Abstract
We propose two linear schemes (1st- and 2nd-order) for the generalized Newtonian flow with the shear-dependent viscosity, which combine the decoupling techniques with the projection methods. The linear stabilization terms mimic \(-k\partial _t \Delta {\varvec{u}}\) and \(-k\partial _{tt} \Delta {\varvec{u}}\) from the PDE point of view. By our schemes, each velocity component can be computed in parallel efficiently using the same solver \((I-\alpha ^{-1}k\Delta )^{-1}\) at every time level. We analyze the convergence rates of the (temporally) semi- and the fully-discrete schemes. The theoretical results are testified by the numerical experiments.
Similar content being viewed by others
Availability of Data and Materials
The author declares that the data used to support the findings of this study are included in the article.
References
Bae, H.-O.: Existence, regularity, and decay rate of solutions of non-Newtonian flow. J. Math. Anal. Appl. 231, 467–491 (1999)
Bae, H.-O.: Regularity criterion for generalized newtonian fluids in bounded domains. J. Math. Anal. Appl. 421, 489–500 (2015)
Baranger, J., Amri, H.E.: Estimateurs a posteriori d’erreur pour le calcul adaptatif d’écoulements quasi-newtonien. Modélisation mathématique et analyse numérique 25(1), 31–47 (1991)
Baranger, J., Najib, K.: Analyse numerique des ecoulements quasi-Newtoniensdont la viscositeobeit a la loi puissanceou la loi de carreau. Numer. Math. 58, 35–49 (1990)
Barrett, J.W., Liu, W.B.: Finite element error analysis of a quasi-Newtonian flow obeying the Carreau or power law. Numer. Math. 64, 433–453 (1993)
Bercovier, M., Engelman, M.: A finite-element method for incompressible non-newtonian flows. J. Comput. Phys. 36, 313–326 (1980)
Berselli, L.C., Diening, L., Růžička, M.: Optimal error estimate for semi-implicit space-time discretization for the equations describing incompressible generalized newtonian fluids. IMA J. Numer. Anal. 35, 680–697 (2015)
Bessonov, N., Sequeira, A., Simakov, S., Vassilevskii, Yu., Volpert, V.: Methods of blood flow modelling. Math. Model. Nat. Phenom. 11(1), 1–25 (2016)
Cea, J., Glowinski, R.: Méthodes numériques pour l’écoulement laminaire d’un fluide rigide viscoplastique incompressible. Int. J. Comput. Math. 3(1), 225–255 (1972)
Chen, T.F., Cox, C.L., Lee, H.C., Tung, K.L.: Least-squares finite element methods for generalized Newtonian and viscoelastic flows. Appl. Numer. Math. 60, 1024–1040 (2010)
Chow, S.-S., Carey, G.F.: Numerical approximation of generalized Newtonian fuids using Powell-Sabin-Heindl elements: I. theoretical estimates. Int. J. Numer. Methods Fluids 41, 1085–1118 (2003)
Diening, L., Prohl, A., Růžička, M.: Semi-implicit Euler scheme for generalized Newtonian fluids. SIAM J. Numer. Anal. 44(3), 1172–1190 (2006)
E, W., Liu, J.: Projection method. i: convergence and numerical boundary layers. SIAM. J. Numer. Anal. 32(4), 1017–1057 (1995)
E, W., Liu, J.: Projection method. ii: Godunov-ryabenki analysis. SIAM. J. Numer. Anal 33(4), 1597–1621 (1996)
Fuchs, M., Seregin, G.: A global nonlinear evolution problem for generalized Newtonian fluids: local initial regularity of the strong solution. Comput. Math. Appl. 53, 509–520 (2007)
Girault, V., Raviart, O.-A.: Finite Element Methods for Navier-Stokes Equations: Theory and Algorithms. Springer-Verlag, Berlin (1986)
Glowinski, R., Wachs, A.: On the numerical simulation of viscoplastic fluid flow. In: Glowinski, R., Xu, J. (eds.) Special Volume: Numerical Methods for Non-Newtonian Fluids, volume XVI of Handbook of Numerical Analysis, pp. 483–718. Elsevier, Amsterdam (2011)
Kaplický, P.: Regularity of flows of a non-Newtonian fluid subject to dirichlet boundary conditions. Z. Anal. Anwend. 24(3), 467–486 (2005)
Kim, J., Moin, P.: Application of a fractional-step method to incompressible Navier-Stokes equations. J. Comput. Phys. 59, 308–323 (1985)
