316 lines
16 KiB
TeX
316 lines
16 KiB
TeX
% ******* PREAMBLE ************
|
|
% ******************************
|
|
\documentclass[a4paper,12pt]{article}
|
|
|
|
%\input{../../../common.tex}
|
|
%\input{manuscript/definitions.tex}
|
|
\usepackage{xcolor}
|
|
\usepackage{amsmath}
|
|
\usepackage{mathtools}
|
|
\usepackage{bm}
|
|
|
|
\newcommand{\journal}{AML}
|
|
\newcommand{\manuscript}{AML-D-20-01616: revision 1}
|
|
|
|
% Editing commands
|
|
\usepackage{ulem}
|
|
\newcommand{\ins}[1]{\textcolor{blue}{#1}}
|
|
\newcommand{\del}[1]{\textcolor{blue}{\sout{#1}}}
|
|
|
|
%\newcommand{\myparagraph}[1]{\paragraph*{#1}\mbox{}\\[6pt]}
|
|
\newcounter{countResp}
|
|
\newcommand{\resp}[1]{\refstepcounter{countResp}\thecountResp)}
|
|
%\myparagraph{\normalsize{WP\theWPcounter: #1}} \\[-20pt]
|
|
%s}
|
|
|
|
|
|
%\newcommand{\ReviewComments}[2]{\textcolor{blue}{\textit{\noindent\resp{}
|
|
%#1}}\\
|
|
%#2\emph{}\\%\vspace{0.5cm}
|
|
%}
|
|
|
|
\usepackage{fullpage}
|
|
|
|
\newcommand{\ReviewComments}[2]{\textcolor{blue}{\\#1}\\[1em]#2}
|
|
|
|
% Revision colors
|
|
\newcommand{\reva}[1]{\textcolor{orange}{#1}}
|
|
\newcommand{\revb}[1]{\textcolor{red}{#1}}
|
|
\newcommand{\revc}[1]{\textcolor{cyan}{#1}}
|
|
\newcommand{\revs}[1]{\textcolor{olive}{#1}}
|
|
|
|
\newcommand{\todo}[1]{\textcolor{red}{\textbf{TODO: #1}}}
|
|
|
|
|
|
\renewcommand{\vec}[1]{\boldsymbol{#1}} % for vectors
|
|
\newcommand{\ten}[1]{\boldsymbol{#1}} % for tensors
|
|
\newcommand{\vecrm}[1]{\boldsymbol{\mathrm{#1}}} % for vectors in mathrm style
|
|
\newcommand{\mat}[1]{\boldsymbol{#1}} % for matrices
|
|
\newcommand{\matrm}[1]{\boldsymbol{\mathrm{#1}}} % for matrices in mathrm style
|
|
\newcommand{\dd}{\mathrm{d}} % differential d
|
|
\newcommand{\determ}[1]{\mathrm{det} \left( #1 \right)} % determinant
|
|
\newcommand{\diver}[1]{\mathrm{div} \, #1} % divergence
|
|
\newcommand{\calli}[1]{\mathcal{#1}} % calligraphic letters
|
|
\newcommand{\pd}{\partial} % partial differentiation d
|
|
\newcommand{\eqnref}[1]{(\ref{#1})} % referencing equations
|
|
% \newcommand{\tsum}{{\textstyle\sum\limits}} % for small sums in large equations
|
|
\newcommand{\pfrac}[2]{\frac{\pd #1}{\pd #2}} % \pfrac{f(x,y)}{x} for partial derivative of f(x,y) with respect to x
|
|
\newcommand{\pfracl}[2]{\pd #1/\pd #2}
|
|
\renewcommand{\dfrac}[2]{\frac{\dd #1}{\dd #2}} % \dfrac{f(x,y)}{x} for total derivative of f(x,y) with respect to x
|
|
% \newcommand{\grad}{\nabla} % for the gradient
|
|
\newcommand{\norm}[2]{\left\| #2 \right\|_{#1}} % \norm{L_2}{x} for the L2-norm of x, \norm{\infty}{x} for the Inf-Norm of x
|
|
% \newcommand{\abs}[1]{| #1 |} % \abs{x} for the absolute value of x
|
|
\newcommand{\comment}[1]{} % put around regions that are not to be compiled
|
|
|
|
% "trademarks"
|
|
\newcommand{\mor}[1]{\textit{MOR}}
|
|
\newcommand{\pmor}[1]{\textit{pMOR}}
|
|
\newcommand{\ppod}[1]{\textit{POD}}
|
|
\newcommand{\fom}[1]{\textit{FOM}}
|
|
\newcommand{\rom}[1]{\textit{ROM#1}}
|
|
\newcommand{\prom}[1]{\textit{pROM#1}}
|
|
|
|
% structural model
|
|
\newcommand{\ds}{\vecrm{d}}
|
|
\newcommand{\rs}{\vecrm{R}^\mathrm{S}}
|
|
\newcommand{\ns}{n_\mathrm{s}}
|
|
|
|
% windkessel model
|
|
\newcommand{\p}{\vecrm{p}}
|
|
\newcommand{\rd}{\vecrm{R}^\mathrm{0D}}
