Publications
104 results found
Bempedelis N, Gori F, Wynn A, et al., 2024, Data-driven optimisation of wind farm layout and wake steering with large-eddy simulations, Wind Energy Science, ISSN: 2366-7443
O'Connor J, Laizet S, Wynn A, et al., 2024, Quantifying uncertainties in direct numerical simulations of a turbulent channel flow, Computers and Fluids, Vol: 268, ISSN: 0045-7930
Direct numerical simulation (DNS) provides unrivalled levels of detail and accuracy for simulating turbulent flows. However, like all numerical methods, DNS is subject to uncertainties arising from the numerical scheme and input parameters (e.g. mesh resolution). While uncertainty quantification (UQ) techniques are being employed more and more to provide a systematic analysis of uncertainty for lower-fidelity models, their application to DNS is still relatively rare. In light of this, the aim of this work is to apply UQ and sensitivity analysis to the DNS of a canonical wall-bounded turbulent channel flow at low Reynolds number (Re = 180). To compute the DNS, Incompact3d – a highly scalable open-source framework based on high-order compact finite differences and a spectral Poisson solver – is used as a black-box solver. Stochastic collocation is used to propagate the input uncertainties through Incompact3d to the output quantities of interest (QOIs). To facilitate the non-intrusive forward UQ analysis, the open-source EasyVVUQ package is used to provide integrated capability for sampling, pre-processing, execution, post-processing, and analysis of the computational campaign. Three separate UQ campaigns are conducted. The first two examine the effect of domain size and the numerical parameters (e.g. mesh resolution, time step, sample time), respectively, and adopt Gaussian quadrature rules combined via tensor products to sample the multi-dimensional input space. Finally, the third campaign investigates the performance of a dimension-adaptive sampling strategy that significantly reduces the computational cost compared to the full tensor product approach. The analysis focuses on the cross-channel statistical moments of the QOIs, as well as local and global sensitivity analyses to assess the sensitivity of each QOI with respect to each individual input. This enables an assessment of the robustness and sensitivity of DNS to the user-defined numerical paramete
Rolfo S, Flageul C, Bartholomew P, et al., 2023, The 2DECOMP&FFT library: an update with new CPU/GPU capabilities, Journal of Open Source Software, Vol: 8, ISSN: 2475-9066
The 2DECOMP&FFT library is a software framework written in modern Fortran to build large-scale parallel applications. It is designed for applications using three-dimensional structured meshes with a particular focus on spatially implicit numerical algorithms. However, the library can be easily used with other discretisation schemes based on a structured layout and wherepencil decomposition can apply. It is based on a general-purpose 2D pencil decomposition for data distribution and data Input Output (I/O). A 1D slab decomposition is also available as a special case of the 2D pencil decomposition. The library includes a highly scalable andefficient interface to perform three-dimensional Fast Fourier Transforms (FFTs). The library has been designed to be user-friendly, with a clean application programming interface hiding most communication details from application developers, and portable with support for modern CPUs and NVIDIA GPUs (support for AMD and Intel GPUs to follow).
Mays M, Lardeau S, Laizet S, 2023, Capturing the drag crisis in the flow around a smooth cylinder using a hybrid RANS-LES model on coarse meshes, International Journal of Heat and Fluid Flow, Vol: 103, Pages: 1-14, ISSN: 0142-727X
A hybrid strategy combining Reynolds-averaged Navier–Stokes (RANS) and large eddy simulation (LES) methods is nowadays seen as an efficient way to simulate turbulent flows of practical relevance. In this work, the scale-resolving hybrid (SRH) model proposed by Manceau (2018) is compared with a conventional unsteady RANS model, and LES and experimental data from the literature for the flow around a smooth cylinder in the flow regime around the drag crisis. Based on a temporal filtering formalism, this approach has seen limited testing for turbulent separated flows. The drag crisis phenomenon is dominated by complex near-wall physics and is challenging to simulate. The predictive accuracy and the robustness to mesh coarsening for the SRH model are assessed for this test case, with the aim to demonstrate that this hybrid approach can be a credible cost-saving alternative to LES for separated turbulent flows. The meshes considered in this numerical study are far coarser than the ones used in the LES reference data, yet, the results for the time mean drag are found in good agreement with the reference data. Other features of the flow, such as the presence and sizes of the laminar separation bubbles and the consequent magnitude of the time mean lift are not as well captured. In general, the qualitative behaviour of the SRH model is good when considered in the context of questions previously raised in the literature about hybrid models. The mesh savings are achieved by coarsening the spatial resolution in the wake, whereas the resolution required in the near-wall area and shear layers remains high, though much reduced compared to the reference LES data. The key point taken from this study is that the SRH model is an attractive option to produce higher-fidelity data on coarse meshes.
