Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Abstract. This paper presents a self-contained parallel unstructured fluid-structure interaction solver
based on a finite volume method. The solver is a part of an open-source finite volume toolbox for solid
mechanics and fluid-solid interaction simulations. The procedure belongs to the group of partitioned
methods, commonly used procedures for FSI problems. The fluid and solid domains are solved separately
and finite-volume methodology is employed for both domains. The fluid part is modelled by the Navier-
Stokes equations in the arbitrary Lagrangian-Eulerian form, whereas solid domain is modelled using St.
Venant-Kirchhoff constitutive material model in the total Lagrangian form. Both methods are second-
order accurate in space and time. Coupling between domains is solved in an iterative manner using
Gauss-Seidel iteration scheme or the interface quasi-Newton procedure. Parallel computing is performed
using the domain decomposition approach. The performance of the solver is evaluated and obtained
numerical results agree well with available data.
Keywords: fluid-structure interaction, finite-volume method, parallelisation, OpenFOAM
1. Introduction
The present work presents a self-contained fluid-structure interaction (FSI) solver for analysis of the
interaction between an incompressible fluid and an elastic compressible structure exhibiting large dis-
placements and rotations. The solver is a significant extension of the previous authors’ contributions (eg.
[1, 2, 3]) and represents the basis for the current attempts to develop a single general solver (toolbox) for
fluid, solid and FSI problems. The solver is based on the cell-centred finite volume discretisation method
for both domains and strongly coupled partitioned solution procedure. It is second-order accurate in both
space and time, applicable to arbitrary polyhedral meshes and parallel computing.
1
The solid part of the model is described by the St.Venant-Kirchhoff constitutive model in the total
Lagrangian form. Apart from being second-order accurate, the model also employs an improved algo-
rithm for the calculation of the displacement gradients. The momentum equation for the solid domain
is discretised using the second-order accurate mid-point rule for the face and volume integrals, and the
second-order accurate three-time-level implicit scheme (so-called backward scheme) for the time. In
order to calculate displacement gradients, the vertex-based Gauss-Green gradient calculation procedure
is applied with a special vertex displacement procedure for parallel computing. The momentum equa-
tion is solved for displacement using segregated algorithm, where components of displacement vector
are temporarily decoupled and solved separately. The system is solved in na iterative manner using the
preconditioned conjugate gradient iterative solver. More details about the solid part of the solver can be
found in [7].
With regard to the FSI interface, the problem is decomposed using Dirichlet-Neumann procedure,
where the fluid flow is solved for a given velocity (displacement), while the solid model is solved for
a given force, at the FSI interface. For the interpolation at the FSI interface two techniques were used:
face-interpolation procedure (from fluid to solid) and vertex-interpolation procedure (from solid to fluid).
The equilibrium of the force and the velocity (displacement) at the FSI interface is enforced in an iterative
manner by a strong coupling between solid and fluid domains; the Gauss-Seidel scheme with fixed or
Aitken relaxation, or the interface quasi-Newton procedure with the approximation for the inverse of the
Jacobian from a least-square model [8].
The FSI solution procedure is summarised in the Algorithm 1.
3. Numerical example
The solver is verified on several numerical problems [7], but here is only presented the well-known
benchmark FSI test case documented in [9]. The geometry and corresponding discretised domain are
shown in Fig.1. It is the case of a horizontal channel with 0.41 m height and 2.5 m length, containing
a rigid cylinder of radius 0.05 m. The cylinder centre is positioned 0.2 m from the bottom and inlet
(left) boundaries of the channel. An elastic plate of 0.35 m length and 0.02 m height is attached to the
right-hand side of the rigid cylinder. The fluid part of the mesh consists of 21 344 quadrilateral finite
volumes, while the solid part consists of 328 finite volumes.
A parabolic velocity profile is specified at the inlet boundary with a constant pressure at the outlet
of the channel and a no-slip boundary condition on the walls. Fluid flow is assumed to be laminar.
Properties of the fluid and elastic plate material are given in Table 1 together with the mean inlet velocity
2
0.41
0.35
(0,0) 2.5
0.02
0.05
A(0.6,0.2)
C(0.2,0.2)
0.35
(a) (b)
Figure 1: The test case: a) geometry, b) discretised mesh. All dimensions are in m.
for two cases, designated as FSI2 and FSI3 in [9]. The problem is 2D with plane strain assumption for
the solid. The numerical solution is obtained with a time step size ∆t = 1.5 × 10−3 s for the FSI2 case
and ∆t = 0.75 × 10−3 s for the FSI3 case.
