ices report 14-14 fluid–structure interaction analysis of ... · fluid–structure interaction...

ICES REPORT 14-14

June 2014

Fluid–structure interaction analysis of bioprosthetic heartvalves: Significance of arterial wall deformation

by

Ming-Chen Hsu, David Kamensky, Yuri Bazilevs, Michael S. Sacks, Thomas J. R. Hughes

The Institute for Computational Engineering and SciencesThe University of Texas at AustinAustin, Texas 78712

Reference: Ming-Chen Hsu, David Kamensky, Yuri Bazilevs, Michael S. Sacks, Thomas J. R. Hughes,"Fluid–structure interaction analysis of bioprosthetic heart valves: Significance of arterial wall deformation,"ICES REPORT 14-14, The Institute for Computational Engineering and Sciences, The University of Texas atAustin, June 2014.

Computational Mechanics manuscript No.(will be inserted by the editor)

Fluid–structure interaction analysis of bioprosthetic heart valves:Significance of arterial wall deformation

Ming-Chen Hsu · David Kamensky · Yuri Bazilevs ·Michael S. Sacks · Thomas J. R. Hughes

Received: date / Accepted: date

Abstract In this paper, we propose a framework whichcombines variational immersed-boundary and arbitraryLagrangian–Eulerian (ALE) methods for fluid–structure in-teraction (FSI) simulation of the bioprosthetic heart valveimplanted in a deformable artery. We find that the variationalimmersed-boundary method for FSI remains robust and ef-fective for heart valve analysis when the background fluidmesh undergoes deformations corresponding to the expan-sion and contraction of the elastic artery. Furthermore, thecomputations presented in this work show that the arterialwall deformation contributes significantly to the realism ofthe simulation results, leading to flow rates and valve mo-tions that more closely resemble those observed in practice.

Keywords Fluid–structure interaction · Bioprosthetic heartvalve · Variational immersed-boundary method · ArbitraryLagrangian–Eulerian formulation · Isogeometric analysis ·Arterial wall deformation

1 Introduction

Heart valves are passive structures that ensure the unidirec-tional blood flow through the heart by opening and clos-ing in response to hemodynamic forcing. Hundreds of thou-sands of diseased valves are replaced by prosthetics annu-

M.-C. Hsu ()Department of Mechanical Engineering, Iowa State University, 2025Black Engineering, Ames, IA 50011, USAE-mail: [email protected]

D. Kamensky ·M. S. Sacks · T. J. R. HughesInstitute for Computational Engineering and Sciences, The Universityof Texas at Austin, 201 East 24th St, Stop C0200, Austin, TX 78712,USA

Y. BazilevsDepartment of Structural Engineering, University of California, SanDiego, 9500 Gilman Drive, Mail Code 0085, La Jolla, CA 92093, USA

ally [1, 2]. Bioprosthetic heart valves (BHV) are prostheticscomposed of thin flexible leaflets that are fabricated frombiological materials and mimic the structure of native heartvalves to avoid pathological hemodynamics [2]. The prin-cipal drawback of this style of prosthetic is its durability,which is limited to 10–15 years [3]. Accurate computationalanalysis of these devices could provide insights into the me-chanical processes that both contribute to and follow fromtheir deterioration, streamlining the design process of newprosthetics.

The biomechanical significance of arterial elasticity wasfirst clearly described by Hales [4] in 1733, after performinga series of pioneering experiments on animals. Hales foundthat arteries expand elastically to store the systolic output ofthe heart, then gradually release this blood during diastole.This is now known as the Windkessel effect.1 Frank [5–7]developed the first mathematical model of the Windkesseleffect in 1899. Frank’s model may be intuitively understoodthrough the electronic–hydraulic analogy [8], which substi-tutes electrical current for volumetric flow and voltage forpressure. In this analogy, Frank’s model—the two-elementWindkessel model—consists of a capacitor and resistor inparallel, downstream of the aortic valve, which acts as a pul-satile current source.

The capacitor models the elastic arteries, which accu-mulate blood to develop pressure, while the resistor mod-els viscous head loss within the circulatory system by anal-ogy to Ohm’s Law. This model allows prediction of thetime-dependent aortic pressure based on the history of flowrate through the aortic valve. Many refinements to Frank’smodel have been proposed since his initial contribution, in-cluding the three- [9], and four- [10] element Windkessel

1 Windkessel translates from German to “air chamber”, and likelyrefers to Hales’ original analogy between arterial compliance and theair-filled cavities used to smooth hose output from 18th-century fireengines.

2 Ming-Chen Hsu et al.

models. Such models are referred to as “lumped-parametermodels”. Lumped-parameter models may be coupled withdetailed computational fluid dynamics (CFD) simulationsof specific arterial sections of interest. The voltage of thelumped-parameter model acts as a pressure boundary condi-tion on the outflow of the CFD domain, and the volumetricflow from the CFD domain acts as a current source for thelumped-parameter model [11]. However, to fully account forthe Windkessel effect of arterial elasticity, fluid–structure in-teraction (FSI) must be incorporated into the detailed modelof the section of interest. In this paper, we demonstrate thatthe elasticity of the section of aorta immediately surround-ing an implanted BHV can have profound effects on the dy-namics of both the valve itself and the surrounding bloodflow.

For the reasons discussed in our earlier work [12,13], wesimulate the BHV leaflets using a non-boundary-fitted (vari-ational immersed-boundary) method, in which the structuraldiscretization is free to move independently through a back-ground fluid mesh. Detailed reviews of non-boundary-fittedmethods for FSI can be found in Sotiropoulos and Yang [14],Mittal and Iaccarino [15], and Peskin [16]. These meth-ods are particularly attractive for applications with complexmoving boundaries, such as heart valve leaflets [17–22].However, they have the inherent disadvantage of uncon-trolled mesh quality near the fluid–structure interface, andmay be unable to resolve important boundary layer featuresthat may globally affect the flow.

More accurate results can be obtained using boundary-fitted approaches by building a fluid mesh that is tailored tothe structure and deforms as the structure moves. In suchcomputations, the fluid subproblem may be posed using anarbitrary Lagrangian–Eulerian (ALE) formulation [23–25],or a space–time formulation [26–28], both of which explic-itly account for the motion of the fluid mechanics domainand mesh. For the relatively mild deformations of a sectionof artery, the problem of mesh deformation may be effec-tively solved using a simple fictitious linear elasticity prob-lem [29–32]. This makes vascular FSI an ideal applicationfor boundary-fitted approaches.

In the analysis of a BHV implanted in a deformingartery, we are faced with the confluence of two problemsthat suggest different computational methods. We thereforeelect to use a hybrid method that leverages the advantages ofboth ALE and immersed-boundary techniques for FSI. Wediscretize the valve leaflets separately, and immerse theminto a deforming boundary-fitted mesh of the artery vol-ume. The proposed technique falls under the umbrella of therecently proposed Fluid–Solid Interface-Tracking/Interface-Capturing (FSITIC) technique [33], a method that targetsFSI problems in which interfaces that are possible to trackare tracked, and those too difficult to track are captured. Re-cently Wick [34,35] made use of the FSITIC approach, cou-

pling a boundary-fitted and immersed-boundary discretiza-tions in a single computation, to compute several 2D FSIbenchmark problems.

Our immersed-boundary approach for FSI was firstdeveloped in Kamensky et al. [12, 13] using the varia-tional framework of augmented Lagrangian methods. Theaugmented Lagrangian approach for FSI was proposed inBazilevs et al. [36] to handle boundary-fitted computa-tions with non-matching fluid–structure interface discretiza-tions. We found in Kamensky et al. [12, 13] that this aug-mented Lagrangian framework can be extended to handlenon-boundary-fitted CFD and FSI problems, and its efficacywas demonstrated using several computations including thecoupling of a BHV and surrounding blood flow at physio-logical pressure levels.

In this work, we take the augmented Lagrangian frame-work for FSI as the starting point of our ALE/immersed-boundary hybrid methodology. A single computation com-bines a boundary-fitted, deforming-mesh treatment of somefluid–structure interfaces with a non-boundary-fitted treat-ment of others. This approach enables us to simulate the FSIof a BHV implanted in an elastic artery through the entirecardiac cycle, at full scale, under realistic physiological con-ditions.

The paper is organized as follows. In Section 2 wepresent the details of our hybrid ALE/immersed-boundarymethod developed for the FSI simulation of a heart valve im-planted in a deformable artery. In Section 3 we provide thesimulation details and report the results of the FSI compu-tations of an actual BHV design. In particular, we comparethe results from the rigid- and elastic-wall simulations andfind that wall elasticity plays an important role in the overallsystem response. In Section 4 we draw conclusions.

2 FSI modeling using a hybridALE/immersed-boundary approach

In this section, we present the computational framework forFSI analysis of a bioprosthetic heart valve implanted in adeformable artery. The blood flow in a deforming artery isgoverned by the Navier–Stokes equations of incompress-ible flow posed on a moving domain. The domain motionis handled using the arbitrary Lagrangian–Eulerian (ALE)formulation, which is a widely used approach for vascularblood flow applications [37–42]. The coupling of the BHVleaflet dynamics to the artery is handled through the recentlyproposed variational immersed-boundary method [12, 13],in which the structural discretization is free to move in-dependently through a background fluid mesh. The hybridALE/immersed-boundary method will be presented and ap-plied to the simulation of an aortic BHV coupled to an elas-tic arterial wall and blood flow over cardiac cycles.