Girayhan Ozbay A, Laizet S, 2023, FR3D: Three-dimensional flow reconstruction and force estimation for unsteady flows around extruded bluff bodies via conformal mapping aided convolutional autoencoders, International Journal of Heat and Fluid Flow, Vol: 103, Pages: 1-15, ISSN: 0142-727X
In many practical fluid dynamics experiments, measuring variables such asvelocity and pressure is possible only at a limited number of sensor locations,for a few two-dimensional planes, or for a small 3D domain in the flow. However, knowledge of the full fields is necessary to understand the dynamicsof many flows. Deep learning reconstruction of full flow fields from sparsemeasurements has recently garnered significant research interest, as a way ofovercoming this limitation. This task is referred to as the flow reconstruction (FR) task. In the present study, we propose a convolutional autoencoderbased neural network model, dubbed FR3D, which enables FR to be carriedout for three-dimensional flows around extruded 3D objects with differentcross-sections. An innovative mapping approach, whereby multiple fluid domains are mapped to an annulus, enables FR3D to generalize its performanceto objects not encountered during training. We conclusively demonstratethis generalization capability using a dataset composed of 80 training and 20testing geometries, all randomly generated. We show that the FR3D modelreconstructs pressure and velocity components with a few percentage pointsof error. Additionally, using these predictions, we accurately estimate theQ-criterion fields as well lift and drag forces on the geometries.
Gori F, Laizet S, Wynn A, 2023, Sensitivity analysis of wake steering optimisation for wind farm power maximisation, Wind Energy Science, Vol: 8, Pages: 1425-1451, ISSN: 2366-7443
Modern large-scale wind farms consist of multiple turbines clustered together, usually in well-structured formations. Clustering has a number of drawbacks during a wind farm's operation, as some of the downstream turbines will inevitably operate in the wake of those upstream, with a significant reduction in power output and an increase in fatigue loads. Wake steering, a control strategy in which upstream wind turbines are misaligned with the wind to redirect their wakes away from downstream turbines, is a promising strategy to mitigate power losses. The purpose of this work is to investigate the sensitivity of open-loop wake steering optimisation in which an internal predictive wake model is used to determine the farm power output as a function of the turbine yaw angles. Three different layouts are investigated with increasing levels of complexity. A simple 2×1 farm layout under aligned conditions is first considered, allowing for a careful investigation of the sensitivity to wake models and operating conditions. A medium-complexity case of a generic 5×5 farm layout under aligned conditions is examined to enable the study of a more complex design space. The final layout investigated is the Horns Rev wind farm (80 turbines), for which there have been very few studies of the performance or sensitivity of wake steering optimisation. Overall, the results indicate a strong sensitivity of wake steering strategies to both the analytical wake model choice and the particular implementation of algorithms used for optimisation. Significant variability can be observed in both farm power improvement and optimal yaw settings, depending on the optimisation setup. Through a statistical analysis of the impact of optimiser initialisation and a study of the multi-modal and discontinuous nature of the underlying farm power objective functions, this study shows that the uncovered sensitivities represent a fundamental challenge to robustly identifying globally optimal solutio
Hodgkin A, Deskos G, Laizet S, 2023, On the interaction of a wind turbine wake with a conventionally neutral atmospheric boundary layer, International Journal of Heat and Fluid Flow, Vol: 102, Pages: 1-16, ISSN: 0142-727X
In this work, we investigate the dynamics of wind turbine tip-vortex breakdown in a conventionally neutral atmospheric boundary layer (ABL). To this end, high-resolution data are collected from large-eddy simulations of awind turbine operating within a neutral ABL and studied by means of proper orthogonal decomposition (POD)and Fourier analysis. The high resolution of the generated data in both space and time allows us to gain insightinto the tip-vortex breakdown mechanisms by (i) capturing the energy modes of the coherent structures, (ii) studying their contribution to the tip-vortex breakdown through their power spectra functions and mean kinetic energy(MKE) flux, and (iii) analysing the growth rate of each contributing perturbation frequency along tip vortices.Our analysis shows that under a fully turbulent scenario, the growth rate of perturbations along the tip vortices islargest for low wave numbers, i.e. long-wave perturbations. Additionally, the MKE flux reaches its highest valueat two diameters downstream of the rotor plane, a behaviour that can be attributed to the coexistence of multipleinteracting POD modes, with the streamwise vortex roller mode being the primary contributor to the total MKEflux budget, contributing approximately 24%. Finally, comparisons with a laminar, uniform flow scenario subjectto a single-frequency perturbation highlight the differences between the two ambient flow conditions. In the nonturbulent, uniform flow scenario, the growth rate attains its maximum value at a wave number corresponding tothe out-of-phase mutual-inductance mechanism, whereas the MKE flux exhibits local minima and maxima alongthe wake and at different downstream locations depending on the perturbation frequency. Our analyses suggestthat the breakdown of the wind turbine tip vortices under a fully turbulent neutral ABL inflow is due to complexinteractions across a range of excitation frequencies, in which the mutual-inductance instability may not be thedominant
