Heriot-Watt University
ROS: The Research Output Service. Heriot-Watt University EdinburghNot a member yet
4689 research outputs found
Sort by
Interphase CO2 transfer in the near wellbore region during subsurface injection
Full text linkThis thesis presents a detailed simulation study to build a fundamental understanding on CO2
distribution between phases in the near wellbore region, whether it is during CCS or
hydrocarbon recovery projects supported by CO2 or WAG injection. A new single-well tracer
test is designed to quantify the residually trapped CO2; this involved a single pass of the tracer
saturated water over the residually trapped zone. The residual gas saturations calculated reflect
the input critical gas saturation values, including the effect of hysteresis, to within 10%
accuracy. However, the tracer arrival times are sensitive to the injected volume of CO2
saturated water. The numerical model may be used to design, optimise, and interpret field tests.
Water injection into a reservoir containing oil with dissolved CO2 is modelled to understand
the component transfer of CO2 between brine and oil phases around the injection well. The
velocity at which a “CO2 desaturation” front displaces away from the well is measured for
different subsurface conditions and various residual oil saturations. The results show that the
CO2 desaturation front in this case investigated travels at one tenth of the phase saturation front
velocity. It also reveals that salinity, pressure and temperature variations have minor impact on
the desaturation front, although they do alter the CO2 solubility and thus, mobilities at the fluid
displacement front. The greater the residual oil saturation the slower the CO2 desaturation front
travelled, due to the greater mass of immobilised CO2 to be displaced. This information is
important in managing the risk of CO2 leakage via injection wells after cessation of injection.
Single phase CO2 injection and CO2 water alternating gas (WAG) injection are considered to
understand component transfer between CO2, brine, and oil. During WAG injection, the water
cycle comes into direct contact with residual oil. Hence, dissolved CO2 in the residual oil will
partition into the injected water. If water is injected for long enough, all the CO2 will be
removed from the contacted oil, and a desaturation front will develop. During the CO2 injection
cycle, some of the CO2 will dissolve into the oil and, thus, there is potential for the next cycle
of water to again be re-saturated with CO2 via the residual oil. By the end of each water cycle
all the CO2 was removed from the contacted oil within the first few meters from the well. The
CO2 desaturation front travels at approximately one twentieth of the phase saturation front
velocity during the first three WAG cycles, but it then stopped for later cycles. The presence
of hydrocarbons in the model buffers and delays the pH drop. The pH reduction during the last
WAG cycles was greater near the injection well as most of the lighter hydrocarbon components
were displaced, meaning the CO2 content of this residual phase was higher than for initial
cycles. This would result in a greater risk of calcite dissolution around the injection well
Compact low repetition rate optical parametric oscillators
Full text linkOptical parametric oscillators (OPO) offer a potential route to a cost effective low
repetition rate ultrafast infrared laser sources. The current favoured technology is
optical parametric amplifiers (OPA), which require high powered pump laser sources
to generate sufficient pulse energies, but these are large and expensive systems. For
an OPO to achieve the same level of performance it would require overcoming some
demanding engineering challenges. The cavity length of the OPO typically must
match that of its pump laser, meaning to achieve sub-50 MHz repetition rates the
cavity length of the OPO needs to be >6 m and for 1 MHz the OPO cavity would
need to be 300 m long.
This thesis explores OPO cavity designs that target the construction of a low
repetition rate OPO within a compact footprint.
The primary method investigated utilises an intracavity Herriott cell to store a large
portion of the cavity length within a compact footprint. A Herriott cell is a type of
multipass cell which is made up of two opposing spherical mirrors usually with a hole
machined in one or both mirrors to allow a beam to enter and exit the cell. Once
the beam enters the cell it is reflected multiple times, forming an elliptical pattern
on the mirrors, with the number of reflections being determined by the separation
distance of the mirrors. Incorporating a Herriott cell into a OPO cavity presents
challenges for optimising the stable resonating mode, maintaining a Boyd-Kleinman
focusing ratio near 1, and achieving the cavity length required for synchronous
pumping. This was demonstrated in a synchronously pumped 49.16 MHz Herriott
cell OPO producing femtosecond pulses from 1440 nm to 1530 nm with average
signal powers up to 312.6 mW when pumped with 1.8W from the Yb:fiber pump
laser, and extended to 12.29 MHz in a 12.2-m cavity.
