• Žitný BC project for Glasgow students 2015 (frozen since September 2015) RZ10 Metzner White convergent/divergent Gauss planar channel flow GAČR = kolagen Motto: It is a big mistake for teachers that they do not ask students about their dreams and ideas before creative genes of pure-minded students came under the tutelage of tutors. And an even bigger mistake is when the students afraid to talk about their crazy ideas. 1 History of project The idea, how to determine the relaxation time of viscoelastic fluid by comparing the pressure drop at elongation and compression flows in a converging/diverging slot, think about everyone after reading the first book about rheology. The problem is in a technical realization of viscoelastic stresses measurement along a channel with a variable cross section: Possible solution was to use the concept of groove strain-gauge sensors for measuring the axial pressure profile in the annular gap of existing extrusion rheometer (the development of these sensors was implemented from the beginning of the seventies). The first practical experiments (convergent/divergent gap between the cylindrical tube with attached sensors and internal double-cone pin) were up in the early nineties. Relevant results were summarized in Pavlovec thesis (theoretical solution assuming the analytic velocity profiles corresponding to a power fluid flow and numerical solutions of viscoelastic fluid type Metzner White) and in thesis by Zeleny (experiment with collagen and a double-cone pin - certain differences in pressure losses in the convergent/divergent slit were recorded). Mr. Pavlovec was an excellent student, but its thesis has not been completed (Fortran programs did not converged and the result was reduced to extremely simplified 1D analysis). The problem is simply too complicated for single student without sufficient background. And that is the true goal of this presentation: to prepare a groundwork for english speaking students from Glasgow. 2 History of project Previous presentation RZ8 was almost duplicated thesis of Pavlovec, programmed in MATLAB with certain exemptions in differencing (but basic simplification consisting in using the power law velocity profile, was preserved). Although solutions converge, the results raise some doubts, especially the occurrence of strange peaks of calculated pressures in the axis of the channel and also the unexpected effect of elasticity: Pressure loss in the convergent section was lower than in the divergent section (see following sheet No. 13). Independent solutions and programming performed by student (Mr. Frouz) reportedly received identical results (presented at the Students Conference). Still, I'd be very happy if he could express his findings in the form of diagrams on a slide 13, so that a consensus was evident. The following diagram is a visualization of what's happening with a loaded fiber. According to this concept the normal stress on the wall should be lower than during the stretching of the fiber during its compression. But it is at odds with the results of the model (see sheet 7). Stretched fiber Compressed fiber Equilibrium fiber This visualization is perhaps misleading (it say that the normal stress in the zone of elongation is less than in the compression zone), because the de-facto ignores the influence of time. I will be grateful to anyone who offers a better interpretation of what was really going on polymer fibers. 