Current PhD Projects

Technical description

Here are just a few of the project outlines we've already worked on, to give you a taste of what you might expect at STOR-i.

Assessing Risk of Damage to Oilrigs
Cargo Revenue Management
Statistical Texture Analysis – the key to great looking hair? 
Short-term Forecasting of Wind Power 
Improving Patient Flows in Accident and Emergency Departments 
Upstream Logistics in the Oil and Gas Industry

Assessing Risk of Damage to Oilrigs

Offshore structures, such as oilrigs, need to be built to withstand the complex forces they may incur out at sea; from waves, wind and other meteorological factors. We are primarily interested in the extreme values of these factors since it is these that lead to structural damage. Accurately quantifying the risk of structural damage to offshore facilities is important for the design of the structures, and for insuring them.

An ongoing project involves the modelling of the extreme values of these factors. This application involves interesting statistical challenges since there may be a high number of factors to consider jointly, or several useful covariates to consider which affect the joint occurrence of extreme values. Furthermore the extreme values at more than one location in the ocean may be of concern, since one may be interested in the probability of multiple structures being affected by the same extreme event.

This research builds on and applies statistical methods in extreme value theory; developing probabilistic theory for multivariate extremes, which in turn motivates new statistical methods, and then applying these methods. It involves a mixture of mathematical and statistical modelling, computing and applied statistical skills.

Back to top »

Cargo Revenue Management   

Cargo Revenue ManagementOperational Research models have been used for many years now to support the decisions taken by airlines concerning how they should price their tickets in order to maximise revenue. Indeed, the area of revenue (or yield) management is one of OR's great success stories. Robert Crandall, former President, Chairman and CEO of American Airlines said, 'I believe that yield management is the single most important technical development in transportation management since we entered the era of airline deregulation'.

A major passenger airline runs a cargo business accounting for 10% of their turnover utilising the remaining capacity in the holds of its aircraft once all passengers' luggage has been loaded. The airline wishes to develop appropriate revenue management approaches for this business.

Cargo revenue management problems have many features which make them more challenging than those which relate to conventional passenger business. For example, while on any given flight the passenger business has a fixed, known number of seats to sell, the cargo business cannot know with certainty (until just before take-off) how much spare capacity there will be in the hold. A current PhD project is concerned with answering such questions as how this business can be divided between long term contracts and spot sales.

This project involves aspects of revenue management.

Back to top »

Statistical Texture Analysis – the key to great looking hair?

Texture is the visual character of an image region whose structure is, in some sense regular, such as a woven material or hair. With ever improving image capture methods, there is increasingly a need to be able to accurately classify textures into different groups. In addition it is often instructive to understand which particular image properties enable the classification.

Whilst the analysis of textures is a distinct area of research within the image processing community it is also instructive to consider the problem from a purely statistical view point. Textured images are very complex data forms with non-stationary covariance structures, typically characterised by

  • Abrupt changes (edges)
  • Local shading (changes in mean)
  • Structure at different scales (e.g. small detail at the finest scales)

Statistical Texture Analysis – the key to great looking hair?

Ongoing work with a leading consumer goods company is concerned with developing the next generation of tools for texture analysis. This work builds on and applies statistical research in areas such as spatial statistics, multiscale methods and multivariate analysis.

Back to top »

Short-term Forecasting of Wind Power

Short-term Forecasting of 
Wind PowerAs wind energy becomes an important supply of electricity, there is an increasing need to be able to predict wind power, and hence the amount of electricity that wind farms will produce. Whilst numerical weather models may be useful, statistical models can help improve the accuracy of predictions. More importantly, statistical approaches quantify the uncertainty in predictions. For example, they can provide a distribution of possible values of the wind power and the likelihood of each value. These distributions are necessary if correct decisions are to be made from the forecasts.

Ongoing work with a wind energy consultancy firm is concerned with developing novel statistical methods for short-term prediction of wind power. These methods need to be computationally efficient, so that they can be updated every few minutes, and can deal with joint prediction of power at a range of turbines within a farm and across different farms.

This work builds on and applies statistical research in areas such as dynamic modelling, sequential Monte Carlo, changepoint detection and spatial modelling.

Back to top »

Improving Patient Flows in Accident and Emergency Departments

Improving Patient Flows in 
Accident and Emergency DepartmentsThe unpredictable nature of much of the workload of healthcare systems means that congestion is a common feature, and reducing or at least managing the congestion is crucial to the quality of service provided. Accident and emergency (A&E) departments in the UK (and emergency rooms internationally) are a particularly important service in this respect, with around 13 million people attending around 200 'major' A&E departments in England every year and with no barrier to attendance at any time.

A key performance target in the NHS is that 98% of patients should be admitted, discharged or transferred, within four hours, and other health systems have set their own equivalent targets. The challenge for management is considerable when managing systems as complex as these, in which patient flows within the network of health service resources are subject to both stochastic and cyclic factors, and solutions involve scheduling a range of expensive staff and equipment 24 hours a day, 7 days a week.

An ongoing stream of work initiated by the Department of Health is looking at:

  • The development of 'generic' simulation models capable of representing the common problems faced in managing patient flows in A&E departments throughout the UK;
  • The application of those models to inform national policies for A&E, and to assist individual hospital trusts to manage their own departments;
  • The development of complementary analytical time-dependent queuing models which can be used to add insight and direction to the experimentation normally associated with the application of the simulation models.

A recent addition to this stream of work involves the Jamaican health service and seeks to evaluate the applicability of this same approach in A&E departments in Jamaica.

This work builds on and applies modelling research in areas such as time-dependent queues, queueing networks and simulation.

Back to top »

Upstream Logistics in the Oil and Gas Industry

Upstream Logistics in the 
Oil and Gas IndustryOrganising the supply of goods (containers as well as bulk) to offshore oil and gas platforms is far from simple. Aside from the inherent complexity in all systems such as inventory management, transportation planning, and purchasing, this logistics chain faces some specific problems. The main obstacle is the weather and others include random logistics demands caused by unpredictable production processes (in particular drilling) and lack of exchange of information, even within companies.

In an industry where the costs caused by late deliveries can be as high as £10 million per day (lost income due to a closed platform), but where there is also pressure on logistics costs, planning becomes particularly important, and the following questions must be considered;

For inventory: How can very limited inventory space on the platform be utilised? What needs to be put onshore? Where (and how many to store) can particularly expensive parts such as turbines be put?

For routing of supply vessels: How can we take account of weather forecasts? How can vessels be routed when there is a chance of bad weather? How will the trade off between the need to deliver important parts before (which is a random point in time) the storm arrives against the need to have short routes (as capacities are measured as tons per day, not tons!) be balanced?

Operational research is needed to analyse the problems separately, in addition to understanding how these problems interact. Deterministic thinking is not the way to solve these problems.

This project involves stochastic programming combined with mathematical modelling related to logistics, inventory management, vehicle routing and scheduling depending on which part of the problem is analysed.

Back to top »

These are just a few of the very real and practical issues our graduates will be well equipped to tackle, giving them the skills and experience to enable their careers to progress rapidly.