The next method demonstrated was a harmonically pumped idler resonant OPO,
in which the OPO cavity mirrors are coated so that the longer wavelength idler
now resonates in the cavity, and the signal leaves the cavity immediately after the
nonlinear crystal. The cavity is made compact by constraining the cavity length to
be a harmonic of the pump cavity length. This increases the repetition rate of the
resonant idler, but the signal remains at the repetition rate of the pump laser. A
294.96 MHz idler resonant cavity was demonstrated producing femtosecond signal
pulses with average powers up to 88 mW when pumped with 1.7 W of pump power.
This is reduces the cavity to just 1/6th the size of a synchronously pumped OPO.
Finally, fiber feedback OPOs have been demonstrated as a method to achieve a
compact low repetition rate OPO, however the additional dispersion from the fiber
can limit operation to the picosecond range. To address this modelling work was
done investigating cascaded fiber systems where the fibers have complimentary
dispersion coefficients, minimising the pulse broadening. A combination of SMF-28
and UHNA7 was used to show that at key wavelengths of 1550 nm, 1700 nm and
2090 nm a 50 fs pulse propagating thorough 1 m of this cascaded system sees minimal
broadening with the shortest pulse seen for 1700 nm at just 59.7 fs
McKean-Vlasov stochastic partial differential equations
Full text linkIn this thesis we study problems that are at the interface between Interacting Particle
Systems (IPS) and both deterministic and stochastic PDEs.
In the first part we consider non-linear McKean-Vlasov PDEs obtained as limit of
appropriate IPS when the number N of particles goes to infinity. We are primarily
interested in the regime in which the limiting PDE exhibits multiple equilibria. We
consider such PDEs and we perturb them with a trace-class infinite dimensional
noise. We show that under some constraints on both the growth and the decay rate
of the eigenvalues of the noise the resulting equation (an SPDE) is well-posed. Moreover, the related semi-group is irreducible and satisfies the strong Feller property.
As a result of that, the existence of at most one invariant measure is deduced.
In the second part we look for a mean-field particle approximation to the SPDE
studied in the first part. The approach taken consisted in considering a paired system
particle-weight with the weight being time dependent and subject to a common
perturbation.
The above results, which are the contents of this thesis, gave rise to two papers [3]
and [4]
Laser-based fabrication and heat treatment of shape memory alloys
Full text linkLaser-based direct-write techniques are a potential route to manufacture miniature
functional devices such as shape memory alloy (SMA) actuators. SMAs belong to a
class of active materials that can remember their predetermined shape or form when
subjected to heat and/or stress. This return to the original shape can be used as the
basis of a mechanical actuator with a high work output density on the micro-scale.
Nickel-titanium (NiTi) SMAs are used in a wide range of applications due to their
biocompatibility, high corrosion resistance and fatigue strength. Macro-scale manufacturing of NiTi has been extensively studied to understand its crystallography
and thermomechanical response. However, there is only limited work carried out to
study the fabrication of NiTi SMAs at the micro-scale.
Laser-induced forward transfer (LIFT) is a 3D micro-fabrication tool to sequentially
print materials from one substrate to another. LIFT has been used to rapidly prototype many devices including electronic circuits, photonic devices and complex 3D
structures. The ability to print materials drop-by-drop offers the opportunity to
manufacture SMA devices with locally modified properties in an otherwise monolithic component. Generally, SMAs require specific thermal treatments to induce
the shape memory effect, typically carried out in a hot-air furnace or an oven. To
induce similar effects in a micro-scale component or only at the surface, lasers have
been employed.
