Heriot-Watt University
ROS: The Research Output Service. Heriot-Watt University EdinburghNot a member yet
4689 research outputs found
Sort by
Reliable image reconstruction techniques : enforcing measurement consistency in deep learning methods
Get PDFSolving linear inverse problems (LIPs) is an integral process within many scientific
fields. These problems typically involve attempting to reconstruct a model or signal
from a set of observations or effects. In practise, LIPs can take many different forms.
These encompass anything from a magnetic resonance imaging (MRI) scan, where
a detailed internal body image is constructed from radio and magnetic waves, to
calculating Earth’s density from measurements of the gravitational field.
Traditional methods to solve LIPs typically involve optimization algorithms, which
have been rigorously studied for decades. However, modern deep learning (DL)
networks have now almost completely taken over, producing results which are far
superior to previous algorithms. Despite this, deep networks have one critical drawback : they cannot ensure measurement consistency in their outputs. This means
that information present in observed measurements is lost after being fed into a deep
network. The aforementioned optimization algorithms do not suffer from this issue,
which is why they continue to see limited use in applications where measurement
consistency is essential (e.g., reconstruction of medical images).
The central theme of this thesis is to design algorithms that exploit the advantages
of both optimization and DL networks for image reconstruction tasks. First, we
proposed a framework that post-processes the output from a deep network via an
optimization problem. As the proposed method includes an optimization algorithm,
it ensures measurement consistency. Simultaneously, it exploits the exceptional quality of DL networks by encoding their output in the optimization problem. We also
developed an algorithm for solving the proposed optimization problem and showed
how measurement consistency is closely connected with generalization errors. Our
framework was applied to three applications: natural image resolution enhancement,
fusion-based hyperspectral image resolution improvement and MRI reconstruction.
Experiments show that our algorithm achieves the new state-of-the-art in all of the
applications considered, both in image quality and measurement consistency. Our
hybrid framework could be the key to the introduction of the superior performance
of DL networks in applications where measurement consistency is paramount
Spectral properties of monopoles and gravitational instantons
Get PDFMotivated by spectral problems arising in gauge theory and gravity, we develop a rigorous proof of infinitely many bound states for a family of radial Laplacians that have
Calogero characteristics at the origin and Coulombic ones at infinity. We first study
the spectrum of the operator obtained by linearising the Yang-Mills-Higgs equations
around a charge one monopole. We then study two Laplace operators on four-dimensional Riemannian manifolds, namely the Laplace operator on the Atiyah-Hitchin moduli space of centred charge two monopoles and the Laplace operator associated with
the Taub-bolt family. We apply our theorem to each operator proving they have infinite
discrete spectrums and numerically compute the eigenvalues. We compare results to
appropriate analytic approximations for each case.Engineering and Physical Sciences Research Council (EPSRC) grant EP/L504774/1
Modeling peer-to-peer negotiations between energy prosumers in developing regions
Get PDFThe past decade has seen a rapid uptake of distributed, variable renewable energy (VRE)
solutions by many end-users to meet their daily electricity requirements. Yet, many
household consumers are unable to afford the high cost of these generating systems, like
solar PV systems and batteries. The problem is particularly acute in developing regions,
such as Sub-Saharan Africa (SSA), where approximately 592 million (more than half of
the total population) live without access to electricity. To address this challenge, this
thesis studies the use of automated negotiations as a local peer-to-peer (P2P) electricity
trading mechanism to improve the accessibility and affordability of electricity in weak
grid and off-grid regions.
The research stems from the need to utilise available local renewable energy sources
(RES) and innovative market structures to improve electricity access in weak and off-grid
settings. It thus studies how the surplus electricity generation from off-grid solar PV
systems (as installed in Nigeria, the most populous electricity-deficit country in the
world) can be utilised to improve and increase access to electricity. To achieve this, we
propose the design of a novel automated negotiation framework for use in such local P2P
electricity markets. This includes the design of an ‘Alternating Offers’ automated
negotiation framework for a fully decentralised P2P electricity market, as well as the
design of an automated negotiation framework for a centralised community-based P2P
electricity market utilising Nash bargaining solution and the Egalitarian social welfare
function. It also includes a study of the appropriate multi-agent systems models (i.e.,
pieces of software that can automatically negotiate the buying and selling of electricity in
micro-transactions on behalf of their human owners).
With a simulated daily minimum surplus PV generation of 2 MWh/day to a maximum
surplus of 31 MWh/day for the Nigerian cities of Abuja and Kano, 20 MWh/day for
Lagos, and 15 MWh/day for Port Harcourt, this research shows how this surplus
unutilised PV generation can potentially power 10,000 – 155,000 Tier-2 households or
2,000 – 31,00 Tier-3 households, daily. This is an estimated annual 3,500 – 30,000 Mt of
CO2 emission savings if utilised instead of fossil-fuel generation. Our proposed
frameworks also show that with an estimated annual revenue of approximately 2,550 depending on the size of the installed system, the prosumer peers would be able
to maximise the utility from their installed systems and reduce their payback period.