3 History of project discussion of simplifying assumptions (assumed velocity profile not affected by elastic forces) RZ8 model assuming linear cross-sectional profile of convergent/divergent slots along with the assumption of a power law velocity profile actually result to a discontinuity of elongation rate (positive constant rate of elongation in the convergent zone while a constant negative rate in the divergent zone). The discontinuity and weak singularity at the transition may cause unrealistic peak pressures and inaccuracies axial profile of normal stresses. These fundamental shortcomings could be eliminated using the model described in this presentation. The axial cross-sectional profile channel (and therefore the rate of elongation) is a smooth function (Gaussian curve), and the velocity field is calculated (based on the concept of the current function and vorticity). Technically speaking, the manufacture of the walls with the Gaussian profile should be no problem (using rapid prototyping - the geometry of the channel varies exchanging inserts, which must anyway be of non-conductive plastic which is then glued strips electrodes). The disadvantage is the computational complexity of relationships and the possibility of an error (because in this presentation are reported in respect of each field checked Kontroloval: .... xxxxx ..). Solution flow of viscoelastic fluid in a convergent / divergent slot can also be simulated by existing commercial software (Polyflow ANSYS) and OpenFOAM. These project can be alternatively solved by other Glasgow students. The ideal would be a comparison of approaches - it is a matter of debate. 4 Literature Crochet M.J., Davies A.R., Walters K.: Numerical simulation of non-Newtonian flow. Elsevier 1984 Pavlovec J.: Výtlačný reometr pro viskoelastické kapaliny. Dipl.práce, ČVUT FS, 1991 Anderson J.D.: Computational Fluid Dynamics. McGraw Hill, Inc., N.Y., 1995 Owens R.G., Phillips T.N.: Computational Rheology. Imperial College Press, London 2002 Additional papers Davidson et al. Velocity and stress fields of polymeric liquids flowing in a periodically constricted channel. J. Non-Newtonian Fluid Mech. 49 (1993) Baird D.G. First normal stress difference measurements…., J. Non-Newtonian Fluid Mech. 148 (2008) Baaijens F. Mixed FEM for viscoelastic flow analysis, a review. J. Non-Newtonian Fluid Mech. 79 (1998) James D.F., Chandler G.M.,Armour S.J. : A converging channel rheometer for the measurement of extensional viscosity. Elsevier 1998 (mam jen papirovou kopii bez bibl.udaju) James D.F., Chandler G.M., Armour S.J. : Measurement of the extensional viscosity of M1 in a converging channel rheometer. J. Non-Newtonian Fluid Mech. 35 (1990) Han Chang Dae: Measurement of the rheological properties of polymer melts with slit rheometer I.Homopolymer systems. J.Applied Polymer Science, 15 (1971) Davies J.M. et al: Theory for normal stresses in slits and capillaries. J. Phys. D. : Appl.Phys, Vol.6 (1973), 2259-2266 Cogswell F.N. Converging flow of polymer melts in extrusion dies. Polymer Engngr. and Sci. (1972) Cogswell F.N. Converging flow and stretching flow a compilation. J. Non-Newtonian Fluid Mech. Vol.4 (1978), 23-38 Binding D.M. An approximate analysis for contraction and converging flows. J. Non-Newtonian Fluid Mech. Vol.27 (1988), 173-189 5 Diploma work J.Pavlovec: Výtlačný reometr pro viskoelastické kapaliny. 1991 Only INFO old presentation RZ8 (skip) Adviser R.Žitný, oponent M.Houška Convergent section with positive elongation Divergent section compression of fibers Constitutive model UCM (upper convective Maxwell) modification White-Metzner 𝜇 𝐼𝐼 𝜏𝑖𝑗 + 𝐺 𝜕𝜏𝑖𝑗 𝜕𝑢𝑖 𝜕𝑢𝑗 𝑢𝑘 − 𝜏 − 𝜏 = 𝜇 𝐼𝐼 𝜕𝑥𝑘 𝜕𝑥𝑘 𝑘𝑗 𝜕𝑥𝑘 𝑖𝑘 Relaxation time 𝜕𝑢𝑖 𝜕𝑢𝑗 + 𝜕𝑥𝑗 𝜕𝑥𝑖 Viscosity 6 Vague idea of Maxwellian model (maxwell.xlsx) Vague because it does not specify the nature of the stress (imagine that it is for example an elongation stress in the direction of flow). The model considers a linear change in velocity u and strain rate along a channel. 