|
|
\newcommand{\pv}{p_\mathrm{v}}
|
|
|
|
% POD-specific abbreviations
|
|
\newcommand{\T}[1]{#1^\mathrm{T}}
|
|
\newcommand{\proj}{\matrm{V}}
|
|
\newcommand{\projt}{\T{\proj}}
|
|
\newcommand{\dr}{\ds_\mathrm{r}}
|
|
|
|
% math
|
|
\DeclareMathOperator*{\argmin}{argmin}
|
|
\DeclareMathOperator*{\argmax}{argmax}
|
|
\DeclareMathOperator{\sign}{sign}
|
|
|
|
% inverse analysis
|
|
\newcommand{\param}{\vecrm{\mu}}
|
|
\newcommand{\jac}{\vecrm{J}}
|
|
|
|
\newcommand{\defeq}{\vcentcolon=}
|
|
|
|
\newcommand{\D}[1]{\textcolor{cyan}{#1}}
|
|
|
|
|
|
\begin{document}
|
|
\parindent = 0pt
|
|
|
|
|
|
%
|
|
\begin{center}
|
|
{\bf Response to the reviewer's comments } \\ (\manuscript)
|
|
\end{center}
|
|
|
|
|
|
Dear Professor Tucker,\\
|
|
|
|
We thank you and the reviewers for the comments and suggestions to improve our manuscript. In the next pages, we replied to each of the comments and included modifications in a new version accordingly.
|
|
|
|
We hope that after our response the manuscript can be accepted for publication in \journal. \\
|
|
|
|
%, in particular for the questions. \\% We will address the comment in the following lines.\\
|
|
|
|
%In response to the first reviewer, indeed updates on the pressure space definition and notation have been done to correctly defined our problem
|
|
%in the continuum setting. Also comments on the space-time regularity and restrictions for the
|
|
%computational domain were added, e.g. in Remark 6. \\
|
|
%
|
|
%Our analysis is done throughtout the article using homogeneous Dirichlet boundary conditions and thus our derivations do not contain the mechanical work as expected.
|
|
%We assume such case for matter of convenience since its extension follows by using standard techniques found in literature. Nevertheless, comments
|
|
%about flow intensification were added in Remark 2. \\
|
|
%
|
|
%Finally, we point out the inaccurate choice of words for our conclusion, which in this new version was updated. \\
|
|
%
|
|
%In response to the second reviewer, regarding the concern about limitations of first order time discretizations, we studied analogously to the work [1] second order
|
|
%approaches such as Crank-Nicolson and BDF(2) schemes for the monolithic iNSE-ALE problem. To our knowledge, only conditionally energy stable discrete formulations can be
|
|
%obtained with the usual techniques and extensions to pressure correction schemes still lack in unconditional energy stable results. \\
|
|
%
|
|
%Further, the question of unconditionally energy stable higher order iNSE-ALE formulation still remains open and further research must be done. \\
|
|
|
|
|
|
%We hope that after our response the manuscript can be accepted for publication in \journal. \\
|
|
|
|
Sincerely yours \\
|
|
|
|
Reidmen Ar\'{o}stica and Crist\'obal Bertoglio
|
|
|
|
%\end{document}
|
|
|
|
%%%% NEW REVIEWER
|
|
|
|
\setcounter{countResp}{0}
|
|
\newpage
|
|
%\setcounter{page}{1}
|
|
|
|
\begin{center}
|
|
{\bf Detailed reply to Referee \#1} \\ (\manuscript)
|
|
\end{center}
|
|
|
|
We thank the reviewer for her/his suggestions and comments and we are happy to read the positive comments about our work. Below, we reply in detail to all the raised issues.
