
Citation: Xin Zhao, Xin Liu, Jian Li. Convergence analysis and error estimate of finite element method of a nonlinear fluid-structure interaction problem[J]. AIMS Mathematics, 2020, 5(5): 5240-5260. doi: 10.3934/math.2020337
[1] | Zuliang Lu, Xiankui Wu, Fei Huang, Fei Cai, Chunjuan Hou, Yin Yang . Convergence and quasi-optimality based on an adaptive finite element method for the bilinear optimal control problem. AIMS Mathematics, 2021, 6(9): 9510-9535. doi: 10.3934/math.2021553 |
[2] | Lingling Sun, Hai Bi, Yidu Yang . A posteriori error estimates of mixed discontinuous Galerkin method for a class of Stokes eigenvalue problems. AIMS Mathematics, 2023, 8(9): 21270-21297. doi: 10.3934/math.20231084 |
[3] | Zuliang Lu, Fei Cai, Ruixiang Xu, Lu Xing . Convergence and proposed optimality of adaptive finite element methods for nonlinear optimal control problems. AIMS Mathematics, 2022, 7(11): 19664-19695. doi: 10.3934/math.20221079 |
[4] | Tiantian Zhang, Wenwen Xu, Xindong Li, Yan Wang . Multipoint flux mixed finite element method for parabolic optimal control problems. AIMS Mathematics, 2022, 7(9): 17461-17474. doi: 10.3934/math.2022962 |
[5] | Zhengang Zhao, Yunying Zheng, Xianglin Zeng . Finite element approximation of fractional hyperbolic integro-differential equation. AIMS Mathematics, 2022, 7(8): 15348-15369. doi: 10.3934/math.2022841 |
[6] | Jiehua Zhang . The finite volume element method for non-stationary Stokes equations with an LC element pair. AIMS Mathematics, 2025, 10(6): 13166-13203. doi: 10.3934/math.2025591 |
[7] | Peng Lv . Structured backward errors analysis for generalized saddle point problems arising from the incompressible Navier-Stokes equations. AIMS Mathematics, 2023, 8(12): 30501-30510. doi: 10.3934/math.20231558 |
[8] | Shujie Jing, Jixiang Guan, Zhiyong Si . A modified characteristics projection finite element method for unsteady incompressible Magnetohydrodynamics equations. AIMS Mathematics, 2020, 5(4): 3922-3951. doi: 10.3934/math.2020254 |
[9] | Jinghong Liu . $ W^{1, \infty} $-seminorm superconvergence of the block finite element method for the five-dimensional Poisson equation. AIMS Mathematics, 2023, 8(12): 31092-31103. doi: 10.3934/math.20231591 |
[10] | Fang Cheng, Ye Hu, Mati ur Rahman . Analyzing the continuity of the mild solution in finite element analysis of semilinear stochastic subdiffusion problems. AIMS Mathematics, 2024, 9(4): 9364-9379. doi: 10.3934/math.2024456 |
Fluid-structure interactions are interactions of some movable or deformable structure with an internal or surrounding fluid flow. The variety of fluid-structure occurrences are abundant and ranges from tent-roofs to micropumps, from parachutes via airbags to blood flow in arteries. It is the most important on both modelling, computational issues and applications, the most challenging multi-physics problems for engineers, mathematicians and physicists. The topic of fluid-structure has recently attracted more and more attention in the scientific community. Different methods have been developed and analyzed such as mathematical model [11,26], mathematical theory [4,13], the weak solution [20,26], spectrum asymptotics [25], Lagrange multiplier method [3] and other method [27]. The literature regarding finite element methods can be found in [8,9,15,23].
In this paper we describe a semi-discrete finite element scheme for the nonlinear fluid-structure interaction problem, which interact between the Navier-Stokes fluids and linear elastic solids. There are lots of literatures on fluid-structure interactions for which the fluid is modeled by viscous fluid models [2,10,19,21]. However, the majority of them applies solid models of lower spatial dimensions or linear interaction problem. Especially, in this paper, we consider the interaction of a nonlinear viscous fluid with elastic body motion in bounder domain. We retain the condition: the interface Γ0 between the fluid structure with continuous velocities and stresses. To some extend, numerical analysis of the fluid-structure interaction problem is more difficult than that of the fluid-fluid interaction problem. Here, we assume that the solid displacements of the linear elastic problem are infinitesimal size is of practical interest. Therefore, the approach provided in fluid-fluid interaction problem can be adapted to the fluid-structure case.
In the past several decades, their motivation, development and theoretical foundations have been presented in lots of literatures [3,5,15,27]. This method can be considered of the extend of the reference [8,9], in the sense that it intends to use the same Galerkin finite element method to analyze the nonlinear fluid-structure interaction problem. Here, we discuss the analysis of finite element method for fluid-structure interaction problem, which couples with the Navier-Stokes equations and linear elastic equations. The analysis of this model is not straightforward even if the data is sufficiently smooth. We must take special care of the nonlinear discrete terms arising from the finite element discretization for the fluid-structure interaction problem; these nonlinear trilinear terms no longer satisfy the anti-symmetry properties. Therefore, compared to the finite element analysis of the linear interaction problem, the most challenging aspect rests in the treatment of the nonlinear convection terms [12,14,16,17,18,22,24,28], which has a significant impact on the analysis. In this paper, we analyze the discrete methods in time of a solution for suitably small data, and uniqueness of a suitably small solution, without smooth solutions. Therefore, the approach presented here is fairly robust and adapts to the important case of interface with fractures or cracks. On the other hand, numerical experiments are also provided for the model presented to confirm the theoretical results.
The rest of paper is organized as follows. In section 2, we introduce the fluid-structure model using the Navier-Stokes equation with the linear elastic equation. In section 3, the finite element method of the fluid-structure model is defined and its existence and uniqueness are provided. The convergence and estimate of the presented method are obtained in sections 4 and 5. Finally, we present several numerical examples to illustrate the features of the proposed methods in section 6.
Let the Lipschitz bounded domain Ω=Ω1⋃Ω2 consist of two subdomains Ω1 and Ω2 of Rd, d=2,3, coupled across an interface I0=∂Ω1⋂∂Ω2. Γ1=∂Ω1∖I0, Γ2=∂Ω2∖I0. Moreover, n1 and n2 denote the outward unit normal vectors for Ω1 and Ω2, respectively. The coupled fluid-structure problem is stated as follows: In the fluid region, the governer problem is
ρ1vt+∇p−μ1∇⋅(∇v+∇vT)+(v⋅∇)v=ρ1f1,inΩ1,divv=0,inΩ1,v=0,onΓ1,v(0,x)=v0,inΩ1, | (2.1) |
where the viscosity μ1>0, the density ρ1>0, the body force f1:[0,T]→H1(Ω1), v0 is the initial value on t=0. v:Ω1×[0,T]→Rd and p:Ω1×[0,T]→R denote the velocity and pressure, respectively.
In the solid region, the solid is assumed to be governed by the following linear elasticity
ρ2utt−μ2∇⋅(∇u+∇uT)−λ2∇(∇⋅u)=ρ2f2,inΩ2,u=0,onΓ2,u(0,x)=u0,ut(0,x)=u1,inΩ2, | (2.2) |
where μ2 and λ2 denote the Lame constants, ρ2 the constant solid density, u:Ω2×[0,T]→Rd the displacement of the solid, f2:[0,T]→H1(Ω2) the given loading force per unit mass, and u0 and u1 the given initial data.
