CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR …€¦ · CO-SIMULATION AND MULTIPHYSICS...
Transcript of CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR …€¦ · CO-SIMULATION AND MULTIPHYSICS...
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
Sharat Prasad, Ph.D.
Engineering Specialist*
Yun Yun Lu
Principal Development Engineer*
Shawn Harwood
Principal Development Engineer*
Karthik Mukundakrishnan
Development Engineer I*
Martin Sanchez Rocha
Development Engineer I*
*Dassault Systèmes Simulia Corp.
THEME
Multiphysics
KEYWORDS
Multiphysics, Fluid-Structure Interaction, Strong Physics Coupling,
Biomechanics.
SUMMARY
Fluid-structure interaction between incompressible fluids and compliant
structures pose a significant numerical challenge in various industrial
applications. One such application is in biomechanics where biofluids such as
blood interact with compliant biological structures such as arterial walls.
Understanding the hemodynamics in conjunction with the compliance of the
arteries has proven to be an important step towards understanding the effect of
mechanical factors in disease progression. The numerical challenges that are
faced in such problems range from non-linear and anisotropic behavior of
arterial tissues, non-Newtonian behavior of blood flow as well as strong
multiphysics coupling between structural and fluid dynamics due to near unity
fluid-solid density ratios with significantly flexible structures.
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
The current paper focuses on utilizing a scalable, parallel multiphysics
framework within Abaqus to solve complex and strongly coupled fluid
structure interaction problems. The framework is combined with an innovative
algorithm for ensuring stability in incompressible fluid-structure interaction
(FSI) problems that have previously required complex and time-consuming
iterative strategies. As an example to demonstrate this methodology and its
efficacy, we study the blood flow and its interaction with the compliant
abdominal aortic aneurysms (AAA) wall. The blood flow is modeled within
the multiphysics computational fluid dynamics capability in Abaqus/CFD. The
structural dynamics of aneurysm walls are modeled with Abaqus/Standard.
The time dependent fully-coupled fluid and structural coupling is enabled in
the background using the co-simulation engine (CSE) within Abaqus without
requiring any user intervention.
1: Introduction
Abdominal aortic aneurysm (AAA) is a prevalent disease that affects a
significant number of people. An aneurysm is a localized bulge of a blood
vessel. AAA occurs in the aortic segment located between the renal arteries
and iliac bifurcation. When the size of AAA increases, there is a significant
risk of rupture causing severe internal bleeding, other complications or even
death. The management and treatment of such conditions require an
understanding of the influence of mechanical factors on AAA growth and its
eventual rupture. It is well known that the arteries exist in a pre-stressed
configuration and the intraluminal pressure due to blood flow acts a normal
force affecting not only its structural behavior but also the biological behavior.
The wall stress is cited as primary mechanical factor affecting the arterial
properties as well as AAA growth and rupture. As a result, many
computational studies have been carried out to study blood flow, structural
dynamics of arterial walls and their coupled interaction with an aim to
understand and quantify the growth of AAA and predicting its rupture. Such a
computational study requires non-linear structural capabilities, computational
fluid dynamics capability as well as coupled multiphysics capabilities to study
the strongly coupled fluid-structure interaction effects.
In the current work, we utilize Abaqus/Standard as the structural solver and
Abaqus/CFD as the incompressible flow solver. Abaqus/CFD incompressible
flow solver uses a hybrid discretization built on the integral conservation
statements for an arbitrary deforming domain. For time-dependent problems,
an advanced second-order projection method is used with a node-centered
finite-element discretization for the pressure. This hybrid approach guarantees
accurate solutions and eliminates the possibility of spurious pressure modes
while retaining the local conservation properties associated with traditional
finite volume methods. The projection method enables legitimate segregation
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
of pressure and velocity fields for efficient solution of the incompressible
Navier-Stokes equations. The projection operators are constructed so that they
satisfy prescribed boundary conditions and are norm-reducing, resulting in a
robust solution algorithm for incompressible flows. An edge-based
implementation is used for all transport equations permitting a single
implementation that spans a broad variety of element topologies. The solution
methods in Abaqus/CFD use a linearly complete second-order accurate least-
squares gradient estimation. This permits accurate evaluation of dual-edge
fluxes for both advective and diffusive processes. All transport equations in
Abaqus/CFD make use of the second-order least-squares operators. The
implementation of advection in Abaqus/CFD is edge-based, monotonicity-
preserving, and preserves smooth variations to second order in space. The
advection relies on a least-squares gradient estimation with unstructured-grid
slope limiters that are topology independent.
