efficient domain space exploration and a significant reduction of high-fidelity evaluations.

This paper is devoted to the assessment of ship maneuvering simulation using different propeller models with the focus on a simplified propeller model that represents the action of the propeller by body force and uses propeller performance curve to determine propeller loading during ship maneuvering. Simulations are also performed with an actual propeller approach with which the propeller rotation is simulated directly with the Reynolds averaged Navier-Stokes equation (RANSE) solver. Both time accurate simulations using sliding grid and rotating frame approximations have been performed for comparison. The zigzag and turning circle maneuvers in calm water have been simulated for two different ship models, namely the ONR tumblehome (ONRT) test case and the KRISO (Korea Research Institute of Ship and Ocean) Container Ship (KCS) test case. Predicted ship motion is compared with measurement data to assess the accuracy of the numerical prediction using RANSE computations with different propeller models.

Cet article a pour premier objectif de faire un point sur les différents domaines de recherche en biohydrodynamique. Après une revue historique et bibliographique, l'accent sera mis sur la modélisation et les études numériques de la fonction locomotrice des poissons. La deuxième partie sera consacrée à la présentation du travail effectué au sein de l'E.M.N. dans le cadre du Projet Interdisciplinaire de Recherche CNRS Robea visant à concevoir un robot-anguille autopropulsé à locomotion anguilliforme. Les résultats des premiers calculs tridimensionnels recensés couplant la résolution de l'écoulement par les équations de Navier-Stokes en moyenne de Reynolds avec le Principe Fondamental de la Dynamique appliqué au corps à déformation imposée seront ainsi présentés. Un bilan et les perspectives de ces études concluront cette contribution.

A variant of the local correlation-based transition model (LCTM) γ is presented to account for the prediction of transition due to stationary crossflow instabilities. The presented T c 1 crossflow criterion is a local approximation of the well-known C1 criterion by Arnal (1984), evaluated locally through auxiliary functions expressed by the solution of the Falkner–Skan–Cooke (FSC) equations. The criterion, originally proposed by Menter and Smirnov (2014), uses the wall-normal directional change of the normalized vorticity as indicator of the crossflow strength. Hereafter, an original calibration FSC-based of Menter & Smirnov criterion is proposed. A local approximation of the sweep angle is included in order to achieve better results on non-wing-like geometries. The capability of this model variant to predict stationary crossflow transition is therefore discussed. A validation study is presented using experimental data on the 6:1 prolate spheroid and the sickle wing. • The study is performed using RANS local transition models. • Falkner–Skan–Cooke equations are used for calibration in default of experiments. • The re–calibration aims to add more physics to the transition correlations. • The sweep angle inclusion widens the range of applications of the Tc1 criterion. • The re–calibrated Tc1 performs very well on non-wing-like geometries. [ABSTRACT FROM AUTHOR]

ABSTRACTThe study describes a synthesis of a numerical and experimental work performed during the European research project EFFORT (European Full-scale FlOw Research and Technology) aiming at studying scale effects for hulls of high geometrical complexity by numerical and experimental means. During this project, the CFD group of Fluid Mechanics Laboratory (UMR6598) has performed a complete study of scale and appendage effects on an hopperdredger “Uilenspiegel”. These computations are discussed and compared with available local wake measurements obtained within EFFORT.

An interface capturing method based on a numerically revisited procedure for velocity and pressure coupling is worked out. The new treatment of density discontinuity is formulated in the framework of the finite volume methodology for arbitrary unstructured grids. A simple analytical pressure-like model case is presented to illustrate the accuracy of the numerical implementation of the treatment of discontinuous variables. Then, the method is implemented in a viscous flow solver and applied to free-surface flows, including the two-dimensional Rayleigh–Taylor instability problem and three-dimensional hydrodynamic flows for the prediction of ship waves around the Series 60 model ship with and without drift angles. These latter simulations show excellent agreement with the experimental results illustrated by comparisons of free-surface elevations and also velocity field. [Copyright &y& Elsevier]

Abstract: The numerical simulation of aerodynamic stall control using a synthetic jet actuator is presented and the automatic optimization of the control parameters is investigated. Unsteady Reynolds-averaged Navier–Stokes equations are solved on unstructured grids using a near-wall low-Reynolds number turbulence closure to simulate the effects of a synthetic jet, located at 12% of the chord from the leading edge of a NACA 0015 airfoil, for a Reynolds number Re =8.96×105 and incidences between 12° and 24°. Then, an automatic optimization procedure coupled with the flow solver is employed to optimize the parameters of the actuator (momentum coefficient, frequency, angle with respect to the wall) at each incidence in order to increase the time-averaged lift. A significant increase of the maximum lift is obtained (+52% with respect to the baseline airfoil) and the stall delayed from 16° to 22° for optimal parameters. The flow characteristics and the influence of the respective control parameters are analysed. [Copyright &y& Elsevier]