Here, we begin as in the case with a fixed interface: the motion of the solid is wholly due to infinitesimal displacements. Again, we assume that the fluid-solid interface is stationary. Although the displacement u is small, the velocity ut is not. Thus, we cannot impose the no-slip condition on the fluid velocity and must retain the interface condition v=ut, along a fixed boundary. Then, across the fixed interface I0 between the fluid and solid, the velocity and stress vector are continuous:
ut=vonI0,μ2(∇u+∇uT)⋅n2+λ2(∇⋅u)n2=pn1−μ1(∇v+∇vT)⋅n1onI0. | (2.3) |
For the mathematical setting of problem (2.1)-(2.2), the following Hilbert spaces are introduced [1]:
Xi=[H10(Ωi)]d={v∈[H1(Ωi)]d:v|Γi=0},i=1,2, |
Q=L2(Ω1), |
Ψ={w∈[H10(Ω)]d:divw=0inΩ1}. |
The fluid-structure interaction problem can be rewritten in variational form as follows: Given
fi∈C([0,T];L2(Ωi)), |
v0∈X1,divv0=0inΩ1, |
u0∈X2,u1∈X2,v0|Γ0=u1|Γ0, |
such that (v,p,u)∈L2([0,T];X1)×L2([0,T];Q)×L2([0,T];X2)
ρ1[vt,w]Ω1+a1[v,w]+b[w,p]+ρ2[utt,w]Ω2+a2[u,w]+c[v,v,w]=ρ1[f1,w]Ω1+ρ2[f2,w]Ω2,∀w∈X∈[H10(Ω)]d, | (2.4) |
b[v,q]=0,∀q∈Q, | (2.5) |
v(0,x)=v0,u(0,x)=u0,ut(0,x)=u1, | (2.6) |
∫t0v(s)|Γ0ds=u(t)|Γ0−u0|Γ0a.e.t. | (2.7) |
Next, the divergence-free weak formulation for (2.4)-(2.7) is defined as follows: Given
fi∈C([0,T];L2(Ωi)),v0∈X1,divv0=0inΩ1,u0∈X2,u1∈X2,v0|Γ0=u1|Γ0, | (2.8) |
seek a pair (v,u)∈L2([0,T];X1)×L2([0,T];X2),divv=0 such that
ρ1[vt,w]Ω1+ρ2[utt,w]Ω2+a1[v,w]+a2[u,w]+c[v,v,w]=ρ1[f1,w]Ω1+ρ2[f2,w]Ω2,∀w∈Ψ, | (2.9) |
v(0,x)=v0,u(0,x)=u0,ut(0,x)=u1, | (2.10) |
∫t0v(s)|Γ0ds=u(t)|Γ0−u0|Γ0a.e.t, | (2.11) |
where the continuous bilinear forms [⋅,⋅]Ωi, ai[⋅,⋅] and b[⋅,⋅] are defined on Xi×Xi and X1×Q, respectively, by
[w1,w]Ωi=∫Ωiw1wdΩi,w1,w∈Xi, |
a1[v,w]=μ12∫Ω1(∇v+∇vT):(∇w+∇wT)dΩ,v,w∈X1, |
a2[u,w]=μ22∫Ω2((∇u+∇uT):(∇w+∇wT)+λ2(divu)(divw))dΩ,u,w∈X2, |
b[v,q]=−∫Ω1divvqdΩ,∀v∈X1,q∈Q. |
Then, the following inequalities hold
a1[v,v]≥μ1‖∇v‖20,Ω1,∀v∈X1,a2[u,u]≥12||u||20,Ω2,∀u∈X2, | (2.12) |
where
||v||0,Ω2≡(μ‖∇v‖20,Ω2+‖divv‖0,Ω2)1/2 |
is equivalent to the classical H1-norm. Moreover, the bilinear term b[⋅,⋅] satisfy the inf-sup condition for the whole system (2.4)-(2.7) and the Navier-Stokes equations:
infq∈Qsupw∈X1b[w,q]‖∇w‖0,Ω‖q‖0,Ω1≥β, | (2.13) |
where the positive constant β is dependent of Ω1. Similarly, the trilinear term c[⋅,⋅,⋅] is defined as follows [28]:
c[v1,v2,w]=(v1⋅∇v2,w),v1,v2,w∈X1. |
Also, the following inequality is valid:
|c[u,v,w]|≤C0‖∇u‖0,Ω1‖∇v‖0,Ω1‖∇w‖0,Ω1,u1,v2,w∈X1. | (2.14) |
Using the auxiliary problem and the results in [8,28], we yield the following existence and uniqueness of the divergence-free weak formulation for (2.9)-(2.11). For convenience, we set fi,t≡∂tfi,i=1,2 in the following.
In order to deal with the nonlinear terms, we have the following lemma.
Lemma 2.1. Assume that both v satisfy the following smallness condition
‖∇v‖0,Ω1≤μ14C0, | (2.15) |
for all t∈[0,T]. Then, we have the estimate
|((v⋅∇)v1,w)|≤μ14‖∇v1‖0‖∇w‖0∀v1,w∈X1. | (2.16) |
Lemma 2.2. Under the hypothesis of (2.8) and (2.15) below, the solution (v,u)∈L2([0,T];X1)×L2([0,T];X2) for (2.9)-(2.11) has the following error estimates:
‖∇v‖L∞([0,T],L2(Ω1))<μ14C0,‖v‖2L∞([0,T],L2(Ω1))+‖ut‖2L∞([0,T],L2(Ω2))+‖v‖2L2([0,T];X1)≤κ0,‖vt‖2L∞([0,T],L2(Ω1))+‖utt‖2L∞([0,T],L2(Ω2))≤κ1, | (2.17) |
where κi,i=0,1 are defined in (2.21) and (2.27), respectively.