In this thesis, the use of LIFT to fabricate NiTi-based SMA components on the
scale of 10-100 µm is extensively studied. Different strategies involving nickel, titanium and NiTi donor layers have been employed. Subsequently, laser-based heat
treatment was studied using XRD, and SEM-EDX to fabricate and/or recrystallise
NiTi. The SMA property of the fabricated component was verified using differential
scanning calorimetry and a cantilever bending test
Interacting particle systems and collective navigation
Full text linkAlmost everywhere we look, we see things interacting. Birdsong, our conversations, and the beep of dial-up internet are all signs of individuals interacting
with the wider world around them. When individuals interact, phenomena
occur far beyond that individuals capabilities. This thesis aims to study these
emergent behaviours using abstract mathematical models. By first developing
models based on individuals’ behaviour, we write equations that describe the
average behaviour across the group. These equations pose many challenges
and are difficult to solve by hand. We develop algorithms to approximately
solve one type of these equations, and apply them to get a better understanding of the models. We also develop an abstract model of collective animal
migration in flowing environments (like the ocean). We study different ways
animals could communicate the direction that they are swimming, and show
it has a large impact on group success. By better understanding emergent
behaviour, we hope to be able to combat some of the major challenges facing
the world today
Design and demonstration of a high-end FTIR spectrometer
No full textAbstract and full text unavailable. Restricted access until 15.08.2026. Please refer to PDF
Predictive modelling for medical morbidity risk related to insurance
Full text linkThe thesis focuses on researching the rates of admission to hospitals (or other health
facilities) due to specific illnesses such as cancer and respiratory diseases among a
United States working-insured population and their dependence on several demographic and health insurance-related factors. The primary objective is the predictive modelling of these admission rates by employing both traditional methodologies
as well as deep learning techniques. As part of this, we develop advanced neural
network-based models using a large and complex dataset of admission counts, and
compare their predictive performance to that of conventional models.
The admissions dataset is obtained from the Commercial Claims and Encounters
Database of Merative MarketScan Research Databases, provided by Merative (US).
The dataset contains individual-level information regarding admissions to hospitals
and other health facilities, linked with patient information (such as enrollment details) and service provider details. Along with socio-demographic details, the data
also consists of geographical information detailing the area of residence.
Among the conventional approaches, classical and Bayesian count regression models with underlying Poisson or negative binomial distributional assumptions are
considered for modelling the admission counts data. Neural network embeddings
of these regression models are additionally being considered by employing generic
Feed Forward Neural Network and specifically designed Combined Actuarial Neural
Network approaches. The predictive performance of the network-based models is
further enhanced by adopting several model improvement approaches, such as nagging predictors and bias regularization techniques. Furthermore, a k-fold validation
process is used to compare the predictive performance of the aforementioned models. The results showcase that the network-based models offer improved predictive
performance over traditional regression models in the context of a real-life complex
health dataset. The enhanced predictive capability of network-based models, which
could be attributed to their capacity to capture potential non-linear interactions
within the dataset, often comes at the cost of ease of implementation when compared to conventional models. Additionally, the absence of a statistically robust
approach for model selection is also a potential drawback of network-based models.
The neural network modelling approach is further extended to develop models that
accommodate the excess zero nature of admissions count data. Towards this, zero-inflated neural network models and zero-inflated combined actuarial neural network
models are developed and demonstrate improved predictive performance over the
earlier-mentioned models. Moreover, integrating additional GLM-like structure into
the neural network models (based on the LocalGLMnet method and its extensions)
facilitates the interpretation of outcomes, in a similar way as that obtained from a
regression model.
The real-life utility of the work in this thesis lies in the fact that, in addition to
facilitating accurate rate setting in the insurance sector, the suite of models and
approaches discussed in the thesis can provide precise predictions that have the
potential to aid in developing personalised and policy-level healthcare interventions
Adaptive hybrid-selling in German mechanical engineering B2B-sales : how can computer-mediated and on-site communication be effectively combined to convince decision-makers?
Full text linkThis research supports B2B-salespeople to select, combine and use on-site and digital
communication channels approaching decision-makers to achieve communication goals
- incorporating different sales process steps - in times of massively changing
communication habits.
Salespersons in german B2B-sales face the huge challenge of having handled by far the
largest part of their customer communication via well-known communication channels
such as on-site visits, telephone and E-mail for decades and now having to handle
additional communication channels (e.g. videoconferences, Social Media or customer
specific videos), which are new territory for many salespeople, in a relatively short
period, since on-site customer appointments were not feasible caused by the Corona
pandemic.
However, skills to communicate professionally through these channels and to decide
which channels are the most appropriate in which situation were not sufficiently
developed. Using a qualitative research strategy and applying a multiple embedded case
study approach, experiences of sales-directors, purchasing managers and salespersons
have been analysed. The data analysis has been conducted via a thematic analysis.