This article deals with Navier–Stokes simulations of multiphase flows around moving bodies coupled with an adaptive mesh refinement strategy. The numerical framework is considered first: the Navier–Stokes solver, the methodologies for handling multiphase flows and moving bodies, the remeshing techniques, and the adaptive procedure are explained and detailed. Then an application involving hydrodynamic impacts is presented in detail and studied to highlight the relevance of the whole global approach. Of particular interest is the accurate computation of pressure peaks arising during impacts.This article deals with Navier–Stokes simulations of multiphase flows around moving bodies coupled with an adaptive mesh refinement strategy. The numerical framework is considered first: the Navier–Stokes solver, the methodologies for handling multiphase flows and moving bodies, the remeshing techniques, and the adaptive procedure are explained and detailed. Then an application involving hydrodynamic impacts is presented in detail and studied to highlight the relevance of the whole global approach. Of particular interest is the accurate computation of pressure peaks arising during impacts.

This paper firstly aims at explaining numerical methods to achieve the integration of body motion into a Navier-Stokes solver. One of the peculiar features is to be able to deal not only with classical solids but also with flexible bodies (with an imposed deformation). One gives details concerning the resolution of the Newton's law (parameterization, use of quaternions to describe rotations), the regridding strategies linked to body motion and deformation and of the fluid-motion coupling methods. In the second part, the studies performed in the framework of the CNRS Project named ROBEA will be presented. This project aims at designing an efficient flexible underwater robot based on fish-like locomotion. The results of the first tridimensional simulations coupling a CFD code and the resolution of the Newton's law for a flexible body are shown. Cet article a pour objectif premier d'exposer les m?thodes num?riques li?es au mouvement de corps mises en ?uvre au sein d'un code de simulation d'?coulements turbulents de fluides visqueux incompressibles. Une des particularit?s est de pouvoir traiter, outre les solides classiques, des corps d?formables ? d?formation impos?e. On d?taillera les techniques utilis?es concernant la r?solution des ?quations du mouvement (param?trage, utilisation d'un quaternion), la gestion du maillage (li? au mouvement et ? la d?formation des corps) et le couplage ?coulement-mouvement. La deuxi?me partie sera consacr?e ? la pr?sentation des travaux effectu?s dans le cadre du Projet Interdisciplinaire de Recherche CNRS ROBEA visant ? concevoir un robot-anguille autopropuls? ? locomotion anguilliforme. Les r?sultats des premiers calculs tridimensionnels recens?s couplant la r?solution de l'?coulement par les ?quations de Navier-Stokes en moyenne de Reynolds avec le Principe Fondamental de la Dynamique appliqu? au corps ? d?formation impos?e seront ainsi expos?s.

Cet article a pour objectif premier d'exposer les méthodes numériques liées au mouvement de corps mises en œuvre au sein d'un code de simulation d'écoulements turbulents de fluides visqueux incompressibles. Une des particularités est de pouvoir traiter, outre les solides classiques, des corps déformables à déformation imposée. On détaillera les techniques utilisées concernant la résolution des équations du mouvement (paramétrage, utilisation d'un quaternion), la gestion du maillage (lié au mouvement et à la déformation des corps) et le couplage écoulement-mouvement. La deuxième partie sera consacrée à la présentation des travaux effectués dans le cadre du Projet Interdisciplinaire de Recherche CNRS ROBEAvisant à concevoir un robot-anguille autopropulsé à locomotion anguilliforme. Les résultats des premiers calculs tridimensionnels recensés couplant la résolution de l'écoulement par les équations de Navier-Stokes en moyenne de Reynolds avec le Principe Fondamental de la Dynamique appliqué au corps à déformation imposée seront ainsi exposés.

Abstract:  The practical use of automated computational fluid dynamics (CFD)-based design tools in the ship-building industry requires powerful flow solvers which are able to take into account realistic geometries as well as complex physical phenomena, such as turbulence. A shape optimization tool is developed in this framework. A derivative-free optimizer, yielding both flexibility and robustness, is preferred to the classical gradient-based method, which is more difficult to implement and is still limited to only moderately complex problems. The flow solver included in the design procedure solves the incompressible Reynolds-averaged Navier–Stokes equations on unstructured grids using a finite-volume formulation involving several near-wall low-Reynolds-number turbulence models. The design tool is used to optimize the stern of a modern hull shape at model and full scale, with different purposes being considered. More precisely, the drag reduction and the homogenization of the flow in the wake are expected by controlling the longitudinal vortex generated. Our interest is particularly focused on the influence of turbulence modeling in the design process. The effects of a two-equation model based on the eddy-viscosity assumption and a second-order closure relying on the Reynolds stress transport equations are compared.