Proof. Let the positive constant γ>0 only depend on the Ω. Then, the following inequalities hold true
‖v‖0,Ω1≤γ‖∇v‖0,Ω1,‖v‖0,Ω1≤γ‖v‖1,Ω1,γ>0. | (2.18) |
Choosing w in (2.9) with
w|Ωi={vif i=1,ut if i=2, |
and using the Young inequality, get
ρ12ddt‖v‖20,Ω1+ρ22ddt‖ut‖20,Ω2+12ddt||u||20,Ω2+μ1‖∇v‖20,Ω1≤C0‖∇v‖30,Ω1+ρ1‖f1‖0,Ω1‖v‖0,Ω1+ρ2‖f2‖0,Ω2‖ut‖0,Ω2≤C0×μ14C0‖∇v‖20,Ω1+μ14‖∇v‖20,Ω1+ρ22‖ut‖20,Ω1+ρ21γ2μ1‖f1‖20,Ω1+ρ22‖f2‖20,Ω2=μ12‖∇v‖20,Ω1+ρ22‖ut‖20,Ω2+ρ21γ2μ1‖f1‖20,Ω1+ρ22‖f2‖20,Ω2. | (2.19) |
Noting that μ12‖∇v‖20,Ω1 can be obsorbed by the left hand of the above inequality and integrating the above inequality with respect to the time from 0 to s∈(0,T], we have
ρ12‖v‖20,Ω1+ρ22‖ut‖20,Ω2+12||u||20,Ω2+μ12∫s0‖∇v‖20,Ω1dt≤ρ12‖v0‖20,Ω1+ρ22‖u1‖20,Ω2+12||u0||20,Ω2+ρ22∫s0‖ut‖20,Ω2dt+∫T∗0(ρ21γ2μ1‖f1‖20,Ω1+ρ22‖f2‖20,Ω2)dt=κ0, | (2.20) |
where
κ0=ρ12‖v0‖20,Ω1+ρ22‖u1‖20,Ω2+12||u0||20,Ω2+ρ21γ2Tμ1‖f1‖2L∞([0,T],L2(Ω1))+ρ2T2‖f2‖2L∞([0,T],L2(Ω2)). | (2.21) |
Using the Gronwall inequality, yields that
‖v‖20,Ω1+‖ut‖20,Ω2+||u||20,Ω2+‖v‖2L2([0,T],X1)≤CeCTκ0≤Cκ0,∀s∈(0,T]. | (2.22) |
Using (2.15), the Young inequality, and choosing appropriate parameter, yields
a1[v,v]+a2[u,u]=−ρ1[vt,v]−ρ2[utt,u]−c[v,v,v]+ρ1[f1,v]+ρ2(f2,u)≤ρ1‖vt‖0,Ω1‖v‖0,Ω1+ρ2‖utt‖0,Ω2‖u‖0,Ω2+C0‖∇v‖30,Ω1+ρ1‖f‖0,Ω1‖v‖0,Ω1+ρ2‖f2‖0,Ω2‖u‖0,Ω2≤2ρ21γ2μ1‖vt‖20,Ω1+μ18‖∇v‖20,Ω1+ρ22γ2‖utt‖20,Ω2+14‖∇u‖20,Ω2+μ14‖∇v‖20,Ω1+2ρ21γ2μ1‖f1‖20,Ω1+μ18‖∇v‖20,Ω1+ρ22γ2‖f2‖20,Ω2+14‖∇u‖20,Ω2 |
which implies
μ12‖∇v‖20,Ω1+12||u||20,Ω2≤2ρ21γ2μ1‖vt‖20,Ω1+ρ22γ2‖utt‖20,Ω2+2ρ21γ2μ1‖f1‖20,Ω1+ρ22γ2‖f2‖20,Ω2. | (2.23) |
In order to estimate the above inequality, we bound the two terms ‖vt‖0 and ‖utt‖0. First, we differentiate the first equations of (2.1) and (2.2) with respect to the time, multiplying (2.1) and (2.2) by the respective test functions vt and utt, respectively, and integrating them on the respective domains Ω1 and Ω2 to obtain the following:
ddt(ρ12‖vt‖20,Ω1+ρ22‖utt‖20,Ω2)+12ddt||ut||20,Ω2+μ1‖∇vt‖20,Ω1≤2C0‖∇v‖0,Ω1‖∇vt‖20,Ω1+ρ1‖ft1‖0,Ω1‖vt‖0,Ω1+ρ2‖ft2‖0,Ω2‖utt‖0,Ω2≤μ12‖∇vt‖20,Ω1+2ρ21γ2μ1‖ft1‖20,Ω1+μ18‖∇vt‖20,Ω1+ρ22γ22‖ft2‖20,Ω2+12||utt||20,Ω2. | (2.24) |
Using (2.15), and integrating the above inequality from 0 to s∈(0,T], we can get
ρ1‖vt‖20,Ω1+ρ2‖utt‖20,Ω2+||ut||20,Ω2+μ1∫s0‖∇vt‖20,Ω1dt≤C(ρ1‖vt(0)‖20,Ω1+ρ2‖utt(0)‖20,Ω2+||u1||20,Ω2)+C∫s0(‖ft1‖20,Ω1+‖ft2‖20,Ω2)dt+∫s0||ut||20,Ω2dt. | (2.25) |
Here, vt(0) and utt(0) can be bounded by the following procedure. Taking t=0 and setting
w|Ωi={vt(0)if i=1,utt(0) if i=2, |
we have
ρ1‖vt(0)‖20,Ω1+ρ2‖utt(0)‖20,Ω2+a1[v0,vt(0)]+b[vt(0),p0]+a2(u0,utt(0))+c[v0,v0,vt(0)]=ρ1[f1(0),vt(0)]+ρ2[f2(0),utt(0)], |
which together with (2.3) implies
ρ1‖vt(0)‖20+ρ2‖utt(0)‖20+(−μ2div(∇u0+uT0)−λ2∇(divu0),utt(0))+(−μ1div(∇u0+uT0)+∇p0+v0⋅∇v0,vt(0))=ρ1[f1(0),vt(0)]+ρ2[f2(0),utt(0)]. |
That is
ρ1‖vt(0)‖20,Ω1+ρ2‖utt(0)‖20,Ω2≤C(ρ1‖f1(0)‖20,Ω1+ρ2‖f2(0)‖20,Ω2). | (2.26) |
Then, setting
κ1=C2∑i=1(ρi‖fi(0)‖20,Ω1+‖fti‖2L2([0,T],L2(Ωi))). | (2.27) |
combining (2.25) with (2.26), and applying the Gronwall inequality, we obtain
‖vt‖20,Ω1+‖utt‖20,Ω2≤C(ρ1‖f1(0)‖20,Ω1+ρ2‖f2(0)‖20,Ω2)+∫s0(‖ft1‖20,Ω1+‖ft2‖20,Ω2)dt)<κ1, | (2.28) |
Using the same approach as above for the solution of (2.9)-(2.11), we can obtain the following stability of the solution to (2.4)-(2.7).
Lemma 2.3. Under the hypothesis of Lemma 2.1 and
fi,t∈L2([0,T];L2(Ωi)),i=1,2,v0∈[H2(Ω1)]d,u1∈[H1(Ω2)]d,u0∈[H2(Ω2)]d, | (2.29) |
the solution (v,p,u) to (2.4)-(2.7) satisfies
v∈L∞([0,T];L2(Ω1))∩L2([0,T];X1),vt∈L∞([0,T];L2(Ω1))∩L2([0,T];X1),u∈L∞([0,T];X2),ut∈L∞([0,T];X2),utt∈L∞([0,T];L2(Ω2)), | (2.30) |
and
‖vt‖2L∞([0,T];L2(Ω1))+‖utt‖2L∞([0,T];L2(Ω2)+‖vt‖2L2([0,T];X1)+‖ut‖2L∞([0,T];X2)+‖p‖L2([0,T];L2(Ω1))≤CeCT(‖f‖2H1([0,T];L2(Ω))+‖u0‖22,Ω2+‖v0‖22,Ω1+‖p0‖1,Ω2+‖u1‖21,Ω2). | (2.31) |
Proof. Using the same approach as for the linear fluid-structure interaction problem [8,9] and Lemma 2.1, we can obtain (2.31).
Given two shape-regular, quasi-uniform triangulationThi of Ωi,i=1,2, the finite element method is to solve (2.4)-(2.7) in a pair of finite dimensional spaces (Xh1,Qh,Xh2)⊂(X1,Q,X2): The triangulations Thi do not cross the interface Γ0 and consist of triangular elements in two dimensions or tetrahedral elements in three dimensions [6,7].
Moreover, we assume that the finite element spaces (Xh1,Qh,Xh2)⊂(X1,Q,X2) satisfies the following approximation properties: For each w∈[H2(Ωi)]d and q∈H1(Ω)∩Q, there exist approximations wh∈Xhi and qh∈Qh such that [28]
‖w−wh‖0,Ωi+h‖∇(w−wh)‖0,Ωi≤Chr+1‖w‖r+1,Ωi,‖q−qh‖0,Ω1≤Chr‖q‖r,Ω1, | (3.1) |
Here, r is the degree of piecewise polynomial of the finite elements. For each wh∈Xhi, we have the inverse inequality
‖∇wh‖0,Ωi≤Ch−1‖wh‖0,Ωi,wh∈Xhi. | (3.2) |
Moreover, the so-called inf-sup condition is valid: for each qh∈Qh, there exists wh∈Xh,vh≠0, such that [9]
infqh∈Qhsupwh∈Xhd(wh,qh)‖∇wh‖0,Ω‖qh‖0,Ω1≥β, | (3.3) |
where β is a positive constant depending on Ω. In the special case, (3.3) is valid for a general choice Xh1 and Qh.