𝜏 + 𝜆 𝑢0 + Δ𝑢 ∙ 𝑥 u0 du g0 dg sig0 lam dx mju 1 0.1 1 0.2 10 0.005 0.01 10 𝑑𝜏 = 𝜇 𝛾0 + Δ𝛾 ∙ 𝑥 𝑑𝑥 𝜆= 𝜇 𝐺 Only INFO old presentation RZ8 (skip) relaxation time [s] Particular solution 𝜇 𝜏𝑃 = 𝛾 + 𝜆 𝛾0 Δ𝑢 − 𝑢0 Δ𝛾 + Δ𝛾𝑥 1 + 𝜆∆𝑢 0 General solution 𝜇 1 𝑢0 𝜏= 𝛾 + 𝜆 𝛾0 Δ𝑢 − 𝑢0 Δ𝛾 + Δ𝛾𝑥 + 𝑐 ∙ 𝑒𝑥𝑝 − ln(𝑥 + ) 1 + 𝜆∆𝑢 0 𝜆∆𝑢 ∆𝑢 5 Effect of relaxation time increase to axial stress profile (, 2, 5) Weissenberg number (elasticity number) 𝑊 = = relaxation time * rate of deformation 7 𝑢 = 𝐺ℎ Pavlovec : dimensionless parameters Transformation into an orthogonal coordinate system Kontroloval:FROUZ Frouz Kontroloval Only INFO old presentation RZ8 (skip) H H LI L L LI L L x 𝑥 − 𝑥0 = ∆𝐿 𝑦 𝑦 𝑦 𝜂= = = 𝑥−𝑥 ℎ(𝑥) 𝐻0 1 + 𝛽 𝐻0 1 + 𝛽 ∆𝐿 0 Inlet section 𝑥0 = 0, ∆𝐿 = 𝐿𝐼 𝛽 = 0, 𝐻0 = 𝐻 Convergent section 𝑥0 = 𝐿𝐼 , ∆𝐿 = 𝐿 ∆𝐻 𝛽=− ,𝐻 = 𝐻 𝐻0 0 Transformation of derivatives 𝜕𝐹 𝜕𝐹 𝜕 𝜕𝐹 𝜕 1 𝜕𝐹 𝜕𝐹 = + = − 𝜕𝑥 𝜕 𝜕𝑥 𝜕 𝜕𝑥 𝐿 𝜕 1 + 𝜕 Divergent section 𝑥0 = 𝐿 + 𝐿𝐼 , ∆𝐿 = 𝐿 ∆𝐻 𝛽= , 𝐻 = 𝐻 − ∆𝐻 𝐻0 0 𝜕𝐹 𝜕𝐹 𝜕 𝜕𝐹 𝜕 1 𝜕𝐹 1 𝜕𝐹 = + = = 𝜕𝑦 𝜕 𝜕𝑦 𝜕 𝜕𝑦 𝐻0 1 + 𝜕 𝐿 1 + 𝜕 8 Power law approximation of velocities Kontroloval FROUZ Only INFO old presentation RZ8 (skip) Axial velocity 𝑢𝑚 𝑢𝑥 = (1 − 1 − 2𝜂 1 + 𝛽𝜉 𝑢𝑚 𝑢𝑥 = 1 − 2𝜂 − 1 1 + 𝛽𝜉 𝑢𝑚 = 𝑛+1 𝑛 ), 0.5 𝑛+1 𝑛 , > 0.5 𝑉 2𝑛 + 1 𝐻0 𝑛 + 1 Transversal component (satisfying continuity equation 𝜀𝛽𝑢𝑚 𝑢𝑦 = 𝜂(1 − 1 − 2𝜂 1 + 𝛽𝜉 𝜀𝛽𝑢𝑚 𝑢𝑦 = 𝜂(1 − 2𝜂 − 1 1 + 𝛽𝜉 𝜀= 𝜕𝑢𝑥 𝜕𝑢𝑦 + = 0) 𝜕𝑥 𝜕𝑦 𝑛+1 𝑛 ), 0.5 𝑛+1 𝑛 ), > 0.5 𝐻0 ∆𝐿 9 Power law approximation of viscosity Kontroloval FROUZ Only INFO old presentation RZ8 (skip) Viscosity as a power function of the second invariant 𝜇 𝐼𝐼 = 𝐾 𝜕𝑢𝑥 2 𝜕𝑥 2 𝜕𝑢𝑦 +2 𝜕𝑦 2 𝜕𝑢𝑥 𝜕𝑢𝑦 + + 𝜕𝑦 𝜕𝑥 2 𝑛−1 Rate of deformation in the dimensionless coordinates (0,1), (0,1) 𝜕𝑢𝑥 1 𝜕𝑢𝑥 𝜕𝑢𝑥 = − 𝜕𝑥 𝐿 𝜕 1 + 𝜕 𝜕𝑢𝑥 1 𝜕𝑢𝑥 = 𝜕𝑦 𝐿 1 + 𝜕 𝜕𝑢𝑦 1 𝜕𝑢𝑦 𝜕𝑢𝑦 = − 𝜕𝑥 𝐿 𝜕 1 + 𝜕 𝜕𝑢𝑦 𝜕𝑢𝑦 1 = 𝜕𝑦 𝐿 1 + 𝜕 10 Constitutive equation White Metzner 𝑥𝑥 = −𝑝 + 𝑥𝑥 𝑦𝑦 = −𝑝 + 𝑦𝑦 𝑥𝑦 = 𝑥𝑦 𝜏𝑥𝑥 + 𝜇 𝐼𝐼 𝐺 𝜇 𝐼𝐼 𝜏𝑦𝑦 + 𝐺 𝜏𝑥𝑦 + 𝜇 𝐼𝐼 𝐺 𝑢𝑥 Kontroloval FROUZ Only INFO old presentation RZ8 (skip) An important conclusion: the lower flat plate is yy = 0, which means that the pressure sensor measures the isotropic pressure p 𝜕𝜏𝑥𝑥 𝜕𝜏𝑥𝑥 𝜕𝑢𝑥 𝜕𝑢𝑥 𝜕𝑢𝑥 + 𝑢𝑦 −2 𝜏𝑥𝑥 − 2 𝜏𝑥𝑦 = 2𝜇 𝐼𝐼 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝜏𝑦𝑦 𝜕𝜏𝑦𝑦 𝜕𝑢𝑦 𝜕𝑢𝑦 𝜕𝑢𝑦 𝑢𝑥 + 𝑢𝑦 −2 𝜏 −2 𝜏 = 2𝜇 𝐼𝐼 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝑥𝑦 𝜕𝑦 𝑦𝑦 𝜕𝑦 𝑢𝑥 𝜕𝜏𝑥𝑦 𝜕𝜏𝑥𝑦 𝜕𝑢𝑦 𝜕𝑢𝑥 + 𝑢𝑦 − 𝜏𝑥𝑥 − 𝜏 = 𝜇 𝐼𝐼 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝑦𝑦 𝜕𝑢𝑥 𝜕𝑢𝑦 + 𝜕𝑦 𝜕𝑥 11 Pressure distribution by integration Cauchy equations from left. Results 7 0 G=150 x 10 diff=-0.16817 Vdot=0.000375 white-metzner figure(1) Only%main INFO old presentation RZ8 (skip) clear all; plot(x(1:ig2),pp(1:ig2)) % white metzner - velocities % rychlosti ux(1:nx,1:ny),uy(1:nx,1:ny), % derivace duxdksi(1:nx,1:ny), duxdeta(1:nx,1:ny), duydksi(1:nx,1:ny), duydeta(1:nx,1:ny) % duxdx(1:nx,1:ny), duxdy(1:nx,1:ny), duydy(1:nx,1:ny), duydx(1:nx,1:ny) % viskozity mju(1:nx,1:ny) % beta, dl, h0, n, vdot, K G koeficient konzistence a modul G -0.5 -1 p [Pa] -1.5 -2 -2.5 -3 -3.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 x [m] PRESSURE 160 140 120 x 100 80 60 40 20 Peek pressure on the axis of the channel is about an approximation error 5 10 15 20 y 25 30 K=100; n=0.5; nx=59; ny=37; dl=0.2; h=0.005; h0=h; dh=0.004; um=0.1; vdot=um*h0*(n+1)/(2*n+1) G=5000 beta=0; uxy; for i=1:nx pp(i)=pav(i); x(i)=(i-1)*dl/(nx-1); end beta=-dh/h0; uxy; ig=nx; ig0=ig; for i=2:nx ig=ig+1; pp(ig)=pav(i); x(ig)=(ig-1)*dl/(nx-1); end ig1=ig; h0=h-dh; beta=dh/h0; uxy; for i=2:nx ig=ig+1; pp(ig)=pav(i); x(ig)=(ig-1)*dl/(nx-1); end ig2=ig; hold on plot(x(ig0),pp(ig0),'ro') plot(x(ig1),pp(ig1),'ro') plot(x(ig2),pp(ig2),'ro') d1=pp(ig0)-pp(ig1); d2=pp(ig1)-pp(ig2); drel=(d1-d2)/max(abs(d1),abs(d2)); disp(sprintf('dp conv=%0.5g dp div=%0.5g dif=%0.5g',d1,d2,drel)); title(sprintf('G=%0.5g diff=%0.5g Vdot=%0.3g',G,drel,vdot)) xlabel('x [m]') ylabel('p [Pa]') 12 Axial pressure profiles p(x) 7 0 G=110 x 10 7 diff=-0.246 Vdot=0.000375 0 -0.5 𝑊= (n=0.5, K=100, V=0.000375 m2/s) G=150 x 10 7 diff=-0.16817 Vdot=0.000375 0 -0.5 -0.5 G=200 x 10 𝐾𝑢 2000 = 𝐺ℎ 𝐺 diff=-0.11166 Vdot=0.000375 Commented presentation RZ8 The graphs show what happens when you grow modulus G (resp. While reducing the relaxation time). Small values -1 -1 -1 G represent great influence elasticity (pressure gradient in the convergent gap is then smaller than the divergent) -1.5 -1.5 -2 p [Pa] p [Pa] p [Pa] -1.5 -2 -2 W=10 W=18 -2.5 -2.5 -2.5 -3 -3 -3 -3.5 0 0.1 0.2 0.3 0.4 0.5 0.6 -3.5 0.7 0 0.1 0.2 0.3 x [m] 7 0.5 0.6 7 diff=0.00025101 Vdot=0.000375 0 W=4 -1 G=1000 x 10 0 W=2 -2 -2.5 -2.5 -3 -3 