|
|
For the sake of convenience, all changes in the revised manuscript version are marked in \textcolor{orange}{orange}.
|
|
|
|
|
|
\ReviewComments{p.1 If a function $p$ belongs to the functional space $L^2$, it is not possibble to fix its value at a point. Pressure should be fixed by zero-ing its mean value.}
|
|
{
|
|
Thanks for pointing this to us, we corrected it in the new version.
|
|
}
|
|
|
|
\ReviewComments{p.2 On Remark 1: if you address spatial discretizations here, you should mention the problem of convective stabilization.
|
|
For heart chambers, for instance, this is the crucial issue. If you address non-zero boundary conditions and refer to the trace theorem, you should mention
|
|
the restrictions of the computational mesh imposed by the trace theorem in finite element spaces.}
|
|
{
|
|
We thank the reviewer for the comments regarding the restrictions on the computational mesh. Remark 6 has been updated including them.
|
|
}
|
|
|
|
\ReviewComments{p.2 I do not understand Proposition 1 for a domain with moving boundary: it states that the change of the kinetic energy is spent to viscous dissipation.
|
|
However, the motion of the boundary performs a mechanical work, and the energy balance should account flow intensification due to the moving boundary, see, e.g. [12].}
|
|
{
|
|
We fully agree with the reviewer's comment. In our case, the mechanical work of the boundary does not appear provided the homogeneous Dirichlet boundary conditions originally imposed to the problem.
|
|
In this version, we added a comment regarding such fact in Remark 2.
|
|
}
|
|
|
|
\ReviewComments{p.2 after Remark 3: check carefully classification of all known methods. I suspect that it contains errors. At least, the paper [12] has another set of indices n+1, n, n, n+1
|
|
and this is important since it is proved to be unconditionally stable whereas Corollary 3 states that it has restrictions on the time step.}
|
|
{
|
|
We thank the reviewer for this comment. We noticed a typo in the discrete formulation (5) which could have induced misundestanding throughout the article.
|
|
We reviewed carefully the discrete formulation in [12], where $k$ there corresponds to $n+1$ in our manuscript.
|
|
Therefore, we updated the indices to (n+1, n, n, n+1) to match the proposed scheme in [12].
|
|
}
|
|
|
|
\ReviewComments{p.4 line -10: moving domain formulation is weird: $X_0 = 0$ is mapped to 1, $X_0$ = 6 is mapped to 1+5.4sin(8$\pi$t), i.e. spans the interval [-4.4, 6.4], which means degeneration
|
|
and tangling of the physical domain.}
|
|
{
|
|
The reviewer is right that $X_0 = 0$ is mapped to 1, nevertheless $X_0 = 6$ is mapped to $6 (1 + 0.9 sin(8 \pi t))$, spanning the interval $[0.6, 11.4]$ thus no mesh degeneration.
|
|
}
|
|
|
|
\ReviewComments{p.4 Fig 1: these curves imply that you lost boundary flow intensification term in the energy balance.}
|
|
{
|
|
We agree that this would be indeed the general case. However, since homogeneous Dirichlet boundary conditions are used in the numerical experiment, the energy balance does not include intensification terms.