The corresponding semi-discrete fluid-structure interaction problem can be rewritten in variational form as follows: seek (vh,ph,uh)∈C1([0,T];Xh1)×C([0,T];Qh)×C1([0,T];Xh2) such that [12,28]
ρ1[vht,wh]Ω1+a1[vh,wh]+b[wh,ph]+ρ2[uhtt,wh]Ω2+a2[uh,wh]+c[uh,uh,wh]=ρ1[f1,wh]Ω1+ρ2[f2,wh]Ω2,∀wh∈Xh, | (3.4) |
b[vh,qh]=0,∀qh∈Qh, | (3.5) |
vh(0,x)=v0h,uh(0,x)=u0h,uht(0,x)=u1h, | (3.6) |
vh|Γ0=uht|Γ0,a.e.t. | (3.7) |
If we define divergence-free finite element space
Ψh={wh∈Xh1:b[wh,qh]=0,∀qh∈Qh}, |
then the corresponding weak formulation of semi-discrete finite element methods for (3.4)-(3.7) is defined as follows: seek a pair (vh,uh)∈C1([0,T];Ψh)×C1([0,T];Xh2) such that
ρ1[vht,wh]Ω1+ρ2[uhtt,wh]Ω2+a1[vh,wh]+a2[uh,wh]+c[uh,uh,wh]=ρ1[f1,wh]Ω1+ρ2[f2,wh]Ω2,∀wh∈Ψh, | (3.8) |
vh(0,x)=v0h,uh(0,x)=u0h,uht(0,x)=u1h, | (3.9) |
vh|Γ0=uht|Γ0,a.e.t. | (3.10) |
Here, the approximation of the initial condition (v0h,p0h,u1h)∈(Xh1,Qh,Xh2) of (v0,p0,u1) is defined as follows: for ∀(wh,qh)∈Xh×Qh
a1[v0h,wh]+[u1h,wh]+b[wh,p0h]=a1[v0,wh]+[u1,wh]Ω2+b[wh,p0], | (3.11) |
b[v0h,qh]=0, | (3.12) |
v0h|Γ0=u1h|Γ0, | (3.13) |
such that
‖∇(v0−v0h)‖0,Ω1+‖p0−p0h‖0,Ω1+‖u1−u1h‖0,Ω2≤Chr(‖v0‖r+1,Ω1+‖p0‖r,Ω1+‖u1‖r+1,Ω2), | (3.14) |
with v0∈Hr+1(Ω1),p0∈Hr(Ω1) and u1∈Hr+1(Ω2),r∈[0,k]. Moreover, we assume that u0h=Phuh is defined by
a2[u0h,wh]=a2[u0,wh],∀wh∈Xh2. | (3.15) |
Based on the results [8,12,28], we can obtain the following existence and uniqueness of the semi-discrete finite element approximation for the fluid-structure interaction.
Lemma 3.1 Under the hypothesis of Lemmas 2.1-2.2, there exists a unique solution to (3.4)-(3.7) such that (vh,ph,uh)∈C1([0,T];Xh1)×C([0,T];Qh)×C([0,T];Xh2) and furthermore has the following bound
‖vh‖2L∞([0,T];L2(Ω1))+‖uht‖2L∞([0,T];L2(Ω2))+‖vh‖2L2([0,T];X1)+‖uh‖2L∞([0,T];X2)≤CeCT(‖f‖2L2([0,T];L2(Ω))+‖u0‖21,Ω2+‖v0‖21,Ω1+‖p0‖0,Ω1+‖u1‖20,Ω2), |
and
‖vht‖2L∞([0,T];L2(Ω1))+‖uhtt‖2L∞([0,T];L2(Ω2))+‖vht‖2L2([0,T];X1)+‖uht‖2L∞([0,T];X2)+‖ph‖L2([0,T];L2(Ω1))≤CeCT(‖f‖2H1([0,T];L2(Ω))+‖u0‖22,Ω2+‖v0‖22,Ω2+‖p0‖21,Ω2+‖u1‖22,Ω2). |
Proof. Using the same approach as for Lemmas 2.1-2.2, we can obtain the proof of Lemma 3.1.
In this section, we mainly consider the convergence of the semi-discrete finite element method for the nonlinear fluid-structure interaction problem. Then, we will provide the stability of the limit of the finite element approximation based on the results on the previous section.
Theorem 4.1 Assume that the finite element meshes are nested and the data (v0,u0,u1,f1,f2) satisfies (2.8) and (2.29). Let (vh,ph,uh)∈Xh1×Qh×Xh2 be the solution to (3.4)-(3.7), then there exists a unique solution (v,p,u)∈X1×Q×X2 satisfying
v∈L∞([0,T];L2(Ω1))⋂L2([0,T];X1),vt∈L∞([0,T];L2(Ω1))⋂L2([0,T];X1),p∈L2([0,T];L2(Ω1)),u∈L∞([0,T];X2),ut∈L∞([0,T];X2),utt∈L∞([0,T];L2(Ω2)), | (4.1) |
vh∗⇀vinL∞([0,T];L2(Ω1)),vh→v,inL2([0,T];X1),vht∗⇀vtinL∞([0,T];L2(Ω1)),vht→vtinL2([0,T];X1),uh∗⇀uinL∞([0,T];X2),uhtt∗⇀uttinL∞([0,T];L2(Ω2)),uht∗⇀utinL∞([0,T];L2(Ω2)),uht∗⇀utinL∞([0,T];X2),ph→pweaklyinL2([0,T];L2(Ω1)). | (4.2) |
Furthermore, (v,p,u) satisfies (2.4)-(2.7).
Proof. Using the boundness of the finite element solution (vh,ph,uh), we may extract a subsequence (vhμ,phμ,uhμ) from it with the mesh scale decreasing to zero as μ→∞, and satisfy (4.1)-(4.2). A proof of the error estimate can be found in [8,9]. For completeness and to show the results of the nonlinear fluid-structure interaction, we will sketch the proof here.
As for the trilinear term, we have for v∈L∞([0,T];L2(Ω1))∩L2([0,T];X1) and w∈C1([0,T];Xhμ1)
∫T0c[vhμ,vhμ,w]dt=∫T0dt∫Ω1∑i,jvihμ∂vjhμ∂xiwjdΩ. | (4.3) |
For all μ>N, passing to the limit as μ→∞, yields that
∫T0dt∫Ω1∑i,jvihμ∂vjhμ∂xiwjdΩ→∫T0dt∫Ω1∑i,jvi∂vj∂xiwjdΩ |
since
|∫T0dt∫Ω1∑i,jvi∂vj∂xiwjdΩ−∫T0dt∫Ω1∑i,jvihμ∂vjhμ∂xiwjdΩ|≤|∫T0dt∫Ω1∑i,j(vi−vihμ)∂vj∂xiwjdΩ+∫T0dt∫Ω1∑i,jvihμ(∂vjhμ∂xi−∂vj∂xi)wjdΩ|≤‖(vi−vihμ)‖L∞([0,T];L2(Ω1))‖vj‖L2([0,T];X1)‖wj‖L2([0,T];L2(Ω1))+‖vihμ‖L∞([0,T];L2(Ω1))‖vj−vjhμ‖L2([0,T];X1)‖wj‖L2([0,T];L2(Ω1)). |
Recalling [8,9], we can obtain the following results. Since uht→ut weak start convergence in L∞([0,T];L2(Ω2)) and L∞([0,T];L2(Ω2)) is dense in L2([0,T];L2(Ω2)), we can obtain the strong convergence for uht→ut in L2([0,T];L2(Ω2)) with its norm. Similarly, we can also have the same result on the strong convergence for vh→v in L2([0,T];L2(Ω1)) with its norm. Thus, all these implies v|Γ0=ut|Γ0. Applying the same approach as above, we can obtain (2.5) since ∞⋃μ=NL2([0,T];Qhμ) is dense in L2([0,T];L2(Ω1)).