The fluid-structure interaction (FSI) capability within Abaqus utilizes
Abaqus/Standard or Abaqus/Explicit as structural solver and Abaqus/CFD as
the fluid solver with a staggered solution methodology. The global elliptic
nature of the pressure in incompressible flows introduces numerical difficulties
for staggered coupling that are realized as instabilities with exponential growth
in the pressure. Incompressible FSI exhibits this behavior for near unity fluid-
solid density ratios with flexible structures (as commonly seen in
biomechanical problems). Abaqus/CFD deploys a stable, staggered, time-
accurate, incompressible FSI algorithm while taking into account added-mass
effect to solve strongly coupled FSI problems. For detailed information about
Abaqus/CFD solution methodology, we refer readers to [1].
We present the results for time-dependent response of blood flow through
AAA when the arterial walls are rigid (Abaqus/CFD analysis with appropriate
fluid boundary conditions). We then present the results for time-dependent
response of aneurysm walls subject to the time varying cardiac pressure pulse
without considering the effect of blood flow. We finally demonstrate a
coupled FSI methodology to investigate the strongly coupled interaction of
compliant arterial walls with blood flow. For each of these cases, we have also
presented reference results for a healthy artery of the same inlet diameter.
2: Methodology
In order to investigate blood flow, arterial dynamics and coupled FSI in AAA,
we utilize the aneurysm geometries proposed in [2]. These virtual aneurysm
models are used in absence of realistic aneurysm geometries. The CAD
models for aneurysms are created in Abaqus/CAE. The fluid domain has an
inlet and outlet diameter of d = 0.02 m and a maximum bulge of diameter D =
0.06 m at the midsection of the AAA sac. The arterial wall has a thickness of
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
W = 1.5x10-3
m. Aneurysm geometries in two different configurations are
studied. The asymmetry of the model is dictated by factor β which is the ratio
of radii r and R (see [2] and Figure 1(d)). A value of β = 1.0 represents an
axisymmetric aneurysm while β = 0.2 represents an asymmetric aneurysm
(Figure 1(b), 1(c)). We also study a healthy artery with the same uniform
diameter as the inlet and outlet diameter of the AAA (Figure 1(a)).
Figure 1: Aneurysm geometry, solid (artery) and fluid (blood) domains (a) healthy
artery (b) β β β β = 0.2 (asymmetric aneurysm) (c) β β β β = 1.0 (axisymmetric aneurysm)
(d) asymmetry ratio, ββββ
In the current work, blood is modeled as a Newtonian fluid with a density ρf =
1050 Kg/m3 and a dynamic viscosity µf
= 0.00385 Pa.s. Although blood is a
non-Newtonian fluid with shear thinning behavior, it is a reasonable
approximation to treat it as a Newtonian fluid in blood vessels of size greater
than 0.5 mm in diameters [3]. In vessels of larger diameters, the viscosity is
relatively constant due to high rates of shear.
The arterial walls typically exhibit non-linear, anisotropic hyperelastic
behavior. Several studies have indicated non-linear viscoelastic behavior as
well [4]. Hence, a wide range of material models have been proposed for
arterial walls. The approximations range from isotropic linear elastic behavior
[2] to isotropic non-linear hyperelastic behavior [5] to fully anisotropic non-
linear hyperelastic behavior [6]. While Abaqus supports different material
models, it is often difficult to obtain accurate material constants. In this current
work, for the sake of illustration, we have assumed an isotropic non-linear
elastic behavior with an elastic modulus E = 2.7 MPa and a Poisson’s ratio ν =
0.45. The arterial walls has a density of ρs = 1200 Kg/m
3.
r
R
(a) (b) (c) (d) Inlet
Outlet
Wall
(FSI region)
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
The boundary of the fluid domain is split into an inlet section where fluid
enters the solution domain; an outlet section where fluid leaves the solution
domain and a wall region which represents the interface with the artery (see
Figure 1(a)). The following boundary conditions are applied to the fluid
domain: (i) a time dependent fully developed velocity boundary condition on
the inlet section (Figure 2), (ii) a time dependent pressure boundary condition
on the outlet section (Figure 2) and (iii) a no-slip/no-penetration boundary
condition on the wall region.
For the laminar flow analysis in Abaqus/CFD with rigid walls, the no-slip/no-
penetration boundary condition dictates that fluid velocity at walls is zero. The
velocity and pressure waveforms conditions utilized here represent the normal
hemodynamic conditions in human abdominal aorta [2]. The Reynolds number
for the flow is 1640 based on the peak inlet velocity and 410 based on time
averaged inlet velocity.
Figure 2: Luminal pulsatile velocity and pressure waveform.
The boundary conditions for the structural model for artery include a zero
displacement condition on the ends of the artery. While more realistic
boundary conditions can be applied including the effect of interaction with
surrounding tissues, a rather simplistic approach is chosen due to lack of data
on such interactions. For studying the dynamics of AAA without including the
effect of fluid flow, we will load the AAA internally with a time dependent
luminal pressure shown in Figure 2. This approach incorporates the effect of
normal forces on the arterial walls but excludes the shear forces exerted by the
blood flow.
