1,721,113 research outputs found
Computational simulation of freely falling water droplets on graphics processing units
Get PDFThis work describes and demonstrates a novel numerical framework suitable for
simulating the behaviour of freely falling liquid droplets. The specific case studied
is designed such that the properties of the system are similar to those of
raindrops falling through air. The study of raindrops is interesting from both
an engineering standpoint and from a standpoint of pure curiosity. As a natural
phenomenon, rainfall is something which is experienced by everybody, yet
its properties are often misunderstood. The primary engineering application is
in improving the ability of radar to determine the characteristics of rainfall for
meteorological purposes.
The significant original contributions to knowledge within this work come from
several areas. The numerical methods used are a unique combination of a high
order incompressible implicit large eddy simulation method, a conservative level
set method, and a pressure projection method. These methods have all been
implemented on a highly parallel GPU architecture, with a resulting performance
increase of approximately ten times when a single GPU was compared to a single
CPU core.
The water droplets were simulated in a regime not previously studied by three
dimensional methods. The results of these simulations confirmed the validity of
the numerical model by reproducing several important experimental results. New
insight was then gained regarding the behaviour of droplet wakes, an area with
little previous research. The results of the test simulations show great promise
for future use of the numerical framework developed. While the simulations todate
have been of air-water interactions, there is little reason the model should
be constrained to such a system. In theory almost any low speed isothermal
interaction of immiscible Newtonain fluids, with length scales of greater than
1mm, could be modeled accurately by these methods
Implicit large-eddy simulation of low-speed separated flows using high-resolution methods
Get PDFMost flows of practical importance are governed by viscous near-wall phenomena
leading to separation and subsequent transition to a turbulent state. This type
of problem currently poses one of the greatest challenges for computational methods
because its characteristics covers a wide range of physical processes that often place
contradictory requirements on the numerics employed.
This thesis seeks to investigate the physics of complex, separated flows pertinent to
aeronautical engineering and to assess the performance of variants of the Implicit
Large-Eddy Simulation approach in predicting this type of problem realistically. For
this purpose, different numerical solution strategies based on high-resolution methods,
distinguished by their order of accuracy, are used in precursor simulations and one
selected approach is applied to a fully three-dimensional wing flow.
In order to isolate the development from laminar to turbulent flow after separation has
occurred, the prototype Taylor-Green Vortex is considered. Here, the behaviour of the
numerical schemes during the linear, non-linear and fully turbulent stages in the flow
evolution is tested for different grid sizes. It is found that the resolution power and
the likelihood of symmetry breaking is increasing with the order of accuracy of the
numerical method. These two properties allow the flow to develop more realistically
on coarse grids if higher order schemes are employed.
In the next step, flow separation from a gently curved surface is included. The fundamental
study of a statistically two-dimensional channel flow with hill-type constrictions
demonstrates the basic applicability of ILES to problems featuring massive separation.
Without specific wall-treatment, high-resolution methods can improve prediction
of the detachment location when compared to classical Large-Eddy Simulations.
Finally, an ILES simulation of three-dimensional flow over a swept wing geometry at
moderate angle of incidence is presented. The results are in excellent agreement with
experiment in the fully separated and turbulent region and they are more accurate than
a classical hybrid RANS/LES approach, using a grid twice the size, over the majority
of the wing. This outcome will probably settle the dispute that has erupted in the past
over the applicability of ILES to complex, wall-bounded flows
Modelling shock-induced instabilities, transition and turbulent mixing using high-order methods
Get PDFHigh-order numerical methods have been considered and implemented in order
to assess their applicability in a range of complex
ows centering on shockinduced
turbulent mixing. Speci cally, Weighted Essentially Non-Oscillatory
(WENO) variable reconstruction schemes of fth and ninth order accuracy
have been investigated within the context of a nite volume Godunov solver.
In addition to this there have been further numerical developments to assess
the HLLC Riemann solver and various quasi-conservative multi-component
models in conjunction with the high-order methods.