Using the estimates of semi-discrete finite element approximation in Lemma 3.1 and analysis above, we have for each fixed N with μ>N
∫T0(ρ1[vhμt,w]Ω1+a1[vhμ,w]+b[w,phμ]+c[vhμ,vhμ,w]+ρ2[utthμ,w]Ω2+a2[uhμ,w])dt=∫T0(ρ1[f1,w]Ω1+ρ2[f2,w]Ω2)dt,w∈C1([0,T];Xhμ). | (4.4) |
Thus, passing to the limit as μ→∞, we conclude that
∫T0(ρ1[vt,w]Ω1+a1[v,w]+b[w,p]+c[v,v,w]+ρ2[utt,w]Ω2+a2[u,w])dt=∫T0(ρ1[f1,w]Ω1+ρ2[f2,w]Ω2)dt,∀w∈C1([0,T],Xhμ) | (4.5) |
and noting ∞⋃μ=NC1([0,T];Xhμ) is dense in L2([0,T];X), yields that
∫T0(ρ1[vt,w]Ω1+a1[v,w]+b[w,p]+c[v,v,w]+ρ2[utt,w]Ω2+a2[u,w])dt=∫T0(ρ1[f1,w]Ω1+ρ2[f2,w]Ω2)dt,w∈X,a.e.t. | (4.6) |
Moreover, we will consider the convergence of the finite element approximate on the initial value and interface boundary. First, setting w(T)=0 and noticing
∫T0[vt,w]Ω1dt=∫T0ddt[v,w]Ω1dt−∫T0[v,wt]Ω1dt=[v(T),w(T)]Ω1−[v(0),w(0)]Ω1−∫T0[v,wt]Ω1dt, | (4.7) |
and
∫T0[utt,w]Ω2dt=∫T0ddt[ut,w]Ω2dt−∫T0[ut,wt]Ω2dt=[ut(T),w(T)]Ω2−[ut(0),w(0)]Ω2−∫T0[ut,wt]Ω2dt, | (4.8) |
we obtain from (4.6) for w∈C1([0,T];X)
∫T0(−ρ1[v,wt]Ω1+a1[v,w]+b[w,p]+c[v,v,w]−ρ2[ut,wt]Ω2+a2[u,w])dt=∫T0(ρ1[f1,w]Ω1+ρ2[f2,w]Ω2)dt+ρ1[v(0),w(0)]Ω1+ρ2[ut(0),w(0)]Ω2. | (4.9) |
On the other hand, by the same reason of (4.4), and passing to the limit as μ→∞, we infer for w∈C1([0,T];Xhμ)
∫T0(−ρ1[v,wt]Ω1+a1[v,w]+b[w,p]+c[v,v,w]−ρ2[ut,wt]Ω2+a2[u,w])dt=∫T0(ρ1[f1,w]Ω1+ρ2[f2,w]Ω2)dt+ρ1[v0,w(0)]Ω1+ρ2[u1,w(0)]Ω2, | (4.10) |
which together with (4.9) implies
ρ1[v(0)−v0,w(0)]Ω1+ρ2[ut(0)−u1,w(0)]Ω2=0,w(0)∈Xhμ. | (4.11) |
Thus, using the bedding theorem that ∞⋃μ=0Xhμ is dense in L2(Ω), we can infer that
v(0)=v0inL2(Ω1),ut(0)=u1inL2(Ω2). | (4.12) |
Using (4.1) and compact embedding theory, we can deduce the following strong convergence for a further subsequence hμ
uhμ,t→utinL2([0,T];L2(Ω2)),uhμ→uinL2([0,T];L2(Ω2)). | (4.13) |
Then, noticing
uhμ=u0hμ+∫t0uhμt(s)ds, | (4.14) |
and ‖u0h−u0‖0,Ω2=0 as h→0, yields that
u=u0+∫t0utds. | (4.15) |
Therefore, we have
u0=u−∫t0ut(s)ds=u(0,x), | (4.16) |
which is the second equation of (2.10).
The goal of this section is to analyze the estimates of the discretization errors for the Galerkin finite element method for the nonlinear fluid-structure interaction problem. The estimates are based on the solution to (3.4)-(3.7). The approach is based on a priori estimates for the solution of Galerkin semi-discrete scheme in space [14,16,18,28], the full set of estimates is obtained by differentiateting the discrete version with the respect to time. Furthermore, the analysis in Theorem 5.1 below, which provide a good global assessment of the discretization error under reasonable assumptions.
First, the projection operator Ph:L2(Ω)→Ψh is introduced as follows:
ρ1[Phv,w]+ρ2[Phu,w]=ρ1[v,w]+ρ2[u,w],∀w∈Ψh, | (5.1) |
which implies
b[Phw,qh]=0,∀qh∈Qh. | (5.2) |
Under the hypotheses of angle condition on two domains Ω1 and Ω2, it holds for ϵ∈(0,1) and integer k>0 that [9]
(‖v−Phv‖0,Ω1+‖u−Phu‖0,Ω2)+h(‖∇(v−Phv)‖0,Ω1+‖∇(v−Phv)‖0,Ω2)≤Ch1+r−ϵ(‖v‖r+1,Ω1+‖u‖r+1,Ω2),∀v∈Hr+1(Ω1),u∈Hr+1(Ω2),i=1,2,r∈[0,k]. | (5.3) |
Then, we have the following result of semi-discrete finite element approximations of the fluid-solid interaction problem.
Theorem 5.1 Under the hypothesis of Theorem 4.1, let (v,p,u)∈X1×Q×X2 and (vh,ph,uh)∈Xh1×Qh×Xh2 be the solution of (2.4)-(2.7) and (3.4)-(3.7), respectively. Then, it holds
‖v−vh‖2L∞([0,T],L2(Ω1))+‖ut−uht‖2L∞([0,T],L2(Ω2))+‖∇(v−vh)‖2L2([0,T],L2(Ω1))+‖∇(u−uh)‖2L2([0,T],L2(Ω2))≤Ch2r(‖v0‖2r+1,Ω1+‖p0‖2r,Ω1+‖u1‖2r+1,Ω2)+Ch2(r−ϵ)(‖v‖2L2([0,T],Hr+1(Ω1))+‖u‖2L2([0,T],Hr+1(Ω2))+‖p‖2L2([0,T],Hr(Ω1))+‖ut‖2L2([0,T],Hr+1(Ω2)))dt. | (5.4) |
Proof. Subtracting (2.4) from (3.4), yields that
ρ1[vt−vht,wh]Ω1+a1[v−vh,wh]+b[wh,p−ph]+c[v,v,wh]−c[vh,vh,wh]ρ2[utt−uhtt,wh]Ω2+a2[u−uh,wh]=0,∀wh∈Xh,a.e.t. | (5.5) |
In order to uniform the variational formulation, we define wh in two different domains. Setting wh=˜ξ−ξh with
ξξh|Ωi={vhifi=1,uhtifi=2, |
and
˜ξ|Ωi={Phvifi=1,Phutifi=2, |
using (5.1), then we can obtain the following result
ρ1[vt−vht,wh]Ω1+ρ2[utt−uhtt,wh]Ω2=ρ1[Phvt−vht,wh]Ω1+ρ2[Phut−uht,wh]Ω2=ρ12ddt‖Phv−vh‖20,Ω1+ρ22ddt‖Phut−uht‖20,Ω2. | (5.6) |
By (3.5) and (5.2),
b[Phv−vh,p−ph]=b[Phv−vh,p−qh],∀qh∈Qh. | (5.7) |
Similarly,
a1[Phv−vh,Phv−vh]+a2[Phu−uh,Phut−uht]=μ1‖∇(Phv−vh)‖20,Ω1+12ddt||Phu−uh||20,Ω2. | (5.8) |
Using the estimates of the trilinear term c[⋅,⋅,⋅], leads to
c[v,v,wh]−c[vh,vh,wh]=c[v−vh,v,wh]−c[vh,v−vh,wh]≤c[v−Phv,v,wh]+c[Phv−vh,v,wh]+c[vh,v−Phv,wh]+c[vh,Phv−vh,wh]=I1+I2+I3+I4, | (5.9) |
Substituting wh=Phv−vh into the above inequalities and using (2.15), we have
|I1+I3|≤C0‖∇(v−Phv)‖0,Ω1(‖∇v‖0,Ω1+‖∇vh‖0,Ω1)‖∇(vh−Phv)‖0,Ω1≤μ12‖∇(v−Phv)‖0,Ω1‖∇(vh−Phv)‖0,Ω1≤μ18‖∇(vh−Phv)‖20,Ω1+μ12‖∇(v−Phv)‖20,Ω1 | (5.10) |
and
|I2+I4|≤C0(‖∇v‖0,Ω1+‖∇vh‖0,Ω1)‖∇(vh−Phv)‖20,Ω1≤μ12‖∇(vh−Phv)‖20,Ω1, | (5.11) |