For fully coupled FSI simulation, we will computationally solve the laminar
blood flow in Abaqus/CFD and couple it with structural response of AAA in
Abaqus/Standard. Such a simulation will enable inclusion of shear forces on
the arterial walls and it will also enable incorporating the effect of arterial
deformations on blood flow. In addition to the boundary conditions mentioned
above, such an analysis would also require additional constraints at the FSI
interface namely (i) continuity of displacements at blood/AAA interface (us =
uf), (ii) continuity of velocity at blood/AAA interface, (vs = vf, no-slip/no-
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
penetration condition) and (iii) equilibrium of tractions at the blood/AAA
interface (σσσσs .n = σσσσf .n). The interfacial constraints are automatically honored
by Abaqus when a coupled fluid-structure interaction is defined at the FSI
interface in structural and fluid models.
3: Computational Mesh
The computational meshes utilized in the simulations for solid and fluid
domains are depicted in Figure 3. The meshes for both solid as well fluid
models are created in Abaqus/CAE.
Figure 3: Computational mesh for fluid and solid domains
For the element choice for the artery, we tested fully integrated (C3D8),
reduced integrated (C3D8R) (with enhanced hourglass control) and
incompatible mode (C3D8I) elements in Abaqus/Standard. Various element
choices yielded nearly identical results. Hence, we have used fully integrated
solid elements for both structural dynamics of arteries and fully coupled fluid-
structure interaction study. 18228 solid elements were used in the structural
analysis of the artery.
The fluid domain is meshed with FC3D8 elements in Abaqus/CFD. Four
different meshes ranging from 30480 to 121920 fluid elements were analyzed
to arrive at a mesh convergent solution. A change of less than 5% in velocities
at the centre of fluid domain is used as an acceptable criterion for mesh
convergence.
For FSI analysis, we chose the coarsest fluid mesh (30480 elements) for sake
of computational efficiency and acceptable solution times.
4: CFD Simulation of Blood Flow in AAA
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
Blood flow is simulated in AAA with boundary conditions mentioned in
section 2. A transient Navier-Stokes solver in Abaqus/CFD is used to obtain
solution for 8 cycles until a time-periodic solution is obtained. Figure 4 depicts
the velocities at five interesting times of the cardiac cycle for β = 0.2.
Figure 4: Velocity contours in AAA, β β β β = 0.2
Figure 5 depicts the mass flow rate at the inlet and outlet and also the error in
mass conservation for AAA with β = 0.2. It can be seen that Abaqus/CFD
obtains an excellent mass conservation with a maximum error of 0.1%.
Figure 5: Mass flow rate and error in mass conservation, β β β β = 0.2
Figure 6 depicts the velocities for axisymmetric aneurysm. The corresponding
values for the healthy artery are depicted in Figure 7.
t = 0.2 s t = 0.3 s t = 0.52 s t = 1.0 s t = 0.4 s
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
Figure 6: Velocity contours in AAA, β β β β = 1.0
Figure 7: Velocity contours in healthy artery
The solution time to simulate eight cardiac cycles for the finest CFD mesh
(121920 elements) was 14 minutes on 16 cores of a Quad Core 2.27 GHz
machine.
At the beginning of the cardiac cycle, a mildly recirculating flow pattern is
seen at the walls for the healthy artery. For AAA, a similar mild recirculation
is seen which extends all the way along the walls of the aneurysm sac. At peak
inlet flow (t = 0.3 s), the effect of strong systolic acceleration results in a
completely attached flow in the healthy artery as well aneurysm models. The
flow starts recirculating again in the aneurysm sac at its intersection with the
inlet section around t = 0.4 s when the cardiac pressure pulse is at its peak. The
recirculation strength (as seen by reversed x velocity component) reaches its
maximum at t = 0.52 s when the inlet flow has also reversed. The recirculation
adheres to the walls of the aneurysm sac extending all the way to its
intersection with the outlet section.
The maximum wall shear stress in healthy artery and AAA is observed at the
peak flow (t = 0.3 s). The maximum wall shear stress seen in AAA is about
12.4% higher than the healthy artery (see Figure 8). The AAA asymmetry
does not have any remarkable effect in the peak wall shear stress.
t = 0.2 s t = 0.3 s t = 0.52 s t = 1.0 s t = 0.4 s
t = 0.2 s t = 0.3 s t = 0.52 s t = 1.0 s t = 0.4 s
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
Figure 8: Maximum wall shear stress (t = 0.3 s) in asymmetric and axisymmetric
AAA
5: Structural Dynamics of Aortic Aneurysm
We now present the result for structural dynamics simulation (without
modeling blood flow) of an aortic aneurysm subjected to a time dependent
intraluminal pressure. Figure 9 depicts the Von Mises stresses and
displacements in the AAA for β = 0.2 and β = 1.0 at minimum and maximum
luminal pressure. Corresponding values for the healthy artery are also
depicted.