Legat, V., Oden, I. T.: An Adaptive \(h-p\) Finite Element Methods for Incompressible Free Surface Flows of Generalized Newtonian Fluids. ICES REPORT 93-11. The Institute for computational engineering and sciences, The university of Texas at Austin, Austin, Texas pp. 78712 (1993)
Li, B., Ueda, Y., Zhou, G.: A second-order stabilization method for linearizing and decoupling nonlinear parabolic systems. SIAM J. Numer. Anal. 58, 2736–2763 (2020)
Lions, P. L.: Quelques Méthodes de Résolution des Problèmes aux Limites Non Linéaries. Dunod (1969)
Málek, J., Nečas, J., Rokyta, R., Růžička, M.: On weak solutions to a class of non-newtonian incompressible fluids in bounded three-dimensional domains. The case \(p \ge 2\). Adv. Differ. Equ. 6, 257–302 (2001)
Málek, J., Necǎs, J., Růžička, M.: On the non-Newtonian incompressible fluids. Math. Models. Methods. Appl. Sci. 3, 35–63 (1993)
Málek, J., Rajagopal, K.R., Růžička, M.: Existence and regularity of solutions and the stability of the rest state for fluids with shear dependent viscosity. Math. Models. Methods. Appl. Sci. 5, 789–812 (1995)
Murakawa, H.: A linear finite volume method for nonlinear cross-diffusion systems. Numer. Math. 136, 1–26 (2017)
Prohl, A., Růžička, M.: On fully implicit space-time discretiaztion of motions of incompressible fluid with shear-dependent viscosities: the case \(p\le 2\). SIAM J. Numer. Anal. 39(1), 214–249 (2002)
Růžička, M.: A note on steady flow of fluid with shear dependent viscosity. Nonlinear Anal. 30, 3029–3039 (1997)
Saito, N.: On the stokes equation with the leak and slip boundary conditions of friction type: regularity of solutions. Publ. RIMS 40, 345–383 (2004)
Sandri, D.: Sur l’approximation numérique des écoulements quasi-newtoniens dont la viscosité suit la loi puissance ou la loi de Carreau. RAIRO - Modélisation mathématique et analyse numérique 27(2), 131–155 (1993)
Sequeira, A.: Hemorheology: Non-newtonian Constitutive Models for Blood Flow Simulations. In: Farina, A., Mikelić, A., Rosso, F. (eds.) Non-Newtonian fluid mechanics and complex flows. Lecture Notes in Mathematics, vol. 2212, pp. 1–44. Springer, Cham (2016)
Shen, J.: On error estimates of some higher order projection and penalty-projection methods for Navier-Stokes equations. Numer. Math. 62, 49–73 (1992)
Shen, J.: On error estimates of the projection methods for the Navier-Stokes equations: second-order schemes. Math. Comput. 65, 1039–1065 (1996)
Funding
The author declares that the research of this study was supported by NSFC General Projects Nos. 12171071 and 12071061.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author declares that he has no competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The research of Guanyu Zhou was partially supported by NSFC General Projects No. 12171071 and No. 12071061.
Appendix A The Proof of (4.35)
Appendix A The Proof of (4.35)
For \(B_{n+1} \in C^s\), the system (4.34) admits a unique solution \((\tilde{{\varvec{e}}}^{n+1}, q^{n+1}) \in H^{s+1} \times H^s\) \((s=1,2)\). The task here is to derive the a-priori estimate (4.35). For \(i=1,2,\cdots ,m\), we reformulate (4.34) into
with \(\tilde{{\varvec{e}}}^{i+1} = \nabla q^{i+1} \cdot {\varvec{n}} = 0\) on \(\Gamma \). By \(\Vert B_i - B_j\Vert _{W^{2,\infty }} \le C|t_i - t_j|\), we have, for \(s=0,1\),
If we prove that
we can conclude (4.35) for \(1 \le i \le m\). Then it is not difficult to proceed to \(m+1 \le i \le 2m\), and furthermore to \(1\le i \le n\), by induction. Therefore, it suffices to prove (A.2). The proof consists of two steps: (i) first we apply the Z-transform to the equations of \(\{(\tilde{{\varvec{e}}}^{i+1}, q^{i+1})\}_{i=1}^m\); (ii) We derive the \(H^2\) and \(H^3\) regularity using the method of translation.
