<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:dc="http://purl.org/dc/elements/1.1/" version="2.0">
  <channel>
    <title>BURA Community:</title>
    <link>https://bura.brunel.ac.uk/handle/2438/23</link>
    <description />
    <pubDate>Wed, 04 Nov 2026 12:26:51 GMT</pubDate>
    <dc:date>2026-11-04T12:26:51Z</dc:date>
    <item>
      <title>An intelligent design framework for multi-principal element alloys using optimized deep learning</title>
      <link>https://bura.brunel.ac.uk/handle/2438/34079</link>
      <description>Title: An intelligent design framework for multi-principal element alloys using optimized deep learning
Authors: Fatimi, Syed Hassan
Abstract: Multi-principal element alloys (MPEAs) are a new type of alloys which have shown promising applications but they lack a comprehensive theoretical framework due to insufficient experimental analysis. To circumvent these issues, machine learn-ing has proven to be effective in assisting design of MPEAs by generating a prediction mechanism of properties of alloys.&#xD;
The core object of this research is to develop a prediction framework for MPEAs via dense neural networks which uses atomic composition to predict alloy phase and its physical properties, making this a supervised learning approach. This framework will then be supplemented with hyperparmeter optimization to improve its prediction power and allow us to get a better theoretical understanding of interactions be-tween atoms in an MPEA. For the purposes of optimization, genetic algorithm (GA) is utilized with novel hyperparameters which serve to include data pre-processing within the framework of hyperparameter optimization. Ablation studies were con-ducted which indicate improvement in prediction by including novel hyperparameters, thus justifying our approach. This model is then tested on recently discovered MPEAs and is able to predict their alloy phases with high accuracy, making this a viable approach for experimental design.&#xD;
Building on our prediction model, we further explored hyperparameter optimization by using teachnig-learning-based optimization algorithm (TLBO), which is a population based heuristic algorithm. TLBO has less control parameters than GA making it more feasible, but it suffers from premature convergence. To this affect, we introduce modifications in conventional TLBO to create balance between local and global search. Each approach is tested and compared with conventional TLBO, and the results depict notable improvement. Combinations of the three modifications are compared with each other and the best combination is termed as enhanced TLBO (eTLBO), which when compared with GA illustrates improvement in prediction accuracy.&#xD;
The outcomes of this reseach provide a deeper understanding of complex mechanisms surounding MPEAs while signifying the importance of optimization in supervised learning models. This thesis leverages the findings to discuss implications regarding high-throughput alloy design, and determine feasible machine learning systems to support it.&#xD;
&#xD;
The outcomes of this reseach provide a deeper understanding of complex mech-anisms surounding MPEAs while signifying the importance of optimization in su-pervised learning models. This thesis leverages the findings to discuss implications regarding high-throughput alloy design, and determine feasible machine learning systems to support it.
Description: This thesis was submitted for the award of Doctor of Philosophy and was awarded by Brunel University London</description>
      <pubDate>Thu, 01 Jan 2026 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://bura.brunel.ac.uk/handle/2438/34079</guid>
      <dc:date>2026-01-01T00:00:00Z</dc:date>
    </item>
    <item>
      <title>Technological protection measures and exceptions in the area of education under the copyright act of Thailand</title>
      <link>https://bura.brunel.ac.uk/handle/2438/34078</link>
      <description>Title: Technological protection measures and exceptions in the area of education under the copyright act of Thailand
Authors: Yimfuang, Chuencheewin
Abstract: Thailand has promulgated the revised anti-circumvention provisions in 2022. This amendment strengthens&#xD;
the protection of technological measures by prohibiting both the direct act of circumvention and the trafficking of&#xD;
circumvention devices. However, there are problems concerning the vagueness of legal wording, which might affect&#xD;
the scope of protection, interpretation, and enforcement. There are limited lists of exceptions to anti-circumvention&#xD;
rules that adversely affect educational uses. Moreover, the administrative procedure for determining exceptions has&#xD;
not been well established in Thailand to monitor the impacts of these laws. There is a lack of agency to facilitate user&#xD;
engagement with digital knowledge. These render the exceptions unenforceable, potentially preventing access to&#xD;
digital knowledge for libraries, students, researchers, and other educational uses. Therefore, the purposes of this&#xD;
research are to demonstrate existing challenges and difficulties in the technological protection measures of Thailand.&#xD;
Then, new models of regulations, exceptions, and other appropriate measures will be proposed. This thesis employs&#xD;
doctrinal legal research and comparative methodology to critically compare and evaluate the scope of anticircumvention&#xD;
rules, exceptions, and other appropriate measures in selected jurisdictions. The common effects of&#xD;
access to digital knowledge will be investigated to provide practical solutions for Thailand and other developing&#xD;
countries. It is arguable that anti-circumvention regulations could prevent copyright infringement in the digital&#xD;
environment and enable the works to be available on the most convenient platforms. However, in the absence of&#xD;
limitation, the restrictive laws allow copyright holders to set prices for passwords or any electronic keys that are&#xD;
unaffordable for developing countries, which results in a pay-per-access society and the exclusion of users from online&#xD;
databases.&#xD;
This thesis suggests the model of legal definitions and prohibitions on circumvention to ensure that the laws&#xD;
effectively protect copyright works. It is important to insert the requirement on the intention of circumventors to&#xD;
reduce the severity of laws and identify the scope of prohibition. The legislative revision for anti-trafficking of&#xD;
circumvention devices is also suggested to prohibit commercial distribution of circumvention means and allow&#xD;
necessary devices to remain available for non-infringing uses. The thesis proposes to introduce exceptions for private&#xD;
study; exceptions concerning libraries and education should be modified and expanded to cover either prohibition of&#xD;
circumvention or trafficking devices to enable these communities to properly access knowledge. Moreover, the thesis&#xD;
recommends legislative reform to the ministerial procedure to determine additional exceptions and assess the impact&#xD;
of anti-circumvention rules. The establishment of a complaint process is proposed to implement the exceptions.&#xD;
Finally, the thesis contends that collective management organisations and the non-printed deposit system should be&#xD;
introduced to Thailand to make the knowledge available for educational communities without endangering the&#xD;
interests of copyright holders. These changes would maintain the balance of interests between all parties.