0.3 0.4 x [m] 0.3 0.5 0.6 0.7 0.4 G=5000 x 10 0.5 0.6 0.7 diff=0.071136 Vdot=0.000375 W=0.4 -1 -1.5 p [Pa] p [Pa] -2 0.2 0.2 -0.5 -1.5 0.1 0.1 7 diff=0.03977 Vdot=0.000375 -1 -1.5 0 0 x [m] -0.5 -0.5 -3.5 -3.5 0.7 p [Pa] 0 G=500 x 10 0.4 x [m] -3.5 -2 G-2.5= 5000 negligible elasticity, different pressures are error of approximation -3 -3.5 0 0.1 0.2 0.3 0.4 x [m] 0.5 0.6 0.7 0 0.1 0.2 0.3 0.4 x [m] 0.5 13 0.6 0.7 Stream function, vorticity, elastic stresses 𝑢= 𝜕 𝜕𝑦 𝜕 𝑣=− 𝜕𝑥 = 𝜕𝑣 𝜕𝑢 − 𝜕𝑥 𝜕𝑦 s 𝜏𝑥𝑥 𝜕𝑢 = 𝑠𝑥𝑥 + 2𝜇 𝜕𝑥 14 Stream function, vorticity, elastic stresses Geometry Approximate profiles u,v, , for power law fluid iteration Elastic stresses sxx, syy, sxy (hyperbolic MW) Poisson equation for Nodal velocities and viscosity. BC for Poisson equation for This is rather complicated project probably too difficult for batchelor or even diploma work. However it can be significantly simplified by skiping the solution of flow , (assuming only velocity approximation) and neglecting viscosity variation. Poisson equation for pressure p 15 Gaussian axial profile of gap The reason for choosing the smooth profile of the upper walls: the exclusion of the weak singularity at a sharp edge between converging and diverging axial channel crossection -3 6 x 10 5 y Physical domain 4 3 2 1 0 x 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 Transformation of coordinates (x,y) (x,y) and metric coefficients 0.4 𝜕 𝜕 𝜕 𝜕 2 𝜕 2 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑥𝜕𝑦 𝜕𝑥2 necessary for transformation of first and second order derivatives i,j+1 i,j i,j-1 Computational domain 16 Gaussian axial profile of gap Metric coefficients of first derivatives Dimensionless geometric parameters 𝐿 𝐿 ℎ 𝐸 = exp(−2 1 − 2 ) = 𝐻𝑘 = 𝑘 =𝐻 -3 6 x 10 5 4 H 3 y 𝜕 >0 𝜕𝑥 2 1 0 0 0.05 0.1 𝜕 <0 𝜕𝑥 =1 h x Lk 𝜕𝜉 1 = 𝜕𝑥 𝐿𝑘 Kontroloval Zitny 0.15 =0 0.2 0.25 0.3 0.35 0.4 Ld 𝜕 2𝐸2 (1 − )(1 − 𝜅) = 𝜕𝑥 (1 − (1 − )𝐸) 𝐿𝑘 Transformation from computational domain (0,2) (0,1) to physical domain x(0,Lk+Ld), y(0,H) 𝑥 = 𝐿𝑘 𝑦 = 𝐻(1 − (1 − )𝐸) 𝜕 1 = 𝜕𝑦 1 − (1 − )𝐸 𝐻 17 Gaussian axial profile of gap Kontroloval: zitny Metric coefficients of second derivatives Because depends upon x only linearly the second derivatives of are zero 𝜕2𝜉 =0 𝜕𝑥 2 𝜕2𝜉 =0 𝜕𝑦 2 𝜕2𝜉 =0 𝜕𝑥𝜕𝑦 Only two derivatives of transversal coordinate are nonzero 𝜕2𝜂 =0 𝜕𝑦 2 𝜕2 2𝐸𝜆2 1 − 𝜅 = 2 (2𝜆2 1 − 𝜉 2 𝜕𝑥 𝐿𝑘 1 − 1 − 𝜅 𝐸 𝜕2 2𝐸2 (1 − )(1 − 𝜅) = 𝜕𝑥𝜕𝑦 1 − 1 − 𝐸 2 𝐻𝐿𝑘 2 1+ 1−𝜅 𝐸 − 1) 1− 1−𝜅 𝐸 Ověřeno numerickými diferencemi 18 INIT calculation of metric coefficients geometry INIT Approximate profiles u,v, , for power law fluid Geometrie h, H, Lk, Ld, , Nx, Ny iteration Elastic stress sxx, syy, sxy (hyperbolic MW) ksi=linspace(0,2,nx); eta=linspace(0,1,ny); Poisson equation for Velocities, µ , BC for Poisson equation for i=1,2..Nx J=1,2..Ny 𝑦𝑖, 𝑗 = 𝐻(1 − (1 − )𝐸) 𝜕 2𝐸2 (1 − )(1 − 𝜅) |𝑖 , 𝑗 = 𝜕𝑥 (1 − (1 − )𝐸) 𝐿𝑘 𝜕2 | = 𝜕𝑥 2 𝑖 , 𝑗 2𝐸𝜆2 1 − 𝜅 𝐿2𝑘 1 − 1 − 𝜅 𝐸 (2𝜆2 Poisson equation for pressure p 𝑥𝑖 = 𝐿𝑘 𝜕 1 |𝑖 , 𝑗 = 𝜕𝑦 1 − (1 − )𝐸 𝐻 21+ 1−𝜅 𝐸 1−𝜉 − 1) 1− 1−𝜅 𝐸 𝜕2 2𝐸2 (1 − )(1 − 𝜅) |𝑖 , 𝑗 = 𝜕𝑥𝜕𝑦 1 − 1 − 𝐸 2 𝐿𝑘 𝐻 19 INIT.m calculation of coefficients (MATLAB) clearvars Matrices defined at INIT um=0.1;n=0.5;K=100; x(nx), y(nx,ny) kartézské souřadnice H=0.005;h=0.001;Lk=0.2;Ls=0.2;delta=0.1; ksi(nx),eta(ny) výpočtová doména lambda=Lk/delta; dksidx(nx,ny),detadx(nx,ny),detady(nx,ny) /x, /x, kappa=h/H; detadxdy(nx,ny),detadxdx(nx,ny) 2/xy, 2/x2, nx=201; gxx(nx,ny),…gx(nx,ny) ny=41; ksi=linspace(0,2,nx); Matrices defined later eta=linspace(0,1,ny); psi(nx,ny), omega(nx,ny) x=Lk*ksi; u(nx,ny),v(nx,ny),dudx(nx,ny),dudy(nx,ny),dvdx(nx,ny),dvdy(nx,by) for j=1:ny mju(nx,ny),dmjudx(nx,ny), dmjudy(nx,ny), dmjudxdy(nx,ny), , /x, /y, for i=1:nx Approximation for power law fluid E=exp(-(Lk/delta)^2*(ksi(i)-1)^2); y(i,j)=eta(j)*H*(1-(1-kappa)*E); psin(nx,ny), omegan(nx,ny) end un(nx,ny),vn(nx,ny),dudxn(nx,ny),dudyn(nx,ny),dvdxn(nx,ny),dvdyn(nx,by) end mjun(nx,ny),dmjudxn(nx,ny), dmjudyn(nx,ny), dmjudxdyn(nx,ny), %derivace transformaci Stresses for i=1:nx