Heriot-Watt University
ROS: The Research Output Service. Heriot-Watt University EdinburghNot a member yet
4689 research outputs found
Sort by
Sequential assimilation of crowdsourced social media data into a simplified flood inundation model
Get PDFFlooding is the most common natural hazard worldwide. Severe floods can cause significant
damage and sometimes loss of life. During a flood event, hydraulic models play an important
role in forecasting and identifying potential inundated areas, where emergency responses
should be deployed. Nevertheless, hydraulic models are not able to capture all of the
processes in flood propagation because flood behaviour is highly dynamic and complex.
Thus, there are always uncertainties associated with model simulations. As a result, near-real
time observations are required to incorporate with hydraulic models to improve model
forecasting skills. Crowdsourced (CS) social media data presents an opportunity for
supporting urban flood management as it can provide insightful information collected by
individuals in near real-time.
In this thesis, approachesto maximise the impact of CS social media data (Twitter) to reduce
uncertainty in flood inundation modelling (LISFLOOD-FP) through data assimilation were
investigated. The developed methodologies were tested and evaluated using a real flooding
case study of Phetchaburi city, Thailand. Firstly, two approaches (binary logistic regression
and fuzzy logic) were developed based on Twitter metadata and spatiotemporal analysis to
assess the quality of CS social media data. Both methods produced good results, but the
binary logistic model was preferred as it involved less subjectivity. Next, the generalized
likelihood uncertainty estimation methodology was applied to estimate model uncertainty
and identify behavioural parameter ranges. Particle swarm optimisation was also carried out
to calibrate for an optimum model parameter set. Following this, an ensemble Kalman filter
was applied to assimilate the flood depth information extracted from the CS data into the
LISFLOOD-FP simulations using various updating strategies. The findings show that the
global state update suffers from inconsistency of predicted water levels due to overestimating
the impact of the CS data, whereas a topography based local state update provides
encouraging results as the uncertainty in model forecasts narrows, albeit for a short time
period. To extend the improvement time span, a combination of state and boundary updating
was further investigated to correct both water levels and model inputs, and was found to
produce longer lasting improvements in terms of uncertainty reduction. Overall, the results
indicate the feasibility of applying CS social media data to reduce model uncertainty in flood
forecasting
Properties of chiral magnetic skyrmions
Get PDFWe study the existence, interaction and stability of solitons within a generalised
chiral magnet model, based on the interpretation of the Dzyaloshinskii-Moriya interaction as a static SO(3) or SU(2) gauge field which is useful throughout. We
review and expand on this picture, including the resulting Bogomol’nyi argument
for the critically coupled chiral magnet [1], and discuss the resulting moduli space of
solutions. This gauge picture is used to establish a correspondence between solutions
of the standard 3D model and solutions of the 2D model.
We calculate the interaction potential between two magnetic solitons in the limit
of large separation, where for certain potentials the SO(3) gauge field reduces to a
U(1) gauge field. With a magnetic field applied at a tilt to the normal of the plane,
the resulting interaction has an unusual oscillating structure.
We model the elliptical instability of magnetic solitons where a soliton decays
into a domain wall by determining (i) when the domain wall energy per unit length
becomes negative and (ii) when the decay lengthscale of solutions of the linearised
Euler-Lagrange equations diverges. We compare these two methods to each other
and to previous numerical results, finding that the methods are complementary, each
providing good fits to numerical evidence in different regions of the phase diagram.
We expand on the viewpoint presented in [2] of chiral magnetic solitons as π-
domain walls hosting kinks, and construct an effective model for the morphology of
these solitons
Investigating the accuracy of error models in history matching of reservoir models
Get PDFReservoir model prediction drive decision making, and the accuracy of the decision-making
process is dependent on the ability of history matched models to forecast true reservoir
performance. Prediction of true reservoir performance is dependent on the sampling of large
number of models with good history match quality and good forecast quality during history
matching. This research investigates the degree of correlation between history match quality
and forecast quality in two 3-Dimensional real field reservoir examples using the conventional
standard least squares history matching approach. The real field examples are a simple field
reservoir with a single well and a single history matching objective, and a complex field
reservoir with multiple wells and multiple history matching objectives. The occurrence of
multiple wells, multiple history matching objectives, measurement, and modelling errors
impact on the correlation between history match quality and forecast quality in real fields. Error
modelling is applied to improve the correlation and also the reliability in prediction of reservoir
performance over the conventional standard least squares history matching approach.