Likewise, the consumer peers would have access to more sustainable, reliable, and
affordable electricity
Statistical natural language processing and sentiment analysis with time-series : embeddings, modelling and applications
Get PDFThis thesis addresses the problem of modelling and understanding the fundamental statistical structure that is present in texts through techniques from natural
language processing (NLP) and statistics. This is achieved by proposing a novel
framework for feature extraction from text data, studying their properties, constructing models for them, and applying these models in real-world applications,
to illustrate how such approaches become relevant and are important, in the era
of Big Data-driven natural language processing.
In the first part of the thesis, the challenge of embedding raw text in a stochastic formulation is addressed, such that the resulting time-series preserve the key
features of natural language: the time-dependent, sequential nature of information, semantics, and the laws of grammar and syntax. As part of this process,
we present mathematically and in detail what noise means in text-based data
sources, how to de-noise, and encode text data for modelling purposes. The
latter is achieved by proposing a novel infrastructure of N-ary relations to construct stochastic text embeddings, based on which a sequence of statistical process
summaries are constructed that study the statistical properties of the created embedding, including long memory and its multifractal extension, stationarity, as
well as behaviour at the extremes.
In the second part of the thesis, we present different ways to construct interpretable sentiment indices from text. We propose lexicon-based models, based on
an entropy measure over varying sentiment supports. We then combine different
sentiment types (positive, negative, neutral) with a number of aggregation rules
and show how to interpret them.
In the third part, we address the task of building models from the constructed
text time-series. The class of models we consider comprises time-series Mixed-Data Sampling (MIDAS) regression models. We develop a novel Autoregressive
Distributed Lag (ARDL) MIDAS model for multimodal covariates sampled at
varying time resolutions. Finally, we show how to incorporate a deep neural
network in the ARDL-MIDAS framework to develop the instrumental variables
necessary to resolve estimation requirements for the ARDL-MIDAS model.
The final part of the thesis demonstrates the application of our theoretical
advancements in multimodal sentiment modelling settings, where one modality
is text-based sentiment. The first application focuses on statistical causal relationships between investor sentiment in cryptocurrency markets, whilst the second
focuses on mixed data modelling and deep neural networks (Transformers) for
forecasting end-of-day investors’ sentiment from intra-daily price and technology indicators. Finally, the third application studies the problem of model risk
in the epidemiology domain, for models predicting the total number of COVID-19 infected cases, which are additionally enhanced with public news sentiment
information via an exposure adjustment
Ultrafast laser welding for permanently mounting optical components
Get PDFThis thesis addresses the application of ultrashort pulse laser welding to the
fabrication of optical components which could be used for the assembly of an optical
system with particular attention to its implications for assembly of laser systems. This
thesis builds upon the welding process demonstrated at Heriot Watt University by
investigating the implementation of this previous work as a means of bonding various
optical components and its effect on optical quality. This work contributes to the shifting
of laser manufacture from a skilled assembly process to a low-cost automated process
by enabling the rapid bonding of optics to (thermal) support structures. To this end, it is
imperative to understand the strengths and weaknesses of ultrafast laser welding and
other bonding techniques currently employed in manufacturing such systems.
Currently, optics are either clamped in bulky assemblies or adhesively bonded to
support structures. Adhesives are broadly used because of their widespread utility for
bonding a plethora of material combinations. However, over time, outgassing from the
adhesive can deposit on optical surfaces, damaging them. In consequence of the
drawbacks of adhesive bonding there has been a concentrated effort to investigate
alternative bonding techniques to address these issues, including ultrashort pulse laser
welding.
To make a substantive comment on the efficacy of ultrashort pulse laser welding a
comparative study was performed for ultrashort pulse laser welded, hydroxide catalysis
bonded (an emerging interlayer free bonding technique in competition to laser welding)
and adhesive bonded 10 mm cubes and 15 mm × 15 mm × 5 mm aluminium coupons.
These were investigated in terms of stress induced birefringence generated by the
bonds in the bulk glass/crystal and their shear strength. This analysis demonstrated a
minimal impact on optical quality resultant of the stress induced birefringence from the
weld structures comparable to hydroxide catalysis bonded samples. Furthermore,
demonstration BK7 optics were successfully welded to aluminium supports with stress
induced birefringence comparable to their pre-welding state. The ultrashort pulse laser welding technique was extended to the welding of 5 mm
diameter by 2 mm thick YAG disks to 10 mm diameter by 2 mm thick aluminium silicon
alloy discs (AlSi); commonly used as heat sinks. Undoped YAG was used in place of
Nd:YAG due to cost considerations and with the anticipation that the lack of doping
would not affect the weld. The effect of the focal plane and average power of the beam
was investigated and optimised in terms of shear strength and yield which were
demonstrated to be comparable to the current state of the art process. In addition, the
impact of an increased applied clamping pressure on the resulting welds was examined.
