Using the HPC Pool to help understand the fundamentals of metallic joints in next generation nuclear reactors

by | Dec 2, 2020 | HPC, Projects | 0 comments

In 2019 the Research Lifecycle Programme made a £1 million investment into computational resources to further enable research across all faculties and institutes. The result of this was the creation of the High Performance Compute (HPC) Pool, a resource to allow the University’s research community to address larger scale computational problems and take on research that previously exceeded our resources.   

Dr Tom Flint, Post-Doctoral Research Associate, Dalton Nuclear Institute, made use of the HPC Pool to carry out research for his recently published paper in Nature Communications: Physics.  

Nature is a multidisciplinary journal family founded in 1869 which has often been a forum for the first publication of scientific breakthroughs (ranging from the discovery of the neutron, pulsars and the structure of DNA to the initial sequencing and analysis of the human genome), and only around 8% of the papers submitted to it are accepted for publication. 

Dr Flint explains:  

“My research is focussed on understanding how the application of high energy density sources of heat to multi-component metallic substrates causes the material to evolve through fusion and vaporisation state transitions; as occur during laser or electron beam welding or additive manufacturing. In this new approach we explicitly consider the volumetric changes due to vaporisation and condensation of each elemental component, and formulate the equations accordingly with the aim of fully capturing the governing physics. This enables us to investigate critical flow phenomena in these processes that more basic mathematical formulations do not permit. 

To do this we have to solve partial differential equations (PDEs) that describes the momentum field evolution in the domain: a PDE for the energy transport in the system, a closure PDE to maintain the mass continuity of the system, and crucially 2 PDEs per chemical component in the system (one for the condensed phase and one for the vaporised phase). So considering a substrate comprised of 4 elements, we will be solving 11 PDEs at every point in the domain. This is incredibly computationally expensive, especially as it takes tens of millions of grid points to resolve the features we are interested in. 

For one of the cases presented in our Communications: physics paper, to simulate a few seconds of fluid/vapour evolution  I would have to run on the computational shared facility (CSF3) for 2-3 months. For context, on a high performance desktop computer this same simulation would take roughly 21 years. In the HPC pool, I can solve this in 3 weeks. The step change in computational performance provided by the HPC Pool has been incredibly valuable for me, with access to 10 times more cores than the CSF3 by which to parallelise my Message Passing Interface (MPI) code. It means I can really start to investigate critical flow phenomena in these advanced manufacturing scenarios over meaningful scales. 

The benefits of these analyses are not purely academic. The drive towards low carbon energy production through nuclear generation in hindered by the large capital expenditure of nuclear power plants. Incorporating power beam welding technologies into the UK manufacturing portfolio will drastically reduce fabrication costs and make adoption of nuclear power more likely. The adoption of these processes is hindered by the lack of understanding of physical mechanisms governing material evolution. Hopefully my work can play a small part in better understanding this.” 

The HPC Pool is aimed at large, multi-node batch jobs running parallel software and in order to use it, computational job sizes must be 128 – 1,024 cores. The HPC Pool is currently free to use for any researcher who has a legitimate HPC need. If you would like access to the HPC Pool for a specific research project, you will first need to apply for access. For more information on the HPC Pool and the application process, please visit:  


Find out more about Dr Tom Flint’s research paper:  






Submit a Comment

Your email address will not be published. Required fields are marked *