Description: This thesis was submitted for the award of Doctor of Philosophy and was awarded by Brunel University London</description>
      <pubDate>Thu, 01 Jan 2026 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://bura.brunel.ac.uk/handle/2438/34078</guid>
      <dc:date>2026-01-01T00:00:00Z</dc:date>
    </item>
    <item>
      <title>Optical measurements of injection and fuel mixing in a dual-fuel internal combustion engine and a constant volume chamber</title>
      <link>https://bura.brunel.ac.uk/handle/2438/34077</link>
      <description>Title: Optical measurements of injection and fuel mixing in a dual-fuel internal combustion engine and a constant volume chamber
Authors: Lauterkorn, Alexander Michael
Abstract: Due to the need to reduce greenhouse gases and prevent global warming, alternative solutions for internal combustion engines are being developed. Since electric engines cannot fully replace internal combustion engines in certain transportation sectors, this thesis explores dual-fuel combustion as a promising alternative. Two experimental investigations were conducted:&#xD;
Dual-Fuel Combustion Experiments at Brunel University A single-cylinder optical dual-fuel engine was used to study the effects of injection timing and fuel quantity on combustion. High-speed imaging captured flame area, ignition delay, and cycle-to-cycle variations. Results showed that a split pilot fuel injection creates an initial flame front, which efficiently ignites the subsequent pilot and main fuel injections, thereby reducing the required Diesel quantity. However, this approach exhibited significant variability in the initial flame formation, attributed to random flame locations, while variations in swirl speeds affected the later combustion stages.&#xD;
Pilot Spray Characteristics and TWIN Nozzle Geometry at Woodward L’Orange Spray experiments in a Constant Volume Chamber examined how TWIN nozzle geometry influences spray penetration, angle, and area. The results indicated that spray hole diameter and distance are the primary factors affecting spray characteristics, while the relative spray hole angle plays a role mainly under low-pressure, short-duration conditions. By comparing the experimental data to established spray penetration formulas, modifications were recommended to better predict the behavior of TWIN nozzles.&#xD;
The findings contribute to a deeper understanding of dual-fuel combustion and nozzle design, paving the way for future studies on optimizing initial flame front formation with premixed main fuel and minimal pilot fuel, as well as exploring additional geometric parameters of TWIN nozzles.