Furthermore, the absolute thermal resistance of these welds was investigated with a
“worst-case” welded sample demonstrated to be comparable to an idealised theoretical
bond using a thermally conductive adhesive commonly used in industry. Thus, combined
with the successful welding of demonstrator optics with low stress induced
birefringence it can be concluded that the assembly of a laser system using ultrashort
pulse laser welding is both feasible and attractive
Properties of hollow-core anti-resonant optical fibres for flexible delivery of high peak power laser light
Get PDFFibre optics play a significant role in the flexible delivery of laser beams in an enormous
number of applications. Due to the limitation of intrinsic material properties, conventional
solid-core fibres are unable to guide ultrafast laser pulses with high peak power. A novel
fibre optic waveguide, the hollow-core anti-resonant fibre (HC-ARF), has attracted great
deal of attention from science and engineering in the last decade. Apart from the
confinement of light in an air core, this fibre with a simple structure exhibits a wider
transmission band and higher damage threshold than any other hollow-core fibres. By
optimising the geometric structure of the fibre, low-loss transmission in a wide range of
wavelength regions from ultraviolet to mid-infrared can be achieved. All these above
advantages mean that HC-ARF is a promising candidate for delivery of high peak power
laser light.
In order to explore the potential of different types of HC-ARFs in high-power laser
applications, essential guiding properties are experimentally measured and analysed in
this thesis. Experiments show that the light guided by HC-ARF is able to exhibit good
output beam quality (M2<1.3) and can deliver ultrafast laser pulses with fine
micromachining results similar to free-space delivery. The minimum lossless bending
diameter of 40 mm can provide sufficient flexibility for many practical applications.
Polarisation-maintaining properties are not observed in HC-ARFs without birefringence
design. Nonlinear effects, which are responsible for the power loss during the propagation
of strong pulses in HC-ARF, can be suppressed by gas filling or evacuating the core,
significantly increasing the maximum transmittable peak power. The result of this work
can also provide feedback and data support for further optimised designs of HC-ARFs
Numerical simulation and experimental investigation of reactive flow in a carbonate reservoir
Get PDFThis thesis presents experimental and numerical investigations of reactive flow
in carbonate reservoirs during CO2 injection for storage. The key aims are to
understand how CO2-brine-rock chemical interactions could cause variations in rock
properties in carbonate formations and identify the key factors controlling these
variations at the core- and inter-well scales.
A series of core flood experiments were conducted using outcrop samples and
representative reservoir samples from a CO2 storage candidate, a depleted carbonate
gas reservoir. The samples were characterized pre- and post-injection for porosity and
permeability changes, and effluent concentrations temporal evolution were monitored.
Reactive transport simulations of the core flood experiments were conducted, and the
simulation results were compared and history-matched against the laboratory data. The
experimentally validated and calibrated parameters obtained from the core-scale
simulations were then used in an inter-well-scale high-resolution heterogeneous
outcrop model to analyze how the introduction of additional geological heterogeneity
changes the evolution of rock properties compared to the core-scale simulations and
experimentation.
This research has shown that combining the experiments and numerical
simulations allows enhanced understanding of chemical interactions during CO2
injection in carbonate reservoir, including the controlling factors. Factors like reservoir
heterogeneities, injection fluid temperature, and type of reservoir fluid at the injection
interval, to a certain extent, influence the mineral reactivity. Information gathered from
the core-scale experiments, and the core-scale models to develop a larger-scale model
is an appropriate approach to predict mineral reactivity and its subsequent impact on
reservoir properties for CO2 storage projects more accurately. A numerical model with
improved accuracy enables more reliable field-scale reactive transport simulations of
CO2 geological storage. Hence, it will guide the optimization of the
injection/production operations and improve the design of CO2 storage processes in a
real target formation
Multi-population and factor-based mortality analytics
Get PDFIn this thesis, I develop a model that uses socio-economic characteristics to explain
differences in the mortality of different populations. This thesis has two main objectives. Firstly we look at mortality data of population from three countries/regions
and test a wide range of stochastic multi-population mortality models with different
sub-population datasets grouped using criteria relevant to socio-economic factors. The
most appropriate model is selected and we can learn the mortality difference over the
sub-populations that is explained by socio-economic differences. Secondly, we take
advantage of a England demographic dataset of large volume and high granularity –
a large number of small geographical areas (i.e. small neighbourhoods) along with
mortality experiences and most relevant socio-economic factors in each of them. With
this granular dataset we implement non-parametric and machine learning models. We
eventually produce a mortality index for the small neighbourhoods in England using
the estimated mortality rates
An investigation of textile sensors and their application in wearable electronics
Get PDFUsing a garment as a wearable sensing device has become a reality. New methods and
techniques in the field of wearable sensors are being developed and can now be
incorporated into the wearer’s everyday attire. This research focuses on two types of
textile based sensors – a wearable textile electrode used for ECG continuous monitoring,
and a stitch sensor for monitoring body movement. These sensors were designed into a
purposely engineered Smart Sports Bra (SSB) which can be regarded as a sensor itself.