Fluid–structure interaction analysis of bioprosthetic heart valves: Significance of arterial wall deformation 3

2.1 Augmented Lagrangian framework for FSI

Let (Ω1)t and (Ω2)t ∈ Rd, d = 2, 3 represent the time-

dependent domains of the fluid and structural mechanicsproblems, respectively, at time t, with (Γ1)t and (Γ2)t repre-senting their corresponding boundaries. Let (ΓI)t ∈ R

d rep-resent the interface between the fluid and structural domains.Let u1 and p denote the fluid velocity and pressure, respec-tively. Let y denote the displacement of structural materialpoints from their positions in a reference configuration, anddefine the structure velocity u2 as the material time deriva-tive of y. We introduce an additional unknown function λλλdefined on (ΓI)t, which takes on the interpretation of a La-grange multiplier. Let Su, Sp, Sd, and S` be the functionspaces for the fluid velocity, fluid pressure, structural ve-locity, and Lagrange multiplier solutions, respectively, andVu, Vp, Vd, and V` be the corresponding weighting func-tion spaces. The variational problem of the augmented La-grangian formulation is: find u1 ∈ Su, p ∈ Sp, y ∈ Sd, andλλλ ∈ S` such that for all test functions w1 ∈ Vu, q ∈ Vp,w2 ∈ Vd, and δλλλ ∈ V`

B1(w1, q, u1, p; u) − F1(w1, q)

+

∫(ΓI)t

w1 · λλλ dΓ +

∫(ΓI)t

w1 · β(u1 − u2) dΓ = 0, (1)

B2(w2, y) − F2(w2)

−

∫(ΓI)t

w2 · λλλ dΓ −∫

(ΓI)t

w2 · β(u1 − u2) dΓ = 0, (2)

∫(ΓI)t

δλλλ · (u1 − u2) dΓ = 0. (3)

In the above, the subscripts 1 and 2 denote the fluid andstructural mechanics quantities, and β is a penalty parame-ter, which we leave unspecified for the moment. B1, B2, F1,and F2 are the semi-linear forms and linear functionals cor-responding to the fluid and structural mechanics problems,respectively, and are given by

B1(w, q, u, p; u) =

∫(Ω1)t

w · ρ1

(∂u∂t

∣∣∣∣∣x

+ (u − u) · ∇∇∇u)

dΩ

+

∫(Ω1)t

εεε(w) : σσσ1 dΩ +

∫(Ω1)t

q∇∇∇ · u dΩ,

(4)

F1(w, q) =

∫(Ω1)t

w · ρ1f1 dΩ +

∫(Γ1h)t

w · h1 dΩ, (5)

B2(w, y) =

∫(Ω2)t

w · ρ2∂2y∂t2

∣∣∣∣∣∣X

dΩ +

∫(Ω2)t

εεε(w) : σσσ2 dΩ, (6)

F2(w) =

∫(Ω2)t

w · ρ2f2 dΩ +

∫(Γ2h)t

w · h2 dΩ, (7)

where ρ1 and ρ2 are the densities,σσσ1 andσσσ2 are the Cauchystresses, f1 and f2 are the applied body forces, h1 and h2 arethe applied surface tractions, (Γ1h)t and (Γ2h)t are the bound-aries where the surface tractions are specified, εεε(·) is thesymmetric gradient operator given by εεε(w) = 1

2 (∇∇∇w+∇∇∇wT ),

u is the velocity of the fluid domain (Ω1)t,∂(·)∂t

∣∣∣∣∣x

is the time

derivative taken with respect to the fixed spatial coordinate xin the referential domain (which does not follow the motion

of the fluid itself), and∂(·)∂t

∣∣∣∣∣X

is the time derivative hold-

ing the material coordinates X fixed. The gradient∇∇∇ is takenwith respect to the spatial coordinate x of the current config-uration. We assume that the fluid is Newtonian with dynamicviscosity µ and Cauchy stress σσσ1 = −pI + 2µεεε(u1).

Bazilevs et al. [36] demonstrate how the Lagrange mul-tiplier, λλλ, may be formally eliminated by substituting an ex-pression for the fluid–structure interface traction in terms ofthe other unknowns. This leads to the following variationalformulation for the coupled problem: find u1 ∈ Su, p ∈ Sp,and y ∈ Sd such that for all w1 ∈ Vu, q ∈ Vp, and w2 ∈ Vd

B1(w1, q, u1, p; u) − F1(w1, q) + B2(w2, y) − F2(w2)

−

∫(ΓI)t

(w1 − w2) ·σσσ1(u1, p) n1 dΓ

−

∫(ΓI)t

δσσσ1(w1, q) n1 · (u1 − u2) dΓ

+

∫(ΓI)t

(w1 − w2) · β(u1 − u2) dΓ = 0. (8)

In the above, δσσσ1(w, q)n1 = 2µεεε(w)n1 + qn1. Equation (8)may be interpreted as an extension of Nitsche’s method [43],which is a consistent, stabilized method for imposing con-straints on the boundaries by augmenting the governingequations with additional constraint equations.

This augmented Lagrangian approach for FSI was origi-nally proposed by Bazilevs et al. [36] and further studied inHsu and Bazilevs [44] to handle boundary-fitted computa-tions with non-matching fluid–structure interface discretiza-tions. In Kamensky et al. [12, 13], we found that this frame-work can be extended to handle non-boundary-fitted FSIproblems and the accuracy and efficiency of the method-ology was examined through several computations. In thiswork, we take the augmented Lagrangian framework for FSIas the starting point of our hybrid ALE/immersed-boundarymethod. A single computation can combine a boundary-fitted, deforming-mesh treatment of some fluid–structure in-terfaces with a non-boundary-fitted treatment of others.

Remark 1 In the above developments we assumed that thetrial and test function spaces of the fluid and structural sub-problems are independent of each other. This approach pro-vides one with the framework that is capable of handling


non-matching fluid and structural interface discretizations.If the fluid and structural velocities and the test functionsare explicitly assumed to be continuous (i.e. u1 = u2 andw1 = w2) at the interface, the FSI formulation given byEq. (8) reduces to: find u1 ∈ Su, p ∈ Sp, and y ∈ Sd suchthat for all w1 ∈ Vu, q1 ∈ Vp, and w2 ∈ Vd

B1(w1, q, u1, p; u)−F1(w1, q)+ B2(w2, y)−F2(w2) = 0.

(9)

This form of the FSI problem is suitable for matching fluid–structure interface meshes. Although somewhat limiting,matching interface discretizations were successfully appliedto cardiovascular FSI in many earlier works [32, 40, 45–50].

2.2 Semi-discrete fluid formulation with weak boundaryconditions

The fluid subproblem may be obtained by setting w2 = 0in Eq. (8). This approach gives a formulation for weak im-position of Dirichlet boundary conditions for the fluid prob-lem, which was first proposed by Bazilevs and Hughes [51]and further refined in Bazilevs et al. [52, 53] to improve theperformance of the fluid mechanics formulation in the pres-ence of underresolved boundary layers. This weak imposi-tion of the Dirichlet boundary conditions is also the start-ing point of the variational immersed-boundary approach[12, 13]. In a non-boundary-fitted method, the elements ofthe fluid discretization may extend into the interior of animmersed object. Imposing Dirichlet boundary conditionsis no longer straightforward given that the basis functionsare non-interpolating at the object boundaries. In order toenforce essential boundary conditions, one can either mod-ify the basis functions so they vanish at the interface [54] oraugment the governing equations with additional constraintequations. In this work we choose the latter approach.

Consider a collection of disjoint elements Ωe, ∪eΩe ⊂

Rd, with closures covering the fluid domain: Ω1 ⊂ ∪eΩe.Note that Ωe is not necessarily a subset of Ω1. Ωe, Ω1, andΓI remain time-dependent, but we drop the subscript t fornotational convenience. The mesh defined by Ωe deformswith a velocity field uh and the boundary ΓI moves with ve-locity u2. The semi-discrete fluid problem is given by: finduh

1 ∈ Shu and ph ∈ Sh

p such that for all wh1 ∈ V

hu and qh ∈ Vh

p

BVMS1

(wh

1, qh, uh

1, ph; uh)− FVMS

1

(wh

1, qh)

−

∫ΓI

wh1 ·

(−phn1 + 2µεεε(uh

1)n1

)dΓ

−

∫ΓI

(2µεεε(wh

1)n1 + qhn1

)·(uh

1 − u2

)dΓ

−

∫(ΓI)−

wh1 · ρ1

((uh

1 − uh)· n1

) (uh

1 − u2

)dΓ

+

∫ΓI

τBTAN

(wh

1 −(wh

1 · n1

)n1

)·((

uh1 − u2

)−

((uh

1 − u2

)· n1

)n1

)dΓ

+

∫ΓI

τBNOR

(wh

1 · n1

) ((uh

1 − u2

)· n1

)dΓ = 0, (10)

where (ΓI)− is the “inflow” part of ΓI, on which (uh1 − uh) ·

n1 < 0, the constants τBTAN and τB

NOR correspond to a split-ting of the penalty term into the tangential and normal di-rections, respectively, and ΓI may cut through element inte-riors. The discrete trial function spaces Sh

u for the velocityand Sh

p for the pressure, as well as the corresponding testfunction spaces Vh

u and Vhp are assumed to be equal order,

and, in this work, are comprised of NURBS functions. Theforms BVMS

1 and FVMS1 are the variational multiscale (VMS)

discretizations of B1 and F1, respectively, given by

BVMS1 (w, q, u, p; u)

=

∫(Ω1)t

w · ρ1

(∂u∂t

∣∣∣∣∣x

+ (u − u) · ∇∇∇u)

dΩ

+

∫(Ω1)t

εεε(w) : σσσ1 dΩ +

∫(Ω1)t

q∇∇∇ · u dΩ

+∑

e

∫Ωe∩Ω1

((u − u) · ∇∇∇w +

∇∇∇qρ1

)· u′ dΩ

+∑

e

∫Ωe∩Ω1

∇∇∇ · w ρ1 p′ dΩ

−∑

e

∫Ωe∩Ω1

w · (u′ · ∇∇∇u) dΩ

−∑

e

∫Ωe∩Ω1

∇∇∇wρ1

:(u′ ⊗ u′

)dΩ

+∑

e

∫Ωe∩Ω1

(u′ · ∇∇∇w

)τ ·

(u′ · ∇∇∇u

)dΩ, (11)

and

FVMS1 (w, q) = F1(w, q), (12)

where

u′ = τM

(ρ1

(∂u∂t

∣∣∣∣∣x

+ (u − u) · ∇∇∇u − f)−∇∇∇ ·σσσ1

), (13)

p′ = τC∇∇∇ · u. (14)

Equations (11)–(14) correspond to the ALE–VMS formula-tion of the Navier–Stokes equations of incompressible flows[55–57]. The additional terms may be interpreted both as


stabilization and as a turbulence model [58–64]. The stabi-lization parameters are

τM =

(s(x, t)

( Ct

∆t2 + (u − u) ·G(u − u) + CIν2G : G

))− 12

,

(15)

τC = (τM tr G)−1 , (16)

τ =(u′ ·Gu′

)− 12 , (17)

where ∆t is the time-step size, ν = µ/ρ1 is the kinematicviscosity, CI is a positive constant derived from an appropri-ate element-wise inverse estimate [65–68], G is the elementmetric tensor defined as

G =∂ξξξ

∂x

T ∂ξξξ

∂x, (18)

where ∂ξξξ/∂x is the inverse Jacobian of the element map-ping between the parametric and physical domain, tr G isthe trace of G, and the parameter Ct is typically taken equalto 4 [60, 63, 69]. The scalar function s(x, t) ≥ 1 in Eq. (15)is a dimensionless scaling factor introduced in Kamensky etal. [12,13] to improve local mass conservation near concen-trated loads. Locally increasing s near thin immersed struc-tures can greatly improve the quality of approximate solu-tions when the concentrated surface force due to the struc-ture induces a significant pressure discontinuity. In most ofthe domain, we keep s = 1, as in the usual VMS formu-lation, but, in an O(h) neighborhood around thin immersedstructures, we increase it to equal the dimensionless constantsshell ≥ 1.

