Skip to navigation Skip to content

Issue
Seventeen

Good-looking models for particle systems

Dr Bannerman's computer-based models not only work well but look good .....…

Good-looking models for particle systems

Good-looking models for particle systems

Stabilising lightweight satellites in outer space and analysing how grain behaves in a hopper, how proteins form or how fluids pass through porous rock, are major challenges for simulation, but Dr Marcus Campbell Bannerman has developed special software and borrowed visualisation techniques from the games industry to make his computer-based models not only work well but look good...

When he was doing research for his PhD in chemical engineering at UMIST (University of Manchester Institute of Science and Technology), Dr Marcus Campbell Bannerman had no access to a high-performance computer, so he built his own cluster. When the simulation software he needed to run on the cluster wasn’t publicly available, he developed his own event-driven software and released it for free. And this “do-it-yourself” approach dates back to his childhood, when he created computer games for himself and his siblings to play, using Gem Basic, an early ancestor of the more modern Visual Basic programming language.

Bannerman, now a lecturer in the School of Engineering at the University of Aberdeen, feels very strongly that learning how to write computer code should be a basic part of the curriculum from the earliest possible age, along with maths and English. But this is not only because it is harder to solve problems and build applications without using code (including simulations), but also because it makes learning more fun.

Having written his first code at eight years of age, creating his own virtual shop and a series of games, as well as a graphical file system to imitate what he had read about the Internet, Bannerman didn’t focus on computing at school or university until his third year as an undergraduate at Manchester. Faced with a particular problem in chemical engineering, his tutor suggested he solve it by writing some code, and Bannerman rekindled his passion for coding. “Until then,” he explains, “I'd never seen an opportunity for coding in chemical engineering. But suddenly I realised I knew how to code, and it worked – I solved the problem and the undergraduate course as a whole became more interesting for me as I realised that programming could help.”

According to Bannerman, the key to the future success of STEM (science, technology, engineering and maths) in schools is to get across the “code first” message to students, and put computer programming on the same footing as other core subjects such as English and maths. Coding not only helps solve problems more quickly than doing the same job on paper, but also frees students to focus on other more interesting tasks. “Some students can’t see the point of coding until they learn it and appreciate its value for themselves,” he adds. “It makes learning much more exciting by removing the tedium of repeating calculations, and everyone learns faster.”

Bannerman also feels strongly that undergraduates and postgraduate students should have easier access to high-performance computers, not just because they crunch numbers much faster, but also because having access to more computational power allows you to explore more. The computer cluster in Manchester cost about £50,000 to build, and was initially designed for a specific application, but other researchers soon started queuing to use it. He also says that high-performance computing (HPC) systems should be considered an “essential overhead” rather than “luxury” items. “We will always need national HPC centres,” he adds, “but SMEs (small to medium-sized enterprises) and students will all need to go beyond the desktop scale.” The recent purchase of an HPC cluster called Maxwell by the University of Aberdeen meets exactly this kind of requirement, providing a free-at-the-point-of-use HPC service to academic researchers, and the new facility could also generate revenues from local industry users.

Current research

Bannerman’s current research focuses on granular, molecular and particle dynamics, “using discrete force models to study both molecular- and macroscopic-scale particle systems, and developing a new class of particle models for solids processing systems.”

Before he came to Aberdeen, he worked in Germany at the Engineering of Advanced Materials excellence cluster at the University of Erlangen, on a project to develop more efficient granular dampers. These passive devices dampen unwanted vibrations and are “perfect” for satellites; large, lightweight structures which unfold once in orbit need something to dissipate vibration from the launch and prevent the misalignment of their sensitive components. The dampers are like small “rattles” attached to the structure, filled with tiny solid particles which collide and absorb the vibrational energy and convert it to heat. As well as performing simulations of the granular dampers, Bannerman also went on a zero-gravity flight (also known as the “vomit comet”) to validate his theoretical models and “fit the single free parameter” of the model – a part of the design equation which could only be worked out by doing an experiment in weightless conditions. “It was lots of fun,” says Bannerman, “but also very nauseous.”