FSI2 FSI3
ρf , kg/m3 1000 1000
ν f , m2 /s 0.001 0.001
u, m/s 1 2
ρs , kg/m3 10000 1000
Es , Pa 1.4 × 106 5.6 × 106
νs 0.4 0.4
Table 1: Fluid and solid properties, and mean inlet velocity u for the test case: ρ f is the fluid density, ν f
is the fluid kinematic viscosity, ρs is the solid density, Es is the Young’s modulus and νs is the Poisson’s
ratio.
Quantitative comparison with the corresponding benchmark solutions is given in Table 2. Mean
values and amplitudes are calculated using maximum and minimum values after the periodic solution is
reached (around 7 s for the FSI2 case and 4 s for the FSI3 case), whereas frequencies are calculated
using the Fast Fourier Transform algorithm.
Table 2: Displacement and force for the test case. The values are given in the same format as in the
benchmark paper [9] (mean ± amplitude [frequency]).
3
4. Conclusions
FSI problems can be successfully simulated using a partitioned method where both fluid and solid do-
mains are solved using finite-volume methodology. In this work, a self-contained solver for FSI problems
is developed in OpenFOAM software. The solver has been tested on the Hron and Turek test cases and
excellent agreement between published and obtained results have been obtained.
It has to be also noted that the solver supports multiprocessing simulations. Parallel computing is
achieved by constructing and decomposing the fluid and structural parts separately. Here, one fluid and
one structural sub-mesh is assigned to each CPU processor and communication between processors is
performed using the Message Passing Interface communication protocol. The interface between the
fluid and structure regions is represented by a pair of so-called global face zones, being presented on all
processors. The global face zone point and face fields are reconstructed individually on each processor
and subsequently global synchronisation is performed for points/faces shared by multiple processors.
More details can be found in [7].
Finally, the presented solver has been used for further developments of the solid mechanics toolbox
for structural analysis in an attempt to generalise calculations of such problems. Here, different fluid and
solid constitutive models as well as solving methodologies are available. First such attempt was a so-
called extendBazaar toolbox available in foam-extend [10], where three separate solvers for fluid, solid
and FSI simulations are available. More recently, extendBazaar has been generalised into solids4Foam
toolbox, where all simulations are carried out using a single solver[11].
References
[1] Greenshields, C.J. and Weller, H.G. and Ivankovic, A., The finite volume method for coupled fluid flow and
stress analysis, Computer Modeling and Simulation in Engineering, 4(3):213-218, 1999.
[2] Kanyanta, V. and Ivankovic, A. and Karac, A.Validation of a fluid-structure interaction numerical model for
predicting flow transients in arteries, Journal of Biomechanics, 42(11):1705-1712, 2009.
[3] Karac, A. and Ivankovic, A., Investigating the behaviour of fluid-filled polyethylene containers under base
drop impact: A combined experimental/numerical approach, International Journal of Impact Engineering,
36(4):621-631, 2009.
[4] Issa, R. I. A front-tracking method for viscous, incompressible, multi-fluid flows, Journal of computational
physics, 100:25, 1992.
[5] Tuković, Ž. and Jasak, H. A Moving Mesh Finite Volume Interface Tracking Method for Surface Tension
Dominated Interfacial Fluid Flow, Computers and fluids, 55:70-84, 2012.
[6] Jasak H, Tukovic Z. Automatic mesh motion for the unstructured finite volume method, Transactions of FA-
MENA, 30(2):1–20, 2006.
[7] Tukovic Z., Karac A., Cardiff P., Jasak H., Ivankovic A. OpenFOAM Finite Volume Solver for Fluid-Solid
Interaction, Transactions of FAMENA, 42(3):1–31, 2018.
[8] Degroote, J. and Bathe, K.-J. and Vierendeels, J., Performance of a new partitioned procedure versus a mono-
lithic procedure in fluid–structure interaction, Computers and structures, 87:793-801, 2009.
[9] Turek, S. and Hron, J., Proposal for Numerical Benchmarking of Fluid-Structure Interaction between an Elas-
tic Object and Laminar Incompressible Flow, Fluid-Structure Interaction, Eds. Bungartz, H-J. and Schäfer,
M., p.371-385, 2006
[10] Tukovic Z., Karac A., Cardiff P., Jasak H., Ivankovic A., OpenFOAM library for Fluid Structure Interaction,
9th OpenFOAM workshop, Zagreb, 2014.
[11] Cardiff, P. and Karač, A. and De Jaeger, P. and Jasak, H. and Nagy, J. and Ivanković, A. and Tuković,
Ž An Open-Source Finite Volume Toolbox for Solid Mechanics and Fluid-Structure Interaction Simulations,
arXiv:1808.10736, 2018.