Remark 2 On the fluid mechanics domain interior, the meshvelocity, uh, may be obtained by solving a linear elasto-statics problem subject to the displacement boundary condi-tions coming from the motion of the boundary-fitted fluid–solid interface [29–32]. This method is effective for rela-tively mild deformations, such as those of the artery. How-ever, for scenarios that involve large translational and/or ro-tational structural motions, such as heart valve dynamics, theboundary-fitted fluid mesh can become severely distorted.Non-boundary-fitted approaches could be an alternative forthis type of problem.

The terms from the second to the last line of Eq. (10)are responsible for the weak enforcement of kinematic andtraction constraints at the non-matching or immersed bound-aries. It was shown in earlier works [51–53, 70, 71] that im-posing the Dirichlet boundary conditions weakly in fluid dy-namics allows the flow to slip on the solid surface when thewall-normal mesh size is relatively large. This effect mim-ics the thin boundary layer that would otherwise need tobe resolved with spatial refinement, allowing more accu-rate solutions on coarse meshes. In the immersed-boundarymethod, the fluid mesh is arbitrarily cut by the structural

boundary, leaving a boundary layer discretization of infe-rior quality compared to the body-fitted case. Therefore, inaddition to imposing the constraints easily in the context ofnon-boundary-fitted approach, we may obtain more accuratefluid solutions as an added benefit of using the weak bound-ary condition formulation (10).

In Eq. (10), the parameters τBTAN and τB

NOR must be suffi-ciently large to stabilize the formulation, but not so large asto degenerate Nitsche’s method into a pure penalty method.Based on previous studies of weakly-enforced Dirichletboundary conditions in fluid mechanics [51–53], we expectthese parameters to scale as

τB(·) =

CBI µ

h(19)

where h is a measure of the element size at the boundaryand CB

I is a dimensionless constant. However, in the case ofan immersed boundary, neither the appropriate definition ofh nor the principle for deriving CB

I is straightforward. As aresult, we chose the penalty-parameter values through nu-merical experiments.

Integrating the fluid formulation (11) over elements thatare only partially contained in Ω1 typically requires spe-cial quadrature techniques, as discussed in Kamensky et al.[12, 13]. In the present work, we do not need these quadra-ture techniques, because fluid elements only overlap spa-tially with thin shell structures, which are modeled geomet-rically as (d − 1)-dimensional surfaces and therefore havezero Lebesgue measure in Rd. To evaluate the surface in-tegrals of Eq. (10) over immersed boundaries, we define aGaussian quadrature rule with respect to a parameterizationof the immersed surface, then locate the quadrature points ofthis rule in the parameter space of the background mesh el-ements to evaluate traces of the fluid test and trial functions.

Unsteady flow computations may sometimes divergedue to flow reversal on outflow boundaries. This is knownas backflow divergence and is frequently encountered incardiovascular simulations. In order to preclude backflowdivergence, an outflow stabilization method originally pro-posed in Bazilevs et al. [46] and further studied in Esmaily-Moghadam et al. [72] is employed in our fluid mechanicsformulation.

2.3 Arterial wall modeling

In this section we show the variational formulation of theboundary-fitted solid problem for the arterial wall modeling.The fluid–solid interface discretization is assumed to be con-forming. Let X be the coordinates of the initial or referenceconfiguration and let y be the displacement with respect tothe reference configuration. The coordinates of the currentconfiguration, x, are given by x = X + y. The deformation


gradient tensor F is defined as

F =∂x∂X

= I +∂y∂X

, (20)

where I is the identity tensor.Let Sd andVd be the trial solution and weighting func-

tion spaces for the solid problem. The arterial wall is mod-eled as a three-dimensional hyperelastic solid and the vari-ational formulation which represents the balance of linearmomentum for the solid is stated as follows: find the dis-placement y ∈ Sd, such that for all weighting functionsw2 ∈ Vd

B2(w2, y) − F2(w2) = 0, (21)

where

B2(w, y) =

∫(Ω2)t

w · ρ2∂2y∂t2

∣∣∣∣∣∣X

dΩ +

∫(Ω2)0

∇∇∇Xw : P dΩ, (22)

F2(w) =

∫(Ω2)t

w · ρ2f2 dΩ +

∫(Γ2h)t

w · h2 dΩ. (23)

In the above, (Ω2)0 is the solid domain in the reference con-figuration, ∇∇∇X is the gradient operator on (Ω2)0, and P = FSis the first Piola–Kirchhoff stress tensor, where S is the sec-ond Piola–Kirchhoff stress tensor given by

S = µJ−2/3(I −

13

tr C C−1)

+12κ(J2 − 1

)C−1. (24)

In Eq. (24), µ and κ are interpreted as the blood vessel shearand bulk moduli, respectively, J = det F is the Jacobian de-terminant, and C = FT F is the Cauchy–Green deformationtensor. Equation (24) is a generalized neo-Hookean modelwith dilatational penalty given in Simo and Hughes [73].Its stress-strain behavior was analytically studied on simplecases of uniaxial strain [32] and pure shear [74]. The modelwas argued in Bazilevs et al. [40] to be appropriate for arte-rial wall modeling in FSI simulations. It was shown that thelevel of elastic strain in arterial FSI problems is sufficientlylarge to preclude the use of infinitesimal (linear) strains, yetnot large enough to be sensitive to the nonlinearity of theparticular material model. However, the current model hasthe advantage of stable behavior for the regime of strongcompression and therefore is selected in this work for themodeling of the arterial wall.

2.4 Immersed shell structures

We model the heart valve as a shell structure immersedinto a deforming background mesh covering the lumen ofthe artery. The exact solution for the pressure around ashell structure may be discontinuous at the structure, which

presents a conceptual difficulty. The fluid discretization can-not be informed by the structure’s position. This means thatour fluid approximation space cannot be selected in such away that the pressure basis functions are themselves discon-tinuous at the immersed boundary. This implies an inher-ent approximation error in the pressure field. This error willconverge slowly for polynomial bases [75]. Nonetheless, webelieve that solutions of sufficient accuracy for engineeringpurposes can be obtained in this fashion and we focus ondeveloping a robust method for obtaining these solutions.

2.4.1 Reduction of Nitsche’s method to the penalty method

Consider integrating the boundary terms of Eq. (10) overboth sides of a thin immersed shell structure. If the velocityand pressure approximation spaces are continuous throughthe vanishing thickness of the shell (and the velocity ap-proximation space is continuously differentiable), then thedependence of the consistency and adjoint consistency termson the normal vector will cause contributions from opposingsides to cancel one another. The only remaining terms willbe the penalty and the inflow stabilization. In the case of animmersed shell structure, we may view the inflow term asa velocity-dependent penalty. The Nitsche-type formulationgiven by Eq. (10) therefore reduces to the following penaltymethod:

BVMS1

(wh

1, qh, uh

1, ph; uh)− FVMS

1

(wh

1, qh)

−

∫(ΓI)−

wh1 · ρ1

((uh

1 − uh)· n1

) (uh

1 − u2

)dΓ

+

∫ΓI

τBTAN

(wh

1 −(wh

1 · n1

)n1

)·((

uh1 − u2

)−

((uh

1 − u2

)· n1

)n1

)dΓ

+

∫ΓI

τBNOR

(wh

1 · n1

) ((uh

1 − u2

)· n1

)dΓ = 0, (25)

when the approximation spaces Vhu and Vh

p are sufficientlyregular around the shell.

To determine the velocity and pressure about an im-mersed valve in its closed state, a method must be capableof developing nearly hydrostatic solutions in the presenceof large pressure gradients. Penalty forces will only exist ifthere are nonzero violations of kinematic constraints. A purepenalty method rules out the desired hydrostatic solutions:every term that could resist the pressure gradient to satisfybalance of linear momentum depends on velocity. Increas-ing β may diminish leakage through a structure, but it is awell-known disadvantage of penalty methods that extremevalues of penalty parameters will adversely affect the nu-merical solvability of the resulting problem. This motivatesus to return to Eqs. (1)–(3) and develop a method that doesnot formally eliminate the multiplier field.