which together with (5.9) yields that
|c[v,v,wh]−c[v,v,wh]|≤5μ18‖∇(vh−Phv)‖20,Ω1+μ12‖∇(v−Phv)‖20,Ω1. | (5.12) |
From the bounds of these inequality and the Young inequality, hence, after inserting (5.5) and noting that
a2[u−Phu,Phut−uht]=ddta2[u−Phu,Phu−uh]−a2[ut−Phut,Phu−uh] |
we have
ρ12ddt‖Phv−vh‖20,Ω+ρ22ddt‖Phut−uht‖20,Ω+μ1‖∇(Phv−vh)‖20,Ω1+12ddt||(Phu−uh)||20,Ω2=−a1[v−Phv,Phv−vh]−b[Phv−vh,p−qh]−a2[u−Phu,Phut−uht]−c[v,v,wh]+c[vh,vh,wh]=−a1[v−Phv,Phv−vh]−b[Phv−vh,p−qh]−ddta2[u−Phu,Phu−uh]+a2[ut−Phut,Phu−uh]−c[v,v,wh]+c[vh,vh,wh]≤7μ18‖∇(Phv−vh)‖20,Ω1+C(‖∇(v−Phv)‖20,Ω1+‖p−qh‖20,Ω1+||ut−Phut||20,Ω2)+12||(Phu−uh)||20,Ω2−ddta2[u−Phu,Phu−uh], | (5.13) |
where the last positive constant C determined by the bounds of the data (μ,f,v0,u0,u1). Recalling the estimate in (3.14), the definition of initial value u0h in (3.15), integrating the above equation from 0 to s, and using the Gronwall inequality, we can achieve the following result:
‖Phv−vh‖20,Ω+‖Phut−uht‖20,Ω+‖∇(Phu−uh)‖20,Ω2)+∫s0(‖∇(Phv−vh)‖20,Ω1dt≤Ch2r(‖v0‖2r+1,Ω1+‖p0‖2r,Ω1+‖u1‖2r+1,Ω2)+Ch2(r−ϵ)∫T0(‖v‖2r+1,Ω1+‖p‖2r,Ω1+‖u‖2r+1,Ω2+‖ut‖2r+1,Ω2)dt, |
which together with a triangle inequality, yields the desired result.
In this section, we present numerical experiment to examine the convergence of the fluid-structure interaction system. The computational domain Ω is designed as [0,2π]×[−1,1], where Ω1=[0,2π]×[0,1], Ω2=[0,2π]×[−1,0], I0=[0,2π]×{0} and Γ1,Γ2 defined as before, i.e. Figure 1.
In order to satisfy the conditions on coupled interface I0, we set μ1=μ2=1−2ν4sin(1)(1−ν) and λ2=ν2sin(1)(1−ν) by Lame formulation, in which 0<ν<12 is Poisson's ratio. Especially, let ν=14, ρ1=ρ2=1 and T=1 in this example. Then the boundary data, initial data and the source terms are chosen such that the exact solution of the fluid-structure interaction system is given by
{v={−cos(x)sin(y−1)et,sin(x)(cos(y−1)−1)et},p=sin(x)cos(y)et,u={−cos(x)sin(y−1)et,sin(x)(cos(y+1)−1)et}. |
For the discretization in space we have considered Taylor-Hood element for the Navier-Stokes equations and P2 for the Elasticity system. It should be noted that the finite element partition in Ω1 and Ω2 must match at the interaction interface. For the discretization in time we combine the central difference scheme for the second-order derivative with the Implicit Euler scheme for the first-order derivative. First, without consider the nonlinear term, we can express the discrete system (3.4)−(3.5) as the following linear algebraic systems
M2d2wdt2+M1dwdt+S1w=F, |
where the matrices M2, M1 and S1 are deduced from the bilinear a1[⋅,⋅],a2[⋅,⋅],b[⋅,⋅],[d⋅dt,⋅],[d2⋅dt2,⋅], F is the variation of the source term and w={v1,v2,ph,u1,u2} are the unknowns. In particular, the matrix M2, M1 and S1, respectively, have the following form
M2=[000000000000000000M2100000M21], |
M1=[M1100M1200M1100M1200000000M1300000M13] |
and
S1=[A1A2B1∗∗AT2A3B2∗∗BT1BT2000∗∗∗∗∗∗∗∗∗∗]. |
Then as a result of treating trilinear by Oseen iteration, i.e. (vnh⋅∇)vn+1h, the time-discrete problem can be read as:
(M2τ2+M1τ+S1+S2)wn+1=Fn+1+(2M2τ2+M1τ)wn−M2τ2wn−1, |
where S2 can be express as
[A400000A4000000000000000000]. |
Obviously, the foregoing matrix systems can be derived from the following equation:
ρ1[vn+1−vnΔt,w1h]Ω1+a1[vn+1h,w1h]+b[w1h,pn+1h]−b[vn+1h,qh]+(vnh⋅∇)vn+1h+ρ2[un+1−2un+un−1Δt2,w2h]Ω2+a2[un+1h,w2h]+∫I0(vn+1h−un+1ht)w1hds+∫I0[μ2(∇uh+∇uTh)n2+λ(divuhn2)]n+1w1hds+∫I0(un+1ht−vn+1h)w2hds−∫I0[pn1−μ1(∇vh+∇vTh)n1]n+1w2hds=ρ1[fn+11,w1h]Ω1+ρ2[fn+12,w2h]Ω2,n=0,1,⋯N, | (6.1) |
where Δt=T/N is the uniform time step size. Moreover, when treating the initial conditions, we use the Implicit Euler scheme as the method of difference.
We partition domain Ω into a uniform matching triangulation. The refined meshes are obtained by dividing primary meshes into four similar cells by connecting the edge midpoints. By matching the time step size Δt with the mesh size O(8h3), Table 1 gives the numerical results through using the monolithic scheme. We see that the convergence rates for the velocity, pressure and displacement of the solid are just about O(h3), O(h2) and O(h3), respectively, as the theoretical prediction.
h | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
2−3 | 5.0313e-02 | 3.7129e-02 | 2.2842e-02 | 7.1309e-02 |
2−4 | 1.1786e-02 | 7.4693e-03 | 2.8885e-03 | 1.3052e-02 |
2−5 | 2.8876e-03 | 1.7159e-03 | 3.7417e-04 | 2.8635e-03 |
2−6 | 7.1868e-04 | 4.1590e-04 | 5.3240e-05 | 6.8775e-04 |
Rate | 2.0431 | 2.1601 | 2.9150 | 2.2320 |
In order to verify the order of time convergence, thought there is no theoretical analysis in this paper, we test the case in which space step size h can be chosen as small as enough. We can observe that the expected order of convergence in τ, i.e. O(τ) in Table 2.
