High performance finite element modeling

SNIC 2018/3-206


SNAC Medium

Principal Investigator:

Johan Jansson


Kungliga Tekniska högskolan

Start Date:


End Date:


Primary Classification:

10105: Computational Mathematics

Secondary Classification:

20306: Fluid Mechanics and Acoustics



Our research in the Numerical Methods group at KTH concerns development of high performance parallel adaptive finite element methods we denote Direct FEM and algorithms we have demonstrated are capable of solving grand challenges in turbulent flow and fluid- structure interaction (FSI), and providing an abstract and general framework for robust and highly effective simulation of multiphysics PDE models for industry, academic research and education. The methods are realized in the FEniCS framework [1, 2, 3], specifically in the Unicorn solver component, which automates the solution of partial differential equations with good scaling on supercomputers, including a recent parallel-in-time development based on an ensemble strategy. We have in 2017 in the AIAA High Lift Prediction Workshop (HiLiftPW3) been able to demonstrate simulation of a full airplane with LES with correct prediction of separation (stall) [4], which is identified by Jameson and NASA as the main challenge in CFD for aerodynamics. Also in 2017 we participated in the 5th AIAA International Workshop on High-Order CFD Methods (HiOCFD5) showing excellent performance of the Unicorn solver component in FEniCS [5]. Our research has been recognized by awards, grants and reviews at the highest level of science and technology, for example: * Guarantor for Spanish Severo Ochoa Center of Excellence (Johan Jansson) * Maximum review score for PRACE Tier-0 * ERC Starting and Proof of Concept Grant (Johan Hoffman) * Priority focus for KTH MOOC-HPFEM project from KTH MOOC Steering Committee, the course had 500 students after 4 days, and the growth of enrollment was ca. 100 students per day the first week (Johan Jansson) * Excellent review of Johan Jansson’s research program on Direct FEM and FEniCS from world-leading mathematicians in BCAM Scientific Advisory Committee. The focus in this project is to: 1. Investigate challenging large-scale industrial problems with great relevance to society, such as vehicle aerodynamics for optimizing fuel consumption and improving control systems, and marine renewable energy such as floating wind turbines and wave energy devices, and biomedicine such as flow in the heart and vessels, the vocal folds and diffusion in neurons. This is in close collaboration with industrial leaders such as Tecnalia in Spain, the International Energy Agency Ocean Energy Systems and cardiologists in Sweden and Spain. 2. Carry out comparison and co-investigation with other leading LES frameworks such as Nek5000 (Schlatter) and PyFR (Vincent), which was already started as part of HiOCFD5. 3. Develop new directions in our methodology such as adaptive 3D compressible flow, new mesh modification methods, new FSI formulations, etc. [1] [2] [3] J Hoffman, J Jansson, N Jansson, FEniCS-HPC: Automated predictive high-performance finite element computing with applications in aerodynamics, PPAM 2015[4] Jansson J.; Krishnasamy E.; Leoni M.; Jansson N.; Hoffman J. , "Time-resolved Adaptive Direct FEM Simulation of High-lift Aircraft Configurations." Springer (accepted 2017). [5]