Fast and scalable finite element algorithms for coupled multiphysicsproblems and non-matching grids


Dr. Martin Kronbichler
Prof. Dr.-Ing. Wolfgang A. Wall
Institute for Computational Mechanics
Technical University of Munich

Project Overview

This proposal aims to develop new scalable algorithms for finite element discretizations of certain classes of multiphysics problems. The algorithms target higher-order discontinuous Galerkin discretizations of fluid flow as one primary field, to which we have developed efficient and scalable implementations previously. The first new development are flexible evaluation routines that can evaluate high-order finite element solutions in arbitrary points, improving root-finding algorithms and local polynomial evaluation to ensure a high node-level performance. The second ingredient is the parallel scalability of the data exchange, implementing new remote-lookup functionality based on consensus algorithms. This work aims to work efficiently also on105and more processor cores, and will therefore rely on point-to-point communication and computations at the owner’s site. From these two ingredients, we will then apply the methods to several coupled multiphysics problems, namely fluid-structure interaction (FSI), fluid flow coupled to particle transport, and cut finite element algorithms to couple fluid flow on a moving domain with the flow on a fixed background mesh in order to solve problems with large deformations and topology changes of the domain. The performance will be evaluated by strong and weak scaling experiments as well as by comparing our equation-specific solutions to the generic coupling code preCICE. The proposed work will be done in collaboration with the CFDLab atLRZ.