Two terms in the history matching objective equation or misfit, account for measurement and
modelling errors and include the mean and standard deviation. In the conventional standard
least squares history matching approach, the mean error is modelled as zero and standard
deviation assumed. In error modelling, mean and standard deviation are learned from model
realisations of reservoir production history obtained by the conventional standard least squares
history matching approach. This thesis compares the predictive performance of three error
modelling techniques with the predictive performance of the conventional standard least
squares history matching approach. These error modelling techniques include the single level
error modelling technique of time varying mean error and constant standard deviation
calculated from the best fitting model; the error modelling technique of time varying mean
error calculated from the best fitting model and variable standard deviation; and the error
modelling technique of the mean and covariance matrix of errors. The accuracy of these error
modelling techniques in improving the reliability of reservoir performance prediction over the
conventional standard least squares history matching approach is investigated in the presence
of multiple wells and multiple history matching objectives, measurement, and modelling errors
Computational respiratory modelling for patient-specific optimisation of inhaled therapeutics
No full textFor the one billion sufferers of respiratory disease, managing their disease with inhalers crucially influences their quality of life. Generic treatment plans could be
improved with aid of models that account for patient-specific features such as breathing and lung morphology. Therefore, we aim to develop and validate an automated
computational framework for patient-specific drug deposition predictions. A novel
image-processing approach is proposed to reconstruct 3D respiratory geometries
from a single 2D X-ray. The 2D-to-3D image processing predicts airway diameter
to 9% median error compared to ground truth segmentations, but produced some
outliers (maximum error 33%). Validation of modelled deposition predicted 5% median error compared to experiments. We also develop a new method for data-driven
stochastic modelling of particles in turbulent flows, which can be extended to general
wall-bounded flows such as airways. The proposed framework is capable of providing patient-specific deposition measurements for varying treatments to determine
which treatment would best satisfy the needs imposed by each patient (such as disease, airway morphology and breathing). Integration of patient-specific modelling
into clinical practice as an additional decision-making tool could optimise patient
treatment plans and lower the socio-economic burden of respiratory diseases
Evaluating the geological case for the natural accumulation and storage of carbon dioxide in the East Irish Sea Basin, UK
No full textThis study aims to improve the understanding of what defines a secure CO2
geological storage site by characterising gas accumulations naturally enriched in
CO2 and nearby fields in the East Irish Sea Basin. To evaluate the geological
elements responsible for natural accumulation, a large well and geophysical
database was compiled from public and proprietary sources. CO2 is present in
low amounts regionally but elevated only within the North Morecambe and Rhyl
gas fields in the Keys Sub-Basin. Palaeogene igneous dykes were mapped
across the northern basin, proximal to the Rhyl Field and accumulations lacking
CO2. Halite-dominated caprock units within the Mercia Mudstone Group are
thickest in the Keys Sub-Basin but variably limited towards the south, where
mudstone-dominated units contain laterally continuous sandstone interbeds and
hydrocarbon shows. Multiple regional pressure trends indicate that the aquifer of
the Keys Sub-Basin is relatively underpressured and isolated from the wider
aquifer by bounding faults. The combination of aquifer isolation, igneous
intrusions, and a thick halite caprock is considered responsible for the local
preservation of natural CO2. As such, the Fylde Halite caprock and Keys Basin
will provide the best containment for storage prospects. However, its isolation
may cause rapid pressurisation from CO2 injection, limiting total capacities.
