14277 research outputs found
Sort by
A data-driven framework for modeling and optimization of industrial hydrocracking units
No full textThis thesis presents a comprehensive framework for modelling and multi-objective
optimization of industrial hydrocracking and hydroprocessing units that e↵ectively
addresses the challenges associated with complex process dynamics and data quality.
The proposed framework combines advanced machine learning techniques, explainability
methods and multi-objective optimization algorithms to improve the operational
performance and efficiency of hydrocracking units.
The framework begins with data preparation based on cycles and modes of operation.
Data preprocessing involves addressing missing data using the iterative imputation
method, followed by detection and removal of anomalies using the 3-! rule
and Isolation Forest algorithm. An extensive comparative analysis was performed
for product yield prediction using a diverse set of machine learning algorithms, including
Decision Trees, Support Vector Regression, Random Forests, Deep Neural
Networks, XGBoost, LightGBM and CatBoost. Through comprehensive performance
assessment, CatBoost emerged as the most robust predictive algorithm, demonstrating
exceptional potential in mitigating overfitting and significantly enhancing predictive
accuracy across datasets.
Subsequently, SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable
Model-agnostic Explanations) analyses were employed to explain model predictions
and identify key influencing input variables to manipulate for optimization.
The multi-objective optimization is then performed using the Non-dominated Sorting
Genetic Algorithm II (NSGA-II), exploring various scenarios to maximize desired
products’ yields or volume swell during cracking.
The framework is implemented and validated through two case studies, achieving
higher prediction accuracy and providing optimization results that significantly
enhance operational performance and profitability. This research highlights the potential
of integrating machine learning and optimization techniques to advance industrial
processes. Future work could explore using reinforcement learning and digital twins
for real-time optimization and control.Includes bibliographical references (pages 150-155
Poseidon's Shield: Secure Marine Data Acquisition System
No full textPublishedThe increasing digitization of the marine industry challenges data acquisition systems capable
of withstanding cybersecurity threats. Poseidon’s Shield is a prototype system designed to
securely collect, transmit, and visualize sensor data from marine vessels in real time. The
prototype addresses security concerns, including data confidentiality, integrity, and availability,
through a combination of symmetric and asymmetric cryptographic techniques layered over
Ethernet and User Diagram Protocol communication. The system is composed of modular
data acquisition, relay, and aggregation nodes that facilitate secure data flow, enhanced with
metadata tracking and provenance visualization. Poseidon’s Shield supports legacy marine
protocols such as NMEA 2000 and Modbus, enabling backward compatibility while providing
a user-friendly interface and real-time security alerts. The prototype developed demonstrates
successful integration of secure communication, data provenance, and interactive visualization
using Google’s Protocol Buffers framework, MongoDB, and modern cryptographic libraries.
While current limitations restrict protection to in-transit data tampering, the system is designed
for extensibility to address broader threats. This open-source solution proves that secure,
interoperable, and practical marine data systems are feasible and can serve as a foundation for
future marine cybersecurity advancements
Reimagining inclusion in clinical trials for neurodiverse populations
No full textWhile much has been done to promote the inclusion of marginalized groups in
health research, little attention has been given to the barriers faced by marginalized
people whose differences are not readily apparent. This thesis critically examines the
inadvertent exclusion of neurodiverse individuals from clinical trials. Drawing on critical
disability theory, crip theory, anti-oppressive practice theory, critical social studies of
medicine, and a neurodiversity paradigm, I argue that clinical trial processes –
recruitment, consent, implementation, and knowledge transfer – are shaped by
neuronormative assumptions that may unintentionally exclude people who are
neurodiverse. To understand the concerns with research practices, I review some of the
historical and contemporary instances of intentional and inadvertent exclusion of
specific socially identifiable groups. Then, informed by interdisciplinary literature on
neurodiversity, including lived experiential accounts of barriers to access in clinical and
other environments, I examine potential barriers to clinical trials research for people who
are neurodiverse. I focus on those whose neurological differences are non-apparent,
specifically those with attention deficit hyperactivity disorder (ADHD), autism spectrum
disorder (ASD), and specific learning disabilities such as dyslexia. I offer practical
recommendations for researchers and research ethics boards (REBs) to adopt more
inclusive and accessible practices in clinical trial design and in ethical review processes.