Understanding the physics of fundamental
ow instabilities and turbulence is
increasingly necessary to the development of a vast range of engineering applications
with relation to
uid dynamics. It is desirable to develop numerical
methods that possess su cient accuracy to capture the detail of such
ows
while remaining robust and viable in terms of cost.
The WENO schemes have been tested on a number of cases in comparison with
more traditional second-order MUSCL schemes. These include two and three
dimensional, single and multi mode Richtmyer-Meshkov instabilities with differing
initial perturbations, a cube of homogeneous decaying turbulence and
two hypersonic geometry cases were simulated. The results from this research
were consistent. The higher-order methods provided measurably greater resolution
of small scale
uctuations. By conducting grid convergence studies it
was seen that the e ect of the higher-order methods was comparable to the
e ect of increasing the number of grid points. The cost analysis repeatedly
showed that despite the additional cost of using a higher-order method they
were much better value as they could resolve
ow features on a signi cantly
coarser grid.
The high-order methods were not only validated for a range of
ow problems
but shown to o er great value for their additional cost; they could potentially
help advance understanding and development in a wide range of elds much
faster than is currently the case
Hybrid molecular and continuum fluid dynamics models for micro and nanofluidic flows
Get PDFFrom molecules to living organisms and from atoms to planets a variety of physical phe-
nomena operate at different temporal and spatial scales. Understanding the nature of those
phenomena is crucial for advancing new technologies in many disciplines. In micro and
nanofluidics as the operational dimensions are downsized to smaller scales the surface-to-
volume ratio increases and the surface phenomena become dominant. Numerical modelling
is the key for obtaining a better insight into the processes involved. The Achilles heel of
fine grain microscopic numerical simulations is their computational cost. Simulating a
multiscale phenomenon with an accurate microscopic description is extremely demand-
ing computationally. On the contrary, simulations of multiscale phenomena based only on
macroscopic descriptions cannot fully capture the physics of the multiscale systems. In
order to confront this dilemma multiscale frameworks, called hybrid codes, have been de-
veloped to couple the microscopic and macroscopic description of a system and to facilitate
the exchange of information.
The aim of this research project is to establish and implement a robust hybrid molecular-
continuum method for micro- and nano-scale fluid flows. Towards that direction a hybrid
multiscale method named as Point Wise Coupling (PWC) has been developed. PWC aims
to circumvent the limitations of the existing hybrid continuum/atomistic approaches and
deliver a modular and applicable methodology. In the PWC, the whole domain is covered
with the macroscopic solver and the microscale model enters as a local refinement. Ad-
ditionally, numerical techniques based on neural networks are employed to minimise the
cost of the molecular solver and reduce the outcomes’ variability induced by the fluctuating
nature of the atomistic data.
Molecular studies have been performed (i) to obtain a better insight of the interfacial
phenomena in the solid/liquid interfaces, and (ii) to study the parametrisation of the molec-
ular models and mapping of atomistic information to hybrid frameworks. Specifically, the
impact of parameters, such as surface roughness and stiffness, to slip process is studied.
PWC framework has been employed to study a number of fundamental test cases in-
cluding Poiseuille flow of polymeric fluids, isothermal slip Couette flow and slip Couette
flow with heat transfer. Attention is drawn to the boundary condition transfer from the
continuum solver to the atomistic description. In the performed hybrid studies the effects
of the numerical optimisation techniques (linear interpolation, neural networks) to simu-
lations’ accuracy, stability and efficiency are studied. The outcomes of the simulations
suggest that the neural networks scheme enhance the simulation’s efficiency by minimising
the number of atomistic simulations and at the same time act as a smoothing operator for
reducing the oscillations’ strength of the atomistic outputs
Modelling noise from rotating sources in subsonic and supersonic regimes
Get PDFNoise is an environmental concern and due to the increasing interest in helicopters
as an alternative inter-city transportation, research for more environment friendly
helicopters is continuously growing. Building on this demand, this study aims at
finding an efficient and accurate noise prediction tool for rotating sources.