O'Connor J, Diessner M, Wilson K, et al., 2023, Optimisation and analysis of streamwise-varying wall-normal blowing in a turbulent boundary layer, Flow, Turbulence and Combustion, Vol: 110, Pages: 993-1021, ISSN: 0003-6994
Skin-friction drag is a major engineering concern, with wide-ranging consequences across many industries. Active flow-control techniques targeted at minimising skin friction have the potential to significantly enhance aerodynamic efficiency, reduce operating costs, and assist in meeting emission targets. However, they are difficult to design and optimise. Furthermore, any performance benefits must be balanced against the input power required to drive the control. Bayesian optimisation is a technique that is ideally suited to problems with a moderate number of input dimensions and where the objective function is expensive to evaluate, such as with high-fidelity computational fluid dynamics simulations. In light of this, this work investigates the potential of low-intensity wall-normal blowing as a skin-friction drag reduction strategy for turbulent boundary layers by combining a high-order flow solver (Incompact3d) with a Bayesian optimisation framework. The optimisation campaign focuses on streamwise-varying wall-normal blowing, parameterised by a cubic spline. The inputs to be optimised are the amplitudes of the spline control points, whereas the objective function is the net-energy saving (NES), which accounts for both the skin-friction drag reduction and the input power required to drive the control (with the input power estimated from real-world data). The results of the optimisation campaign are mixed, with significant drag reduction reported but no improvement over the canonical case in terms of NES. Selected cases are chosen for further analysis and the drag reduction mechanisms and flow physics are highlighted. The results demonstrate that low-intensity wall-normal blowing is an effective strategy for skin-friction drag reduction and that Bayesian optimisation is an effective tool for optimising such strategies. Furthermore, the results show that even a minor improvement in the blowing efficiency of the device used in the present work will lead to meaningful
Laizet S, Revell A, Emerson D, 2023, The UK Turbulence Consortium, Flow, Turbulence and Combustion, Vol: 110, Pages: 773-774, ISSN: 0003-6994
Bempedelis N, Laizet S, Deskos G, 2023, Turbulent entrainment in finite-length wind farms, Journal of Fluid Mechanics, Vol: 955, ISSN: 0022-1120
In this article, we present an entrainment-based model for predicting the flow and power output of finite-length wind farms. The model is an extension of the three-layer approach of Luzzatto-Fegiz & Caulfield (Phys. Rev. Fluids, vol. 3, 2018, 093802) for wind farms of infinite length, and assumes dependence of key flow quantities, such as the wind farm bulk velocity, on the streamwise distance from the farm entrance. To assist our analysis and validate the proposed model, we undertake a series of large-eddy simulations with different turbine spacing arrangements and layouts. Comparisons are also made with the top-down model with entrance effects of Meneveau (J. Turbul., vol. 13, 2012, N7) and data from the literature. The finite-length entrainment model is shown to be capable of capturing the power drop between contiguous rows of turbines as well as describing the advection and turbulent transport of kinetic energy in both the entrance and fully developed regions. The fully developed regime is approximated only deep in the wind farm, after approximately 15 rows of turbines. Our data suggest that for the cases considered in this study, the empirical coefficients that can be used to describe turbulent entrainment and transfers above the wind farm exhibit little dependence on the farm layout and may be considered constant for modelling purposes. However, the flow field within the wind farm layer can be strongly modulated by the turbine density (spacing) as well as the array layout, and to that extent it can be argued that they are both primary factors determining the wind farm power output.
Gori F, Wynn A, Laizet S, 2023, Sensitivity of wind farm wake steering strategies to analytical wake models, Pages: 669-677
The aerodynamic interactions between wind turbines arranged in farm layout lead to annual energy production losses ranging from 10% to 30%. Wake steering represents a promising strategy in wind farm control for power loss mitigation. The purpose of this work is to assess the sensitivity of optimal wake steering strategies to both analytical wake model choice and optimisation parameters. Using the FLOw Redirection and Induction in Steady State (FLORIS) framework, different wake models are employed to optimise a 4 × 4 farm layout for power maximisation. Model comparison findings indicate significant discrepancies in absolute power predictions for optimal set-points, as well as in optimal decision variables, with different or even opposite optimal yaw angle settings. Initialisation sensitivity results show that solutions corresponding to local extrema lead to potential power losses up to 14% compared to the global maximum for power production. Moreover, wind farm power function is observed to be multi-modal and discontinuous, suggesting that care must be taken when using gradient-based methods in wake steering optimisation.
