“What a great boost!We completed our simulation in just one-third of the originally planned time.”
– Mirco Magnini, associate professor at the University of Nottingham, shares his experience with blastAMR, an NHR4CES tool that accelerates OpenFOAM simulations.
NHR4CES tool makes simulations three times faster
Mirco Mangini specializes in numerical simulation and modeling of multiphase flows, heat transfer, boiling, microscale flows. While working on a 3D simulation of nucleate boiling of water in the microlayer regime, he saw a promising LinkedIn post regarding AMR in OpenFOAM by Holger Marschall, PI of our SDL Energy Conversion. They had a private Zoom call and Holger pointed Mirco and his team in the direction of blastAMR and its GitHub release page (online since October 10, 2023, constantly being developed further). blastAMR is an NHR4CES library that accelerates OpenFOAM simulations and is useful for CFD engineers using OpenFOAM for their physics simulations. The library for Load-balanced AdaptiveMesh Refinement in OpenFOAM was developed by Mohammed Elwardi Fadeli from our SDL Energy Conversion.
Mirco saw the potential in this tool to speed up his science, and the results speak for themselves. The library enables faster OpenFOAM simulations on generic grid types. In general, blastAMR manages to run simulations in at least 60 % of the original simulation time – or as in Mirco’s case: around 33 % of the original simulation time!
Less simulation time means less energy used and less cost to run simulations: You can run more simulations with the same budget. Since access to HPC is limited to a given number of CPU hours per period, faster running time is nowadays key in CFD. Besides the faster running time, the possibility to extend AMR to two-dimensional and axisymmetric simulations – both not supported in the official OpenFOAM release – is a great benefit.
“I would recommend blastAMR to all CFD practitioners, be them from academia or industry, who need to increase the speed of their simulations”, says Mirco, who is glad that he got in contact with our SDL Energy Conversion. “OpenFOAM users are of course facilitated by having prior experience in OpenFOAM, but also users of other CFD software in the long term may benefit from moving to an opensource package such as OpenFOAM and then use blastAMR to speed up their simulations.“
Mirco also found handling it user-friendly and is very satisfied: “The Github repository is exemplary, and it was very easy to adapt the case files to my own solver and simulation setup.”
One of NHR4CES’ key skills is to support scientists in making their simulations more efficient. Personal contact with our Simulations and Data Labs (SDL) and Cross-Sectional Groups (CSG), as well as NHR4CES-developed research tools, guarantee this support – confirms Mirco Magnini, associate professor at the University of Nottingham.
The simulation
Mirco’s simulation shows the growth of a steam bubble within a pool of liquid where the bottom surface is heated up above 100 °C. The simulation is used to study the physics of bubble growth and, most importantly, the heat transfer that this provides between fluid and hot surface. Boiling is a very efficient cooling technique and used in many technologies, from nuclear power to thermal control of electronic devices. High-fidelity simulations can effectively complement experiment insights and give a key contribution to reveal the physics of boiling. This simulation was one of the last tasks of the 3-year project BONSAI.
You can see a steam bubble growing over a heated wall (Tsat=373 K; Twall=385 K). The simulation is done on openfoam v2112 with the opensource solver boilingFOAM and uses the powerful library blastAMR, developed at NHR4CES, for adaptive mesh refinement with load balance. The domain is 3D and they model 1/4 of the geometry with a 90 degree symmetry. The domain is a box of 0.8 mm, meshed with 80x80x80 cubic cells and 3 levels of refinement at the interface. The initial bubble radius is 10 um. They follow the growth of the bubble until its radius reaches about 800 um at t=0.15 ms, an 80X growth from t=0. These are extremely challenging simulations. Without AMR, ~260 millions cells would be necessary to have 1.25 um cells everywhere; AMR allows to reduce this to ~4 million and with blastAMR this simulation runs in only 10 hours on 128 cores (Sulis HPC).
About Mirco Magnini
Mirco Magnini is an associate Professor in the Department of Mechanical, Materials and Manufacturing Engineering at the University of Nottingham since 2019 and specializes in numerical simulation and modeling of multiphase flows, heat transfer, boiling, microscale flows. He has a master degree in Mechanical Engineering from the University of Bologna and a PhD in Energy Engineering from the same institution. Prior to joining Nottingham, he was a research associate at EPFL and Imperial College London.
Mohammed Elwardi Fadeli is a PhD candidate in the Department of Mechanical Engineering at TU Darmstadt since 2022. He works on load-balanced adaptive meshing methods for reactive flows with applications in areas such as combustion modelling, turbulent mixing, multiphase flows and heat transfer. He has master degree in Petroleum Engineering from the university of Boumerdes – Algeria and some background into DevOps and cluster infrastructure. He worked over 100 hours developing blastAMR! Contact Mohammed Elwardi Fadeli for more information about the tool or see how we can help you speed up your science!