(i) For \(\zeta \in {\mathbb {S}} = \{ \zeta \in {\mathbb {C}} \mid |\zeta |=1\}\), we introduce the Z-transform of \(({\varvec{{\tilde{e}}}}^i,q^i,\bar{{\varvec{F}}}^i,p(t_i))\):
where \(\{({\varvec{{\tilde{e}}}}^i, q^i)\}_{i=2}^\infty \) is the solution to the system (A.1) with \((\bar{{\varvec{F}}}^{i+1}, p(t_{i+1})) = ({\varvec{0}},0)\) for \(i > m\). Multiplying (A.1) by \(\zeta ^{i+1}\), we see that \(({\varvec{{\tilde{e}}}}(\zeta ), q(\zeta ))\) satisfies the system
with \(\tilde{{\varvec{e}}}= \nabla q \cdot {\varvec{n}} = 0\) on \(\Gamma \). By the Parseval equality
we reduce the proof of (A.2) to the derivation of the a-priori estimate
The validation of (A.4) will be carried out in step (ii). As a previous step, we write out the variational form of (A.3) and investigate the ellipticity. For all \(({\varvec{v}}, r) \in H_0^1(\Omega )^d \times H^1(\Omega )\),
By Korn and Poincaré’s inequalities, there exist two constants \(C_{B1}\) and \(C_{B2}\) such that
Note that \(\zeta = \cos \theta + {\mathfrak {i}} \sin \theta \) with \(\theta \in [0,2\pi )\), and
For a suitable constant \(\kappa \in (0,1)\) (will be specified later), we consider the following three cases:
-
(1)
\(\cos \theta \in (1 -\kappa , 1]\). Since \(|1-\zeta |^2 = 2(1-\cos \theta )\), under the assumption that \(\alpha > 2\kappa /C_{B1}\), the real part of \(a({\varvec{v}},r; {\varvec{v}},r)\) is coercive:
$$\begin{aligned} \begin{aligned} \mathrm {Re}[a({\varvec{v}},r; {\varvec{v}},r)]&\ge \left( (2\cos \theta + 1 - 2\cos ^2\theta )C_{B1} - \alpha ^{-1}2\cos \theta (1-\cos \theta ) \right) \Vert \nabla {\varvec{v}}\Vert _{L^2}^2 \\&\quad + \beta k (1-\cos \theta )\Vert \nabla q \Vert _{L^2}^2 \\&\ge ( C_{B1} - \alpha ^{-1} 2\kappa ) \Vert \nabla {\varvec{v}}\Vert _{L^2}^2 + \beta k |1-\zeta |^2 \Vert \nabla r \Vert _{L^2}^2/2 \ge 0. \end{aligned} \end{aligned}$$ -
(2)
\(\cos \theta \in [-1, -1 + \kappa )\). If \(\alpha < 2(2-\kappa )(1-\kappa )/(3C_{B2})\), then the coercivity of the real part holds true:
$$\begin{aligned} \begin{aligned} \mathrm {Re}[a({\varvec{v}},r; {\varvec{v}},r)]&\ge \left( \alpha ^{-1}2(1-\kappa )(2-\kappa ) - 3C_{B2} \right) \Vert \nabla {\varvec{v}}\Vert _{L^2}^2 + \beta k |1-\zeta |^2 \Vert \nabla q \Vert _{L^2}^2/2 \ge 0. \end{aligned} \end{aligned}$$ -
(3)
\(\cos \theta \in [-1+\kappa , 1-\kappa ]\). We have \(|1-\zeta |^2 \in [2\kappa , 4-2\kappa ]\). When \(\sin \theta \in [\sqrt{\kappa (2-\kappa )},1]\), if \(\alpha < 1/(2C_{B2})\), we find that