By confronting systemic ableism and reimagining research practices and processes,
this thesis contributes to broader conversations about justice, equity, and accessibility in health research, and aims to support meaningful participation for neurodiverse
communities
Comprehensive risk analysis framework for CO₂-enhanced oil recovery and geological storage
No full textThis PhD thesis presents a novel research framework to develop a comprehensive risk analysis model for enhanced oil recovery (EOR) and carbon dioxide (CO₂) utilization and storage (CO₂U&S). The primary objective is to create an integrated risk management strategy that optimizes oil recovery, enhances CO₂ storage efficiency, and minimizes environmental and operational risks. The research combines deterministic risk assessment based on spatio-temporal reservoir and aquifer models with advanced data-driven technologies from the industry 4.0 or fourth industrial revolution (4IR), including least squares support vector machines (LSSVM), artificial neural networks (ANN), genetic algorithms (GA), particle swarm optimization (PSO), deep learning networks like long short-term memory (LSTM), and probabilistic methods such as Bayesian networks (BN), and dynamic Bayesian networks (DBN). This approach addresses the complexities of CO₂-EOR operations and CO₂ storage in geological formations such as hydrocarbon reservoirs and aquifers. A key component is the development of a dynamic risk modeling strategy for CO₂ storage in onshore/offshore reservoirs and aquifers, emphasizing short to medium-term periods and integrating hybrid connectionist models with Bayesian probabilistic analysis. The research also incorporates dynamic economic risk assessments and loss function models to understand and mitigate financial impacts. Alongside economic evaluations, the analysis shows that policy like Section 45Q tax credits make carbon utilization and storage projects more financially viable and sustainable. The integration of real-time monitoring tools, such as the early warning index system (EWIS), enhances operational reliability by identifying potential risks proactively and facilitating timely interventions. The framework also emphasizes adaptive pressure management and enhanced stratigraphic trapping mechanisms to improve containment stability and ensure the long-term reliability of CO₂ storage sites. By addressing associated risks and optimizing EOR and CO₂ storage, this research provides a multidisciplinary and comprehensive approach to sustainable energy practices. The outcomes aim to equip policymakers, industry stakeholders, and environmental advocates with a robust decision-making framework that aligns with energy security, economic efficiency, and environmental stewardship goals. In addition, the novel methodologies developed in this thesis contribute to advancing the understanding and practical application of risk assessment and mitigation strategies in the field of CO₂ storage and utilization
On the Minkowski type problem for unbounded convex hypersurfaces
No full textThe main goal of this thesis is to study the Minkowski type problems for C-compatible
sets in a pointed closed convex cone C with nonempty interior.
Since the beginning of 20th century, we have witnessed the rapid development toward
the classical Brunn-Minkowski theory for convex bodies and its extensions. However,
analogous theory for unbounded closed convex sets just started from the pioneer
work of Schneider [85]. Important results include the Brunn-Minkowski inequality for
C-coconvex sets, the existence and uniqueness of solutions to the Minkowski problem
for C-coconvex sets, and the existence of solutions to the logarithmic Minkowksi
problem for C-coconvex sets characterizing the cone-volume measure. The Brunn-
Minkowski theory for C-close sets was extended non-trivially to the Lₚ setting by
Yang, Ye and Zhu in [104] and the dual setting by Li, Ye and Zhu in [62]. Corresponding
Minkowski type problems have been raised and partially solved in [62, 104].