Schuch FN, Silvestrini JH, Meiburg E, et al., 2023, The Plunging of Hyperpycnal Plumes on Tilted Bed by Three-Dimensional Large-Eddy Simulations, Pages: 41-55, ISSN: 2195-4356
Theoretical and experimental interest in transport and deposition of sediments from rivers to oceans has increased rapidly over the last two decades. The marine ecosystem is strongly affected by mixing at river mouths, with, for instance, anthropogenic actions like pollutant spreading. Particle-laden flows entering a lighter ambient fluid (hyperpycnal flows) can plunge at a sufficient depth, and their deposits might preserve a remarkable record across a variety of climatic and tectonic settings. Numerical simulations play an essential role in this context since they provide information on all flow variables for any point of time and space. This work offers valuable spatio-temporal information generated by turbulence-resolving 3D simulations of poly-disperse hyperpycnal plumes over a tilted bed. The simulations are performed with the high-order flow solver Xcompact3d, which solves the incompressible Navier–Stokes equations on a Cartesian mesh using high-order finite-difference schemes. Five cases are presented, with different values for flow discharge and sediment concentration at the inlet. A detailed comparison with experimental data and analytical models is already available in the literature. The main objective of this work is to present a new dataset that shows the entire three-dimensional spatio-temporal evolution of the plunge phenomenon and all the relevant quantities of interest.
Jane-Ippel C, Bempedelis N, Palacios R, et al., 2023, High-fidelity simulations of wake-to-wake interaction in an atmospheric boundary layer over a complex terrain, 8th Wake Conference, Publisher: IOP PUBLISHING LTD, ISSN: 1742-6588
Diessner M, O'Connor J, Wynn A, et al., 2022, Investigating Bayesian optimization for expensive-to-evaluate black box functions: Application in fluid dynamics, FRONTIERS IN APPLIED MATHEMATICS AND STATISTICS, Vol: 8
- Author Web Link
- Cite
- Citations: 4
Giannenas A, Bempedelis N, Schuch F, et al., 2022, A Cartesian immersed boundary method based on 1D flow reconstructions for high-fidelity simulations of incompressible turbulent flows around moving objects, Flow, Turbulence and Combustion, Vol: 109, Pages: 931-959, ISSN: 0003-6994
The aim of the present numerical study is to show that the recentlydeveloped Alternating Direction Reconstruction Immersed BoundaryMethod (ADR-IBM) [1] can be used for Fluid-Structure Interaction (FSI)problems and can be combined with an Actuator Line Model (ALM)and a Computer-Aided Design (CAD) interface for high-fidelity simulations of fluid flow problems with rotors and geometrically compleximmersed objects. The method relies on 1D cubic spline interpolations to reconstruct an artificial flow field inside the immersed objectwhile imposing the appropriate boundary conditions on the boundariesof the object. The new capabilities of the method are demonstratedwith the following flow configurations: a turbulent channel flow withthe wall modelled as an immersed boundary, Vortex Induced Vibrations (VIVs) of one-degree-of-freedom (2D) and two-degree-of-freedom(3D) cylinders, a helicopter rotor and a multi-rotor unmanned aerialvehicle in hover and forward motion. These simulations are performed with the high-order fluid flow solver Incompact3d which is based ona 2D domain decomposition in order to exploit modern CPU-basedsupercomputers. It is shown that the ADR-IBM can be used for thestudy of FSI problems and for high-fidelity simulations of incompressible turbulent flows around moving complex objects with rotors.
Hamzehloo A, Lusher DJ, Laizet S, et al., 2022, Direct numerical simulations of shocklet-containing turbulent channel counter-flows, 12th International Symposium on Turbulence and Shear Flow Phenomena (TSFP12), Publisher: International Symposium series on Turbulence and Shear Flow Phenomena, Pages: 1-6
Counter-flow or counter-current configurations can maintain high turbulence intensities and exhibit a significant level of mixing. We have previously introduced a wall-bounded counter-flow turbulent channel configuration (Physical Review Fluids, 6(9), p.094603.) as an efficient framework to study compressibility effects on turbulence. Here, we extend our previous direct numerical simulation study to a relatively higher Mach number (M = 0.7) to investigate strong compressibility effects (also by reducing the Prandtl number from Pr = 0.7 to 0.2), and the formation and evolution of unsteadyshocklet structures. It is found that the configuration is able to produce highly turbulent flows with embedded shocklets and significant asymmetry in probability density functions of dilatation. A peak turbulent Mach number close to unity is obtained, for which the contribution of the dilatational dissipation to total dissipation is nevertheless found to be limited to 6%.