sxx(nx,ny), syy(nx,ny), sxy(nx,ny), E=exp(-(lambda*(ksi(i)-1))^2); for j=1:ny k1=1-(1-kappa)*E; dksidx(i,j)=1/Lk; detadx(i,j)=eta(j)*(1-kappa)*E*2*lambda^2*(1-ksi(i))/(k1*Lk); detady(i,j)=1/(k1*H); detadxdy(i,j)=(1-kappa)*2*lambda^2*(1-ksi(i))*E/(H*Lk*k1^2); detadxdx(i,j)=2*lambda^2*eta(j)*(1-kappa)*E/(Lk^2*k1)*(2*lambda^2*(1-ksi(i))^2*(1+(1-kappa)*E)/k1-1); gxx(i,j)=dksidx(i,j)^2; gyy(i,j)=detadx(i,j)^2+detady(i,j)^2; gxy(i,j)=2*detadx(i,j)*dksidx(i,j); gx(i,j)=detadxdx(i,j); end end %figure(2) %surf(eta,ksi,detadx) %title('deta/dx') geometry Approximate profiles u,v, , for power law fluid iteration Elastic stress sxx, syy, sxy (hyperbolic MW) Poisson equation for Velocities, µ , BC for Poisson equation for Poisson equation for pressure p Transformation of derivatives of quantity F (e.g. , , sxx, syy, sxy) Kontroloval: zitny Transformation of first derivative 𝜕𝐹 𝜕𝐹 𝜕𝜉 𝜕𝐹 𝜕𝜂 = + 𝜕𝑥 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝐹 𝜕𝐹 𝜕𝜂 = 𝜕𝑦 𝜕𝜂 𝜕𝑦 Transformation of second derivatives (derivation is a bit lengthy) 𝜕 2 𝐹 𝜕𝐹 𝜕 2 𝜂 𝜕 2 𝐹 𝜕𝜉 2 𝜕 2 𝐹 𝜕𝜂 2 𝜕 2 𝐹 𝜕𝜂 𝜕𝜉 = + ( ) + 2 ( ) +2 𝜕𝑥 2 𝜕𝜂 𝜕𝑥 2 𝜕𝜉 2 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝜉𝜕𝜂 𝜕𝑥 𝜕𝑥 𝜕 2 𝐹 𝜕 2 𝐹 𝜕𝜂 2 = ( ) 𝜕𝑦 2 𝜕𝜂2 𝜕𝑦 𝜕2𝐹 𝜕 2 𝐹 𝜕𝜂 𝜕𝜂 𝜕𝐹 𝜕 2 𝜕 2 𝐹 𝜕𝜂 𝜕𝜉 = + + 𝜕𝑥𝜕𝑦 𝜕𝜂2 𝜕𝑦 𝜕𝑥 𝜕𝜂 𝜕𝑥𝜕𝑦 𝜕𝜉𝜕𝜂 𝜕𝑦 𝜕𝑥 𝜕2𝐹 𝜕 𝜕𝐹 𝜕𝜂 𝜕 2 𝐹 𝜕𝜂 𝜕𝐹 𝜕 2 𝜕 2 𝐹 𝜕𝜉 𝜕 2 𝐹 𝜕 𝜕𝜂 𝜕𝐹 𝜕 2 = = + =( + 2 ) + 𝜕𝑥𝜕𝑦 𝜕𝑥 𝜕𝜂 𝜕𝑦 𝜕𝑥𝜕 𝜕𝑦 𝜕𝜂 𝜕𝑥𝜕𝑦 𝜕𝜕 𝜕𝑥 𝜕 𝜕𝑥 𝜕𝑦 𝜕𝜂 𝜕𝑥𝜕𝑦 ... it would be possible to express the transformation of derivatives also in the reverse direction, but usually we do not need this 21 Cauchy equations with consideration of inertia forces (𝑢 Kontroloval: zitny 𝜕𝑢 𝜕𝑢 𝜕𝑝 𝜕𝜏𝑥𝑥 𝜕𝜏𝑥𝑦 +𝑣 )=− + + 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑣 𝜕𝑣 𝜕𝑝 𝜕𝜏𝑥𝑦 𝜕𝜏𝑦𝑦 (𝑢 +𝑣 )= − + + 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑦 Elimination of pressure by introducing the vorticity = 𝜕𝑣 𝜕𝑢 − 𝜕𝑥 𝜕𝑦 and using the continuity equation The result is the transport equation of vorticity (right side resultant stress). At zero Re (creeping flow, 0) the left side is zero and vorticity disappears from this equation. 𝜕 𝜕 2 2 2 (𝑢 +𝑣 )= (𝜏 − 𝜏𝑥𝑥 ) + ( 2 − 2 ) 𝜏𝑥𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥𝑦 𝑦𝑦 𝜕𝑥 𝜕𝑦 Fulfill the continuity equation by introducing the current function 𝜕 𝑢= 𝜕𝑦 𝜕 𝑣=− 𝜕𝑥 2 2 − = 2 + 2 𝜕𝑥 𝜕𝑦 22 Initial conditions - stream function (for a power law fluid) geometry Stream function (,) x 10 Approximate profiles u,v, , for power law fluid -4 4 iteration 3 Elastic stress sxx, syy, sxy (hyperbolic MW) 2 Poisson equation for 1 0 2 1.5 Velocities, µ , BC for 1 0.8 1 0.6 0.4 0.5 0 Poisson equation for 0.2 0 Velocity components u,v (,) Poisson equation for pressure p v [m/s] x 10 -3 0.5 6 0.4 4 0.3 0.2 2 0.1 0 0 2 -2 1.5 1 0.8 1 0.6 -4 1 0.4 0.5 0 0.2 0 -6 2 0.5 1.5 1 0.5 0 0 Velocities calculated in this way satisfy continuity equation and correspond to the steady creeping flow of power law fluid (without elastic effects). It is assumed that these initial conditions can be applied also as boundary conditions at inlet and outlet (far from the disturbance of throat). 23 Initial conditions - stream function (for a power law fluid) Section corresponding to the distance from the entrance =x/Lk Kontroloval: zitny 𝐻 = 𝐻 1 − 1 − 𝑒 − 2 1− 2 = 𝐻(1 − 1 − 𝐸) 𝜕𝐻 𝐸 2 = 2𝐻 (1 − )( − 1) 𝜕𝑥 𝐿𝑘 Stream function (um is maximum velocity for the gap thickness H). It holds umH=umH . 𝐻 2𝑦 𝑛 2𝑦 2𝑦 2𝑛+1 𝑥, 𝑦 = 𝑢𝑚 ( − (1 − 𝑠𝑖𝑔𝑛(1 − )|1 − | 𝑛 )) 2 𝐻 2𝑛 + 1 𝐻 𝐻 Corresponding velocity components 2𝑛+1 𝐻 𝑛 = 𝑢𝑚 (2 − (1 − 𝑠𝑖𝑔𝑛(1 − 2)|1 − 2| 𝑛 )) 2 2𝑛 + 1 𝜕 𝑢𝑚 𝐻 2𝑦 𝑛+1 𝑢 𝑥, 𝑦 = = (1 − |1 − | 𝑛 ) 𝜕𝑦 𝐻 𝐻 𝜕 𝐻 𝑣 𝑥, 𝑦 = − = −2𝑢𝑚 𝜕𝑥 𝐻 2 𝐻 𝑦 2𝑦 𝑛+1 (1 − ) (1 − )(1 − |1 − | 𝑛 ) 𝐻 𝐿𝑘 𝐻 24 Initial conditions – vorticity (for a power law fluid) Vorticity 𝜕𝑣 𝜕𝑢 = − 𝜕𝑥 𝜕𝑦 Auxilliary variables corresponds to the direction anti-clockwise 𝑍 =1− 𝐻 𝐻 𝐹= ( − 1)(1 − 𝑍 𝐻 𝐻 𝐻 = 𝐻 1 − 1 − 𝑒 − 2 1− 2 = 𝐻(1 − 1 − 𝐸) 𝜕𝐻 𝐸 2 = 2𝐻 (1 − )( − 1) 𝜕𝑥 𝐿𝑘 2𝑦 𝐻 𝑛+1 𝑛 ) Kontroloval: zitny 𝜕𝐹 𝐻 𝐻 = 2 ((1 − 2 )(1 − 𝑍 𝜕𝐻 𝐻 𝐻 𝑛+1 𝑛 ) dudy-num 𝐻 𝑛+1 − 𝑠𝑖𝑔𝑛 𝑍 ( − 1) 𝑍 