|
|
}
|
|
|
|
\ReviewComments{p.4, first line of Conclusion: the authors are too ambitious to claim that they theoretically analyzed time discretization schemes.
|
|
Actually, they suggest hints to a future stability analysis.}
|
|
{
|
|
We agree that our choice of words was inaccurate. This has been updated.
|
|
}
|
|
|
|
\ReviewComments{General: in mathematical journals, integrals are completed by dx, dt, ds, etc.}
|
|
{
|
|
Thanks for this remark. It has been updated.
|
|
}
|
|
|
|
\ReviewComments{p1, line 4 of section 2: $J^t$; p2. (4): $w^{**}$ no subscript $_h$ here!; p.4. Boundary conditions are not set in section 5.; p.5 Fig. 1: use $M \star\star$ rather than M**.
|
|
p.1. Jacobian, is the standard, p.4. Schwarz, Monolithic.}
|
|
{
|
|
We acknowledge the comments regarding figure styling and typos. They have been updated in this version.
|
|
}
|
|
\pagebreak
|
|
|
|
|
|
%%%%%%%%%%%%%%%% NEXT REVIEWER
|
|
\newpage
|
|
\setcounter{countResp}{0}
|
|
%\setcounter{page}{1}
|
|
|
|
\begin{center}
|
|
{\bf Detailed reply to Referee \#2} \\ (\manuscript)
|
|
\end{center}
|
|
|
|
We thank the reviewer for her/his positive impression of the manuscript and for his/her suggestions and comments. Below, we reply in detail to all the raised issues.
|
|
For the sake of convenience, all changes in the revised manuscript version are marked in \textcolor{red}{red}.
|
|
|
|
%\vspace{-.5cm}
|
|
|
|
\ReviewComments{My main concern however is the limitation to first order method. In applications I do not think that first order schemes are of much interest.
|
|
At least in FSI, mostly higher BFD schemes [1] (also includin a stability analysis), the fractional step theta scheme [2] or variants of the Crank-Nicolson method [3]
|
|
are applied since they all give second order. Is it possible to include at least one of these schemes in the analysis?}
|
|
{
|
|
We thank the reviewer for the comment. Analogous to [1] second-order monolithic schemes for the iNSE-ALE problem can be proposed, e.g. using Crank-Nicolson and BDF(2) approaches. However, as obtained in [1] for the advection-diffusion problem, both CN and BDF(2) are only conditionally stable, depending on the domain velocity. % and thus out of the scope of this work.
|
|
We believe that obtaining a higher order in time, unconditionally energy stable scheme for the advection-diffusion equations in ALE form (what includes the iNSE) still remains an open problem.
|
|
We have now extended Remark 6 in order to explain so and refer to [1] for more details.
|
|
}
|
|
|
|
\ReviewComments{Also, there is much research on pressure correction schemes of higher order, (I guess not for ALE), but it would be worth to extend these
|
|
ideas to ALE formulations.}
|
|
{
|
|
We agree with the reviewer that there is a lack of higher order fractional-step schemes for the iNSE problem in ALE formalism. In this article, we provide
|
|
a Chorin-Temam approach using our finding for the first order monolithic schemes since unconditional energy stability can be archieved.
|
|
|
|
As mentioned in the previous comment, already with second order schemes, e.g. Crank-Nicolson and BDF(2), unconditional energy stability is an open problem for the monolithic case,
|
|
thus its extension to pressure correction schemes can only be expected with conditional energy stability if current techniques are used, a topic beyond the scope of this paper.
|
|
}
|
|
|
|
\ReviewComments{Notation: the index 't' is sometimes a subscript, sometimes a superscript, e.g. in Jt in line 4 of Section 2, but $J^t$ in (1). Maybe this could be unified?}
|
|
{
|
|
Thanks for pointing that inconsistency out, we acknowledge the lack of unification in the notation. We have updated the notation to consider all time indexes as superscripts.