Giannenas A, Laizet S, Rigas G, 2022, Harmonic forcing of a laminar bluff body wake with rear pitching flaps, Journal of Fluid Mechanics, Vol: 945, ISSN: 0022-1120
A numerical study on the response of a 2D bluff body wake subjected to harmonicforcing imposed by two rear pitching flaps is performed. The wake is generated by arectangle at a height-based Reynolds number Re = 100, characterised by laminar vortexshedding. Two forcing strategies are examined corresponding to in-phase “snaking” andout-of-phase “clapping”. The effects of the bluff body aspect ratio (AR = 1, 2, 4), flappingfrequency, flapping amplitude, flap length and Reynolds number are investigated. For thesnaking motion, a strong fundamental resonance of the root mean square (RMS) drag isobserved when the wake is forced near the vortex shedding frequency. For the clappingmotion, a weak subharmonic resonance is observed when the forcing is applied neartwice the vortex shedding frequency resulting in an increase of the lift RMS whereasthe drag RMS remains unaffected. Both resonances intensify the vortex shedding and aconcomitant mean drag increase is observed for the snaking motion. Forcing away fromthe resonant regimes, both motions result in considerable drag reduction through wakesymmetrisation and propulsion mechanisms. The formation of two vortex dipoles peroscillation period due to the flapping motion, which weaken the natural vortex shedding,has been identified as the main symmetrisation mechanism. A single scaling parameter isproposed to collapse the mean drag reduction of the forced flow for both motions over awide range of flapping frequencies, amplitudes and flap lengths. Finally, the assessmentof the performance of the forcing strategies has revealed that clapping is more effectivethan snaking.
Hodgkin A, Laizet S, Deskos G, 2022, Numerical investigation of the influence of shear and thermal stratification on the wind turbine tip-vortex stability, Wind Energy, Vol: 25, Pages: 1270-1289, ISSN: 1095-4244
The interaction between wind turbine wakes and atmospheric turbulence is characterisedby complex dynamics. In this study, two major components of the atmospheric boundarylayer dynamics have been isolated, namely, the mean velocity profile shear and the thermalstratification, to examine their impact on the near-wake development by undertaking aseries of highly resolved large-eddy simulations. Subsequently, instantaneous flow fields areextracted from the simulations and used to conduct Fourier analysis and proper orthogonaldecomposition (POD) and compute the mean kinetic energy fluxes by different POD modesto better understand the tip-vortex instability mechanisms. Our findings indicate that shearcan significantly affect the breakup of the wind turbine tip-vortices as well as the shapeand stable length of the wake, whereas thermal stratification seems to only have limitedcontribution to the spatial development of the near-wake field. Finally, our analysis showsthat the applied perturbation frequency determines the tip-vortex breakup location as itcontrols the onset of the mutual inductance instability.
Hodgkin A, Laizet S, Deskos G, 2022, Do ambient shear and thermal stratification impact wind turbine tip-vortex breakdown?, ISSN: 1742-6588
Modern wind turbines experience uneven inflow conditions across the rotor, due to the ambient flow's shear and thermal stratification. Such conditions alter the shape and length of turbine wakes and thus impact the loads and power generation of downstream turbines. To this end, understanding the spatial evolution of the individual wakes under different atmospheric conditions is key to controlling and optimising turbine arrays. With this numerical study we aim to obtain a better understanding of the fundamental physics governing the near-wake dynamics of wind turbines under shear and thermal stability, by examining their tip-vortex breakup mechanisms. Our approach considers scale-resolving simulations of a single turbine wake under a linear shear profile as well as the application of harmonic tip perturbations to trigger flow instabilities. For the subsequent analysis we use the proper orthogonal decomposition (POD) method to extract coherent structures from the flow, and we also calculate mean kinetic energy fluxes to quantify each coherent structure's contribution to wake recovery. The wake's helical spiral is found to hinder wake recovery for all studied ambient flow conditions, whereas the mutual inductance instability has positive MKE flux leading to an enhanced wake recovery. Finally, the ambient shear has the largest impact on the local MKE flux with respect to downstream location by changing the shape of the curve and location of extrema, whereas thermal stratification has only a minimal impact on the magnitude of the near-wake local MKE flux distribution.
Ozbay A, Laizet S, 2022, Deep learning fluid flow reconstruction around arbitrary two-dimensional objects from sparse sensors using conformal mappings, AIP Advances, Vol: 12, ISSN: 2158-3226
The usage of neural networks (NNs) for flow reconstruction (FR) tasks from a limited number of sensors is attracting strong research interest owing to NNs’ ability to replicate high-dimensional relationships. Trained on a single flow case for a given Reynolds number or over a reduced range of Reynolds numbers, these models are unfortunately not able to handle flows around different objects without re-training. We propose a new framework called Spatial Multi-Geometry FR (SMGFR) task, capable of reconstructing fluid flows around different two-dimensional objects without re-training, mapping the computational domain as an annulus. Different NNs for different sensor setups (where information about the flow is collected) are trained with high-fidelity simulation data for a Reynolds number equal to ∼300 for 64 objects randomly generated using Bezier curves. The performance of the models and sensor setups is then assessed for the flow around 16 unseen objects. It is shown that our mapping approach improves percentage errors by up to 15% in SMGFR when compared to a more conventional approach where the models are trained on a Cartesian grid and achieves errors under 3%, 10%, and 30% for predictions of pressure, velocity, and vorticity fields, respectively. Finally, SMGFR is extended to predictions of snapshots in the future, introducing the Spatiotemporal MGFR (STMGFR) task. A novel approach is developed for STMGFR involving splitting deep neural networks into a spatial and a temporal component. We demonstrate that this approach is able to reproduce, in time and in space, the main features of flows around arbitrary objects.
