Advanced Research Computing @ Cardiff

Cardiff University is renowned for its research quality, having been rated 7th out of 106 UK universities in the most recent Research Assessment Exercise (RAE).

The University is known worldwide for the quality and impact of its advanced research computing and e-research output. A full list of Cardiff University Schools currently pursuing some advanced research computing is given below, with links to some research examples:

The above list is growing all the time, with new applications frequently being developed which use e-research techniques.

Moreover, all four of the Graduate Schools are represented, namely

• Graduate School of Biomedical and Life Sciences

• Graduate School in Humanities

• Graduate School of Physical Sciences and Engineering

• The Research and Graduate School in the Social Sciences.

With one or more academic Schools in each Graduate School already using e-research techniques, and a total of 15 out of 28 academic schools involved, advanced research computing at Cardiff University is now fundamental to the University's world-class research portfolio. Some examples are given below.

Dentistry

Biomaterials and Biomechanics Research Centre

Principal Investigator: Professor John Middleton

Research Summary: Prof Middleton’s research group undertakes interdisciplinary research in the areas of computational biology and medicine, with particular interest in simulations of the movements of teeth, dental implants, appliances and the surrounding soft tissue. This leads to better understanding of treatment options, patient outcomes and better preventive care for teeth and gums.

Technical summary: The group use applications software, including finite element modelling and engineering applications and techniques, to model and predict the movement of teeth, prosthetics and soft tissue in the mouth - both over years of use and wear, and also as caused by an impact on the mouth and teeth.

A computer simulation of a dental implant, showing the areas of high stress in the surrounding bone. Courtesy of Middleton.

Software used: Proprietary software traditionally used for engineering and impact analysis, including Abaqus and LS-Dyna.

Facilities used: Powerful desktop PCs with one or two processors and large amounts of memory, together with sophisticated display and visualization tools. Prof Middleton’s team plan to increase this dramatically by exploiting the new ARCCA cluster computer when it becomes available in 2008. This will enable more advanced research in this area.

Prof John Middleton and the Biomaterials and Biomechanics Research Centre (BBRC)

Earth, Ocean and Planetary Sciences

Computational Geoscience

Principal Investigator: Dr J. Huw Davies

Research Summary: Dr Davies models Earth’s plate tectonics which result from mantle flow beneath the earth’s crust. This will underpin a better understanding of earthquakes and volcanic eruptions, which generally take place on plate boundaries.

Technical Summary: Dr Davies’s research is computationally very intensive, due to the size and complexity of the problem he is investigating. He simulates the fluid dynamics of the whole mantle using finite element modelling.  This allows him to simulate mantle convection including subduction where one plate moves under another into the planetary mantle.

A simulation of mantle convection. The yellow plumes are practically steady. Courtesy of Davies.

Software used: Originating with colleagues in USA. Has been further developed in-house over the past 10 years at both Liverpool and Cardiff Universities, together with colleagues in Germany and USA.

Facilities used: Large, high-memory cluster computer, with up to 1000 processors. Also using stereo visualisation of 100 million data points in world model, for both interpreting science and demonstration.

Recent Results: Directly tested models of mantle circulation with results of mantle seismic tomography. Investigated layered mantle convection and demonstrated that this favoured mode by the Geochemistry community is very unlikely.

Computational Geoscience Group homepage.

History and Archaeology

Dr Alex Bayliss is a researcher based at English Heritage and is collaborating with Professor Alasdair Whittle and colleagues in Cardiff University's School of History and Archaeology. For further information please see the External Research page.

Manufacturing Engineering Centre

Director of MEC and Head of the Systems Division: Prof. Duc Truong Pham OBE

The Bees Algorithm: A Novel Tool for Optimisation Problems

A new optimisation procedure based on the behaviour of honey bees is delivering sweet results for industry, enabling companies to maximise profits by changing basic elements of their processes. Researchers at the Manufacturing Engineering Centre (MEC) developed the procedure after the "waggle dance" of bees foraging for nectar. 

When a bee finds a source of nectar, it returns to the hive and performs a dance to show the other bees the direction and distance of the source and how plentiful it is. How many of the bees observing the dance will fly off to find this new source will depend on its distance and quality. 