𝐻 𝑛 1 𝑛 1−𝑍 ) OMEGA 3000 4000 2000 1000 2000 0 1 𝜕𝑢 𝜕 2 2𝑢𝑚 𝐻 𝑛 + 1 = = ( )𝑠𝑖𝑔𝑛(𝑍)|𝑍|𝑛 𝜕𝑦 𝜕𝑦 2 𝐻 2 𝑛 -1000 0 -2000 -2000 -3000 2 dvdx-num 1.5 1 0.8 1 0.6 0.4 0.5 0 0.8 -4000 2 1.5 0.2 0 1 0.8 1 0.6 0.4 0.5 𝜕𝑣 𝜕2 2𝑢𝑚 2 𝑦 𝜕𝐹 𝜕𝐻 𝐾 =− 2 =− ( 1− − ) 𝜕𝑥 𝜕𝑥 𝐿𝑘 𝜕𝐻 𝜕𝑥 𝐿𝑘 0.6 0 0.2 0 Verified by three methods: analytically, direct numerical difference and differential transformed derivatives (it took me few days to match the threee procedures 0.4 0.2 0 -0.2 2 1.5 1 0.8 1 0.6 0.4 0.5 0 0.2 0 25 Initial conditions – viscosity (for a power law fluid) Auxilliary variables 𝑍 =1− 𝜕𝑢 𝜕2 𝐻 = = −2𝑢𝑚 𝜕𝑥 𝜕𝑥𝜕𝑦 𝐻 2 𝜕𝑣 𝜕2 𝐻 =− = −2𝑢𝑚 𝜕𝑦 𝜕𝑥𝜕𝑦 𝐻 2 𝜕𝑢 𝜕𝑥 2 𝜇 𝐼𝐼 = 𝐾 2 2𝑦 𝐻 Kontroloval: zitny 𝐻 = 𝐻 1 − 1 − 𝑒 − 2 −1 (1 − )𝐸 (1 − 𝑍 𝐿𝑘 𝑛+1 𝑛 𝑛+1 + 𝑠𝑖𝑔𝑛 𝑍 𝑍 𝑛 1 𝑛 1− 2 = 𝐻(1 − 1 − 𝐸) 1−𝑍 ) Verified numerically 𝐻 1 − (1 − ) (1 − 𝑍 𝐻 𝐿𝑘 +2 𝜕𝑣 𝜕𝑦 2 + 𝜕𝑢 𝜕𝑣 + 𝜕𝑦 𝜕𝑥 𝑛+1 𝑛 2 Analytical expressions for first derivatives are too complicated and will be approximated numerically 𝑛+1 + 𝑠𝑖𝑔𝑛 𝑍 𝑍 𝑛 1 𝑛 1−𝑍 ) 𝑛−1 Dangerous peeks correspond to regions with smal velocity gradients (this is disadvantage of power law model) 𝜕µ 𝜕µ 𝜕 𝜕µ 𝜕 µ𝐸 − µ𝑊 𝜕 µ𝑁 − µ𝑆 𝜕 = + + 𝜕𝑥 𝜕 𝜕𝑥 𝜕 𝜕𝑥 2ℎ𝑥 𝜕𝑥 2ℎ𝑦 𝜕𝑥 𝜕µ 𝜕µ 𝜕 µ𝑁 − µ𝑆 𝜕 = 𝜕𝑦 𝜕 𝜕𝑦 2ℎ𝑦 𝜕𝑦 26 Initial conditions – POWER.m (MATLAB code 1/2) geometry Velocities, stream function, vorticity, viscosity for power law fluids Approximate profiles u,v, , for power law fluid iteration Elastic stress sxx, syy, sxy (hyperbolic MW) for i=1:nx Ek=exp(-lambda^2*(1-ksi(i))^2); Hk=H*(1-(1-kappa)*Ek); for j=1:ny Z=1-2*y(i,j)/Hk; HK=H/Hk; psi(i,j)=um*H/2*(1-Z-n/(2*n+1)*(1-sign(Z)*abs(Z)^((2*n+1)/n))); u(i,j)=um*H/Hk*(1-abs(Z)^((n+1)/n)); v(i,j)=-2*um*(lambda*H/Hk)^2*(1-Hk/H)*(1-ksi(i))*y(i,j)/Lk*(1-abs(Z)^((n+1)/n)); dudy(i,j)=2*um*H/Hk^2*(n+1)/n*sign(Z)*abs(Z)^(1/n); F=(HK^2-HK)*(1-abs(Z)^((n+1)/n)); dHkdx=2*H*lambda^2*(1-kappa)*(ksi(i)-1)*Ek/Lk; dFdHk=H/Hk^2*((-2*HK+1)*(1-abs(Z)^((n+1)/n))-sign(Z)*(HK-1)*(n+1)/n*abs(Z)^(1/n)*(1-Z)); dvdx(i,j)=-2*um*lambda^2*y(i,j)/Lk*(dFdHk*dHkdx*(1-ksi(i))-F/Lk); omega(i,j)=dvdx(i,j)-dudy(i,j); dudx(i,j)=-2*um*(HK*lambda)^2*(1-kappa)*(ksi(i)-1)*Ek/Lk*(1-abs(Z)^((n+1)/n)+sign(Z)*(n+1)/n*abs(Z)^(1/n)*(1-Z)); dvdy(i,j)=-2*um*(HK*lambda)^2*(1-1/HK)*(1-ksi(i))/Lk*(1-abs(Z)^((n+1)/n)+sign(Z)*(n+1)/n*abs(Z)^(1/n)*(1-Z)); mju(i,j)=K*(2*dudx(i,j)^2+2*dvdy(i,j)^2+(dudy(i,j)+dvdx(i,j))^2)^((n-1)/2); end end Poisson equation for Velocities, µ , BC for Poisson equation for Poisson equation for pressure p 27 Initial conditions – POWER.m (MATLAB code 2/2) geometry Approximation of viscosity derivatives Approximate profiles u,v, , for power law fluid iteration Elastic stress sxx, syy, sxy (hyperbolic MW) dmjudx(1:nx,1:ny)=0; dmjudy(1:nx,1:ny)=0; for i=2:nx-1 for j=2:ny-1 dmjudx(i,j)=(mju(i+1,j)-mju(i-1,j))/(2*hx)*dksidx(i,j)+(mju(i,j+1)-mju(i,j-1))/(2*hy)*detadx(i,j); dmjudy(i,j)=(mju(i,j+1)-mju(i,j-1))/(2*hy)*detady(i,j); end end for i=1:nx dmjudx(i,1)=dmjudx(i,2); dmjudx(i,ny)=dmjudx(i,ny-1); end for j=1:ny dmjudx(1,j)=dmjudx(2,j); dmjudx(nx,j)=dmjudx(nx-1,j); end Poisson equation for Velocities, µ , BC for Poisson equation for Poisson equation for pressure p % copy matrices psin=psi; omegan=omega; un=u; vn=v; dudxn=dudx; dudyn=dudy; dvdxn=dvdx; dvdyn=dvdy; mjun=mju; dmjudxn=dmjudx(nx,ny); dmjudyn=dmjudy; 28 PSI Transformation of stream function into computational domain Kontroloval: zitny 2 2 + = − 𝜕𝑥 2 𝜕𝑦 2 Poisson equation for stream function can be transformed into computational domain using previously calculated metric coefficients 𝜕 2 𝜓 𝜕 2 𝜕 2 𝜓 ( ) + 2 𝜕𝜉 2 𝜕𝑥 𝜕 𝜕 𝜕𝑥 2 𝜕 + 𝜕𝑦 2 𝜕 2 𝜓 𝜕 𝜕 𝜕 𝜕 2 +2 + = − 𝜕 𝜕𝑥 𝜕𝑥 𝜕 𝜕𝑥 2 𝜕2𝜓 𝜕2𝜓 𝜕2𝜓 𝜕 𝑔 + 2 𝑔𝑦𝑦 + 𝑔 + 𝑔 = − 𝜕𝜉 2 𝑥𝑥 𝜕 𝜕 𝑥𝑦 𝜕 𝑥 𝑔𝑥𝑥 𝜕 = ( )2 𝜕𝑥 𝜕 𝑔𝑦𝑦 = 𝜕𝑥 2 𝜕 + 𝜕𝑦 2 𝜕 𝜕 𝑔𝑥𝑦 = 2 𝜕𝑥 𝜕𝑥 𝜕2 𝑔𝑥 = 2 𝜕𝑥 29 PSI Discretization of stream function Kontroloval: xxxxxx =1 i,j+1 i,j i,j-1 =0 =1 =0 hy =2 hx 𝜕2𝜓 𝜕2𝜓 𝜕2𝜓 𝜕 𝑔 + 2 𝑔𝑦𝑦 + 𝑔 + 𝑔 = − 𝜕𝜉 2 𝑥𝑥 𝜕 𝜕 𝑥𝑦 𝜕 𝑥 𝑔𝑦𝑦 𝑔𝑥𝑥 − 2 + + 𝑖−1,𝑗 𝑖,𝑗 𝑖+1,𝑗 2 2 𝑖,𝑗−1 − 2𝑖,𝑗 + 𝑖,𝑗+1 ℎ𝑥 ℎ𝑦 𝑔𝑥𝑦 𝑔𝑥 + + 𝑖−1,𝑗−1 − 𝑖−1,𝑗+1 − 𝑖+1,𝑗−1 + − 𝑖,𝑗−1 = −𝑖𝑗 4ℎ𝑥 ℎ𝑦 𝑖+1,𝑗+1 2ℎ𝑦 𝑖,𝑗+1 30 PSI by alternation