Francisco E, Espath LF, Laizet S, et al., 2022, Direct numerical simulations of intrusive density- and particle-driven gravity currents, Physics of Fluids, Vol: 34, Pages: 1-19, ISSN: 1070-6631
In the present study, mesopycnal flows are investigated using Direct Numerical Simulations (DNS). In particular, intrusive density- and particle-driven gravity currentsin the lock exchange set-up are simulated with the high-order finite-difference framework Xcompact3d. To account for the settling velocity of particles, a customisedFick’s law for the particle-solution species is used with an additional term incorporating a constant settling velocity proportional to the concentration of particles. Ageneral energy budget equation is presented, for which the energy can migrate acrossthe domain’s boundaries. The relevant main features of intrusive gravity currents,such as front velocity, energy exchanges, sedimentation rate, deposit profile, anddeposit map are discussed with comparison between two and three-dimensional simulations. In particular, the influence of the Grashof number, the interface thickness,the energy exchanges, the sedimentation process, and how the presence of more thanone particle fraction may change the flow dynamics are investigated. The results arein good agreement with previous experiments and theoretical work, in particular forthe prediction of the front velocity. For the particle-driven case, the suspended massevolution along with the sedimentation rate suggests the occurrence of three differentstages. In the first stage after the lock release, the particle mixture tends to suspenditself due to gravitational forces. Once most of the particle-mixture mass is suspended, the current intrudes while increases its velocity, reaching its kinetic energypeak. In the last stage, the particles are deposited at a nearly constant sedimentationrate. As a results, the front velocity constantly decelerates.
Khojasteh AR, Laizet S, Heitz D, et al., 2022, Lagrangian and Eulerian dataset of the wake downstream of a smooth cylinder at a Reynolds number equal to 3900., Data in Brief, Vol: 40, Pages: 1-9, ISSN: 2352-3409
The dataset contains Eulerian velocity and pressure fields, and Lagrangian particle trajectories of the wake flow downstream of a smooth cylinder at a Reynolds number equal to 3900. An open source Direct Numerical Simulation (DNS) flow solver named Incompact3d was used to calculate the Eulerian field around the cylinder. The synthetic Lagrangian tracer particles were transported using a fourth-order Runge-Kutta scheme in time and trilinear interpolations in space. Trajectories of roughly 200,000 particles for two 3D sub-domains are available to the public. This dataset can be used as a test case for tracking algorithm assessment, exploring the Lagrangian physics, statistic analyses, machine learning, and data assimilation interests.
Özbay AG, Laizet S, 2022, UNSTEADY TWO-DIMENSIONAL FLOW RECONSTRUCTION AND FORCE COEFFICIENT ESTIMATION AROUND ARBITRARY SHAPES VIA CONFORMAL MAPPING AIDED DEEP NEURAL NETWORKS
In many practical fluid dynamics experiments, measuring variables such as velocity and pressure is possible only at a limited number of sensor locations. However, knowledge of the full fields is necessary to understand the dynamics of many flows. Deep learning reconstruction of full flow fields from sparse measurements as a way of overcoming this limitation has recently garnered significant research interest, referred to as the flow reconstruction (FR) task. We extend existing FR models by enabling such models to make predictions on flows around arbitrary 2D geometries without the need for re-training. This geometry flexibility is achieved through an innovative mapping approach, whereby multiple fluid domains are mapped to an annulus. Using this mapping approach, we explore the performance of a novel FR model trained on 64 geometries and tested on a further 16 different geometries. We demonstrate that the model trained using the mapping approach reconstructs the flow fields well even on geometries not present in the training data.