In this thesis, we will continue the research on the Minkowski type problem for
unbounded convex hypersurfaces. In Chapter 3, we show that there is unique Cclose
set solving the Lₚ Minkowski problem for p ∈ [0, 1], based on which, we then
study the continuity of solutions under several different cases. The Lₚ Minkowski
problem will be extended to the Lₚ dual Minkowski problem, where the volume is
generalized to the q-th dual volume. The key ingredient is the (p, q)-th dual curvature
measure derived from the variational formula of the q-th dual volume in terms of
p-co-sum of C-coconvex sets. We solve the related Minkowski problem and provide
the uniqueness and continuity of its solutions. The Lₚ dual Minkowski problem will
be further extended to the Orlicz setting, which involves two non-trivial and nonhomogeneous
functions. In particular, we will derive the crucial variational formula
of the general dual volume of C-compatible sets in terms of the Orlicz-co-sum. Such
a formula naturally gives general dual Orlicz curvature measure and motivates the
related Minkowski problem, which will be solved under certain conditions by the
variational approach and method of Lagrange multipliers. Note that each Minkowski
type problem is related to a second order elliptical partial differential equation. Thus,
our solutions to the related Minkowski problems indeed provide weak solutions to
such Monge-Amp`ere type equations. We believe our results not only give motivations
to the Monge-Amp`ere equations for unbounded convex functions, but also prompts
the solvability of these equations and hence possibly their regularities. As its natural
connections with differential geometry, complex geometry etc, we believe our works
could be helpful in those areas, as well
SNIT: a modified TLS handshake protocol for censorship circumvention
No full textInternet censorship is a global problem. Many countries censor the internet for different
reasons. This threatens internet freedom and access to information. 82.8%
of websites use the Transport Layer Security (TLS) protocol, which significantly enhances
security. However, weaknesses exposed by TLS can still be exploited for internet
censorship. For example, the unencrypted Server Name Indication (SNI) directly
reveals the website’s identity. We propose a modified handshake protocol, SNIT,
for both TLS 1.2 and TLS 1.3, making it difficult to conduct SNI-based censorship.
SNIT has high resistance to active probing. On average, the performance loss is 31.69
ms per TLS connection, and there is no effect on subsequent traffic. Compared to
competitive approaches, SNIT has decent overall security and performance.Includes bibliographical references (pages 74-85
Numerical studies on effects of leading-edge manufacturing defects on marine propeller cavitation performance
No full textDuring the propeller manufacturing process, grinding of propeller surfaces can introduce imperfections and deviations from the desired geometry. These defects
could lead to degradation of propeller performance in terms of efficiency, cavitation,
vibration and noise. However, there is a lack of scientific literature available that
specifically addresses the subject of manufacturing tolerances of propellers.
In this dissertation, numerical simulations were conducted for foils with constant
DTMB modified NACA-66 a = 0.8 sections using the steady Reynolds-Averaged
Navier-Stokes(RANS) solvers in Star-CCM+ to investigate the effects of manufacturing tolerances. 2-D simulations were first performed for the modified NACA-66
(a = 0.8, t/c = 0.0416and f/c = 0.014) foils without and with the leading-edge(LE)
defects in infinite flow. Convergence studies were carried out to examine the effects of
domain size, grid distribution, grid resolution, and turbulence model on the solution.
Using the best-practice settings for2-D simulation, verification studies were carried
out for the cavitation buckets of NACA-66(a = 0.8, t/c = 0.2and f/c = 0.02) foils
without defect. CFD simulations with best-practice settings were then extended to
themodifiedNACA-66(a = 0.8, t/c = 0.0416and f/c = 0.014) foils with three different sizes of LE defects, representing three levels of manufacturing tolerances within
International Standards Organization(ISO) 484 Class S. The results showed that the
LE defects have significant effects on the cavitation performance of 2-D foils in terms
of reduced cavitation inception speed in the typical design range of angle of attack.