Applying this research strategy, communication experiences from the real-world context
in B2B-sales have been elicited to conclude which combination of different
communication channels is valued by interlocutors.
Via a practical model and a checklist, which can be applied in real-world
communications with prospects and customer’s decision makers, the research’s
objective is to provide a pragmatic help for B2B-salespersons in order to increase deal
closing rates via a convincing communication with customer’s decision-makers
Immiscible fingering in porous media under different wetting conditions and its role in polymer flooding
No full textImmiscible viscous fingering occurs when a low viscosity fluid immiscibly displaces a
high viscosity fluid. In the field of geoenergy, this is typically a major problem whether
in gas storage or in oil recovery. When water is injected into the reservoir to aid recovery,
it can finger through a viscous oil, leaving large volumes bypassed and giving early water
breakthrough – neither of which is ideal from an economic or carbon footprint viewpoint.
Three major questions present themselves with regard to viscous fingering in such
systems: how can fingering be modelled correctly?; how can fingering be evaluated in
the laboratory?; and how can it be remedied? These are the 3 main areas of research that
will be addressed in this thesis.
A novel simulation methodology is used to directly model viscous fingers using standard,
commercial numerical simulators. In this work, this approach is validated against
literature experiments at a range of unstable viscosity ratios (μo/μw ~400 to 7,000). It is
then applied to model conventional core flood experiments, conducted as part of this
thesis, where μo/μw = 100. The simulation method is then used to upscale the core flood
results using scaling theory to a series of conceptual and sector models of the Captain
reservoir, which is currently undergoing polymer flooding in the North Sea.
The same numerical method is used to demonstrate how laboratory scale unstable
displacement experiments are sensitive to the suppression of viscous fingering by
capillary dispersion. This is then shown to occur even under extremely weak wetting
conditions. Using scaling theory, it is then shown how fingering “remerges” as the system
size is increased towards the field scale. These observations are then further supported by
carrying out laboratory 2D slab flood experiments under different wetting conditions for
an unstable immiscible displacement with viscosity ratio μo/μw = 100. The systems studied
include a weakly water-wet case which shows an apparently stable front, while the
equivalent weakly oil-wet system is highly fingered. By applying scaling theory, it is
demonstrated that capillary forces must be made negligible at the laboratory scale in order
to maintain the same viscous-capillary force balance which applies at the field scale
system.
Finally, the well-established enhanced oil recovery technique of polymer flooding is re-evaluated in the context of these findings. It is demonstrated both by simulation and
experiment that the principal increased recovery mechanism of the polymer is through viscous crossflow. This mechanism is shown to be responsible for the large and very rapid
response in oil recovery on polymer injection – even in highly viscous systems (>2,000
mPa.s) - as bypassed oil crossflows into established water channels (fingers). This
mechanism is evident in the laboratory when viscous fingers are allowed to form (viscous-dominated) and supports the conjecture that both polymer flooding and water flooding
are best examined without the stabilising effect of capillarity.
In addition, the findings of this thesis cast doubt on the conventional methods of
“measuring” relative permeability in the laboratory for application in adverse viscosity
ratio immiscible displacements in the field
Ultrafast ultraviolet : hollow capillary fibres for time-resolved photoelectron imaging
No full textThis thesis explores the highly differential energy- and angular-resolved time-resolved
photoelectron imaging (TRPEI) measurement investigating excited state molecular
dynamics, in combination with high-level ab initio calculations, and coupled with the
extremely short time resolution (10 fs full-width-half-maximum at 250 nm central
wavelength) offered by a novel hollow capillary fibre (HCF)-based resonant dispersive
wave (RDW) scheme. This method was shown to be extraordinarily powerful for
unravelling ultrafast excited state excess energy redistribution dynamics operating in
various small saturated molecular systems. It is anticipated that the fully wavelength
tunable, ultrashort pulse durations and high infra-red to deep-ultraviolet conversion
efficiency of the HCF-based RDW scheme will soon attract wider uptake in the ultrafast
research community.Heriot-Watt University fundin
Access Repository Dashboard
Do you manage ROS: The Research Output Service. Heriot-Watt University Edinburgh ? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!