Hodgkin A, Laizet S, Deskos G, 2022, IMPLICATIONS OF SHEAR AND THERMAL STRATIFICATION ON WIND TURBINE TIP-VORTEX STABILITY
The interaction between wind turbines in a wind farm through their wakes is a phenomenon that has been studied for decades and is still relevant today. Turbines clustered together in arrays will often operate in the wake of other upstream turbines which may lead to significant power losses and fatigue loads. For modern large-scale wind turbines, the mean shear velocity profile and thermal stratification are major components of the atmospheric boundary layer so it is important to understand their impact on near-wake development. Additionally, veer is present due to the rotation of the Earth. The impact of shear, thermal stratification and veer on the stable wake length of turbines with a dynamic control strategy is studied numerically in this work using a suite of highly resolved large-eddy simulations. Instantaneous flow fields are extracted from the simulations and used to conduct proper orthogonal decomposition (POD) and compute the mean kinetic energy fluxes by different POD modes to better understand the tip-vortex instability mechanisms. Our findings show that the dynamic pitch control scheme is able to shorten the stable wake length to about 1.5R in uniform flow. Shear can significantly affect the break up of wind turbine tip-vortices as well as the shape and stable length of the wake, whereas thermal stratification seems to only have limited contribution to the spatial development of the near-wake field. Veer causes the wake boundary to skew but has a limited impact on the wake length.
Balogh GD, Flynn T, Laizet S, et al., 2021, Scalable many-core algorithms for tridiagonal solvers, Computing in Science and Engineering, Vol: 24, ISSN: 1521-9615
We present a novel distributed memory Tridiagonal solver library, targeting large-scale systems based on modern multi-core and many-core processor architectures. The library uses methods based on both approximate and exact algorithms. Performance comparisons with the state-of-the-art, using both a large Cray EX system and a GPU cluster show the algorithmic trade-offs required at increasing machine scale to achieve good performance, particularly considering the advent of exascale systems.
Giannenas A, Laizet S, 2021, A simple and scalable Immersed Boundary Method for high-fidelity simulations of fixed and moving objects on a Cartesian mesh, Applied Mathematical Modelling: simulation and computation for engineering and environmental systems, Vol: 99, Pages: 606-627, ISSN: 0307-904X
A simple and scalable Immersed Boundary Method based on cubic spline reconstructionsis presented for high-fidelity simulations of immersed objects in a turbulent flow on aCartesian mesh. The novelty of the proposed IBM lies in its simplicity, accuracy, scalabilityand its ability to simulate both fixed and moving immersed objects. The new IBM is thoroughly validated against a 1D benchmark, with the 2D flow around a cylinder at Re = 40and 300 and the 3D flow around a sphere at Re = 300 and Re = 3700. Convergence studies and detailed error maps showing the spatial distribution of the velocity L2-Norm errorcompared to a spectral reference solution for the cylinders at Re = 40 show the robustnessof the proposed method. The cost and performance of the method are also presented formulti-billion mesh node simulations with up to 65,536 computational cores. The potentialof the method in handling multiple moving objects for practical applications is demonstrated with the control of a square bluff body wake by two rear pitching flaps.
Mahfoze O, Laizet S, 2021, Non-explicit Large Eddy Simulations of turbulent channel flows from Reτ= 180 up to Reτ= 5,200, Computers and Fluids, Vol: 228, Pages: 1-19, ISSN: 0045-7930
This numerical study based on high-order finite-difference schemes presents LES-NWR (Large Eddy Simulation with near-wall resolution) of turbulent channel flows up to Reτ = 5200 using non-explicit approaches for which numerical dissipation is introduced via the discretisation of viscous terms of the Navier-Stokes equations. These models are cheaper than explicit LES models as no extra terms are needed in the equations toaccount for the contribution of the unresolved scales. The Approximate Deconvolution Method with Relaxation Term (ADM-RT) approach is also assessed and the present LES data are compared with reference Direct Numerical Simulations (DNS) data for first and second order moments as well as for the turbulent kinetic energy budget. Even if the viscous sublayer is not resolved, the proposed non-explicit LES approaches are in excellent agreement with the reference DNS data and to a certain extent with the ADM-RT model, for a fraction of the cost of the DNS. The proposed non-explicit models, for which the possibility to locally adjust the added numerical dissipation is investigated, are straightforward to implement and come with a negligible additional computational cost while the ADM-RT model is 30% more expensive than the non-explicit models.The parameters of the models are defined before the simulations and no modifications of the parameters are needed when the Reynolds number and the mesh resolution are changed. It is shown that a modulation of the magnitude of the numerical dissipation in time and in space is not necessarily needed, at least for the mesh resolutions and Reynolds numbers considered in the present study. The main conclusion is that non-explicit models can replace advantageously explicit models when high-order finite-difference methods are used. They can generate accurate LES-NWR of turbulent channel flows over a wide range of Reynolds numbers at a fraction of a cost of DNS.