2.4.2 Reintroducing the multipliers

Since the introduction of constraints tends to make discreteproblems more difficult to solve, we will only reintroduce ascalar multiplier field to strengthen enforcement of the no-penetration part of the FSI kinematic constraint, rather thanthe vector-valued multiplier field of Eqs. (1)–(3). The vis-cous, tangential component of the constraint will continue tobe enforced by only the penalty τB

TAN. This may be thoughtof as a formal elimination of just the tangential componentof the multiplier field, which also retains the ability to al-low the flow to slip at the boundary, which tends to producemore accurate fluid solutions as discussed in Section 2.2.For clarity, we redefine the FSI boundary terms on the mid-surface of the shell structure, Γt, rather than considering thefull boundary, ΓI. This means that constants in the currentformulation may differ from those of Eqs. (1)–(3) by factorsof two. We arrive, then, at the formulation

B1(w1, q, u1, p; u) − F1(w1, q)

+

∫Γt

w1 · (λnn2) dΓ +

∫Γt

w1 · β(u1 − u2) dΓ = 0, (26)

B2(w2, y) − F2(w2)

−

∫Γt

w2 · (λnn2) dΓ −∫Γt

w2 · β(u1 − u2) dΓ = 0, (27)

∫Γt

δλnn2 · (u1 − u2) dΓ = 0, (28)

where λn is the new scalar multiplier field and, to empha-size the relation to Eqs. (1)–(3), the penalty force has notbeen split into normal and tangential components. The con-sistency and adjoint consistency terms associated with elim-inating the tangential component of the multiplier have beenomitted under the assumption that they will vanish after in-tegrating over both sides of the thin shell, as discussed inSection 2.4.1.

We discretize the multiplier field by collocating kine-matic constraints at points of the quadrature rule for inte-grals over Γt. This entails adding a scalar multiplier un-known at each quadrature point. In discrete evaluations ofintegrals, these multiplier unknowns are treated like pointvalues of a function defined on Γt. Because the spatial res-olution of the discrete multiplier representation is not con-trolled relative to the background fluid mesh, we must relaxthe collocated constraints to ensure stability of the numericalscheme. We accomplish this through the time-discrete algo-rithm given in Section 2.5. The algorithmic constraint relax-ation is interpreted at the time-continuous level by Kamen-sky et al. [12,13], through an analogy to Chorin’s method ofartificial compressibility [76], in which the Lagrange multi-plier solves an auxiliary differential equation in time.

2.4.3 Treatment of shell structure mechanics

We assume that the structure is a thin shell, representedmathematically by its mid-surface. Further, we assumethis surface to be piecewise C1-continuous and apply theKirchhoff–Love shell formulation and isogeometric dis-cretization studied by Kiendl et al. [77–79]. The spatial co-ordinates of the shell mid-surface in the reference and cur-rent configurations are given by X(ξ1, ξ2) and x(ξ1, ξ2), re-spectively, parameterized by ξ1 and ξ2. Assuming the range1, 2 for Greek letter indices, we define the covariant sur-face basis vectors

gα =∂x∂ξα

, (29)

g3 =g1 × g2

||g1 × g2||, (30)

and

Gα =∂X∂ξα

, (31)

G3 =G1 ×G2

||G1 ×G2||, (32)

in the current and reference configurations, respectively. Us-ing kinematic assumptions and mathematical manipulationsgiven in Kiendl [79], we split the in-plane Green–Lagrangestrain Eαβ into membrane and curvature contributions

Eαβ = εαβ + ξ3καβ, (33)

where

εαβ =12

(gα · gβ −Gα ·Gβ

), (34)

καβ =∂Gα

∂ξβ·G3 −

∂gα∂ξβ· g3, (35)

are the membrane strain and change of curvature tensors,respectively, at the shell mid-surface. In Eq. (33), ξ3 ∈

[−hth/2, hth/2] is the through-thickness coordinate and hth

is the shell thickness. The forms B2 and F2 appearing in thestructure subproblem then become, in the case of a thin shellstructure,

B2(w, y) =

∫Γt

w · ρ2hth∂2y∂t2

∣∣∣∣∣∣X

dΓ +

∫Γ0

∫hth

δE : S dξ3dΓ,

(36)

F2(w) =

∫Γt

w · ρ2hthf dΓ +

∫Γt

w · hnet dΓ, (37)

where S is the second Piola–Kirchhoff stress, δE is the vari-ation of the Green–Lagrange strain, Γ0 and Γt are the shellmid-surface in the reference and deformed configurations,respectively, hnet = h(ξ3 = −hth/2) + h(ξ3 = hth/2) sumstraction contributions from the two sides of the shell. For the


purposes of this paper, we assume a St. Venant–Kirchhoff

material, in which S is computed from a constant elastic-ity tensor, , applied to E. For isotropic materials, the con-stitutive material tensor may be derived from a Young’smodulus, E, and Poisson ratio, ν, and the integral over ξ3

in Eq. (36) can be computed analytically. The St. Venant–Kirchhoff material model can become unstable when sub-jected to strongly compressive stress states [80], but suchstates are not encountered in the present application, becausetransverse normal stress is ignored by the thin-shell formu-lation and in-plane stresses within heart valve leaflets areprimarily tensile.

Isogeometric analysis [81,82] is employed for modelingthe shell structure. We use C1-continuous quadratic B-splinefunctions to represent both the geometry and displacementsolution field. The details of this discretization are given inKiendl et al. [77–79]. A noteworthy aspect of this discretiza-tion is the fact that it requires no rotational degrees of free-dom. The C1-continuous approximation space (for a singlepatch) is in H2, so we may directly apply Galerkin’s methodto the forms defined in Eqs. (36) and (37).

2.5 Time integration and FSI solution strategy

We complete the discretization of the coupled FSI for-mulation by using finite differences to approximate thetime derivatives appearing therein. In particular, we employthe Generalized-α technique [32, 83, 84], which is a fully-implicit second-order accurate method with control over thedissipation of high-frequency modes. This produces a non-linear algebraic system of equations relating the unknowncoefficients of the fluid, solid structure, mesh-movement,shell structure, and multiplier solutions at time level tn+1

to the known solutions from time level tn. An attempt tosolve this system with a monolithic approach (e.g., by New-ton’s iteration with a consistent tangent) would encounterthe following difficulties: 1) The sparsity pattern of the non-linear residual’s Jacobian matrix would change as the im-mersed shell structure moves through the background mesh.2) Fluid, structure, and mesh solvers would become moredifficult to interchange. 3) The potential for drastically-different multiplier and fluid resolutions could lead to in-stability.

To circumvent the third issue, at each time step, we com-pute the solution using the following two-step procedure:

1. Solve for the fluid, solid structure, mesh displacement,and shell structure unknowns, holding λn fixed. Note thatthe fluid and shell structure are still coupled in this prob-lem, due to the penalty term.

2. Update the multiplier λn, by adding the normal compo-nent of penalty forces present in the solution from Step1.

The solution from Step 1 will not satisfy the kinematic con-straints exactly at all quadrature points on Γt. This is a de-liberate weakening of the constraints to improve stability, asmentioned in Section 2.4.2. The two-step solution proceduremay be interpreted as penalization of an implicitly-evaluatedtime integral of the velocity difference between the fluidand shell structure, as detailed in Kamensky et al. [12, 13],and is conceptually-similar to the method of artificial com-pressibility [76] for incompressible flow problems. Note thatthe time integral of the velocity difference is a displace-ment: we effectively implement spring-like sliding contactelements between the fluid and shell structure. This preventsthe steady creeping flow through shell structures that can oc-cur when only the current velocity difference is penalized, asin the penalty approach coming from Nitsche’s method.

To solve the nonlinear coupled problem in Step 1, weapply a fixed-point iteration based on Newton’s method. Thelinear system to be solved within each iteration of Newton’smethod would have the form

∂Rfl

∂Ufl

∂Rfl

∂Uso

∂Rfl

∂Ume

∂Rfl

∂Ush

∂Rso

∂Ufl

∂Rso

∂Uso

∂Rso

∂Ume

∂Rso

∂Ush

∂Rme

∂Ufl

∂Rme

∂Uso

∂Rme

∂Ume

∂Rme

∂Ush

∂Rsh

∂Ufl

∂Rsh

∂Uso

∂Rsh

∂Ume

∂Rsh

∂Ush

∆Ufl

∆Uso

∆Ume

∆Ush

= −

Rfl

Rso

Rme

Rsh

, (38)

where R(·) and U(·) are the nonlinear residuals and discreteunknowns of the fluid (fl), solid structure (so), mesh (me),and shell structure (sh). ∆U(·) are the corresponding solu-tion increments. To avoid the aforementioned disadvantagesof assembling the full consistent tangent, we approximate itwith the block-diagonal matrix

∂Rfl

∂Ufl

∂Rfl

∂Uso0 0

∂Rso

∂Ufl

∂Rso

∂Uso0 0

0 0∂Rme

∂Ume0

0 0 0∂Rsh

∂Ush

, (39)

then assemble and solve each block of equations in se-quence (from top to bottom). We use a number of furtherapproximations within each of the left-hand side blocks, butmaintain the original nonlinear residuals, R(·), of the fully-coupled problem. Converging these residuals to zero solvesthe original problem, regardless of any approximations usedin the tangent matrix. The procedure that we apply at eachstep of the fixed-point iteration is not equivalent to a linear


leftventricle

aortic sinus

ascendingaorta

aortic valve leaflets

artery wall

Fig. 1: A schematic drawing illustrating the position of theaortic valve relative to the left ventricle of the heart and theascending aorta.

Fig. 2: B-spline heart valve mesh comprised of 1,404quadratic elements. The pinned boundary condition is ap-plied to the leaflet attachment edge.

solve with matrix (39). To accelerate convergence, we usethe updated solutions from previous blocks to assemble theequations for subsequent ones. We repeat this fixed-pointiteration to converge R(·) toward zero and obtain a fully-coupled solution of the fluid-solid-mesh-shell system. Inpractice, we use a fixed number of iterations, chosen to yieldtypically-satisfactory convergence at the selected time stepsize. This algorithm combines the quasi-direct and block-iterative FSI coupling approaches outlined in Tezduyar etal. [85–87] and Bazilevs et al. [88].

3 Bioprosthetic heart valve simulations

In this section, we use the proposed hybrid ALE/immersed-boundary method to simulate the FSI of an aortic BHV im-planted in an elastic artery over cardiac cycles. The aorticvalve regulates flow between the left ventricle of the heartand the ascending aorta. Figure 1 provides a schematic de-piction of its position in relation to the surrounding anatomy.The valve leaflets are discretized separately and immersed

n1

x1

x2

d

S1

S2

n2

Fig. 3: Illustration of contact notation.

into a deforming boundary-fitted background mesh of theartery lumen.