As can be seen from the results, the maximum Von Mises stress seen in the
artery is 156% more in an asymmetric aneurysm and 135% more in an
axisymmetric compared to the healthy case at maximum luminal pressure (t =
0.4 s). The presence of aneurysm invokes higher stresses on arterial walls
making it prone to rupture. The effect of asymmetry in aneurysm geometry is
to increase the stresses even further.
Healthy
β β β β = 0.2
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
Figure 9: Von Mises stresses and displacements in the artery at minimum (t = 0.2
s) and maximum luminal pressure (t = 0.4 s) for healthy artery and AAA with β β β β =
0.2, β β β β = 1.0
6: Coupled Blood Flow/Arterial Wall Dynamics Using FSI Approach
We now demonstrate the coupled FSI analysis for the AAA. Figure 10 shows
the Von Mises stresses in the AAA at peak systolic acceleration and peak
cardiac pulse. The corresponding values for the healthy artery are also shown.
The FSI analysis incorporates the effect of both the normal as well as shear
forces due to blood flow. During the cardiac cycle, the time-dependent fluid
forces will deform the inner wall which in turn will affect the fluid velocity
field. The sudden application of fluid forces at t = 0 can cause convergence
issues in the FSI analysis since a sudden pressure pulse causes large
accelerations in the structure. In order to circumvent this problem, an equal
and opposite pressure load is applied to the structure at t = 0. This dummy load
is ramped down over a 0.1s period to load the structure more smoothly. The
coupled FSI simulation is run for six cycles to obtain a time-periodic solution.
t = 0.2 s t = 0.4 s
β β β β = 1.0
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
Figure 10: Von Mises stresses in the artery at peak systolic acceleration (t = 0.3 s)
and peak cardiac pulse (t = 0.4 s) in the FSI analysis
As can be seen from the results, the coupled simulation is still dominated by
the normal forces exerted by the cardiac pressure pulse.
7: Conclusion
In this study, the methodology for investigating the coupled fluid-structure
interaction of pulsatile blood flow through a compliant artery (with and without
an aneurysm sac) was demonstrated. The staggered FSI solution approach was
implemented using Abaqus/Standard as the structural solver and Abaqus/CFD
as the incompressible Navier-Stokes solver. The case study successfully
demonstrates the efficacy of multiphysics capabilities within Abaqus in
studying structural, fluid and strongly coupled dynamics of blood flow and
compliant arteries with nearly identical fluid-solid density ratios. The stability
of the strongly coupled FSI problem is ensured with a single step algorithm
within a staggered solution paradigm without resorting to iterative strategies.
Future work will involve incorporating the effect of material properties such as
hyperelasticity and viscoelasticity of the artery and the aneurysm sac in a
coupled FSI simulation.
t = 0.3 s t = 0.4 s
CO-SIMULATION AND MULTIPHYSICS TECHNOLOGIES FOR
COUPLED FLUID-STRUCTURE INTERACTION PROBLEMS
REFERENCES
[1] Abaqus version 6.11 Analysis User’s Manual, Dassault Systemes Simulia
Corp, 2011
[2] Scotti, C., Shkolnik, A., Muluk, S. and Finol, E. - Fluid-structure
interaction in abdominal aortic aneurysms: effects of asymmetry and wall
thickness, BioMedical Engineering OnLine, Vol 4, pp. 64, 2005.
[3] McDonald, D. – Blood flow in arteries, Wilkins & Wilkins, 1960
[4] Kuchařová, M., Ďoubal, S., Klemera, P., Rejchrt, P. and Navrátil, M. - Viscoelasticity of biological materials – Measurement and practical impact on
biomedicine, Physiol. Res. 56 (Suppl. 1), pp. S33, 2007
[5] Raghavan, M. L. and Vorp, David A. – Towards a biomechanical tool to
evaluate rupture potential of abdominal aortic aneurysm: identification of a
finite strain constitutive model and evaluation of its applicability, Journal of
Biomechnics, Vol 33, pp. 475, 2000
[6] Vorp, David A. – Biomechanics of abdominal aortic aneurysm, Journal of
Biomechanics, Vol 40, pp. 1887, 2007.
[7] Scotti, C. and Finol, E. – Compliant biomechanics of abdominal aortic
aneurysm: A fluid-structure interaction study, Computers and Structures, Vol
85, pp. 1097, 2007.