To investigate the differences in 2-D and 3-D simulations and further quantify the
effect of LE defect in future validation studies, 3-D simulations were carried out for
the modified NACA-66(a = 0.8, t/c = 0.0416and f/c = 0.014) foils in 1.0m and
0.525m spans with and without LE defects in cavitation tunnel. Effects of RANS
modelling parameters, such as domain size, grid aspect ratio, first-grid spacing, y⁺,
and turbulence model, on the solutions were carefully examined. Using the corresponding recommended settings, the cavitation buckets, the reduction of cavitation
inception speed and the efficiency due to LE defect were predicted. Additionally,
preliminary validation studies were performed on two sections of 0.525 m span with
no and 0.5mm defects.
Furthermore, this dissertation extended 3-DRANS studies on the foils in cavitation tunnel to full-scale propellers, based on the geometry of David Taylor Model
Basin(DTMB) 5168 propeller, with and without LE defects. Effects of simulation
parameters, including domain size, grid size, stretch ratio, first-grid spacing, y⁺, and
turbulence model on the solutions were carefully examined and the best modelling
practices for the full-scale propeller was developed. Since there is no full-scale data
available, convergence studies were performed for the model-scale propeller followed
by validation studies in order to develop the best-modelling practices for model-scale
propellers. The wakefield, open-water and cavitation performance were presented and
compared with the experimental data. The best-practice settings for the model-scale
and full-scale propellers were compared. Using the best-practice modelling settings
for full-scale propellers, simulations were carried out to full-scale propellers without
and with 0.10mm, 0.25mm, and 0.50mm LE defects. The results showed that the
LE defects within Class S tolerances narrow the cavitation buckets. As a consequence,
such LE defects can result in more than 40% reduction in cavitation inception speed,
which is similar to the conclusions drawn from earlier 2-D studies.Includes bibliographical references (pages 232-247
Integrated reservoir simulation and machine learning for enhanced reservoir characterization and performance prediction
No full textThis thesis presents a comprehensive study on reservoir simulation and machine learning techniques for improved understanding and prediction of reservoir behavior. The research focuses on the Sarir C-Main field and utilizes various data sources including seismic cubes, well logs, base maps, check shot data, and production history. The methodology involves the development of static and dynamic models through processes such as data quality control, log interpretation, seismic interpretation, horizon and surface interpretation, fault interpretation, gridding, domain conversion, property and petrophysical modeling. Additionally, well completion, fluid model definition, and rock physics functions are established. History matching and prediction are performed using simulation cases, and machine learning techniques including data gathering, cleaning, dynamic time warping (DTW), long short-term memory (LSTM), and transfer learning are applied. The results obtained through Petrel simulation demonstrate the effectiveness of depletion strategy, history matching, and completion in capturing reservoir behavior. Furthermore, machine learning techniques, specifically DTW and LSTM, exhibit promising results in predicting oil production. The study concluded that machine learning approaches, such as the LSTM model, offer distinct advantages. They require significantly less time and can yield reliable predictions. By leveraging the power of transfer learning, accurate predictions can be achieved efficiently when limited data are available, offering a more streamlined and practical alternative to traditional reservoir simulation methods.Includes bibliographical references (pages 113-115
Identifying diverse infrastructure needs, barriers, and opportunities for enhancing food security: a case study of the island of Newfoundland
No full textHunger and food insecurity are on the rise globally. A study of food infrastructures has the
potential to offer sustainable solutions embedded in communities and networks. This thesis
examined the extent to which an infrastructural lens may provide insight and understanding that
could help inform the development of more sustainable solutions to systemic problems with food
security on the island of Newfoundland.
The purpose of this research was to analyze food (in)security in Newfoundland using an
infrastructure lens, and to identify infrastructural needs, barriers, and opportunities to improve
overall capacity for improving food security.
This research combined data from literature reviews, content analysis and semi-structured
interviews with key informants, and utilized a diverse infrastructures analytical lens which enabled
the researcher to describe the state of current food infrastructures, identify existing infrastructural
barriers to food security, and suggest infrastructural solutions and recommendations. Data was
organized into themes with trends, commonalities, and differences using NVivo. All of the data
points were then re-organized into three overarching categories: needs, barriers, and opportunities.