τ | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
1/5 | 2.7888e-01 | 2.9410e-01 | 2.8505e-01 | 7.0934e-01 |
1/10 | 1.3926e-01 | 1.5211e-01 | 1.4489e-01 | 3.6428e-01 |
1/20 | 6.9729e-02 | 7.7373e-02 | 7.3046e-02 | 1.8516e-01 |
1/40 | 3.4912e-02 | 3.9019e-02 | 3.6674e-02 | 9.3457e-02 |
Rate | 0.9993 | 0.9714 | 0.9861 | 0.9747 |
Then in Table 3 and Table 4, we test the different decoupling order scheme for this interaction system. Table 3 shows the results by first solving Navier-Stokes equation, then solving Elasticity equation; Table 4 shows the results in reverse order. In both of them, we use the Nitsche's trick to treat the Dirichlet interface condition, i.e. the first equation of (2.3). That is to say, we use ∫I0(vnh−un−1ht)w1hds or ∫I0(unht−vnh)w2hds to weakly treat the Dirichlet interface condition.
h | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
2−3 | 8.3133e-02 | 4.3960e-02 | 2.2426e-02 | 7.2486e-02 |
2−4 | 1.4350e-02 | 7.8258e-03 | 2.8314e-03 | 1.3255e-02 |
2−5 | 3.0401e-03 | 1.7028e-03 | 3.6418e-04 | 2.8794e-03 |
2−6 | 7.2433e-04 | 4.7933e-04 | 5.1454e-05 | 6.8676e-04 |
Rate | 2.2809 | 2.1730 | 2.9226 | 2.2406 |
h | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
2−3 | 5.0113e-02 | 4.1194e-02 | 2.6434e-02 | 9.3836e-02 |
2−4 | 1.1725e-02 | 7.7002e-03 | 3.2815e-03 | 1.5731e-02 |
2−5 | 3.5428e-03 | 2.0901e-03 | 4.1104e-04 | 3.6282e-03 |
2−6 | 7.1862e-04 | 4.4337e-04 | 5.3622e-05 | 7.3375e-04 |
Rate | 2.0413 | 2.1793 | 2.9818 | 2.3329 |
Concretely, we take the decoupling method used in Table 3 as an example, i.e. first solving Navier-Stokes equation then solving Elasticity equation, to explain the decoupling computational process. First, we give the following fully discrete scheme:
ρ1[vnht,w1h]Ω1+a1[vnh,w1h]+b[w1h,pnh]−b[vnh,qh]+(vn−1h⋅∇)vnh+∫I0(vnh−un−1ht)w1hds=ρ1[fn1,w1h]Ω1−∫I0[μ2(∇uh+∇uTh)n2+λ(divuhn2)]n−1w1hds |
ρ2[unhtt,w2h]Ω2+a2[unh,w2h]+∫I0(unht−vnh)w2hds=ρ2[fn2,w2h]Ω2+∫I0[pn1−μ1(∇vh+∇vTh)n1]nw2hds, |
then the corresponding matrix representations of foregoing equations can be written by imitating the process before. What we need emphasize is that we substitute the coupled Neumann interface ∫I0[pn1−μ1(∇vh+∇vTh)n1]nw1hds by the known value of n−1, i.e, −∫I0[μ2(∇uh+∇uTh)n2+λ(∇⋅uhn2)]n−1w1hds in Navier-Stokes equation; then we substitute the coupled Neumann interface ∫I0[μ2(∇uh+∇uTh)n2+λ(∇⋅uhn2)]nw2hds by the updated value of Navier-Stokes equation, i.e, ∫I0[pn1−μ1(∇vh+∇vTh)n1]nw2hds in the Elasticity equation.
In the end, we need to explain the second method in Table 4. When treating the coupled interface I0 in the Elasticity equation, we substitute ∫I0[pn1−μ1(∇vh+∇vTh)n1]n−1w2hds for corresponding term; When treating the coupled interface I0 in the Navier-Stokes equation, we substitute ∫I0[μ2(∇uh+∇uTh)n2+λ(∇⋅uhn2)]nw1hds for corresponding term.
It can be seen that the convergence rates for the velocity, pressure of the fluid and displacement of the solid are the same as previous coupled method.
This work is supported by NSF of China (Grant No. 11771259), special support program to develop innovative talents in the region of Shaanxi province.
The authors would like to thank the editor and reviewers for the constructive comments and useful suggestions, which improved the manuscript.
All authors contributed equally to the manuscript and read and approved the final manuscript.
The authors declare that they have no competing interests.
[1] | R. A. Adams, Sobolev Spaces, Academic press, New York, 1975. |
[2] |
M. Astorino and C. Grandmont, Convergence analysis of a projection semi-implicit coupling scheme for fluid-structure interaction problems, Numer. Math., 116 (2010), 721-767. doi: 10.1007/s00211-010-0311-x
![]() |
[3] |
D. Boffi, L. Gastaldi, A fictitious domain approach with Lagrange multiplier for fluid-structure interactions, Numer. Math., 135 (2017), 711-732. doi: 10.1007/s00211-016-0814-1
![]() |
[4] |
J. Boujot, Mathematical formulation of fluid-structure interaction problems, RAIRO Model. Math. Anal. Numer., 21 (1987), 239-260. doi: 10.1051/m2an/1987210202391
![]() |
[5] |
T. Chacon-Rebollo, V. Girault, F. Murat, et al. Analysis of a Coupled Fluid-Structure Model with Applications to Hemodynamics, SIAM J. Numer. Anal., 54 (2016), 994-1019. doi: 10.1137/140991509
![]() |
[6] | Z. Chen, Finite Element Methods and Their Applications, Spring-Verlag, Heidelberg, 2005. |
[7] | P. G. Ciarlet, The Finite Element Method for Elliptic Problems, North-Holland, Amsterdam, 1978. |
[8] |
Q. Du, M. Gunzburger, L. Hou, et al. Analysis of a linear fluid-structure interaction problem, Disc. Cont. Dyn. Syst., 9 (2003), 633-650. doi: 10.3934/dcds.2003.9.633
![]() |
[9] |
Q. Du, M. Gunzburger, L. Hou, et al. Semidiscrete finite element approximations of a linear fluidstructure interaction problem, SIAM J. Nmer. Anal., 42 (2004), 1-29. doi: 10.1137/S0036142903408654
![]() |
[10] |
M. Fernández, Incremental displacement-correction schemes for incompressible fluid-structure interaction, Numer. Math., 123 (2013), 21-65. doi: 10.1007/s00211-012-0481-9
![]() |
[11] |
F. Flori and P. Orenga, Fluid-structure interaction: Analysis of a 3-D compressible model, Ann. Inst. H. Poincare Anal. Non Lineaire, 17 (2000), 753-777. doi: 10.1016/S0294-1449(00)00119-0
![]() |
[12] | V. Girault and P. A. Raviart, Finite Element Method for Navier-Stokes Equations: Theory and Algorithms, Springer-Verlag, Berlin, Heidelberg, 1987. |
[13] |
C. Grandmont and Y. Maday, Existence for an unsteady fluid-structure interaction problem, M2AN Math. Model. Numer. Anal., 34 (2000), 609-636. doi: 10.1051/m2an:2000159
![]() |