The MEC's Bees Algorithm mimics this behaviour. A computer can be set up to calculate the results of different settings on a manufacturing process. More efforts are then devoted to searching around the most successful settings, in the same way as more bees are sent to the most promising flower patches.

The Algorithm has been shown to cope with up to 3,000 variables and can be faster than existing calculations. By entering basic data about all or part of a company, or even just one machine, the MEC 'Bees' team can calculate the best outcome for a wide range of business processes. The team has already used the Bees Algorithm to work out the most efficient settings on welding systems, the best dimensions for mechanical springs, the most efficient job schedule for a production machine and the best sequence for placing electronic components on a printed-circuit board.

The Algorithm was unveiled by the team at the Internet-based Innovative Production and Machines and Systems (IPROMS) Conference hosted by the MEC as part of its work with the EU-funded IPROMS Network of Excellence. The team's article was one of 110 papers presented to 4,000 delegates from 73 countries at the conference, which was held entirely on-line.

Further information can be found on the MEC website.

Physics and Astronomy

There are four research groups currently using advanced research computing techniques:

• Galaxy formation and evolution: led by Dr Alistair Nelson
• Star formation: led by Prof Anthony Whitworth
• Gravitational Physics Group: led by Prof B.S. Sathyaprakash
• Condensed Matter Group: led by Prof John Inglesfield

Two examples are given below.

Gravitation Physics Group

Principal Investigator: Prof B.S. Sathyaprakash

Research Summary: Black holes, big bang and gravitational waves are outrageous predictions of Einstein’s theory of gravity. Although we have indirect evidence for all, none has been observed directly. However, this is soon going to change. A worldwide network of interferometric gravitational wave detectors (three in the US, two in Europe and one in Japan) are now taking data that can directly observe formation of black holes, their collisions, big bang and other catastrophic events. Prof. Sathyaprakash’s group is among prime data analysis centres in the international partnership, which seeks to detect gravitational waves for the first time. The partnership aims ultimately to use observations of gravitational waves to shed light on many of the unanswered questions in astrophysics and cosmology.

Technical summary: Prof. Sathyaprakash’s research is computationally very intensive, in part due to the huge amount of live data being generated by the network of gravitational wave detectors across the world, and also because of the large parameter space in which the search is carried out.

The sensitivity of interferometric gravitational wave detectors depends on the direction of the source. Courtesy of Sathyaprakash.

Software used: In many respects, analysis of data from gravitational wave detectors is unique and no legacy software can be used in the searches. The group, in partnership with other Universities around the world, has developed algorithms and written software over the past 15 years. This software contains efficient pattern recognition algorithms capable of detecting extremely weak gravitational wave signals buried in noisy data.

Facilities used: Large cluster computers, as well as access to the international Einstein@Home collaboration.

Recent Results: The new generation of very large gravitational wave detectors are now being tuned to optimum sensitivity, sufficient to detect gravitational waves for the first time. In parallel, the data analysis algorithms are being tuned, refined and tested to the required level. As a result, the first, historic detection of gravitational waves is expected to take place in the next 4-8 years.

Gravitational Physics Group homepage.

Star Formation

Principal Investigator: Prof Anthony Whitworth

Research summary: Prof Whitworth’s team use computer simulations to replicate the formation of stars, dwarf stars and large planets, starting from their initial conditions as large gas clouds. They create accurate simulations of the physical conditions using well-understood equations including gravity and hydrodynamics. The experiments shed light on the origins of stars like our own sun, including why so many stars are formed in binary pairs, and what determines whether a star has a planetary system.

Technical summary: Prof Whitworth’s research requires specialist ‘shared memory’ cluster computers due to the large number of timesteps in the simulation and the close physical dependencies between different regions in the star forming region. The situation is highly complex due to the huge range in density and temperature and the complex behaviour of large numbers of gravitationally-attracted particles. 

A binary star is formed in a simulation of star formation. Courtesy of Kitsionas & Whitworth.

Software used : The software has been locally developed from the ground up as part of a 15 year development project.

Facilities used: Prof Whitworth currently uses two older shared memory machines which are being replaced with much faster, state-of-the-art equipment as part of the SRIF 3 investment in ARCCA facilities. When appropriate, he also occasionally makes use of national facilities such as UKAFF.

Star Formation Group homepage.