The data in each category was re-analyzed into codes within each category.
The results were categorized into infrastructural solutions and recommendations including
building processing facilities, enhancing clean energy infrastructure, overcoming ecological
factors using technology and innovation, public education campaigns, government incentives,
institutional leadership, cooperation and sharing, and a poverty reduction strategy.
This thesis concluded that globally recognized challenges to food security, including access
and distribution, are intrinsically tied to infrastructure. Infrastructural analyses and solutions can
enhance capacity for improving food security and offer sustainable solutions.Includes bibliographical references (pages 96-105
In search of a development philosophy for the fledgling African economies
No full textThe socio-economic underdevelopment which has been the bane of the post-colonial African states has often been attributed to a failure of leadership.¹ The big question, though, is: Why have virtually all African states failed in leadership?
While the failure of leadership in Africa can be attributable to many factors, colonialism, in many ramifications, created African states that were designed to fail. To begin with, the clash of cultures between traditional African societies, on the one hand, and the foreign Western ideals, on the other hand, dealt a heavy blow to the Africans. It effectively turned the African into a schizophrenic creature who, as it were, is trapped between ‘the anvil and the hammer,² neither here nor there. He has lost a grip on his traditional values yet can’t get a hold of the foreign ones. This picture of the modern African man shows why he is destined to fail. Moreover, the political structure created by colonialism did not take cognizance of the traditional/cultural peculiarities within the various local communities. The infamous scramble for Africa which started with the Berlin conference of 1884, and the subsequent colonial rule, saw countries created by mapping out land areas, without taking into consideration the varying and often polarized ideologies of the different ethnicities in these new countries. In Nigeria, for instance, there are over 250 ethnic nationalities/cultures/languages all lumped into one country and named so by the colonialists. Most of these constituent groups do not agree on anything. Little wonder that shortly after independence, the country had to break into a bloody civil war that claimed over 3 million lives. This is the story of not just Nigeria, but indeed most of the sub-Saharan African states. These states were plunged into bloody coups and countercoups, genocidal wars, and power tussles in the absence of true nationalism.
In the face of these socio-political realities, whither Africa? Colonialism came with a predominantly capitalist system of economic development, and its concomitant democracy, but many traditional African communities were socialist states that didn’t have a centralized government. This thesis attempts to explore some of the reasons why most of the post-colonial African democracies have not fared well, as well as examine some possible alternatives/ solutions to the problem. This is done with a view that the African intellectual elite, more than any other class, should be saddled with this responsibility.
This thesis, therefore, discusses politics and economics in sub-Saharan Africa. Although a large part of the essay references Africa in general, since there are similar problems of under-development, the peculiarities of each region however require that the scope be narrowed down to Africa south of the Sahara, where the countries share a lot more in common in terms of history and heritage. The thesis is an attempt at forging a way forward in dealing with the crises of governance in sub-Saharan Africa. Furthermore, the thesis takes liberal democracy and doctrinaire socialism, the two major competing political theories in the world today, as reference points in this discourse. The reason for this is not only due to the fact that most African states currently practice some form of democracy, but also because the pre-colonial structures in these states had some elements of either or both of these theories. I, therefore, try to figure out how we can build a stronger, more progressive state structure that could lean on, but is not entirely dependent on either/ or both liberal democracy and doctrinaire socialism. The thesis will be divided into three parts. The first part attempts to establish that there is a nexus between development and economics., and to show that most African states are in fact under-developed. The second part discusses why the post-independent African democracies have not thrived, while the third part explores alternatives/ solutions to the problem.
¹ Cf. Chinua Achebe, The Trouble with Nigeria, 1984.
² In the words of the famous Ghanaian poet, Kofi Awoonor.Includes bibliographical references (pages 78-87