Therefore, the large South Morecambe Field is ideally located and the most
prospective site
Models for efficient control and fair sharing of assets in energy communities
Get PDFIn recent times, energy communities have gained significant interest. These communities
empower citizen prosumers by leveraging their own renewable energy generation and
storage assets to manage their energy requirements and engage in the broader energy
market. Such communities offer a promising solution for sustainable energy systems,
promoting renewable integration and active user involvement. Within energy communities,
members can engage in energy trading and invest in shared assets like production units,
energy storage, and network infrastructure. However, efficiently controlling these assets in
real-time and equitably distributing energy outputs among diverse members with varying
needs remains a vital challenge. Addressing this concern is of both research and practical
importance. It is essential to consider technical constraints like local low-voltage network
characteristics and power ratings during this process.
To tackle these challenges, this thesis presents a model that examines the techno-economic benefits of community-owned versus individually-owned energy assets, accounting for physical asset degradation and network constraints. Employing cooperative game
theory principles, the thesis proposes a redistribution model for community benefits based
on the marginal contribution of each household. This redistribution mechanism utilizes
the concept of marginal value from coalitional game theory and distributed AI (specifically the multi-agent system). Study results demonstrate that the proposed marginal cost
redistribution mechanism is fairer and more computationally manageable than existing
state-of-the-art methods, thus providing a scalable approach for economic sharing of joint
assets in community energy systems.
However, integration of centrally shared community-owned energy assets may face
limitations due to network/grid constraints. To address this issue, the thesis proposes a
novel framework for a local peer-to-community (P2C) market mechanism as an alternative
solution to investing in community-owned assets. The dynamics of the P2C market
mechanism are studied for three different types of P2C sellers with non-uniform pricing
schemes and tested across various community settings (comprising a mix of prosumers
and consumers) and different rates of renewable energy adoption. All proposed models are
validated and applied to a real case study from a large-scale smart energy demonstration
project in the UK, using a substantial dataset of real renewable generation and demand. This
practical case study provides confidence in the robustness of the experimental comparison
results presented in the thesis.Engineering and Physical Sciences
Council (EPSRC) Doctoral Training Programme (DTP) grant (EP/R513040/1
Development of small molecule EPAC activators
No full textAbstract unavailable. Please refer to PDF. Restricted access until 31.08.2025
Joint inversion of 4D AVO and time-shifts for geomechanical evaluation
Get PDFThe research outlined in this thesis discusses the significance of geomechanical
deformation within and around producing reservoirs worldwide and its implications for
reservoir behaviour during production. It emphasizes the role of geomechanical
monitoring in understanding the changes in subsurface stress and strain caused by pore
pressure decline due to reservoir production. The resulting deformation can lead to
various issues such as compaction, fluid flow performance changes, and extension or
subsidence in the overburden or seafloor.
In this thesis, I begin with highlighting the use of travel timeshifts as valuable indicators
of geomechanical deformation in reservoirs and overburden. These timeshifts, caused by
changes in pore pressure, saturation, and physical displacements in rocks, can provide
insights into reservoir pressure models, areas for infill drilling, and overburden behaviour.
Overburden timeshifts can also indicate casing failures, making monitoring crucial to
mitigate wellbore instability risks. I then discuss the use of time-lapse amplitude variation-with-offset (AVO) analysis and timeshifts to distinguish between changes in
fluid saturation and reservoir pressure, facilitating reservoir characterization.
Geomechanical modelling is emphasized as an important component in the inversion
process for accurate estimation of reservoir pore pressure changes, with considerations
for density, overburden strains, and anisotropy. The study also introduces new analytic
approximations for time-lapse reflectivity, shedding light on the effects of geomechanics
on time-lapse AVO. The results from field-realistic synthetic modelling highlight the
potential of overburden amplitudes to estimate geomechanical contrasts across interfaces,
particularly in cases of localized reservoir depletion or steeply dipping reservoirs. This
information is valuable for mapping drilling hazards in the extending overburden. The
intercept term in time-lapse AVO can directly estimate vertical strain in the reservoir,
complementing time-shift information. The approach is applicable to compacting fields
with large overburden strains, where the AVO signature is defined at a discrete interface
determined by rock properties. I then highlight the importance of using field values
instead of laboratory values for accurate results, given the limitations of using laboratory
values in real field depletion scenarios. Neglecting density, overburden strains, and anisotropy in 4D AVO inversion can lead to significant errors in estimating reservoir pore
pressure changes.
