Showcase Conference 2019

Fire occurrence is one of the most devastating events in residential buildings, among other civil engineered structures. The importance of providing mathematical tools that support fire risk assessments is imperative to improve fire containment measurements as well as accident prevention. We propose a novel probabilistic method based on credal networks to assess the impact on the expected risk of the variables involved in the cause and prevention of fire events. We present the Grenfell Tower catastrophe as a case study to demonstrate the capabilities of our method in the hope to provide decision makers with a probabilistic tool to reduce the probability of such unfortunate events.

Uveal Melanoma is the most common intraocular cancer in adults. Around 50% of patients diagnosed with Uveal Melanoma go on to develop metastasis, which is often fatal. There have been numerous prognostic indicators found in recent years to help determine whether patients will go on to develop Metastasis. Vascular Lakes are a new area of interest for Uveal Melanoma patients, it is currently unknown whether they provide nutrients to tumours or a pathway for tumours to escape.

Being able to predict the limits of a production line can be beneficial in conditions with increased environmental and political uncertainty. Currently, most production flow models consider uncertainty either for the delivery of resources or the internal structure of the production flow (failure of production equipment). The aim is to present a tool that can combine sources of uncertainty and simulate the production flow under a demand function to help determine the theoretical limits of the system.

Antimicrobial peptides (AMPs) are naturally occurring macromolecules that demonstrate a potent antimicrobial activity against a broad range of microbes, including viruses, bacteria, and fungi. AMPs are part of every organism’s innate immune response and act as a first line of defence against infection. This study investigates two potent and FDA approved AMPs (nisin and lactoferrin (LF)) encapsulated in gelatin microparticles in a facile one-pot synthesis for the treatment of infections.

Uncertainty analysis is too important to be ignored, it is better to compute with what we know that make untenable assumptions we don’t. Often, when working with uncertainties engineers often treat their code as though it is a black box and use Monte Carlo, however if we know all the calculations we could instead consider the code as a crystal box. A compiler, Puffin, has been developed that is able to read crystal-box codes and insert calls to a library of intrusive uncertainty operations, it can work either automatically or with end user input.

Computational methods in reliability analysis have been challenged with developing accurate methods with the least computational expense. As network topology increases in size, exponential increase in computational time is required. Methods developed include applying the survival signature to estimate reliability based on cutsets and minimal pathsets in the network or developing surrogate models for existing models such as Artificial Neural Networks.

Norovirus is the most common cause of gastrointestinal illness and is known for mild symptoms, but high contagiousness which can result in sudden outbreaks. The Global Knowledge Graph (GKG) is a media aggregator available within the Global Database of Events, Locations and Tone (GDELT). The project evaluates the predictive value of GKG in the context of weekly norovirus laboratory reports in England and Wales.

The presentation will highlight the logical progression followed along our PhD, that took us from cargo loading and ship scheduling optimisation to the prediction of nonlinear dynamical systems described by imperfect and computationally expensive models, and its application to weather forecasting.After demonstrating the economic potential of taking into account more rigorously and more systematically the uncertainty of sea level predictions in shallow seas, we develop our recent works about the interpretation of ensemble predictions for nonlinear systems and imperfect models, and conclude with some possible applications.

VRdose is a tool designed to optimise the radiological protection of nuclear workers via prior radiation transport modelling and work planning. This study aims to determine the level of conservatism in the results via Monte Carlo comparison and assess the applicability of the software for risk assessment of complex nuclear environments.

The installation and wind tunnel testing of a camber-morphing trailing edge system on an aeroelastic wing is presented. Such morphing system, called High Bandwidth Morphing Actuator (HBMA), is capable of achieving actuation frequencies up to 25 Hz with varying amplitudes. The installation of the morphing actuator in the aeroelastic rig is firstly achieved. Then the aeroelastic behaviour of the entire system is assessed and an active controller is designed, by using the Receptance Method, with the aim of increasing the damping of the first bending and torsional modes. The HBMA proved to be capable of introducing the desired control input that resulted in an increase the flutter velocity up to 10%.

Nuclear reactors simulations are of vital importance both for the nuclear regulator and the rest of the nuclear community. A software development known as Smart High Fidelity Coupled Core Simulator is proposed to answer the demands of the nuclear regulator and the rest of the community in the UK as well as operate under reasonable computational power. The Linking of Subchannel Analysis Tools with Advanced Multiscale Core Simulations focuses on the thermal hydraulics of LWRs and their coupling to neutronics and fuel performance in reactor physics within this software development.

The spatio-temporal dynamics of reward processing in competitive environments, such as auctions, have received little attention in neuroscience. In this pilot study, we analysed the electrophysiological responses to loss and gain outcomes in a novel Vickrey Auction paradigm. By manipulating the perceived level of social competition, we examined the effects of social vs monetary reward in the brain, and how these compared to the predictions of Game Theory.

Predicting Mixed Martial Arts (MMA) contests presents several problems not associated with other sports. Due to these challenges, and perhaps the recency with which MMA has become a mainstream sport, there is very little academic literature surrounding the subject. In this paper, we present a new methodology for predicting MMA contests. Our approach utilises data scraped from freely available websites to estimate the skills of athletes in various key aspects of the sport in order to simulate the contest as an actual fight using Markov chains, rather than just a binary outcome. We compare the model’s accuracy to that of the bookmakers using their historical odds, and even show that the model can be used as the basis for a successful betting scheme.

This talk presents an initial investigation into utilising a Sequential Monte Carlo (SMC) sampler for the determination of an object's orbit from optical observations. By first employing the concept of the admissible region to generate initial candidate orbits from angle only data, an SMC sampler is then used to refine the orbit estimates as more observations are processed. We demonstrate success in accurately determining the orbital parameters of a simulated dataset using an SMC sampler with a Gaussian defensive mixture proposal.

Exposure to aggressive environments is one of the most critical problems of reinforced concrete (RC) structures, which can affect both their static and dynamic behaviour. This study presents the linear and non-linear performance of existing corroded RC framed structures through an advanced numerical model. A new approach is presented for the evaluation of the ultimate capacity of RC elements. Such an approach has been compared and validated against a set of the experimental results from the literature. Two different case-studies are presented.

P values are a widely recognised means of communicating the evidence against a proposed hypothesis, and meta-analysis commonly involves aggregating evidence across studies from which a p-value can be extracted. Methods such as inverse variance are routinely employed for this purpose, but this approach relies on an assumption of independence. This research aims to develop an alternative approach without these assumptions. P-values are combined as a conjunction of events, and this conjunction is bounded by the Frechet inequality. This interval maps to an interval defining the bounds of probability against the hypothesis in question. This is demonstrated for both point and interval p-values, allowing uncertainty to be preserved through the meta-analysis.

This research seeks to apply techniques of converting Fault-tree into Credal Network and making the use of the latter to perform the necessary Probabilistic Safety Assessment of a Nuclear Power Plant. As a case study, we will be looking into the Three-Mile Island accident.