This study therefore investigates the modelling of noise from rotating sources such
as helicopter rotors by addressing noise propagation in both subsonic and transonic/
supersonic regimes. The aim of this research is to explore the field of aeroacoustics
prediction for rotor generated noise and to develop a noise prediction tool for sources
moving in subsonic or supersonic flow regimes. The aeroacoustics predictions presented have been obtained using a hybrid approach. With such an approach the near
field noise generation process is simulated by means of an aerodynamics prediction
tool while the noise propagation to the near field is computed by mean of the Ffowcs
Williams-Hawkings equation in time domain.
For the near-field aerodynamic calculations di erent CFD tools have been exploited.
More precisely, three test cases have been analysed. For the first test case of 2D
aerofoil-vortex interaction, reproducing the experimental campaign of Lee et al., the
near-field is computed via the commercial software Fluent. The unsteady implicit
Euler solver with second order discretisation both in space and in time is exploited.
This uses the ROE FDS scheme for the fluxes calculation. The same solver is used
in the near-field simulations of the third test case, where the analysis of a non-lifting
hovering rotor is carried out in delocalised conditions, reproducing the experiments
of Purcell on the UH1H model rotor. The second test case analysed is based on
the HELISHAPE experimental campaign for the ONERA model rotor in BVI conditions. Two comprehensive codes, from Agusta-Westland and Roma Tre, are used
to simulate the complex aeromechanics of the rotor in low speed descent.
The noise propagation phase has been performed via the new noise prediction tool
developed during this study, named HelicA (for Helic-opter A-coustics). This tool
is based on the Emission Surface formulation and exploits a novel root finder and
Emission Surface construction algorithms. It can use control surfaces which are in
subsonic or transonic/supersonic conditions. Verification and validation processes
have been performed on the noise prediction tool before using this code in the aforementioned test cases. These processes are based on the comparison of the tool’s
predictions with available analytical and numerical results. The verification and
validation cases include sources moving at Mach numbers ranging from MT = 0
to MT > 1. The noise prediction tool is applied to the three aforementioned test
cases and the results are in very good agreement with the measurements even for
the strong shock delocalisation cases
Low-mach number effects and late-time treatment of Richtmyer-Meshkov and Rayleigh-Taylor instabilities
Get PDFThe Richtmyer-Meshkov instability appears when the mixing between two fluids is triggered
by the passage of a shock wave. It occurs in a range of different applications,
such as astrophysics, inertial confinement fusion and supersonic combustion. Due to
the extreme complexity of this phenomenon to be reproduced in a controlled environment,
its study heavily relies on numerical methods. The presence of a shock wave
as a triggering factor requires the use of compressible solvers, but once the shock has
started the mixing process, the flow field freely decays and becomes incompressible.
The dynamics of this instability is still to be fully understood, especially its long-time
behaviour. One of the hypothesis is that the mixing layer achieves a self-similar development
at some point during its evolution. However, the low-Mach flow at late-times
does not always allow to push compressible simulations so far in time and when it is
possible, they become extremely demanding from a computational point of view. In
fact, it is known that standard compressible methods fail when the Mach number of
the numerical field is low and moreover they lose time-marching efficiency.
In this thesis, a new approach to the study of the very late-stage of the instability
through the use of ILES is presented. The technique consists in starting the simulation
by using the compressible model and to initialise the incompressible solver when
the compressibility of the numerical field becomes sufficiently low. This allows to bypass
the issues previously mentioned and study the very late-stage of the instability
at reasonable computational costs. For this purpose, a new incompressible solver that
employs high-resolution methods and which is based on the pressure-projection technique
is developed. A number of different Riemann-solvers and reconstruction schemes
are tested against experiments using the incompressible, impulsive version of RMI as
test case. Two alternative methods are considered for triggering the mixing: velocity
impulse and gravity pulse. Excellent results were obtained by using the former, whereas
discrepancies were noticed when the latter was employed. Comparisons against numerical
simulations in the literature allowed to identify the inviscid nature of the solver as
the cause of these differences. However, this did not affect the capability of the solver
to correctly compute multi-mode cases, in which viscosity is negligible. A preliminary
study on the compressibility of the numerical field in time proved the feasibility of
the numerical transition and a switching criterion based on the Mach number was established.