$$\begin{aligned} \begin{aligned} \mathrm {Re}[{\mathfrak {i}}a({\varvec{v}},r; {\varvec{v}},r)]&\ge ( \alpha ^{-1} \sin \theta (1-\cos \theta ) - \sin \theta (1-\cos \theta )2C_{B2} ) \Vert \nabla {\varvec{v}}\Vert _{L^2}^2 + \beta k \sin \theta \Vert \nabla q\Vert _{L^2}^2 \\&\ge \sin \theta (1-\cos \theta )(\alpha ^{-1} - 2C_{B2}) \Vert \nabla {\varvec{v}}\Vert _{L^2}^2 + \beta k |1-\zeta |^2 \Vert \nabla q\Vert _{L^2}^2/4 \ge 0. \end{aligned} \end{aligned}$$When \(\sin \theta \in [-1, -\sqrt{\kappa (2-\kappa )}]\), we have
$$\begin{aligned} \begin{aligned} \mathrm {Re}[-{\mathfrak {i}}a({\varvec{v}},r; {\varvec{v}},r)]&\ge -\sin \theta (1-\cos \theta )(\alpha ^{-1} - 2C_{B2}) \Vert \nabla {\varvec{v}}\Vert _{L^2}^2 + \beta k |1-\zeta |^2 \Vert \nabla q\Vert _{L^2}^2/4 \ge 0. \end{aligned} \end{aligned}$$
In summary, for sufficiently small \(\alpha \) and \(\kappa \in (0,1)\) satisfying \(2\kappa /C_{B1}< \alpha < \min ((2C_{B2})^{-1}, 2(2-\kappa )(1-\kappa )/(3C_{B2}))\), there exists a constant \(C>0\) such that
(ii) In this step, we prove (A.4) by the method of translation. Without loss of generality, we consider two simple cases: (1) \(\Omega = {\mathbb {R}}^d\) and (2) \(\Omega = {\mathbb {R}}^d_+ := {\mathbb {R}}^d \cap \{x_d >0\}\), and assume that \(({\varvec{{\tilde{e}}}},q)\) has compact support. The case of the general smooth domain can be reduced to these two cases essentially (cf. [29]). We denote by \(({\varvec{\xi }}_1,\cdots ,{\varvec{\xi }}_d) = I \in {\mathbb {R}}^{d\times d}\).
-
(1)
\(\Omega = {\mathbb {R}}^d\). For \(i=1,2,\cdots ,d\) and \(D_h w({\varvec{x}}) = (w({\varvec{x}} + h{\varvec{\xi }}_i) - w({\varvec{x}}))/h\), we substitute \(({\varvec{v}},q) = (D_{-h}D_h {\varvec{{\tilde{e}}}}({\varvec{x}}), D_{-h}D_h q({\varvec{x}}))\) into (A.5) and obtain (by (A.6)):
$$\begin{aligned} \begin{aligned}&C(\Vert \nabla D_h{\varvec{{\tilde{e}}}} \Vert _{L^2}^2 + \beta k |1-\zeta |^2 \Vert \nabla D_h q\Vert _{L^2}^2) - C\Vert D_h B_2\Vert _{L^\infty } \Vert {\mathbb {D}}(D_h{\varvec{{\tilde{e}}}}) \Vert _{L^2} \Vert {\mathbb {D}}({\varvec{{\tilde{e}}}}) \Vert _{L^2} \\&\le C\Vert {\varvec{G}}\Vert _{L^2} \Vert D_{-h}D_h {\varvec{{\tilde{e}}}}\Vert _{L^2} + k( \Vert \Delta D_{-h}D_h {\bar{p}}\Vert _{L^2} + \Vert \Delta D_{-h}D_h p^1\Vert _{L^2} )\Vert q\Vert _{L^2}, \end{aligned}\nonumber \\ \end{aligned}$$(A.7)which implies (A.4) with \(s=0\). Differentiating (A.3) with respect to \(x_i\), we can prove (A.4) with \(s=1\) in a similar manner.