After a thorough investigation, two optimum textile electrodes were created; a plain
electrode using cut and sew method (CSM) and a net type knitted electrode using knitting
method (KM). The CSM electrode was made with conductive fabric (MedTexTM P-130)
and the KM electrode was made with conductive thread (silver-plated nylon 234/34 four-ply), these materials having the lowest tested contact impedance; 450Ω and 500Ω,
respectively. Both electrodes demonstrated a level of noise and baseline drift comparable
with standard commercial wet-gel electrodes, which was corrected by optimising their
size to 20x40 mm, holding pressure of 4 kPa (30 mmHg) and the electrode position at the
6th intercostal space on the right and left mid-clavicular, with one placed at the scapular
line in the rear side (i.e. back horizontal formation) which gives clear and reliable ECG
signal. These optimum electrodes were integrated directly into SSBs, in which a novel
high shear, net structure, acting as a shock absorber to body movement that shows more
stable electrode to skin contact by reducing the body motion artefact.
During the investigation of the stitch stretch sensor the single jersey nylon fabric (4.44
tex two-ply) with 25% spandex (7.78 tex) had the highest elastic recovery (93%). Using
this fabric, the work went on to show that the stitch type 304 (Zig-zag lock stitch) using
the 117/17 two-ply thread demonstrated the best results i.e., maximum working range
50%, gauge factor 1.61, hysteresis 6.25% ΔR, linearity (R2
) is 0.98, and good
repeatability (drift in R2
is -0.00). The stitch stretch sensor was also incorporated into a
sports bra SSB and positioned across the chest for respiration monitoring.
This thesis contributes to a growing body of research in wearable E -textile solutions to
support health and well-being, with fully functional sensors and easy-to-use design, for
continues health monitoring
Bioinspired approaches for coordination and behaviour adaptation of aerial robot swarms
Get PDFBehavioural adaptation is a pervasive component in a myriad of animal societies.
A well-known strategy, known as Levy Walk, has been commonly linked to such
adaptation in foraging animals, where the motion of individuals couples periods of
localized search and long straight forward motions. Despite the vast number of
studies on Levy Walks in computational ecology, it was only in the past decade
that the first studies applied this concept to robotics tasks. Therefore, this Thesis
draws inspiration from the Levy Walk behaviour, and its recent applications to
robotics, to design biologically inspired models for two swarm robotics tasks, aiming
at increasing the performance with respect to the state of the art.
The first task is cooperative surveillance, where the aim is to deploy a swarm so
that at any point in time regions of the domain are observed by multiple robots simultaneously. One of the contributions of this Thesis, is the Levy Swarm Algorithm
that augments the concept of Levy Walk to include the Reynolds’ flocking rules and
achieve both exploration and coordination in a swarm of unmanned aerial vehicles.
The second task is adaptive foraging in environments of clustered rewards. In
such environments behavioural adaptation is of paramount importance to modulate
the transition between exploitation and exploration. Nature enables these adaptive
changes by coupling the behaviour to the fluctuation of hormones that are mostly
regulated by the endocrine system. This Thesis draws further inspiration from Nature and proposes a second model, the Endocrine Levy Walk, that employs an Artificial Endocrine System as a modulating mechanism of Levy Walk behaviour. The
Endocrine Levy Walk is compared with the Yuragi model (Nurzaman et al., 2010),
in both simulated and physical experiments where it shows its increased performance in terms of search efficiency, energy efficiency and number of rewards found.
The Endocrine Levy Walk is then augmented to consider social interactions between
members of the swarm by mimicking the behaviour of fireflies, where individuals attract others when finding suitable environmental conditions. This extended model,
the Endocrine Levy Firefly, is compared to the Levy+ model (Sutantyo et al., 2013)
and the Adaptive Collective Levy Walk Nauta et al. (2020). This comparison is also
made both in simulated and physical experiments and assessed in terms of search
efficiency, number of rewards found and cluster search efficiency, strengthening the
argument in favour of the Endocrine Levy Firefly as a promising approach to tackle
collaborative foragin
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!