The approach was therefore tested on a single-mode perturbation case and
compared against compressible simulation. Very good agreement was found in the prediction
of the growth of the instability and the analysis of the divergence of velocity of
the numerical field proved the incompressibility of the solution generated by the hybrid
solver. Finally, the approach was applied to multi-mode test cases. Excellent agreement
with the theory was found. The turbulent kinetic energy presented a modified
subinertial range and the growth exponent was very close to fully compressible predictions
and experiments. Deeper results analysis showed against compressible simulations
showed very good agreement on the flow physics. In fact, the instability settled to a
self-similar regime with the same time-scale predicted by compressible analysis, but
the simulated time reached by the hybrid solver was three times longer. The results
obtained proved the applicability of the approach, opening to new possibilities for the
study of the instability
Eddy viscosity turbulence models for compressible mixing
Get PDFThe K - L and K - ϵ turbulence models are used to simulate the turbulent mixing induced
by the Rayleigh-Taylor and Richtmyer-Meshkov instabilities. The models contain
additional source terms for the turbulence kinetic energy which depend on the type of the
instability. A new criterion based on ratio of the averaged flow and turbulence time scales
is introduced for differentiating between the two types of instabilities. The original formulation
of the turbulence kinetic energy source present in the K - ϵ model is modified
in order to accurately capture the evolution of the Richtmyer-Meshkov instability in both
heavy/light and light/heavy configurations. Additional constraints are imposed to the
models in order to prevent non-physical solutions when strong gradients are present in the
flow.
Three test problems are considered and the performance of the turbulence models is assessed
by comparing their solutions with the results obtained by high resolution Implicit
Large Eddy Simulations (ILES). First, the classical Rayleigh-Taylor and Richtmyer-Meshkov
problems are solved. A new approach for initializing the turbulence models in proposed
for the Rayleigh-Taylor problem. It is found that both turbulence models describe successfully
the self similar growth of the Rayleigh-Taylor and Richtmyer-Meshkov instabilities
and can predict accurately the spatial distribution of the fluid concentrations and of the
turbulent kinetic energy. The last problem involves the mixing induced at two planar interfaces
by multiple shock reflections and refractions. The turbulence models estimate correctly
the evolution of the mixing and of the total kinetic energy in the mixing zones.
The transport equations of the turbulence models are solved numerically and the influence
of the numerical schemes on the results is investigated. It is concluded that the numerical
schemes do not have an important influence on the results in the case of the classical
Rayleigh-Taylor problem (provided that grid convergence has been achieved and the turbulence
models have been initialized using the method proposed here). However, in the
presence of shocks (such as in the case of the Richtmyer-Meshkov instability), the HLLC
Riemann solver should be used together with a reconstruction scheme of third or higher
order of accuracy
High-order methods on mixed-element unstructured meshes for aeronautical applications
Get PDFHigher resolution and reliability are the desiderata for Computational Fluid Dynamics
and main drivers for the development, implementation and validation of highorder
accurate methods. Complex fluid dynamic phenomena such as shock-wave
boundary-layer interactions, turbulent separated flows and fluid problems involving
multiple scales are adequately resolved with high-order schemes. The spatial representation
of the flow field by an unstructured mesh provides flexibility, automation,
fast and effortless grid generation and exceptional load balance on multiple processor
computers. This plethora of advantages is mirrored by the unprecedented popularity
of unstructured-based schemes.
The objective of this PhD project is the implementation of two high-order schemes
for the compressible Navier-Stokes equations in the context of the finite volume “kexact”
framework: the MUSCL-TVD and WENO. The schemes are formulated in
two and three space dimensions for mixed-element unstructured meshes; in addition,
the Spalart-Allmaras turbulence model is implemented into the developed numerical
framework. A wide range of applications are considered spanning from low-speed
flows (M = 0.08) to supersonic conditions (M = 5.0); inviscid and viscous simulations
in a broad spectrum of Reynolds numbers ranging from Re = 500 up to Re = 37×106.