-
(2)
\(\Omega = {\mathbb {R}}^d_+\). For \(i=1,2,\cdots ,d-1\), we substitute \(({\varvec{v}},q) = (D_{-h}D_h {\varvec{{\tilde{e}}}}({\varvec{x}}), D_{-h}D_h q({\varvec{x}}))\) into (A.5) and obtain (A.7), which yields: for \(i=1,2,\cdots ,d-1\),
$$\begin{aligned} \Vert \partial _{x_i} {\varvec{{\tilde{e}}}} \Vert _{H^1}^2 + \beta k |1-\zeta |^2 \Vert \nabla \partial _{x_i} q\Vert _{L^2}^2 \le C(\Vert {\varvec{G}}\Vert _{L^2}^2 + k^2 \Vert {\bar{p}}\Vert _{H^4}^2 + k^2 \Vert p^1\Vert _{H^4}^2). \end{aligned}$$(A.8)For \(i=1,\cdots ,d-1\), differentiating (A.3b) with \(x_i\), we get
$$\begin{aligned} -k\beta (1-\zeta )\Delta \partial _{x_i} q = k \partial _{x_i} g - \nabla \cdot \partial _{x_i} {\varvec{{\tilde{e}}}}, \end{aligned}$$(A.9)with \(\nabla \partial _{x_i} q \cdot {\varvec{n}} = 0\) on \(\Gamma = {\mathbb {R}}^d \cap \{x_d= 0\}\). Multiplying (A.9) by \((1-{\bar{\zeta }})/|1-\zeta |\) and testing by \(D_{-h}D_h \partial _{x_i} q\), we derive the estimate:
$$\begin{aligned} \beta k |1-\zeta | \Vert \partial _{x_i} q \Vert _{H^2} \le C( k \Vert {\bar{p}}\Vert _{H^3} + k \Vert p^1\Vert _{H^3} + \Vert \nabla \cdot \partial _{x_i} {\varvec{{\tilde{e}}}}\Vert _{L^2}) \quad (i=1,\cdots ,d-1). \nonumber \\ \end{aligned}$$(A.10)Setting
$$\begin{aligned} {\varvec{U}} = {\varvec{{\tilde{e}}}} - \beta k (1-\zeta ) \nabla q, \quad Q = q + \beta k (1-\zeta ) (\alpha ^{-1} (1-\zeta )^2 + (1-(1-\zeta )^2)b_2)\Delta q, \end{aligned}$$we find that \(({\varvec{U}}, Q)\) satisfies the system:
$$\begin{aligned} -(1-(1-\zeta )^2)\nabla \cdot [b_2 {\mathbb {D}}({\varvec{U}})] - \alpha ^{-1}(1-\zeta )^2 \Delta {\varvec{U}} + \nabla Q = \hat{{\varvec{G}}} , \quad \quad \nabla \cdot {\varvec{U}} = kg, \nonumber \\ \end{aligned}$$(A.11)where \(\hat{{\varvec{G}}} := {\varvec{G}} + (1-(1-\zeta )^2) \Big ( \nabla \cdot [(B_2 - b_2 {\mathbb {I}}) {\mathbb {D}}({\varvec{U}}-{\varvec{{\tilde{e}}}})] + \beta k (1-\zeta ) (\nabla b_2 \cdot \nabla ^2 q + \Delta q \nabla b_2) \Big )\). Because \(\Vert {\varvec{{\tilde{e}}}}\Vert _{H^\frac{3}{2}(\Gamma )}\) and \(\Vert {\varvec{U}}\Vert _{H^\frac{3}{2}(\Gamma )}\) only involve \(\Vert \partial _{x_i} {\varvec{{\tilde{e}}}}\Vert _{H^1}\) and \(\Vert \partial _{x_i} \nabla q\Vert _{H^1}\) with \(i=1,\cdots ,d-1\),
$$\begin{aligned} \begin{aligned} \Vert {\varvec{{\tilde{e}}}}\Vert _{H^\frac{3}{2}(\Gamma )} + \Vert {\varvec{U}}\Vert _{H^{3/2}(\Gamma )}&\le C(\Vert {\varvec{G}}\Vert _{L^2} + k \Vert {\bar{p}}\Vert _{H^3} + k \Vert p^1\Vert _{H^3} + \Vert \nabla \cdot \partial _{x_i} {\varvec{{\tilde{e}}}}\Vert _{L^2}) \quad \text {(by} (A.10))\\&\le C(\Vert {\varvec{G}}\Vert _{L^2} + k \Vert {\bar{p}}\Vert _{H^4} + k \Vert p^1\Vert _{H^4}). \quad \text {(by} (A.8)) \end{aligned}\nonumber \\ \end{aligned}$$(A.12)Therefore, we achieve (by the regularity of Stokes equations)
