PhD Studentship: Assessment of ship powering performance using machine learning techniques

PhD Studentship: Assessment of ship powering performance using machine learning techniques

Fluid Structure Interactions Group

Location:  Highfield Campus
Closing Date:   Wednesday 22 March 2017
Reference:  711016AT

Project Reference:  CMEES-FSI-131

Project Themes:  Fluid Dynamics

Shipping accounts for approximately 90% of intercontinental trade. Over 90,000 ships ply the oceans, crewed by about 1.2M people. Shipping accounts for around 3% of global CO2 emissions, equivalent to an industrial economy the size of Germany. Improving shipping safety and reducing its emissions are key goals for the international maritime community.

Measuring changes in ship fuel consumption caused by specific energy-saving technologies is difficult, particularly for data acquired in a sea-state and where the sea-state itself is not separately recorded. It is also challenging to predict the behaviour of similar vessels in a fleet, where data is not measured. This limits the ability of shipping companies to monitor emissions, inhibits introduction of new technologies and reduces scope for optimisation of fleet operations.

This project will develop and apply soft computing and machine-learning techniques to develop robust and reliable models of ship powering performance. Such models must be accurate enough to detect small changes in performance arising from use of, say, energy-saving devices and to have predictive capability for similar vessels from one (or a few) instrumented vessels across a fleet.

It is intended that a machine learning approach be applied to model ship performance both over time and in different operational conditions. Non-linearities in vessel response, changes in ship condition and noise in the data are likely to make such approaches difficult if great accuracy is required (i.e. for assessing energy-saving devices). It may therefore be necessary to adopt a mix of statistical analysis, naval architecture understanding and modelling and machine learning techniques – in other words a ‘hybrid’ model. This project is thus aimed at a machine learning approach supplemented by knowledge of the physics of the problem where possible. This builds on experience with the University’s Performance Sports Engineering Laboratory analysis of swimming, hockey and cricket performance using these techniques. The challenge in the ship application is the variation in operational conditions and the level of accuracy required where absolute performance measures are necessary.

The successful PhD candidate will be expected to work closely with Shell Shipping and Maritime, Technology, who will provide data for analysis, together with colleagues across the University through the Southampton Marine and Maritime Institute (SMMI) with expertise in related fields. You will have a good first degree in a related discipline and be enthusiastic about conducting high quality academic research of direct value to the shipping industry.

If you wish to discuss any details of the project informally, please contact Professor Dominic Hudson, fluid-structure interactions group, Email: dominic@soton.ac.uk, Tel: +44 (0) 23 8059 2306.

To apply please use the following link http://www.southampton.ac.uk/engineering/postgraduate/research_degrees/apply.page? and select Faculty of Engineering and the Environment.

Further details:

• Job Description and Person Specification