3.1 Heart valve model

The BHV leaflet geometry used in this study is based ona 23-mm design by Edwards Lifesciences. We model eachleaflet using a C1-continuous B-spline patch, which com-prises 468 quadratic B-spline elements. The pinned bound-ary condition is applied to the leaflet attachment edge asshown in Figure 2. An isotropic St. Venant–Kirchhoff ma-terial with E = 107 dyn/cm2 and ν = 0.45 is applied to theBHV. The thickness and density of the leaflets are 0.0386 cmand 1.0 g/cm3, respectively. There is no damping applied tothe valve dynamics in this study.

3.2 Leaflet–leaflet contact

Contact between leaflets is an essential feature of a func-tioning heart valve. BHV leaflets contact one another dur-ing the opening, and especially during the closing to blockflow. An advantage of immersed-boundary methods for FSIis that pre-existing contact algorithms from structural anal-ysis [89–93] may be incorporated directly into the struc-tural subproblem without affecting the fluid subproblem. Weadopt a penalty-based approach for sliding contact and em-ploy contact elements associated with the quadrature pointsof the shell structure.

As detailed in Kamensky et al. [12, 13], a contact ele-ment activates when its associated quadrature point, locatedon a particular BHV leaflet designated S 1, is found to pen-etrate through another leaflet, designated S 2. Penalties arecomputed using a signed distance, d, from S 2 to the quadra-ture point on S 1, and their activation is controlled by sev-eral geometrical conditions omitted from the current paperfor brevity. Opposing concentrated loads are applied at thequadrature points on S 1 and their closest points on S 2. This


Fig. 4 A view of the arterial walland lumen into which the valveis immersed.

notation is illustrated for a pair of contacting points in Fig-ure 3. The designation of one leaflet as S 1 and another asS 2 is arbitrary, and to preserve geometrical symmetries, wesum the forces resulting from both choices.

3.3 Artery model

The BHV model mentioned earlier is immersed into apressure-driven incompressible flow through a deformableartery. The fluid density and viscosity are ρ1 = 1.0 g/cm3

and µ = 3.0 × 10−2 g/(cm s), respectively, which model thephysical properties of human blood.

The artery is modeled as a 16 cm long elastic cylindri-cal tube with a three-lobed dilation near the BHV, as shownin Figure 4. This dilation represents the aortic sinus, whichis known to play an important role in heart valve dynam-ics [94]. The cylindrical portion of the artery has an insidediameter of 2.3 cm and a thickness of 0.15 cm. It is com-prised of quadratic NURBS patches, allowing us to repre-sent the circular portions exactly. The sinus is generated bydisplacing control points radially from an initial cylindricalconfiguration, so the normal thickness of the sinus varies.We use a multi-patch design to avoid including a singular-ity at the center of the cylindrical sections. Cross-sectionsof this multi-patch design are shown in Figure 5. The meshof this artery, which includes the fluid-filled interior andsolid arterial wall, consists of 69,696 quadratic B-spline el-ements. For analysis purposes, basis functions are made C0-continuous at the fluid–solid interface and the discretizationis conforming.

Mesh refinement is focused near the valve and sinus, asshown in Figure 4. Figure 5 shows that the mesh is clusteredtoward the wall to better capture the boundary-layer solu-tion. As shown in Figure 6, we extend the pinned edges ofthe valve leaflets with a rigid stent. The stent extends out-side of the fluid domain and intersects with the solid region,to properly seal the gap between the pinned edge of the valveand the arterial wall.

The arterial wall is modeled as a hyperelastic material—a neo-Hookean model with dilatational penalty (see Simoand Hughes [73] and Section 2.3 of the present paper)—withYoung’s modulus and Poisson’s ratio set to 107 dyn/cm2

and 0.45, respectively. The density of the arterial wall is

Fig. 5: Cross-sections of the fluid and solid meshes, takenfrom the cylindrical portion and from the sinus.

1.0 g/cm3. Mass-proportional damping is added to modelthe interaction of the artery with surrounding tissues and in-terstitial fluids. In this case the inertial term in Eq. (22) isreplaced as follows:

ρ2∂2y∂t2 ← ρ2

∂2y∂t2 + aρ2

∂y∂t, (40)

and the damping coefficient, a, is set to 1.0 × 104 s−1.

3.4 Boundary conditions and parameters of the numericalscheme

The solid wall is subjected to zero traction boundary condi-tions at the outer surface. The inlet and outlet branches areallowed to slide in their cut planes as well as deform radi-ally in response to the variations in the blood flow forces


Fig. 6: The sinus, magnified and shown in relation to thevalve leaflets (pink) and rigid stent (blue).

(see Bazilevs et al. [40] for details). This gives more real-istic arterial wall displacement patterns than fixed inlet andoutlet cross-sections.

Because the BHV stent is assumed to contain aneffectively-rigid metal frame [95], the dynamics of the arteryand BHV leaflets are coupled primarily through the fluidrather than the sutures connecting the stent to the artery. Wetherefore constrain the stent to be stationary, and likewisefix the displacement unknowns of any control point of thesolid portion of the artery mesh whose corresponding basisfunction’s support intersects the stent.

The nominal outflow boundary is 11 cm downstream ofthe valve, located at the right end of the channel, based onthe orientation of Figure 4. The nominal inflow is located 5cm upstream at the left end of the channel. The designationsof inflow and outflow are based on the prevailing flow direc-tion during systole, when the valve is open and the majorityof flow occurs. In general, fluid may move in both directionsand there is typically some regurgitation during diastole. Aphysiologically-realistic left ventricular pressure profile ob-tained from Yap et al. [96] and shown in Figure 7 is appliedas a traction boundary condition at the inflow. The durationof a single cardiac cycle is 0.86 s.

The traction −(p0 + RQ)n1 is applied at the outflow,where p0 is a constant physiological pressure level, Q is thevolumetric flow rate through the outflow (with the conven-tion that Q > 0 indicates flow leaving the domain), R > 0 isa resistance constant, and n1 is the outward facing normal ofthe fluid domain. This resistance boundary condition and itsimplementation are discussed in Bazilevs et al. [46]. In thepresent computation, we use p0 = 80 mmHg and R = 70(dyn s)/cm5. These values ensure a realistic transvalvularpressure difference of 80 mmHg in the diastolic steady state

Time (s)

LV p

ress

ure

(mm

Hg)

LV p

ress

ure

(kPa

)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8-20

0

20

40

60

80

100

120

140

0

3

6

9

12

15

18

Fig. 7: Physiological left ventricular (LV) pressure profileapplied at the inlet of the fluid domain. The duration of asingle cardiac cycle is 0.86 s. The data is obtained from Yapet al. [96]

(where Q is nearly zero) while permitting a reasonable flowrate during systole. At both inflow and outflow boundarieswe apply backflow stabilization with γ = 0.5 (see Esmaily-Moghadam et al. [72] for details).

The time-step size is set to ∆t = 1.0 × 10−4 s and theτM scaling factor is sshell = 106. For the immersed heartvalve, we find that results are relatively insensitive to thetangential-velocity penalty-parameter values, while condi-tioning and nonlinear convergence improve when the valuesare lower. We therefore set a lower value for τB

TAN = 2.0×102

g/(cm2 s) and a higher value for τBNOR = 2.0× 103 g/(cm2 s),

also because the no-penetration condition is more critical foraccuracy.

3.5 Results and discussion

We compute both the rigid- and elastic-wall cases to studythe importance of including arterial wall elasticity in theheart-valve FSI simulations. Starting from homogeneousinitial conditions, we compute several cardiac cycles untila time-periodic solution is achieved. Figure 8 shows the vol-umetric flow rate through the top of the tube throughout thecardiac cycle. The flow rates computed using rigid and elas-tic arteries differ primarily in the period immediately follow-ing valve closure. The rigid-wall results show large oscilla-tion in the flow rate, as well as in the valve movement (seeFigure 9). The oscillation is much smaller when arterial wallelasticity is included.

In the rigid-wall case, the energy of the fluid hammerstriking the closed valve is initially converted to elastic po-tential in the leaflets, transferred back to kinetic energy asthe valve rebounds, converted into potential as the fluid


Time (s)

Flow

rate

(mL/

s)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

-200

-100

0

100

200

300

400

500 Rigid wallElastic wall

Fig. 8: Computed volumetric flow rate through the top ofthe fluid domain, during a full cardiac cycle of 0.86 s, forthe rigid and elastic arterial wall cases.

moves through an adverse pressure gradient, then convertedonce again to kinetic energy as the blood reverses direction,forming a new fluid hammer and restarting a cyclic rever-beration. This oscillation is gradually damped by the resis-tance outflow condition and viscous forces in the fluid be-ing directly modeled. The reverberation of the fluid ham-mer impact on the closing valve is the source of the S2 heartsound, marking the beginning of diastole [97,98]. However,the flow rate oscillation that follows from the rigid arteryassumption is observed to be much smaller or completelyabsent in human aortas [99, 100]. This is consistent withour elastic-wall computations, which show that an elasticartery has a compliance effect and can distend to absorb afluid hammer impact and dissipate the initial kinetic energyto surrounding tissues and interstitial fluids (modeled herethrough damping). The artery’s absorption of fluid hammerimpacts on the valve greatly reduces the maximum strains(and thus stresses) observed in the leaflets, as shown in Fig-ure 9.

Remark 3 The strains shown in Figure 9 are the maximumin-plane principal Green–Lagrange strain (MIPE, the largesteigenvalue of E). We choose to plot the strains on the aor-tic side of the leaflets to include contributions from bothstretching and bending. Evaluation of strain at the shell mid-surface, ξ3 = 0, would only display the membrane contribu-tion.