Field-realistic synthetic modelling demonstrates that overburden amplitudes can be used
to estimate geomechanical contrasts, particularly in localized depletion or steeply dipping
reservoirs. Extracting vertical and lateral strains from time-lapse AVO attributes and
calibrating the difference between gradient and intercept may provide additional
attributes to monitor well deformation and assess risk.
A joint inversion scheme is presented, linking time-lapse P-wave AVO to geomechanical
strains. It is applied to both synthetic seismic data and field data from clastic and
carbonate fields in the North Sea. The time-lapse intercept attribute proves useful in
estimating vertical strain in the reservoir, while horizontal strains can be extracted from
the gradient term of time-lapse AVO. The distribution of horizontal strains may serve as
an additional attribute for monitoring overburden and reservoir deformation.
Incorporating time-lapse amplitude changes in 4D seismic response modelling is
highlighted as crucial for robust interpretation
The mechanical properties of wounded skin for a quantitative monitoring of healing
Get PDFChronic wounds are a global socio-economic burden. Mechanical biomarkers show
promise for the early diagnosis of chronicity, but their clinical application is hindered by
the difficulty in obtaining reliable measurements, due to the inherent intra and inter-subject variability, and due to the lack of gold standards or common guidelines in the
mechanical testing techniques.
In this thesis, a review of the current and emerging wound monitoring systems has been
carried out, to select state-of-the-art approaches that would allow a full-field
characterisation of the wounded tissues in a non-invasive way. We hypothesised that the
physical changes that occur in each healing phase (i.e, due to haemostasis, inflammation,
proliferation and remodelling) can be measured at physiologically relevant force levels,
and that these measurements will allow the stratification of wounds into healing or non-healing categories.
To explore this, we first employed digital image correlation (DIC) and uniaxial tensile
testing on different skin models (silicone phantom, pig, and mouse), in order to map the
local tissue strains around different wound types, whilst also obtaining global stress-strain
measures. Then, we employed compressional optical coherence elastography (c-OCE) to
further characterise the elastic properties below the surface of wounded skin. And last,
we studied the histomorphological features of skin, to correlate the observed mechanical
properties with biostructural changes throughout healing, which were quantified with an
alignment coefficient.
Whilst global measurements did not allow the differentiation between health phases, they
were useful to validate the consistency of the implemented systems, and were also used
to feed a computer simulated model, which allowed us to compare the ideal or expected
mechanical behaviour of the samples under study versus the experimental case. On the
other hand, local measurements displayed clear patterns that correlated with the degree
of healing of each wound, and in the case of artificial wounds, pointed at the location of
the defects even when these were superficial. This work provides insights into how strain
evaluations could support decisions such as wearable sensor placement for healing
monitoring or clinical surgical repair
Grassmannian flows : applications to PDEs with local and nonlocal nonlinearities
Get PDFIn this thesis we present a method for linearising certain classes of nonlinear partial
differential equations. Specifically, we linearise PDEs with nonlocal nonlinearities,
whose particular form characterises each of these classes of PDEs. After providing an overview of its background and original construction, we explore two areas
of application of this method. The first one concerns Smoluchowski’s coagulation
equation. We discuss the constant, additive and multiplicative kernel cases, as well
as more general variants, and show how these can be linearised. The second area
of application we focus on relates to integrable systems. There, we extend our approach in a non-commutative manner that accommodates local nonlinearities too,
thus enabling us to linearise (matrix) integrable systems. In fact, we formulate a
unified programme that entails all cases of (matrix) nonlocal integrable PDEs considered, along with their local analogues. In particular, within the context of this
unified scheme, we derive the decompositions for the nonlinear Schrodinger (NLS)
and the Korteweg–de Vries (KdV) equations, as well as those for a coupled cubic
diffusion/anti-diffusion system and the modified KdV (mKdV) equation
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!