Hamzehloo A, Lusher D, Laizet S, et al., 2021, Direct numerical simulation of compressible turbulence in acounter-flow channel configuration, Physical Review Fluids, Vol: 6, Pages: 1-21, ISSN: 2469-990X
Counter-flow configurations, whereby two streams of fluid are brought together from oppositedirections, are highly efficient mixers due to the high turbulence intensities that can be maintained.In this paper, a simplified version of the problem is introduced that is amenable to direct numericalsimulation. The resulting turbulent flow problem is confined between two walls, with one non-zeromean velocity component varying in the space direction normal to the wall, corresponding to asimple shear flow. Compared to conventional channel flows, the mean flow is inflectional and themaximum turbulence intensity relative to the maximum mean velocity is nearly an order of magnitude higher. The numerical requirements and turbulence properties of this configuration are firstdetermined. The Reynolds shear stress is required to vary linearly by the imposed forcing, witha peak at the channel centreline. A similar behaviour is observed for the streamwise Reynoldsstress, the budget of which shows an approximately uniform distribution of dissipation, with largecontributions from production, pressure-strain and turbulent diffusion. A viscous sublayer is obtained near the walls and with increasing Reynolds number small-scale streaks in the streamwisemomentum are observed, superimposed on the large-scale structures that buffet this region. Whenthe peak local mean Mach number reaches 0.55, turbulent Mach numbers of 0.6 are obtained,indicating that this flow configuration can be useful to study compressibility effects on turbulence.
Khojasteh AR, Yang Y, Heitz D, et al., 2021, Lagrangian coherent track initialisation, Physics of Fluids, Vol: 33, Pages: 1-13, ISSN: 1070-6631
Advances in time-resolved three-dimensional Particle Tracking Velocimetry (4D-PTV)techniques have been consistently revealed more accurate Lagrangian particle motions.A novel track initialisation technique as a complementary part of 4D-PTV, based on localtemporal and spatial coherency of neighbour trajectories, is proposed. The proposed Lagrangian Coherent Track Initialisation (LCTI) applies physics-based Finite Time LyapunovExponent (FTLE) to build four frame coherent tracks. We locally determine LagrangianCoherent Structures (LCS) among neighbour trajectories by using the FTLE boundaries(i.e., ridges) to distinguish clusters of coherent motions. To evaluate the proposed technique, we created an open-access synthetic Lagrangian and Eulerian dataset of the wakedownstream of a smooth cylinder at a Reynolds number equal to 3900 obtained fromthree-dimensional (3D) Direct Numerical Simulation (DNS). Performance of the proposedmethod based on three characteristic parameters, temporal scale, particle concentration(i.e., density), and noise ratio, showed robust behaviour in finding true tracks compared tothe recent initialisation algorithms. Sensitivity of LCTI to the number of untracked andwrong tracks are also discussed. We address the capability of using the proposed methodas a function of a 4D-PTV scheme in the Lagrangian Particle Tracking (LPT) challenge.We showed that LCTI prevents 4D-PTV divergence in flows with high particle concentrations. Finally, the LCTI behaviour was demonstrated in a jet impingement experiment.LCTI was found to be a reliable tracking tool in complex flow motions, with a strengthrevealed for flows with high velocity and acceleration gradients.
Özbay AG, Hamzehloo A, Laizet S, et al., 2021, Poisson CNN: Convolutional neural networks for the solution of the Poisson equation on a Cartesian mesh, Data-Centric Engineering, Vol: 2, Pages: 1-31, ISSN: 2632-6736
<jats:title>Abstract</jats:title> <jats:p>The Poisson equation is commonly encountered in engineering, for instance, in computational fluid dynamics (CFD) where it is needed to compute corrections to the pressure field to ensure the incompressibility of the velocity field. In the present work, we propose a novel fully convolutional neural network (CNN) architecture to infer the solution of the Poisson equation on a 2D Cartesian grid with different resolutions given the right-hand side term, arbitrary boundary conditions, and grid parameters. It provides unprecedented versatility for a CNN approach dealing with partial differential equations. The boundary conditions are handled using a novel approach by decomposing the original Poisson problem into a homogeneous Poisson problem plus four inhomogeneous Laplace subproblems. The model is trained using a novel loss function approximating the continuous <jats:inline-formula> <jats:alternatives> <jats:inline-graphic xmlns:xlink="http://www.w3.org/1999/xlink" mime-subtype="png" xlink:href="S2632673621000071_inline1.png" /> <jats:tex-math>$ {L}^p $</jats:tex-math> </jats:alternatives> </jats:inline-formula> norm between the prediction and the target. Even when predicting on grids denser than previously encountered, our model demonstrates encouraging capacity to reproduce the correct solution profile. The proposed model, which outperforms well-known neural network models, can be included in a CFD solver to help with solving the Poisson equation. Analytical test cases indicate that our CNN architecture is capable of predicting the correct solution of a Poisson problem with mean percentage errors below 10%, an improvement by comparison to the first step of conventional iterative methods. Predictions from our model, used as the initial guess to iterative algorithms like Multigrid, can reduce the root mean square error af
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.