|
|
}
|
|
|
|
\ReviewComments{Problem description: Do you not require any temporal regularity of the domain velocity w(t)?}
|
|
{
|
|
We fully agree with the reviewer, a temporal regularity assumption for the domain velocity was lacking. We have updated the regularity of the deformation mapping $\mathcal{X}$ to be at least $\mathcal{C}^1$ is space and time,
|
|
thus for $\frac{\partial \mathcal{X}}{\partial t} := \mathbf{w}$ to be $\mathcal{C}^1$ in space and $\mathcal{C}^0$ in time.
|
|
}
|
|
|
|
\ReviewComments{I think the assumption of an initial pressure p(0)=0 might cause problems. Is this compatible with the configuration that you use in your test case, i.e.
|
|
does it fit to the initial velocity and the right hand side, the Dirichlet data? How do you check that?}
|
|
{
|
|
Indeed it was a typo. It was removed and updated to the correct
|
|
pressure space of $L^2(\Omega^0)$ functions with zero-mean.
|
|
}
|
|
|
|
\ReviewComments{The notation based on stars is difficult to undestand. I see that it is very general and has advantages. But, if you refer to specific shcemes, I could not undestand
|
|
what exactly is meant without wrtiting it down. Maybe you can give details on the most relevant schemes that you are considering.}
|
|
{
|
|
We thank the reviewer for the comment and agree with it. Nevertheless it provides advantages when describing a wide range of schemes.
|
|
In particular, the items after Remark 4 where we adress specific choices of indices $(\star, \star\star, \ast, \ast\ast)$ are based on relevant schemes applied in different scenarios.
|
|
}
|
|
|
|
\ReviewComments{Proposition 4. What exactly is the role of (7). You derive it by testing with proper functions, thats fine. And I see that many terms cancel for the specific choice of alpha, beta
|
|
and star-star. But where does it go to?}
|
|
{
|
|
Thanks for pointing that out. We notice the lack of a detailed description which has been corrected in this version of the article. (7) was intended to describe the unconditional energy stability obtained
|
|
for the monolithic case whenever suitable conditions are applied (in this case: $\alpha = \beta = 1, \star\star = n$). Now, Proposition 4 includes the specific bound obtained from the analysis.
|
|
}
|
|
|
|
\ReviewComments{Numerical examples: Is u(0) a scalar? I guess, 2nd component is zero? Does this initial value fit to a zero pressure at time t=0?}
|
|
{
|
|
We thank the reviewer for the question. It was a typo in the definition of $u(0)$ that has been updated to the correct expression.
|
|
As guessed by the reviewer, the 2nd component indeed is 0.\\
|
|
As clarified above, there is no initial condition for the pressure since it was a typo.
|
|
%For the monolithic scheme, only the initial velocity is required for the time iteration whereas for the Chorin-Temam scheme, the initial pressure is computed solving $(FVS)_{0}$.
|
|
%Thus, no direct computation of $p_0$ is required.
|
|
}
|
|
|
|
\ReviewComments{Reference [1] has some html code.}
|
|
{
|
|
Thanks for pointing that out. It has been corrected.
|
|
}
|
|
|
|
\ReviewComments{Reference [9] has been published in 2018 in Mathematics and Computer in Simulation.}
|
|
{
|
|
Thanks, it has been updated.
|
|
}
|
|
|
|
\vspace{1cm}
|
|
|
|
{
|
|
[1] L. Formaggia, F. Nobile. Stability analysis of second-order time
|
|
accurate schemes for ALE-FEM, Comp. Meth. Appl. Math. Engrg. 193, 2004
|
|
|
|
[2] M. Razzaq, H. Damanik, J. Hron, A. Ouazzi, S. Turek, FEM multigrid
|
|
techniques for fluid-structure interaction with application to
|
|
hemodynamics, Applied Numerical Mathematics, 2012
|
|
|
|
[3] T. Richter and T. Wick. In T. Carraro and M. Geiger and S. Körkel
|
|
and R. Rannacher (Eds.), On time discretizations of fluid-structure
|
|
interactions, Multiple Shooting and Time Domain Decomposition
|
|
Methods. Springer, 2015
|
|
}
|
|
\end{document} |