The applications include: the Taylor-Green vortex, the ONERA-M6 wing, flat plate,
the NACA-0012 and the MD 30P-30N aerofoils, and a shock-wave boundary-layer
interaction.
For the examined cases, WENO schemes demonstrate superior accuracy, numerical
dissipation and non-oscillatory behaviour over the MUSCL-TVD. High-order schemes
inherit low numerical dissipation properties while turbulence models induce dissipation,
this disequilibrium has adverse effects on the stability, convergence and accuracy
of the simulation; therefore, turbulence model re-calibration would be required in order
to accommodate high-order discretisation methods
Molecular dynamics simulations of confined liquids in nanochannels with rough walls
Get PDFDuring the past few decades Micro-Electromechanical systems (MEMS) have been increasingly used in various engineering domains ranging from electronics to biological sciences as nowadays they can be massively produced in numerous shapes and with various compositions. Additionally, the development of the manufacturing techniques has allowed MEMS to be easily integrated into devices and expand their applications as sensors and actuators. The future of MEMS seems to be more than promising; however the small scales involved in this type of devices give rise to phenomena that cannot be treated by continuum simulations such as Computational Fluid Dynamics (CFD) or Computational Structural Dynamics (CSD). On the contrary, Molecular Dynamics (MD) Simulations are considered to be an effective approach in investigating the flow behaviour and the rheological properties of liquids in the nanoscale.
The aim of this PhD project is to establish and implement Molecular Dynamics Models for the investigation of nano-scale liquid flows and the fluid properties in nanochannels with rough walls. This thesis uses MD to investigate the effect of nano-scale roughness on the slip length, the fluid viscosity and the Kapitza resistance. Rough nanochannel walls have been modelled with the help of the multivariate Weierstrass - Mandelbrot (W-M) function which has been used in the past to describe fractally rough surfaces being common in nature.
A number of different approaches have been used to extract the aforementioned thermodynamic and flow properties including Equilibrium Molecular Dynamics (EMD) and Non-Equilibrium Molecular Dynamics (NEMD) Simulations. The outcomes of this research suggest that surface roughness can greatly affect the flow behaviour of highly confined liquids as well as their thermodynamic behaviour. Therefore they could potentially be used as a first step for the selection of the surface treatment and finishing techniques of MEMS devices according to the desired fluid behaviour
Molecular modelling of meso- and nanoscale dynamics
Get PDFMolecular modelling of meso- and nanoscale dynamics is concerned with length and time
scales that are in the transition zone from molecular to continuum models. Molecular
simulation methods, in particular molecular dynamics (MD), only allow the simulation of
relatively small nanoscale systems. Continuum methods, such as computational fluid dynamics
(CFD), are applicable at macroscopic scales but cease to be valid for nanoscales.
This thesis is focused on hybrid MD-CFD methods with geometrical decomposition that
seek to bridge the gap between molecular and continuum modelling.
The hybrid solution interface (HSI) establishes the coupling between the molecular and
the continuum domain. In this work, different realisation approaches for the HSI, flux and
state coupling, are discussed and compared. A detailed investigation on MD flux boundary
conditions, the most crucial part of a flux based HSI, is carried out. Different schemes
for the imposition of mass, momentum and energy fluxes through convective and viscous
transport are presented: direct and indirect flux imposition for convective fluxes; the imposition
of momentum fluxes through reflective walls, external forces and the reverse velocity
scheme; and imposition of energy fluxes through external forces and an energy transfer
scheme. Different combinations of these schemes are compared for standard flow situations.
The momentum and energy transfer by an external force creates a relaxation zone at
the MD boundary. The characteristics of this zone is investigated in detail and a theoretical
model for the density profile has been derived. The reverse velocity scheme has been created
as part of this work to avoid the problems encountered when using the external force
for the momentum transfer. It is shown that indirect convective flux imposition in conjunction
with the reverse velocity scheme gives the best results for the standard flow situations.
The scheme is also tested for liquid flow past Carbon nanotubes
- …