direction implicit method (ADI) Time evolution approximation geometry Approximate profiles u,v, , for power law fluid PSI iter=1,2,… i=2..Nx-1 Sweep of tridiagonal system for in horizontal direction (only smoothing) This horizontal sweep doesn’t work correctly. Maybe because is almost independent on iteration Elastic stress sxx, syy, sxy (hyperbolic MW) Poisson equation for Velocities, µ , BC for Poisson equation for Poisson equation for pressure p i=2..Nx-1 Sweep of tridiagonal system for in vertical direction (sections i) Norm of difference new-old and repetition the PSI loop 31 PSI Discretization of stream function (time marching) gxx … coefficients were calculated during INIT.m 1 − 2 1 2 =+ + 𝑡1 𝜕𝑥 2 𝜕𝑦 2 𝑔𝑥𝑥 𝜕 2 𝜕 = ( ) 𝑔𝑦𝑦 = 𝜕𝑥 𝜕𝑥 Kontroloval: xxxxxx 2 𝜕 + 𝜕𝑦 2 𝜕 𝜕 𝑔𝑥𝑦 = 2 𝜕𝑥 𝜕𝑥 𝜕2 𝑔𝑥 = 2 𝜕𝑥 2 − 1 2 1 2 2 =+ + 𝑡2 𝜕𝑥 2 𝜕𝑦 2 -implicit (horizontal lines) This part (-implicit step) does not work at tests 1𝑖𝑗 𝑡1 It was substituted by a simple filter, see the source code 𝑖𝑗 𝑔𝑦𝑦 𝑔𝑥𝑥 𝑔𝑥𝑥 1 𝑔𝑥𝑥 1 + 2 − = + 2 𝑖,𝑗−1 − 2𝑖,𝑗 + 𝑖,𝑗+1 𝑖,𝑗 𝑖+1,𝑗 2 2 2 𝑡 ℎ𝑥 ℎ𝑥 ℎ𝑥 ℎ𝑦 1 𝑔𝑥 𝑖+1,𝑗+1 + 𝑖−1,𝑗−1 − 𝑖−1,𝑗+1 − 𝑖+1,𝑗−1 + − 𝑖,𝑗−1 + 𝑖𝑗 2ℎ𝑦 𝑖,𝑗+1 − 1𝑖−1,𝑗 𝑔𝑥𝑦 + 4ℎ𝑥 ℎ𝑦 -implicit (vertical lines) 2𝑖𝑗 𝑔𝑦𝑦 𝑔𝑦𝑦 2 𝑔𝑦𝑦 𝑔𝑥 𝑔𝑥 2 − 𝑖,𝑗−1 − + 2 2 𝑖,𝑗 − + 2𝑖,𝑗+1 2 2 𝑡2 2ℎ𝑦 2ℎ𝑦 ℎ𝑦 ℎ𝑦 ℎ𝑦 1𝑖𝑗 𝑔𝑥𝑥 𝑔𝑥𝑦 1 1 1 = + 2 𝑖−1,𝑗 − 2 𝑖,𝑗 + 𝑖+1,𝑗 + 1𝑖+1,𝑗+1 + 𝑖−1,𝑗−1 − 𝑖−1,𝑗+1 − 𝑖+1,𝑗−1 + 𝑖𝑗 𝑡2 ℎ𝑥 4ℎ𝑥 ℎ𝑦 32 PSI Discretization of stream function PSIE.m (MATLAB) Kontroloval: xxxxxx Estimated dt2=(H/ny)^2*1e3; dt1=(Lk/nx)^2*1e3; psiold=psi; time step for iter=1:20 % horizontal lines (only smoothing) for j=2:ny-1 aa(1)=0;bb(1)=1;cc(1)=0;dd(1)=psi(1,j); aa(nx)=0;bb(nx)=1;cc(nx)=0;dd(nx)=psi(nx,j); for i=2:nx-1 aa(i)=-gxx(i,j)/hx^2; bb(i)=1/dt1+2*gxx(i,j)/hx^2; cc(i)=-gxx(i,j)/hx^2; dd(i)=psi(i,j)/dt1; end Soution of psi(1:nx,j)=tridag(aa,bb,cc,dd,nx); end tridiagonal system % vertical lines by factorization for i=2:nx-1 aa(1)=0;bb(1)=1;cc(1)=0;dd(1)=psi(i,1); aa(ny)=0;bb(ny)=1;cc(ny)=0;dd(ny)=psi(i,ny); for j=2:ny-1 aa(j)=-gyy(i,j)/hy^2+gx(i,j)/(2*hy); bb(j)=1/dt2+2*gyy(i,j)/hy^2; cc(j)=-gyy(i,j)/hy^2-gx(i,j)/(2*hy); ddxx=gxx(i,j)/hx^2*(psi(i-1,j)-2*psi(i,j)+psi(i+1,j)); ddxy=gxy(i,j)/(4*hx*hy)*(psi(i+1,j+1)+psi(i-1,j-1)-psi(i-1,j+1)-psi(i+1,j-1)); dd(j)=psi(i,j)/dt2+ddxx+ddxy+ddomega; end psi(i,:)=tridag(aa,bb,cc,dd,ny); end % norm of differences sum=0; for i=1:nx for j=1:ny sum=sum+abs(psi(i,j)-psiold(i,j)); end end sdev=sum/(nx*ny)/psiw; disp(sprintf(' iter=%d relsum=%.5g',iter,sdev)) if sdev<0.00001 break end end function u = tridag(a,b,c,r,n) bet=b(1); u(1)=r(1)/bet; for j=2:n gam(j)=c(j-1)/bet; bet=b(j)-a(j)*gam(j); u(j)=(r(j)-a(j)*u(j-1))/bet; end for j=n-1:-1:1 u(j)=u(j)-gam(j+1)*u(j+1); end ddomega=omega(i,j); 33 Post processing (velocities, viscosities, BC for vorticity) UVMBC.m for i=2:nx-1 geometry for j=2:ny-1 Approximate profiles u,v, , v(i,j)=-(psi(i+1,j)-psi(i-1,j))/(2*hx)*dksidx(i,j)-(psi(i,j+1)-psi(i,j-1))/(2*hy)*detadx(i,j); for power law fluid u(i,j)=(psi(i,j+1)-psi(i,j-1))/(2*hy)*detady(i,j); end iteration end Elastic stress sxx, syy, sxy for i=2:nx-1 (hyperbolic MW) for j=2:ny-1 Poisson equation for dudx(i,j)=(u(i+1,j)-u(i-1,j))/(2*hx)*dksidx(i,j)+(u(i,j+1)-u(i,j-1))/(2*hy)*detadx(i,j); dudy(i,j)=(u(i,j+1)-u(i,j-1))/(2*hy)*detady(i,j); dvdx(i,j)=(v(i+1,j)-v(i-1,j))/(2*hx)*dksidx(i,j)+(v(i,j+1)-v(i,j-1))/(2*hy)*detadx(i,j); dvdy(i,j)=(v(i,j+1)-v(i,j-1))/(2*hy)*detady(i,j); Velocities, µ , BC for end end Poisson equation for for i=2:nx-1 for j=2:ny-1 mju(i,j)=K*(2*dudx(i,j)^2+2*dvdy(i,j)^2+(dudy(i,j)+dvdx(i,j))^2)^((n-1)/2); end Poisson equation for pressure p end for i=2:nx-1 for j=2:ny-1 dmjudx(i,j)=(mju(i+1,j)-mju(i-1,j))/(2*hx)*dksidx(i,j)+(mju(i,j+1)-mju(i,j-1))/(2*hy)*detadx(i,j); dmjudy(i,j)=(mju(i,j+1)-mju(i,j-1))/(2*hy)*detady(i,j); end end for i=1:nx dmjudx(i,1)=dmjudx(i,2); dmjudx(i,ny)=dmjudx(i,ny-1); end for j=1:ny dmjudx(1,j)=dmjudx(2,j); dmjudx(nx,j)=dmjudx(nx-1,j); end rlx=0.5; for i=1:nx omega(i,1)=omega(i,1)*(1-rlx)+rlx*2*psi(i,2)/hy^2*detady(i,1)^2; omega(i,ny)=omega(i,ny)*(1-rlx)+rlx*2*(psi(i,ny-1)-psiw)/hy^2*(detadx(i,ny)^2+detady(i,ny)^2; end 34 Decomposition extrastress on the elastic and viscous component Kontroloval: zitny Decomposition viscous stress tensor for purely viscous component (described by model GNF - Generalised Newtonian Fluid) and the elastic component sij is apparently purposeless, but has the advantage that it leads to an equation in which the vorticity dominates even at zero Re. 