Remark 4 Note that the effect we demonstrate here is not thefull Windkessel effect. Direct simulation of the Windkesseleffect would require a much larger network of arteries down-stream of the valve, and the final outflow from these arteriesshould be relatively constant [8]. We instead demonstratethat the elasticity of the arteries directly adjacent to a heartvalve can significantly impact its dynamics, especially at the

t = 0.36–0.49 s t = 0.36 s

t = 0.355–0.465 s. t = 0.375 s

Fig. 9: Leaflet oscillation and the highest MIPE during thecardiac cycle for the rigid and elastic arterial wall cases. Thestrains are evaluated on the aortic side of the leaflets. Themaximum MIPE on the plots are 0.766 for the rigid-wallcase and 0.483 for the elastic-wall case.

point of valve closure, where maximum strains occur, andshould therefore not be neglected in simulations. We recom-mend combining this technology with a lumped-parameterWindkessel model of arteries further downstream, but wehave applied a simple resistance boundary condition in thispresent work to more clearly highlight the effect of arterialFSI within the directly-simulated domain.

We now examine the details of the fluid and structuresolutions obtained from the elastic-artery computation. Fig-ure 10 shows several snapshots of the details of the fluidsolution fields and Figure 11 shows the deformations andstrain fields of the leaflets at several points during the car-diac cycle. As the valve opens during systole, we see tran-sition to turbulent flow. We also see that the leaflets remainpartially in contact while opening. The snapshot at t = 0.35 sillustrates the fluid hammer effect that is evident in the flowrate. After 0.62 s, the solution becomes effectively hydro-static. The strain near the commissure points at t = 0.35 sis slightly higher than at t = 0.7 s. This is due to the effectof the fluid hammer striking the valve as it initially closes.This phenomenon is usually neglected by both quasi-staticand pressure-driven dynamic computations, as neither ac-counts for the inertia of the fluid [95,101]. The FSI solutionalso shows that the geometrical symmetry of the initial datais not preserved, which is typical for turbulent flow. This re-sult underscores the importance of computing FSI for the


t = 0.03 s t = 0.05 s t = 0.11 s t = 0.26 s

t = 0.33 s t = 0.34 s t = 0.35 s t = 0.62 s

Fig. 10: Volume rendering of the velocity field at several points during a cardiac cycle. The time t is synchronized withFigure 7 for the current cycle.

t = 0.03 s t = 0.26 s t = 0.35 s

t = 0.05 s t = 0.33 s t = 0.62 s

t = 0.11 s t = 0.34 s t = 0.86 s

Fig. 11: Deformations of the valve from the FSI computation, colored by the MIPE evaluated on the aortic side of the leaflet.Note the different scale for each time.


Fig. 12: Relative wall displacement between opening (t =

0.24 s) and closing (t = 0.345 s) phases.

entire valve, without symmetry assumptions. In Figure 12,the models are superposed in the configurations correspond-ing to the opening (t = 0.24 s) and closing phases (t = 0.345s) for better visualization of the relative arterial wall dis-placement results.

4 Conclusions

We presented a computational framework for FSI whichcombines a recently proposed variational immersed-boundary method [12, 13] and the traditional ALE tech-nique. We applied this hybrid ALE/immersed-boundaryframework to simulate a bioprosthetic heart valve implantedin a deformable artery. Our computations demonstrate thatthe variational immersed-boundary method for FSI remainseffective for heart valve analysis when the background fluidmesh undergoes relatively mild deformations, correspond-ing to the expansion and contraction of an elastic artery. Fur-ther, we find that arterial wall deformation contributes sig-nificantly to the realism of BHV simulation results. It dampsout oscillations in the flow rate and valve deformation duringthe closing phase, leading to flow profiles that more closelyresemble those observed in practice [99, 100].

The highest strain on the valve, occurring at the point ofvalve closure, is much lower when wall elasticity is consid-ered. This difference in peak strain between the rigid-arteryand elastic-artery computations suggests a potential futureresearch direction: it indicates that arterial stiffness could be

an important variable to consider in computational studies ofstructural fatigue in BHVs. Atherosclerosis and BHV leafletdeterioration are known to be correlated [102], although theprevailing hypothesis, which we do not purport to refute inthis work, is that these phenomena have a shared etiologyrather than a cause-and-effect relationship.

One conspicuous shortcoming of our simulations is therelatively simple material model of the valve leaflets. TheSt. Venant–Kirchhoff material used in this work does not ac-curately reflect some of the properties of biological materi-als [80,103]. In the present application, the largest strains areprimarily tensile, avoiding the St. Venant–Kirchhoff mate-rial’s most significant pathology: instability under compres-sion. However, its tensile behavior does not exhibit the ex-ponential stiffening characteristic of soft tissues [104, 105].The introduction of a more realistic soft tissue materialmodel will allow for meaningful comparison of the valve’sdeformations with detailed geometrical data collected in theflow loop experiments of Iyengar et al. [106] and Sugimotoet al. [107].

Acknowledgements Y. Bazilevs was supported by the NSF CAREERAward No. 1055091. T. J. R. Hughes was supported by grants fromthe Office of Naval Research (N00014-08-1-0992), the National Sci-ence Foundation (CMMI-01101007), and SINTEF (UTA10-000374)with the University of Texas at Austin. M. S. Sacks was supported byNIH/NHLBI grants R01 HL108330 and HL119297, and FDA contractHHSF223201111595P. D. Kamensky was partially supported by theCSEM Graduate Fellowship. We thank the Texas Advanced Comput-ing Center (TACC) at the University of Texas at Austin for providingHPC resources that have contributed to the research results reported inthis paper.

References

1. F. J. Schoen and R. J. Levy. Calcification of tissue heart valvesubstitutes: progress toward understanding and prevention. Ann.Thorac. Surg., 79(3):1072–1080, 2005.

2. P. Pibarot and J. G. Dumesnil. Prosthetic heart valves: selectionof the optimal prosthesis and long-term management. Circula-tion, 119(7):1034–1048, 2009.

3. R. F. Siddiqui, J. R. Abraham, and J. Butany. Bioprosthetic heartvalves: modes of failure. Histopathology, 55:135–144, 2009.

4. S. Hales. Statical Essays: Containing Haemastaticks: Or, an Ac-count of Some Hydraulick and Hydrostatical Experiments Madeon the Blood and Blood-Vessels of Animals. W. Innys and R.Manby; T. Woodward, London, 1733.

5. O. Frank. Die Grundform des arteriellen Pulses. Erste Abhand-lung. Mathematische Analyse. Zeitschrift fur Biologie, 37:485–526, 1899.

6. K. Sagawa, R. K. Lie, and J. Schaefer. Translation of OttoFrank’s paper “Die Grundform des Arteriellen Pulses” Zeitschriftfur Biologie 37: 483-526 (1899). Journal of Molecular and Cel-lular Cardiology, 22(3):253–254, 1990.

7. O. Frank. The basic shape of the arterial pulse. first treatise:Mathematical analysis. Journal of Molecular and Cellular Car-diology, 22(3):255–277, 1990.

8. N. Westerhof, J.-W. Lankhaar, and B. E. Westerhof. The arterialWindkessel. Medical & Biological Engineering & Computing,47(2):131–141, 2009.


9. N. Westerhof, F. Bosman, C. J. De Vries, and A. Noordergraaf.Analog studies of the human systemic arterial tree. Journal ofBiomechanics, 2(2):121–143, 1969.

10. N. Stergiopulos, B. E. Westerhof, and N. Westerhof. Total ar-terial inertance as the fourth element of the windkessel model.American Journal of Physiology, 276:H81–88, 1999.

11. I. E. Vignon-Clementel, C. A. Figueroa, K. E. Jansen, and C. A.Taylor. Outflow boundary conditions for three-dimensional fi-nite element modeling of blood flow and pressure in arter-ies. Computer Methods in Applied Mechanics and Engineering,195:3776–3796, 2006.

12. D. Kamensky, M.-C. Hsu, D. Schillinger, J. A. Evans, A. Ag-garwal, Y. Bazilevs, M. S. Sacks, and T. J. R. Hughes. A vari-ational immersed boundary framework for fluid–structure inter-action: Isogeometric implementation and application to biopros-thetic heart valves. ICES Report 14-12, May 2014.

13. D. Kamensky, M.-C. Hsu, D. Schillinger, J. A. Evans, A. Ag-garwal, Y. Bazilevs, M. S. Sacks, and T. J. R. Hughes. A vari-ational immersed boundary framework for fluid–structure inter-action: Isogeometric implementation and application to biopros-thetic heart valves. Computer Methods in Applied Mechanics andEngineering, 2014. In review.

14. F. Sotiropoulos and X. Yang. Immersed boundary methods forsimulating fluid–structure interaction. Progress in AerospaceSciences, 65:1–21, 2014.

15. R. Mittal and G. Iaccarino. Immersed boundary methods. AnnualReview of Fluid Mechanics, 37:239–261, 2005.

16. C. S. Peskin. The immersed boundary method. Acta Numerica,11:479–517, 2002.

17. J. De Hart, G. W. M. Peters, P. J. G. Schreurs, and F. P. T.Baaijens. A three-dimensional computational analysis of fluid–structure interaction in the aortic valve. Journal of Biomechanics,36:103–112, 2003.

18. J. De Hart, F. P. T. Baaijens, G. W. M. Peters, and P. J. G.Schreurs. A computational fluid–structure interaction analysisof a fiber-reinforced stentless aortic valve. Journal of Biome-chanics, 36:699–712, 2003.

19. M. Astorino, J.-F. Gerbeau, O. Pantz, and K.-F. Traore. Fluid–structure interaction and multi-body contact: Application to aor-tic valves. Computer Methods in Applied Mechanics and Engi-neering, 198:3603–3612, 2009.

20. M. Astorino, J. Hamers, S. C. Shadden, and J.-F. Gerbeau. A ro-bust and efficient valve model based on resistive immersed sur-faces. International Journal for Numerical Methods in Biomedi-cal Engineering, 28(9):937–959, 2012.

21. B. E. Griffith. Immersed boundary model of aortic heart valvedynamics with physiological driving and loading conditions. In-ternational Journal for Numerical Methods in Biomedical Engi-neering, 28:317–345, 2012.