[14] | Y. He, Y. Lin, W. Sun, Stabilized finite element method for the non-stationary Navier-Stokes problem, Disc. Cont. Dyn. Syst. Series B, 6 (2006), 41-68. |
[15] |
F. Hecht and O. Pironneau, An Energy Stable Monolithic Eulerian Fluid-Structure Finite Element Method, Int. J. Numer. Meth. Fl., 85 (2017), 430-446. doi: 10.1002/fld.4388
![]() |
[16] |
J. G. Heywood and R. Rannacher, Finite element approximation of the nonstationary Navier-Stokes problem I: Regularity of solutions and second-order error estimates for spatial discretization, SIAM J. Numer. Anal., 19 (1982), 275-311. doi: 10.1137/0719018
![]() |
[17] |
J. G. Heywood and R. Rannacher, Finite-element approximations of the nonstationary NavierStokes problem. Part IV: Error estimates for second-order time discretization, SIAM J. Numer. Anal., 27 (1990), 353-384. doi: 10.1137/0727022
![]() |
[18] |
A. T. Hill and E. Süli, Approximation of the global attractor for the incompressible Navier-Stokes equations, IMA J. Numer. Anal., 20 (2000), 633-667. doi: 10.1093/imanum/20.4.633
![]() |
[19] |
G. Hou, J. Wang, A. Layton, Numerical methods for fluid-structure interaction-A review, Commun. Comput. Phys., 12 (2012), 337-377. doi: 10.4208/cicp.291210.290411s
![]() |
[20] |
G. Hsiao, R. Kleinman and G. Roach, Weak solutions of fluid-solid interaction problems, Math. Nachr., 218 (2000), 139-163. doi: 10.1002/1522-2616(200010)218:1<139::AID-MANA139>3.0.CO;2-S
![]() |
[21] |
R. A. Khurram and A. Masud, A multiscale/stabilized formulation of the incompressible NavierStokes equations for moving boundary flows and fluid-structure interaction, Comput. Mech., 38 (2006), 403-416. doi: 10.1007/s00466-006-0059-4
![]() |
[22] | W. Layton, Introduction to the Numerical Analysis of Incompressible Viscous Flows, Comput. Sci. Eng., SIAM, Philadelphia, 2008. |
[23] |
P. LeTallec and S. Mani, Numerical analysis of a linearized fluid-structure interaction problem, Numer. Math., 87 (2000), 317-354. doi: 10.1007/s002110000183
![]() |
[24] | J. Li, Y. He and Z. Chen, A new stabilized finite element method for the transient Navier-Stokes equations, Comput. Methods Appl. Mech. Engrg., 197 (2009), 22-35. |
[25] |
S. Micu and E. Zuazua, Asymptotics for the spectrum of a fluid/structure hybrid system arising in the control of noise, SIAM J. Math. Anal.,29 (1998), 967-1001. doi: 10.1137/S0036141096312349
![]() |
[26] | H. Morand and R. Ohayon, Fluid Structure Interaction: Applied Numerical Methods, Academic press, New York, 1975. |
[27] |
T. Richter, A fully Eulerian formulation for fluid structure interaction problems, J. Comput. Phys., 233 (2013), 227-240. doi: 10.1016/j.jcp.2012.08.047
![]() |
[28] | R. Temam, Navier-Stokes Equations, Theory and Numerical Analysis, Third ed., North-Holland, Amsterdam, 1984. |
1. | Jian Li, Jiawei Gao, Yu Shu, A parallel, non-spatial iterative, and rotational pressure projection method for the nonlinear fluid-fluid interaction, 2021, 165, 01689274, 119, 10.1016/j.apnum.2021.02.002 |
h | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
2−3 | 5.0313e-02 | 3.7129e-02 | 2.2842e-02 | 7.1309e-02 |
2−4 | 1.1786e-02 | 7.4693e-03 | 2.8885e-03 | 1.3052e-02 |
2−5 | 2.8876e-03 | 1.7159e-03 | 3.7417e-04 | 2.8635e-03 |
2−6 | 7.1868e-04 | 4.1590e-04 | 5.3240e-05 | 6.8775e-04 |
Rate | 2.0431 | 2.1601 | 2.9150 | 2.2320 |
τ | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
1/5 | 2.7888e-01 | 2.9410e-01 | 2.8505e-01 | 7.0934e-01 |
1/10 | 1.3926e-01 | 1.5211e-01 | 1.4489e-01 | 3.6428e-01 |
1/20 | 6.9729e-02 | 7.7373e-02 | 7.3046e-02 | 1.8516e-01 |
1/40 | 3.4912e-02 | 3.9019e-02 | 3.6674e-02 | 9.3457e-02 |
Rate | 0.9993 | 0.9714 | 0.9861 | 0.9747 |
h | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
2−3 | 8.3133e-02 | 4.3960e-02 | 2.2426e-02 | 7.2486e-02 |
2−4 | 1.4350e-02 | 7.8258e-03 | 2.8314e-03 | 1.3255e-02 |
2−5 | 3.0401e-03 | 1.7028e-03 | 3.6418e-04 | 2.8794e-03 |
2−6 | 7.2433e-04 | 4.7933e-04 | 5.1454e-05 | 6.8676e-04 |
Rate | 2.2809 | 2.1730 | 2.9226 | 2.2406 |
h | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
2−3 | 5.0113e-02 | 4.1194e-02 | 2.6434e-02 | 9.3836e-02 |
2−4 | 1.1725e-02 | 7.7002e-03 | 3.2815e-03 | 1.5731e-02 |
2−5 | 3.5428e-03 | 2.0901e-03 | 4.1104e-04 | 3.6282e-03 |
2−6 | 7.1862e-04 | 4.4337e-04 | 5.3622e-05 | 7.3375e-04 |
Rate | 2.0413 | 2.1793 | 2.9818 | 2.3329 |
h | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
2−3 | 5.0313e-02 | 3.7129e-02 | 2.2842e-02 | 7.1309e-02 |
2−4 | 1.1786e-02 | 7.4693e-03 | 2.8885e-03 | 1.3052e-02 |
2−5 | 2.8876e-03 | 1.7159e-03 | 3.7417e-04 | 2.8635e-03 |
2−6 | 7.1868e-04 | 4.1590e-04 | 5.3240e-05 | 6.8775e-04 |
Rate | 2.0431 | 2.1601 | 2.9150 | 2.2320 |
τ | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
1/5 | 2.7888e-01 | 2.9410e-01 | 2.8505e-01 | 7.0934e-01 |
1/10 | 1.3926e-01 | 1.5211e-01 | 1.4489e-01 | 3.6428e-01 |
1/20 | 6.9729e-02 | 7.7373e-02 | 7.3046e-02 | 1.8516e-01 |
1/40 | 3.4912e-02 | 3.9019e-02 | 3.6674e-02 | 9.3457e-02 |
Rate | 0.9993 | 0.9714 | 0.9861 | 0.9747 |
h | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
2−3 | 8.3133e-02 | 4.3960e-02 | 2.2426e-02 | 7.2486e-02 |
2−4 | 1.4350e-02 | 7.8258e-03 | 2.8314e-03 | 1.3255e-02 |
2−5 | 3.0401e-03 | 1.7028e-03 | 3.6418e-04 | 2.8794e-03 |
2−6 | 7.2433e-04 | 4.7933e-04 | 5.1454e-05 | 6.8676e-04 |
Rate | 2.2809 | 2.1730 | 2.9226 | 2.2406 |
h | ‖vn−vnh‖1 | ‖pn−pnh‖0 | ‖un−unh‖0 | ‖un−unh‖1 |
2−3 | 5.0113e-02 | 4.1194e-02 | 2.6434e-02 | 9.3836e-02 |
2−4 | 1.1725e-02 | 7.7002e-03 | 3.2815e-03 | 1.5731e-02 |
2−5 | 3.5428e-03 | 2.0901e-03 | 4.1104e-04 | 3.6282e-03 |
2−6 | 7.1862e-04 | 4.4337e-04 | 5.3622e-05 | 7.3375e-04 |
Rate | 2.0413 | 2.1793 | 2.9818 | 2.3329 |