His recent work in Aberdeen focuses on creating new particle models, and he continues to drive development of DynamO – an open-source event-driven particle simulator which is designed “to drive forward the study of discrete particle models” and which has already found application in a wide range of systems such as granular dampers, nano-colloidal fluids, and protein folding.

His recent work in Aberdeen focuses on creating new particle models, and he continues to drive development of DynamO – an open-source event-driven particle simulator which is designed “to drive forward the study of discrete particle models” and which has already found application in a wide range of systems such as granular dampers, nano-colloidal fluids, and protein folding.

The algorithms behind DynamO use an alternative approach to the more traditional “time-stepping” method of modelling particle dynamics. Bannerman explains this new, simpler approach by comparing it to studying a ball as it moves through the air to bounce off a wall (much like particles or molecules in motion). To simulate what happens, time-stepping performs a calculation of forces at regular intervals, like shooting a time-lapse film, while event-driven models take the starting point and calculate the end point in a single step, skipping what happens in between. “This means we are looking at impulses rather than forces,” Bannerman says, “and in certain applications we can use it for modelling fluids and folding proteins to significantly increase the performance of our models.”

A wide range of particle models is available for the new package, which can also be used for molecular dynamics (to study model fluids), or for granular dynamics (to investigate solid particle processes). Bannerman publicly released the package three years ago and was keen to make it available to all, not only because he himself uses open-source software, but also because it “popularises” the package and encourages other researchers to contribute and adapt it for a wider range of applications. “There was a gap in the market,” he says, “and I also needed something for my own research. It was a problem that clicked with my personal interests – and the new approach worked. There was literature available on the method, but no-one had released the tools before.” The software is currently being used by 20 research groups worldwide.

Versions of the models used for particle dynamics have also helped to simulate battles and what happens in buildings when people panic during an emergency. “We see our own research problems everywhere,” says Bannerman, describing the important role of particle dynamics in everyday life. “Using novel algorithms, it is now possible to simulate the processes that take place in large-scale equipment, modelling millions of particles in near real-time.”

Seeing is understanding

Visualisation is a critical factor in all research that uses simulation as a tool, and Bannerman has borrowed a range of techniques from the games industry for his own scientific research. Once again, because he wasn't happy with the visualisation tools available, he decided to develop his own – now distributed as part of the DynamO package. He uses these games-based techniques “to render large data-sets interactively and in real-time, to help reveal hidden relationships within the simulation data as well as providing visually pleasing renderings of simulations for publication.”

 “It is all about visualisation,” he says, “not only to communicate the results of the models to a wider community, but also because it is useful in debugging code – you can see at a glance whether something is working or not, and whether the simulation is physical.”

As well as providing a “reality check” for the coding, the rendering tools used in games enable a more rapid understanding of the data – for example, when simulating large numbers of molecules, it not only looks better when the structures have shadows, but the shadows themselves also help your eye distinguish the shape, dimensions and depth of the structure. These enhancements are not just cosmetic, Bannerman explains. If you can highlight data fields as subtle changes in colour, this helps you to notice anomalies or interesting results more quickly.

Having shifted his focus from applied chemical engineering to more “theoretical” work in computational science, Bannerman has now rediscovered the “joy of applications” through his recent research on granular dampers and greener, low-carbon cement. But if it hadn’t been for coding and developing games for his siblings, who knows how his career would have evolved? Perhaps the only way to answer this would be to run another simulation...

 

DynamO in action

The event-driven software developed by Dr Campbell Bannerman at the University of Aberdeen is used to model a wide range of particle systems, from molecular to granular dynamics, including everything from simple fluids and crystals to the movement of grain in a silo and how fluids flow through porous media.

According to Bannerman, DynamO is “is one of the fastest and most efficient event-driven particle simulators available,” using 500 bytes per particle and running at about 75,000 events per second, and also the most stable of its type.

Click here for more details

 

 

 

 

 

 

 

 

 

 

 

"Good-looking models for particle systems". Science Scotland (Issue Seventeen)
Printed from http://www.sciencescotland.org/feature.php?id=258 on 22/08/17 03:54:16 PM

Science Scotland is a science & technology publication brought to you by The Royal Society of Edinburgh (www.rse.org.uk).