$$\begin{aligned} \begin{aligned} \Vert {\varvec{U}}\Vert _{H^2} + \Vert Q\Vert _{H^1}&\le C(\Vert \hat{{\varvec{G}}} \Vert _{L^2} + k \Vert g\Vert _{H^1} + \Vert {\varvec{U}}\Vert _{H^{3/2}(\Gamma )}) \\&\quad \le C(\Vert {\varvec{G}} \Vert _{L^2} + \Vert B_2 - b_2 {\mathbb {I}}\Vert _{W^{1,\infty }} (\Vert {\varvec{U}}\Vert _{H^2} + \Vert {\varvec{{\tilde{e}}}}\Vert _{H^2}) + k\Vert q\Vert _{H^2} \\&\quad + k \Vert {\bar{p}}\Vert _{H^4} + k \Vert p^1\Vert _{H^4}). \end{aligned} \end{aligned}$$Under the assumption that \(\Vert B_i - b_i {\mathbb {I}}\Vert _{L^\infty } \le \epsilon \) \((\forall i \ge 2)\) with sufficiently small \(\epsilon \), and in view of \(k\Vert q\Vert _{H^2} \le C\Vert {\varvec{{\tilde{e}}}}\Vert _{H^1} + k\Vert g\Vert _{L^2}\) (by (A.3b)), we conclude
$$\begin{aligned} \Vert {\varvec{U}}\Vert _{H^2} + \Vert Q\Vert _{H^1} \le C(\Vert {\varvec{G}}\Vert _{L^2} + \epsilon \Vert {\varvec{{\tilde{e}}}}\Vert _{H^2} + k \Vert {\bar{p}}\Vert _{H^4} + k \Vert p^1\Vert _{H^4}). \end{aligned}$$(A.13)Setting \(Z(\zeta ) = \alpha ^{-1} (1-\zeta )^2 + (1-(1-\zeta )^2)b_2\), we obtain from (A.3a\(- \nabla Z(\zeta )\) (A.3b)) that
$$\begin{aligned} -(1-(1-\zeta )^2)\nabla \cdot [B_2 {\mathbb {D}}({\varvec{{\tilde{e}}}})] - \alpha ^{-1}(1-\zeta )^2 \Delta {\varvec{{\tilde{e}}}} - \nabla [Z(\zeta ) \nabla \cdot {\varvec{{\tilde{e}}}}] = -\nabla Q + {\varvec{G}} + k \nabla [Z(\zeta ) g], \end{aligned}$$which implies (by the coercivity of \(a({\varvec{v}},r ;{\varvec{v}},r) + (Z(\zeta )\nabla \cdot {\varvec{v}}, \nabla \cdot {\varvec{v}})\))
$$\begin{aligned} \begin{aligned} \Vert {\varvec{{\tilde{e}}}} \Vert _{H^2}&\le C (\Vert -\nabla Q + {\varvec{G}} + k \nabla [Z(\zeta ) g] \Vert _{L^2} + \Vert {\varvec{{\tilde{e}}}} \Vert _{H^{3/2}(\Gamma )}) \\&\le C ( \Vert {\varvec{G}}\Vert _{L^2} + k \Vert {\bar{p}}\Vert _{H^4} + k \Vert p^1\Vert _{H^4} ) \quad \text {(by }(A.13), (A.8) \text { and } (A.12)). \end{aligned} \end{aligned}$$We conclude (A.4) for \(s=0\). The case of \(s=1\) can be handled similarly.
Rights and permissions
About this article
Cite this article
Zhou, G. The Convergence Analysis of Semi- and Fully-Discrete Projection-Decoupling Schemes for the Generalized Newtonian Models. J Sci Comput 91, 57 (2022). https://doi.org/10.1007/s10915-022-01828-5
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10915-022-01828-5