OpenFOAM offers widely used open-source tools for computational fluid dynamics, but scalability can be a challenge. Research indicates that Intel® processors may help users scale up their OpenFOAM workloads.
If you’re not in a line of work that revolves around computational fluid dynamics, you may have never heard of OpenFOAM. But chances are you have benefitted from this free, open-source software toolkit for CFD, because OpenFOAM is used in the design and engineering of everything from airplanes and automobiles to propellers, pumps and pipes.
In the view of one academic research team — composed of researchers from Texas A&M university, Iowa State University, Pennsylvania State University and Virginia Tech University — the rise of OpenFOAM signals a major transformation in our approaches to CFD.
“There is a revolution going on, impacting and transforming how computational mechanics and the associated design and optimization are done: the emergence, availability, and large-scale use of OpenFOAM,” the team notes in a paper published in the journal Notices of the AMS. “It belongs to the contemporary open-source trend not unlike the roles played by the Linux operating system or the Internet encyclopedia Wikipedia.”[1]
OpenFOAM, which stands for Open Field Operation and Manipulation, is increasingly used by researchers and engineers working in both academic and industrial settings. A few examples:
- Audi uses OpenFOAM for highly standardized CFD simulations in underhood air flow, climatization and external aerodynamics.[2]
- Caterpillar uses OpenFOAM for a number of CFD marine applications, such as hull resistance and propeller cavitation.[3]
- A research team affiliated with the Pittsburgh Supercomputing Center (PSC), Texas A&M University and Texas Advanced Computing Center uses OpenFOAM to better understand coolant and heat transfer in high-temperature-jet reactors.[4]
OpenFOAM at the University of Cambridge
Researchers at the University of Cambridge are leveraging OpenFOAM on high-performance computing clusters for multiple projects, including simulations for vehicle drag reduction and full-annular simulation of a gas turbine combustor.
The researchers like OpenFOAM for multiple reasons, according to Jeffrey Salmond, the university’s head of research software engineering. Among the advantages that he cites:
- OpenFOAM is a comprehensive toolkit that contains the numerical methods and algorithms to cover simulations over a huge range of length scales, flow velocities and materials.
- OpenFOAM is very flexible. It can be used with complex geometries (i.e., to simulate flow around a complex shape) and it incorporates lots of different physics, including large-eddy-type turbulence models, compressible flows and radiative transfer.
- OpenFOAM is very modular. The structure of the package makes it relatively easy to write new features (e.g., a new numerical method or a new model for a better representation of a material) and to build new capabilities that leverage all the previous investments in the package.
Putting OpenFOAM to the test
Let’s dive down into a more technical view. For all its benefits, OpenFOAM is known to have challenging strong- and weak-scaling behaviour, according to a paper authored by Ravi Ojha from Tata Consultancy Services and Christoforos Hadjigeorgiou from the University of Cambridge. [5] To see if they could overcome this challenge, this research duo put OpenFOAM to the test on a hardware platform based on the Intel® Xeon Phi™ processor.
For this effort, Ojha and Hadjigeorgiou focused on investigating and optimizing OpenFOAM for scaling a realistic model of up to 16 processor nodes. The main model used for the work was a freely available Land Rover Evoque.
While the technical details are intriguing, we will keep the story at a higher level. With the support of professionals from Tata Consultancy Services and the Dell HPC Engineering Group, and with access to Dell internal compute facilities, the researchers demonstrated that, for large simulations, the Intel processor “could be a good candidate for scaling-up the relevant industry-scale workloads and reducing simulation time-to-solution.”
If you appreciate the how-to technical details, you can download the paper to get the full story. Or you can just keep these takeaways in mind: OpenFOAM has a tremendous amount of momentum among CFD users, and Intel and Dell EMC have the computational platforms that may help your organization take greater advantage of this groundbreaking software toolkit.
[1] Goong Chen, Qingang Xiong, Philip J. Morris, Eric G. Paterson, Alexey Sergeev, Yi-Ching Wang. “OpenFOAM for
Computational Fluid Dynamics,” Notices of the AMS. 2014;61(4):354–363.
[2] Richard Borris, “Customized developments for AUDI’s OpenFOAM processes,” abstract for the fourth OpenFOAM User Conference, 2016.
[3] T. Huuva, S. Törnros, “CFD and OpenFOAM at Caterpillar with a Main Focus on Marine Applications,” abstract for the fourth OpenFOAM User Conference, 2016.
[4] Pittsburgh Supercomputing Center news release, “PSC Wins a Record Five HPCwire Readers’, Editors’ Choice Awards,” November 14, 2017.
[5] Ravi Ojha, Christoforos Hadjigeorgiou. “OpenFOAM on Intel® XEON Phi™ Processors,” January 12, 2018.