In this article, the finite element simulation of cables is investigated for future applications to robotics and hydrodynamics. The solution is based on the geometrically exact approach of Cosserat beams in finite transformations, as initiated by Simo in the 1980s. However, the internal basic kinematics of the beam theory is not those of Reissner–Timoshenko but rather those of Kirchhoff. Based on these kinematics, the dynamic model adopted is a nonlinear extension of the so-called linear model of twisted and stretched Euler–Bernoulli beams. In agreement with the investigated applications, one or both of the ends of the cable are submitted to predefined motions. This model is also implemented into a computational fluid dynamics code, which solves the Reynolds-averaged Navier–Stokes equations. Regarding this last point, an implicit/iterative algorithm including a conservative load transfer for the variable hydrodynamic forces exerted all along the beam length has been used to reach a stable coupling. The relevance of the approach is tested through three advanced examples. The first is related to the prediction of cable motion in robotics. Then, the two last illustrations deal with fluid-structure interaction (FSI). A 2D classical benchmark in FSI is first investigated, and, at last, a computation illustrates the procedure in a 3D case.

Generating meshes with the right resolution is crucial for hybrid RANS/LES simulations of high Reynolds number flow with complex physical phenomena and geometries. This makes automatic mesh generation through adaptive refinement an interesting option. However, since the behavior of these turbulence models depends on the local grid size, mesh changes in time as a result of adaptive refinement may affect the production and destruction of turbulence. Therefore, grid adaptation with refinement criteria based on time-averaged quantities is proposed here to produce meshes that do not evolve rapidly in time and that are suitable for capturing the unsteady flow at each instant. Two averaging approaches, averaging over the instantaneous flow field, and averaging over the refinement criterion, are tested to simulate a turbulent flow behind a backward-facing step with a Detached Eddy Simulation type turbulence model. Compared to grid adaption based on instantaneous solutions, the computational cost is reduced, the accuracy of the solutions improves and an adapted mesh which has a generally static topology based on the main flow features is obtained. The investigation of this refinement process for two realistic test cases, a ship and a hybrid delta wing, confirms the reliability of the average-based adaptation and shows that automatic meshing for hybrid RANS/LES simulations of complex flows is possible. • Adaptive grid refinement is achieved for hybrid RANS/LES simulations. • Static mesh topologies based on time-averaged refinement criteria help to preserve the turbulence. • Accurate and efficient simulation of complex realistic cases is confirmed by tests. [ABSTRACT FROM AUTHOR]

We develop and analyze error estimators and mesh adaptation strategies within a discontinuous Galerkin formulation. The basic idea of the study is to reduce the computational cost of the simulation by employing mesh adaptation as a better alternative to the use of uniform grids. The novelty of the study resides in the use of multiwavelets and how their remarkable properties may shed new light on driving the adaptation process. This is motivated by the fact that multiwavelets break any input apart into a hierarchy of low resolution data and subsequently finer details. Our error estimator makes use of multiwavelets' properties while being local to the element, thereby maintaining the parallel efficiency of the solver. Early tests on the one-dimensional viscous Burgers equation have shown convincing results (García Bautista et al. [1]). This work is focused on the laminar backward-facing step configuration to assess the performance of the method in higher dimensions. • Mesh adaptation high-order discontinuous Galerkin schemes via multiwavelets. • Multiwavelet-guided adaptation enables substantial memory savings and speedup times. • Multiwavelet expansion yields detailed information; roadmap to hp-adaptation. [ABSTRACT FROM AUTHOR]

Highlights: [•] The paper studies grid refinement for realistic free-surface hydrodynamic flows. [•] Good refinement criteria are a combination of free-surface and derivative-based ones. [•] Refinement criteria based on the pressure Hessian need to contain smoothing. [•] Ratio of two criteria is set independent of scaling, Froude or Reynolds number. [•] Fine meshes of good quality are generated entirely with automatic grid refinement. [ABSTRACT FROM AUTHOR]

The paper presents a finite-volume-based Detached-Eddy Simulation for the prediction of flow around a passenger vehicle. The flow solver used is ISIS–CFD, developed by the CFD Department of the Fluid Mechanics Laboratory of Ecole Centrale de Nantes. This article presents a cross wind simulation around the square-back Willy model for several yaw angles. The model was designed in order that separations are limited to the region of the base for moderate yaw angles. This model without sharp corners on the fore body and a square base is designed to facilitate the analysis of separations which are, in that case, limited to its leeward side and base. The angle between the upstream velocity and the direction of the model varies from 0° and 30°. The results are compared to a previous numerical study based on a RANS simulation and experimental data at the Reynolds number Re=0.9×106. All comparisons (aerodynamic forces, wall pressures, and total pressure) show that DES simulations provide a better agreement with experimental data than isotropic or anisotropic statistical models, particularly for large yaw angles. [ABSTRACT FROM AUTHOR]

Abstract: An adaptive grid refinement method is presented for hydrodynamic flow simulation. It is meant for application to a wide range of realistic flow problems, so generality and flexibility of the method are essential. Directional refinement is developed to be used with unstructured hexahedral meshes, tensor-based refinement allows the implementation of many different refinement criteria. Good grid quality is assured by creating buffers of refined cells around relevant flow features. Tests are performed with two refinement criteria, based on the free surface and the Hessian matrix of the solution respectively; these show great increases in efficiency with respect to non-refined grids. [Copyright &y& Elsevier]