22. I. Borazjani. Fluid–structure interaction, immersed boundary-finite element method simulations of bio-prosthetic heart valves.Computer Methods in Applied Mechanics and Engineering,257(0):103–116, 2013.

23. T. J. R. Hughes, W. K. Liu, and T. K. Zimmermann. Lagrangian–Eulerian finite element formulation for incompressible viscousflows. Computer Methods in Applied Mechanics and Engineer-ing, 29:329–349, 1981.

24. J. Donea, S. Giuliani, and J. P. Halleux. An arbitrary Lagrangian–Eulerian finite element method for transient dynamic fluid–structure interactions. Computer Methods in Applied Mechanicsand Engineering, 33(1-3):689–723, 1982.

25. J. Donea, A. Huerta, J.-P. Ponthot, and A. Rodriguez-Ferran. Ar-bitrary Lagrangian–Eulerian methods. In Encyclopedia of Com-putational Mechanics, Volume 3: Fluids, chapter 14. John Wiley& Sons, 2004.

26. T. E. Tezduyar, M. Behr, and J. Liou. A new strategy for fi-nite element computations involving moving boundaries and in-terfaces – the deforming-spatial-domain/space–time procedure:I. The concept and the preliminary numerical tests. ComputerMethods in Applied Mechanics and Engineering, 94(3):339–351,1992.

27. T. E. Tezduyar, M. Behr, S. Mittal, and J. Liou. A new strat-egy for finite element computations involving moving boundariesand interfaces – the deforming-spatial-domain/space–time pro-cedure: II. Computation of free-surface flows, two-liquid flows,and flows with drifting cylinders. Computer Methods in AppliedMechanics and Engineering, 94(3):353–371, 1992.

28. T. E. Tezduyar, M. Behr, S. Mittal, and A. A. Johnson. Compu-tation of unsteady incompressible flows with the finite elementmethods – space–time formulations, iterative strategies and mas-sively parallel implementations. In New Methods in TransientAnalysis, PVP-Vol.246/AMD-Vol.143, pages 7–24, New York,1992. ASME.

29. A. A. Johnson and T. E. Tezduyar. Mesh update strategies in par-allel finite element computations of flow problems with movingboundaries and interfaces. Computer Methods in Applied Me-chanics and Engineering, 119:73–94, 1994.

30. K. Stein, T. Tezduyar, and R. Benney. Mesh moving techniquesfor fluid–structure interactions with large displacements. Journalof Applied Mechanics, 70:58–63, 2003.

31. K. Stein, T. E. Tezduyar, and R. Benney. Automatic mesh up-date with the solid-extension mesh moving technique. ComputerMethods in Applied Mechanics and Engineering, 193:2019–2032, 2004.

32. Y. Bazilevs, V. M. Calo, T. J. R. Hughes, and Y. Zhang. Isoge-ometric fluid–structure interaction: theory, algorithms, and com-putations. Computational Mechanics, 43:3–37, 2008.

33. T. E. Tezduyar, K. Takizawa, C. Moorman, S. Wright, andJ. Christopher. Space–time finite element computation of com-plex fluid–structure interactions. International Journal for Nu-merical Methods in Fluids, 64:1201–1218, 2010.

34. T. Wick. Coupling of fully Eulerian and arbitrary Lagrangian–Eulerian methods for fluid-structure interaction computations.Computational Mechanics, 52(5), 2013.

35. T. Wick. Flapping and contact FSI computations with the fluid–solid interface-tracking/interface-capturing technique and meshadaptivity. Computational Mechanics, 53(1):29–43, 2014.

36. Y. Bazilevs, M.-C. Hsu, and M. A. Scott. Isogeometric fluid–structure interaction analysis with emphasis on non-matchingdiscretizations, and with application to wind turbines. ComputerMethods in Applied Mechanics and Engineering, 249–252:28–41, 2012.

37. L. Formaggia, J. F. Gerbeau, F. Nobile, and A. Quarteroni. On thecoupling of 3D and 1D Navier-Stokes equations for flow prob-lems in compliant vessels. Computer Methods in Applied Me-chanics and Engineering, 191:561–582, 2001.

38. J.-F. Gerbeau, M. Vidrascu, and P. Frey. Fluid–structure inter-action in blood flows on geometries based on medical imaging.Computers and Structures, 83:155–165, 2005.

39. F. Nobile and C. Vergara. An effective fluid–structure interactionformulation for vascular dynamics by generalized Robin condi-tions. SIAM Journal on Scientific Computing, 30:731–763, 2008.

40. Y. Bazilevs, M.-C. Hsu, Y. Zhang, W. Wang, T. Kvamsdal,S. Hentschel, and J. Isaksen. Computational fluid–structure inter-action: Methods and application to cerebral aneurysms. Biome-chanics and Modeling in Mechanobiology, 9:481–498, 2010.

41. M. Perego, A. Veneziani, and C. Vergara. A variational approachfor estimating the compliance of the cardiovascular tissue: An in-verse fluid–structure interaction problem. SIAM Journal on Sci-entific Computing, 33:1181–1211, 2011.


42. K. Takizawa, Y. Bazilevs, and T. E. Tezduyar. Space–time andALE–VMS techniques for patient-specific cardiovascular fluid–structure interaction modeling. Archives of Computational Meth-ods in Engineering, 19:171–225, 2012.

43. J. Nitsche. Uber ein variationsprinzip zur losung von Dirichlet-problemen bei verwendung von teilraumen, die keinen randbe-dingungen unterworfen sind. Abh. Math. Univ. Hamburg, 36:9–15, 1971.

44. M.-C. Hsu and Y. Bazilevs. Fluid–structure interaction modelingof wind turbines: simulating the full machine. ComputationalMechanics, 50:821–833, 2012.

45. Y. Bazilevs, V. M. Calo, Y. Zhang, and T. J. R. Hughes. Iso-geometric fluid–structure interaction analysis with applicationsto arterial blood flow. Computational Mechanics, 38:310–322,2006.

46. Y. Bazilevs, J. R. Gohean, T. J. R. Hughes, R. D. Moser, andY. Zhang. Patient-specific isogeometric fluid–structure interac-tion analysis of thoracic aortic blood flow due to implantation ofthe Jarvik 2000 left ventricular assist device. Computer Methodsin Applied Mechanics and Engineering, 198:3534–3550, 2009.

47. Y. Bazilevs, M.-C. Hsu, D. Benson, S. Sankaran, and A. Mars-den. Computational fluid–structure interaction: Methods and ap-plication to a total cavopulmonary connection. ComputationalMechanics, 45:77–89, 2009.

48. Y. Bazilevs, M.-C. Hsu, Y. Zhang, W. Wang, X. Liang, T. Kvams-dal, R. Brekken, and J. Isaksen. A fully-coupled fluid–structureinteraction simulation of cerebral aneurysms. ComputationalMechanics, 46:3–16, 2010.

49. Y. Zhang, W. Wang, X. Liang, Y. Bazilevs, M.-C. Hsu, T. Kvams-dal, R. Brekken, and J.G. Isaksen. High-fidelity tetrahedral meshgeneration from medical imaging data for fluid–structure inter-action analysis of cerebral aneurysms. Computer Modeling inEngineering & Sciences, 42:131–150, 2009.

50. M.-C. Hsu and Y. Bazilevs. Blood vessel tissue prestress mod-eling for vascular fluid–structure interaction simulations. FiniteElements in Analysis and Design, 47:593–599, 2011.

51. Y. Bazilevs and T. J. R. Hughes. Weak imposition of Dirichletboundary conditions in fluid mechanics. Computers and Fluids,36:12–26, 2007.

52. Y. Bazilevs, C. Michler, V. M. Calo, and T. J. R. Hughes.Weak Dirichlet boundary conditions for wall-bounded turbulentflows. Computer Methods in Applied Mechanics and Engineer-ing, 196:4853–4862, 2007.

53. Y. Bazilevs, C. Michler, V. M. Calo, and T. J. R. Hughes. Isogeo-metric variational multiscale modeling of wall-bounded turbulentflows with weakly enforced boundary conditions on unstretchedmeshes. Computer Methods in Applied Mechanics and Engineer-ing, 199:780–790, 2010.

54. K. Hollig. Finite Element Methods with B-Splines. SIAM,Philadelphia, Pennsylvania, USA, 2003.

55. Y. Bazilevs, M.-C. Hsu, K. Takizawa, and T. E. Tezduyar. ALE–VMS and ST–VMS methods for computer modeling of wind-turbine rotor aerodynamics and fluid–structure interaction. Math-ematical Models and Methods in Applied Sciences, 22:1230002,2012.

56. K. Takizawa, Y. Bazilevs, T. E. Tezduyar, M.-C. Hsu, O. Øiseth,K. M. Mathisen, N. Kostov, and S. McIntyre. Engineeringanalysis and design with ALE–VMS and Space–Time meth-ods. Archives of Computational Methods in Engineering, 2014.doi:10.1007/s11831-014-9113-0.

57. Y. Bazilevs, K. Takizawa, T. E. Tezduyar, M.-C. Hsu, N. Kostov,and S. McIntyre. Aerodynamic and FSI analysis of wind turbineswith the ALE–VMS and ST–VMS methods. Archives of Com-putational Methods in Engineering, 2014. doi:10.1007/s11831-014-9119-7.

58. A. N. Brooks and T. J. R. Hughes. Streamline upwind/Petrov-Galerkin formulations for convection dominated flows with par-ticular emphasis on the incompressible Navier-Stokes equa-tions. Computer Methods in Applied Mechanics and Engineer-ing, 32:199–259, 1982.

59. T. E. Tezduyar. Stabilized finite element formulations for incom-pressible flow computations. Advances in Applied Mechanics,28:1–44, 1992.

60. T. E. Tezduyar and Y. Osawa. Finite element stabilization param-eters computed from element matrices and vectors. ComputerMethods in Applied Mechanics and Engineering, 190:411–430,2000.

61. T. J. R. Hughes, L. Mazzei, A. A. Oberai, and A. Wray. Themultiscale formulation of large eddy simulation: Decay of ho-mogeneous isotropic turbulence. Physics of Fluids, 13:505–512,2001.