Description: This thesis was submitted for the award of Doctor of Philosophy and was awarded by Brunel University London</description>
      <pubDate>Thu, 01 Jan 2026 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://bura.brunel.ac.uk/handle/2438/34077</guid>
      <dc:date>2026-01-01T00:00:00Z</dc:date>
    </item>
    <item>
      <title>Dynamic data reduction for foundational AI models</title>
      <link>https://bura.brunel.ac.uk/handle/2438/34071</link>
      <description>Title: Dynamic data reduction for foundational AI models
Authors: Sanderson, Dominic
Abstract: Arti cial Intelligence (AI) fundamentally relies on data to generate useful outputs. The accuracy&#xD;
of an AI system is not only determined by the e ectiveness of the model, but also both&#xD;
the quantity and quality of the data it is trained on. Therefore, a relatively simple way to&#xD;
maximise AI performance is to use more data. However, as we continue to push the boundaries&#xD;
of AI, the demand for larger and more diverse datasets has only intensi ed, highlighting the&#xD;
importance of high quality data acquisition and management in the development of advanced&#xD;
AI technologies. This causes a growing concern about the environmental impact associated with&#xD;
these methods. The escalating energy requirements for training and operating sophisticated AI&#xD;
models contribute to a larger carbon footprint, prompting discussions on sustainable practices&#xD;
within the  eld of arti cial intelligence. Addressing these concerns necessitates a balance between&#xD;
pursuing technological advancements and implementing strategies to mitigate their environmental&#xD;
repercussions. This thesis explores dynamic data reduction methods as a means of exploring&#xD;
data quality and usefulness. For example, by reducing the amount of training data used, less&#xD;
computation is needed, reducing the environmental cost of the algorithm.&#xD;
Random dynamic data reduction is explored  rst by reducing the amount of data used for&#xD;
training at di erent stages of training, for image classi cation tasks. Three dynamic data reduction&#xD;
methods are introduced to observe di erent strategies of data reduction: data step, data&#xD;
increment, and data cut. These strategies present varying amounts of training data application&#xD;
which means that fewer image evaluations are performed to train the model than if the model&#xD;
is trained on the full dataset. Meanwhile, the strategies ensure that the full dataset is used, to&#xD;
ensure maximum data coverage. Results show that the data step method is best performing, and&#xD;
for datasets MNIST and smallNORB, model accuracy may be increased by 0.04% and 0.593%&#xD;
respectively, while 12% and 6% fewer evaluations image are performed.&#xD;
Random data reduction is so far limited by uncertainty of the quality of data that is removed&#xD;
for a portion of the training. This may be alleviated by introducing a method to determine speci c&#xD;
data to remove. By  rst applying dimensionality reduction algorithms, data may be ordered and&#xD;
then removed based on distance from each classes' centroid. This allows for more intelligent data&#xD;
selection. Data step reduction is performed on larger image datasets, namely Imagenette and&#xD;
the Micro-PCB dataset, yielding inferior accuracy for the Imagenette dataset with an accuracy&#xD;
decrease of 0.114%, and superior accuracy for the Micro-PCB dataset by 0.013%.&#xD;
The data step method has shown promising results as a dynamic data reduction strategy. An&#xD;
alternative method using a sliding window of data is explored, which ensures unique data each&#xD;
epoch throughout training, instead of a step-up in data use. This method is combined with the&#xD;
previously used non-random data reduction, on image data of di ering sizes. Results show that&#xD;
model accuracy is worse for MNIST, CIFAR-10, Imagenette and the Micro-PCB datasets by up&#xD;
to 0.16%, but accuracy increases for the smallNORB dataset by 0.654%. However, statistical&#xD;
signi cance against benchmark experimentation informs us that after reducing data, results show&#xD;
no statistically signi cant di erence from the benchmark, even though training data has been&#xD;
reduced. Still, despite the underwhelming performance, insight is given into which data is more&#xD;
useful at the beginning or end stages of training by comparing the direction of the sliding data&#xD;
window. Results show that for unstructured datasets MNIST, CIFAR-10 and Imagenette, the&#xD;
forward data  ow direction yields superior accuracies, while for the highly structured smallNORB&#xD;
and Micro-PCB datasets the opposite is true. This supports the theory that some data is more&#xD;
useful in the  ne-tuning stages of training, depending on how structured the data is.&#xD;
Having explored dynamic data reduction in image data, data reduction in imbalanced, timeseries&#xD;
occupancy datasets is investigated. Occupancy detection from non-invasive sensor readings&#xD;
is useful for building interior management, while maintaining personal privacy. However, realworld&#xD;
data is often imbalanced, often giving bias to training. By using data reduction techniques inspired by those used for image data, results show that data reduction on the larger class only&#xD;
can maintain and even improve AUC-ROC scores 0.483% with as much as 39% of the dataset&#xD;
removed. Dataset fusion is also explored as a means to improve the generalisability of the&#xD;
classi er. Results of experimentation show a decrease in AUC-ROC of as much as 48% with&#xD;
the full dataset, and 33% with the reduced dataset. This shows that by attempting to increase&#xD;
the available data, overall performance is decreased as the model struggles to specialise. Also,&#xD;
data reduction on a fused dataset can minimise the detrimental e ects of too much data. Class&#xD;
density is also calculated to be a metric that measures class balance and data su ciency, which&#xD;
is vital for determining whether a dataset may be reduced to improve classi cation performance&#xD;
prior to any model training.&#xD;
The above investigations show that the exclusion of some data may improve model performance,&#xD;
as well as reduce training cost. The novel data reduction strategies that are introduced&#xD;
also give insight into which data is more important to training at the earlier and later stages.&#xD;
These  ndings prompt us to consider not just the quantity of data, but the quality and relevance&#xD;
of data during both collection, where focus can be aimed at collecting most useful data, and&#xD;
training, where most relevant data may be applied at the most bene cial stage of training.
Description: This thesis was submitted for the award of Doctor of Philosophy and was awarded by Brunel University London</description>
      <pubDate>Thu, 01 Jan 2026 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://bura.brunel.ac.uk/handle/2438/34071</guid>
      <dc:date>2026-01-01T00:00:00Z</dc:date>
    </item>
  </channel>
</rss>

