Keywords: dynamic fluid structure interaction, geometric conservation, parallelisation, scalability, cantilever, extrusion.

The most common approach to closely coupled dynamic fluid-structure interaction (DFSI) involves a three phase strategy involving an augmented fluids solver, a dynamic structures solver and a mesh adaptation tool. Conventionally, the fluids and structural solvers are distinct codes frequently using completely different discretisation schemes, solver strategies and mesh structures. Overcoming these hurdles to yield solutions that do not degrade with time is non-trivial.

DFSI is a significant computational challenge for a number of reasons. First, in order for the interactions to be fully captured the governing equations for computational fluid dynamics (CFD) must obey the geometric conservation law (GCL) and the governing equations for computational structural dynamics (CSD) must take into account the fluid traction load; second, the procedure needs to operate in such a manner that it preserves the appropriate levels of accuracy in exchanging load information between the 'fluids' and 'structural' solvers and, third, the mesh deformation within the fluid domain needs to be managed to avoid the generation of warped elements which can severely compromise the convergence behaviour of the flow algorithm.

DFSI is significantly more compute demanding than either the CFD or CSD on its own, and hence it is a natural candidate to exploit high performance parallel cluster systems. However, the fact that most procedures use distinct codes makes this extremely difficult to achieve.

This paper describes an approach to the parallelisation of DFSI that employs:

a)

A conventional three phase numerical procedure for Generalised Navier Stokes flow and elastic solids

b)

A single software framework embedding modules for flow, dynamic structures and mesh adaptation which work from the one single mesh database

c)

A solver strategy which uses a 'group' domain decomposition approach that solves only for the physical involved in each sub-domain

d)

A multi-phase mesh partitioning strategy that ensures that equal portions of each sub-domain are located on each processor, thus leading to a high quality load balance

Parallel scalability for computational mechanics solvers and codes is fairly straightforward to conceive of for problems where the load per mesh node or element is fairly uniform. Many single discipline codes have realised high degrees of scalability and so the omens would appear to be good for more complex solvers for multi-disciplinary (MD) and closely coupled multi-physics (MP) codes. The problem with MD/MP codes is that their compute loads are not homogeneous across the whole mesh, which gives rise to problems in realising parallel scalability because of the challenge of achieving a load balance across the partitioned mesh and ensuring that the inter-processor communications remains minimised.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £95 +P&P)