62. T. J. R. Hughes, G. Scovazzi, and L. P. Franca. Multiscale andstabilized methods. In E. Stein, R. de Borst, and T. J. R. Hughes,editors, Encyclopedia of Computational Mechanics, Volume 3:Fluids, chapter 2. John Wiley & Sons, 2004.

63. Y. Bazilevs, V. M. Calo, J. A. Cottrel, T. J. R. Hughes, A. Re-ali, and G. Scovazzi. Variational multiscale residual-based tur-bulence modeling for large eddy simulation of incompressibleflows. Computer Methods in Applied Mechanics and Engineer-ing, 197:173–201, 2007.

64. M.-C. Hsu, Y. Bazilevs, V. M. Calo, T. E. Tezduyar, and T. J. R.Hughes. Improving stability of stabilized and multiscale formu-lations in flow simulations at small time steps. Computer Meth-ods in Applied Mechanics and Engineering, 199:828–840, 2010.

65. C. Johnson. Numerical solution of partial differential equationsby the finite element method. Cambridge University Press, Swe-den, 1987.

66. S. C. Brenner and L. R. Scott. The Mathematical Theory of FiniteElement Methods, 2nd ed. Springer, 2002.

67. A. Ern and J. L. Guermond. Theory and Practice of Finite Ele-ments. Springer, 2004.

68. J. A. Evans and T. J. R. Hughes. Explicit trace inequalitiesfor isogeometric analysis and parametric hexahedral finite ele-ments. Computer Methods in Applied Mechanics and Engineer-ing, 123:259–290, 2013.

69. M.-C. Hsu, I. Akkerman, and Y. Bazilevs. Finite element simula-tion of wind turbine aerodynamics: Validation study using NRELPhase VI experiment. Wind Energy, 17:461–481, 2014.

70. Y. Bazilevs and I. Akkerman. Large eddy simulation of tur-bulent Taylor–Couette flow using isogeometric analysis and theresidual–based variational multiscale method. Journal of Com-putational Physics, 229:3402–3414, 2010.

71. M.-C. Hsu, I. Akkerman, and Y. Bazilevs. Wind turbine aerody-namics using ALE–VMS: Validation and the role of weakly en-forced boundary conditions. Computational Mechanics, 50:499–511, 2012.

72. M. Esmaily-Moghadam, Y. Bazilevs, T.-Y. Hsia, I. E. Vignon-Clementel, A. L. Marsden, and Modeling of Congenital HeartsAlliance (MOCHA). A comparison of outlet boundary treat-ments for prevention of backflow divergence with relevance toblood flow simulations. Computational Mechanics, 48:277–291,2011.

73. J. C. Simo and T. J. R. Hughes. Computational Inelasticity.Springer-Verlag, New York, 1998.

74. S. Lipton, J. A. Evans, Y. Bazilevs, T. Elguedj, and T. J. R.Hughes. Robustness of isogeometric structural discretizationsunder severe mesh distortion. Computer Methods in Applied Me-chanics and Engineering, 199:357–373, 2010.

75. L. N. Trefethen. Gibbs phenomenon. In Approximation The-ory and Approximation Practice, chapter 9. SIAM, Philadelphia,Pennsylvania, USA, 2012.


76. A. J. Chorin. A numerical method for solving incompress-ible viscous flow problems. Journal of Computational Physics,135(2):118–125, 1967.

77. J. Kiendl, K.-U. Bletzinger, J. Linhard, and R. Wuchner. Isoge-ometric shell analysis with Kirchhoff–Love elements. ComputerMethods in Applied Mechanics and Engineering, 198:3902–3914, 2009.

78. J. Kiendl, Y. Bazilevs, M.-C. Hsu, R. Wuchner, and K.-U.Bletzinger. The bending strip method for isogeometric anal-ysis of Kirchhoff–Love shell structures comprised of multiplepatches. Computer Methods in Applied Mechanics and Engi-neering, 199:2403–2416, 2010.

79. J. Kiendl. Isogeometric Analysis and Shape Optimal Design ofShell Structures. PhD thesis, Lehrstuhl fur Statik, TechnischeUniversitat Munchen, 2011.

80. G. A. Holzapfel. Nonlinear Solid Mechanics: A Continuum Ap-proach for Engineering. Wiley, Chichester, 2000.

81. T. J. R. Hughes, J. A. Cottrell, and Y. Bazilevs. Isogeometricanalysis: CAD, finite elements, NURBS, exact geometry, andmesh refinement. Computer Methods in Applied Mechanics andEngineering, 194:4135–4195, 2005.

82. J. A. Cottrell, T. J. R. Hughes, and Y. Bazilevs. IsogeometricAnalysis: Toward Integration of CAD and FEA. Wiley, Chich-ester, 2009.

83. J. Chung and G. M. Hulbert. A time integration algorithm forstructural dynamics with improved numerical dissipation: Thegeneralized-α method. Journal of Applied Mechanics, 60:371–75, 1993.

84. K. E. Jansen, C. H. Whiting, and G. M. Hulbert. A generalized-αmethod for integrating the filtered Navier-Stokes equations witha stabilized finite element method. Computer Methods in AppliedMechanics and Engineering, 190:305–319, 2000.

85. T. E. Tezduyar, S. Sathe, and K. Stein. Solution techniques for thefully-discretized equations in computation of fluid–structure in-teractions with the space–time formulations. Computer Methodsin Applied Mechanics and Engineering, 195:5743–5753, 2006.

86. T. E. Tezduyar, S. Sathe, R. Keedy, and K. Stein. Space–timefinite element techniques for computation of fluid–structure in-teractions. Computer Methods in Applied Mechanics and Engi-neering, 195:2002–2027, 2006.

87. T. E. Tezduyar and S. Sathe. Modeling of fluid–structure inter-actions with the space–time finite elements: Solution techniques.International Journal for Numerical Methods in Fluids, 54:855–900, 2007.

88. Y. Bazilevs, K. Takizawa, and T. E. Tezduyar. ComputationalFluid–Structure Interaction: Methods and Applications. Wiley,Chichester, 2013.

89. P. Wriggers. Finite element algorithms for contact problems.Archives of Computational Methods in Engineering, 2:1–49,1995.

90. P. Wriggers. Computational Contact Mechanics, 2nd ed.Springer-Verlag, Berlin Heidelberg, 2006.

91. T. A. Laursen. Computational Contact and Impact Mechanics:Fundamentals of Modeling Interfacial Phenomena in NonlinearFinite Element Analysis. Springer-Verlag, Berlin Heidelberg,2003.

92. L. De Lorenzis, I. Temizer, P. Wriggers, and G. Zavarise. Alarge deformation frictional contact formulation using NURBS-based isogeometric analysis. International Journal for Numeri-cal Methods in Engineering, 87:1278–1300, 2011.

93. R. Dimitri, L. De Lorenzis, M. A. Scott, P. Wriggers, R. L. Tay-lor, and G. Zavarise. Isogeometric large deformation frictionlesscontact using T-splines. Computer Methods in Applied Mechan-ics and Engineering, 269:394–414, 2014.

94. B. J. Bellhouse and F. H. Bellhouse. Mechanism of closure ofthe aortic valve. Nature, 217(5123):86–87, 1968.

95. W. Sun, A. Abad, and M. S. Sacks. Simulated bioprosthetic heartvalve deformation under quasi-static loading. J Biomech Eng,127(6):905–914, 2005.

96. C. H. Yap, N. Saikrishnan, G. Tamilselvan, and A. P. Yoganathan.Experimental technique of measuring dynamic fluid shear stresson the aortic surface of the aortic valve leaflet. Journal of Biome-chanical Engineering, 133(6):061007, 2011.

97. J. M. Felner. The second heart sound. In Clinical Methods:The History, Physical, and Laboratory, 3rd edition, chapter 23.Butterworths, Boston, USA, 1990.

98. H. N. Sabbah and P. D. Stein. Relation of the second sound todiastolic vibration of the closed aortic valve. American Journalof Physiology - Heart and Circulatory Physiology, 234(6):H696–H700, 1978.

99. M. E. Kendall, J. C. Rembert, and J. C. Greenfield Jr. Pressure-flow studies in man: The nature of the aortic flow pattern in bothvalvular mitral insufficiency and the prolapsing mitral valve syn-drome. American Heart Journal, 86(3):359–365, 1973.

100. J. B. Uther, K. L. Peterson, R. Shabetai, and E. Braunwald. Mea-surement of ascending aortic flow patterns in man. Journal ofApplied Physiology, 34(4):513–518, 1973.

101. H. Kim, J. Lu, M. S. Sacks, and K. B. Chandran. Dynamic sim-ulation of bioprosthetic heart valves using a stress resultant shellmodel. Annals of Biomedical Engineering, 36(2):262–275, 2008.

102. G. Nollert, J. Miksch, E. Kreuzer, and B. Reichart. Risk factorsfor atherosclerosis and the degeneration of pericardial valves af-ter aortic valve replacement. Journal of Thoracic and Cardiovas-cular Surgery, 126(4):965–968, 2003.

103. J. D. Humphrey. Cardiovascular Solid Mechanics: Cells, Tis-sues, and Organs. Springer-Verlag, New York, 2002.

104. P. Tong and Y.-C. Fung. The stress-strain relationship for theskin. Journal of Biomechanics, 9(10):649 – 657, 1976.

105. Y. C. Fung. Biomechanics: Mechanical Properties of Living Tis-sues. Springer-Verlag, New York, second edition, 1993.

106. A. K. S. Iyengar, H. Sugimoto, D. B. Smith, and M. S. Sacks. Dy-namic in vitro quantification of bioprosthetic heart valve leafletmotion using structured light projection. Annals of BiomedicalEngineering, 29(11):963–973, 2001.

107. H. Sugimoto and M. S. Sacks. Effects of leaflet stiffness on invitro dynamic bioprosthetic heart valve leaflet shape. Cardiovas-cular Engineering and Technology, 4(1):2–15, 2013.

ices report 14-14 fluid–structure interaction analysis of ... · fluid–structure interaction...

Documents