𝜕𝑢 𝜏𝑥𝑥 = 𝑠𝑥𝑥 + 2𝜇 𝜕𝑥 𝜕𝑣 𝜕𝑢 𝜏𝑦𝑦 = 𝑠𝑦𝑦 + 2𝜇 = 𝑠𝑦𝑦 − 2𝜇 𝜕𝑦 𝜕𝑥 𝜏𝑥𝑦 𝜕𝑢 𝜕𝑣 = 𝑠𝑥𝑦 + 𝜇( + ) 𝜕𝑦 𝜕𝑥 It is only an approximation of the corresponding constant viscosity Transport equations for vorticity (right side of the resultant elastic strain). From a purely hyperbolic transport equation is zero Re turned out nice for elliptic equations (LAD - LAplacian Driven) 𝜕 𝜕 2 𝜔 2 𝜔 2 2 2 𝑢 +𝑣 −µ + = 𝑠 − 𝑠𝑥𝑥 + − 𝑠 𝜕𝑥 𝜕𝑦 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥𝑦 𝑦𝑦 𝜕𝑥 2 𝜕𝑦 2 𝑥𝑦 For comparison 𝜕 𝜕 𝑢 +𝑣 𝜕𝑥 𝜕𝑦 2 2 2 = (𝜏 − 𝜏𝑥𝑥 ) + ( 2 − 2 ) 𝜏𝑥𝑦 𝜕𝑥𝑦 𝑦𝑦 𝜕𝑥 𝜕𝑦 35 Kontroloval: Decomposition extrastress on the elastic and viscous component (variable µ9zitny The variable viscosity (non-zero first and second derivatives of viscosity). In addition Laplacian vorticity there appears a number of terms (partial derivatives of velocity) 2 2 2 𝜏 − 𝜏𝑥𝑥 + − 𝜏 𝜕𝑥𝑦 𝑦𝑦 𝜕𝑥 2 𝜕𝑦 2 𝑥𝑦 2 2 2 2 2 = 𝑠 − 𝑠𝑥𝑥 + − 𝑠 + + 𝜕𝑥𝑦 𝑦𝑦 𝜕𝑥 2 𝜕𝑦 2 𝑥𝑦 𝜕𝑥 2 𝜕𝑦 2 𝜕 𝜕 𝜕 𝜕 2 𝜕𝑣 2 2 +2 +2 +4 + − 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥𝑦 𝜕𝑦 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑣 𝜕𝑢 + 𝜕𝑥 𝜕𝑦 Neglecting second derivatives of viscosity (even the first derivatives are highly discontinuous and could be causing problems during solution) 𝜕 𝜕 𝜕 𝜕 2 𝜔 2 𝜔 2 2 2 (𝑢 − 2 ) + (𝑣 − 2 ) −µ + = 𝑠 − 𝑠𝑥𝑥 + − 𝑠 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥𝑦 𝑦𝑦 𝜕𝑥 2 𝜕𝑦 2 𝑥𝑦 36 Transformation vorticity into computational domain Kontroloval: zitny Transport equation for vorticity (variable viscosity) 𝜕 2 𝜕 −µ 𝜕𝜉 2 𝜕𝑥 2 𝜕2 + 2 𝜕 𝜕 𝜕𝑥 2 𝜕 + 𝜕𝑦 2 𝜕 2 𝜕 𝜕 +2 + 𝜕 𝜕𝑥 𝜕𝑥 𝜕 𝜕µ 𝜕 𝜕 𝜕2 𝜕µ 𝜕 𝜕µ 𝜕 + (𝑢 − 2 ) + −µ 2 + (𝑣 − 2 ) + (𝑢 − 2 ) 𝜕 𝜕𝑥 𝜕𝑥 𝜕 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑥 𝜕 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕 𝜕 𝜕𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕 2 𝜕 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕 𝜕 𝜕𝑠𝑥𝑦 𝜕 2 = + + + 𝜕2 𝜕𝑦 𝜕𝑥 𝜕 𝜕𝑥𝑦 𝜕 𝜕𝑦 𝜕𝑥 𝜕 𝜕𝑥 2 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝜕 2 𝑠𝑥𝑦 𝜕 𝜕 𝜕 2 𝑠𝑥𝑦 𝜕 2 + ( ) + ( ) +2 − ( ) 𝜕2 𝜕𝑥 𝜕2 𝜕𝑥 𝜕 𝜕𝑥 𝜕𝑥 𝜕2 𝜕𝑦 𝜕µ 𝜕 𝑣𝑥 = (𝑢 − 2 ) 𝜕𝑥 𝜕𝑥 𝐷 𝐷𝑡 material derivative of streamline 𝜕2 𝜕µ 𝜕 𝜕µ 𝜕 𝜕2 𝐷 𝜕µ 𝜕 𝜕µ 𝜕 𝑣𝑦 = −µ 2 + 𝑣 − 2 + 𝑢 − 2 = −µ 2 + − 2( + ) 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝐷𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕2 𝜕2 𝜕2 𝜕 𝜕 µ 𝑔 + 𝑔 + 𝑔 + 𝑣 + 𝑣 𝜕𝜉 2 𝑥𝑥 𝜕2 𝑦𝑦 𝜕 𝑥𝑦 𝜕 𝑥 𝜕 𝑦 𝜕 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 = 𝑞1 + 𝑞2 + 𝑞3 + 𝑞 + 𝑞 + 𝑞 + 𝑞 − 𝑞 𝜕2 𝜕 𝜕 𝜕 4 𝜕2 5 𝜕2 6 𝜕 7 𝜕2 8 37 Discretized vorticity – central differences Kontroloval: xxxxxx 𝜕2 𝜕2 𝜕2 𝜕 𝜕 −µ 𝑔 + 𝑔 + 𝑔 + 𝑣 + 𝑣 𝜕𝜉 2 𝑥𝑥 𝜕2 𝑦𝑦 𝜕 𝑥𝑦 𝜕 𝑥 𝜕 𝑦 𝜕 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 = 𝑞1 + 𝑞2 + 𝑞3 + 𝑞 + 𝑞 + 𝑞 + 𝑞 − 𝑞 𝜕2 𝜕 𝜕 𝜕 4 𝜕2 5 𝜕2 6 𝜕 7 𝜕2 8 𝑔𝑦𝑦 𝑔𝑥𝑦 𝑔𝑥𝑥 µ 𝑖−1,𝑗 − 2𝑖,𝑗 + 𝑖+1,𝑗 + 2 𝑖,𝑗−1 − 2𝑖,𝑗 + 𝑖,𝑗+1 + + 𝑖−1,𝑗−1 − 𝑖−1,𝑗+1 − 𝑖+1,𝑗−1 4ℎ𝑥 ℎ𝑦 𝑖+1,𝑗+1 ℎ𝑥2 ℎ𝑦 𝑣𝑦 𝑣𝑥 + (𝑖+1,𝑗 − 𝑖−1,𝑗 ) + ( − 𝑖,𝑗−1 ) = 𝐺𝑖𝑗 2ℎ𝑥 2ℎ𝑦 𝑖,𝑗+1 Right hand size Gij is nonzero only for nozero stresses 𝐺𝑖𝑗 𝑞1 𝑞2 = 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝑖,𝑗+1 − 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝑖,𝑗 + 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝑖,𝑗−1 + 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝑖,𝑗+1 − 𝑠𝑦𝑦 − 𝑠𝑥𝑥 2ℎ𝑦 ℎ𝑦 𝑞3 + 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝑖+1,𝑗+1 + 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝑖−1,𝑗−1 − 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝑖+1,𝑗−1 − 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝑖−1,𝑗+1 4ℎ𝑥 ℎ𝑦 𝑞4 𝑞5 𝑞6 − 𝑞8 + 𝑠𝑥𝑦𝑖,𝑗+1 − 𝑠𝑥𝑦𝑖,𝑗−1 + 2 𝑠𝑥𝑦𝑖+1,𝑗 − 2𝑠𝑥𝑦𝑖,𝑗 + 𝑠𝑥𝑦𝑖−1,𝑗 + 𝑠𝑥𝑦𝑖,𝑗+1 − 2𝑠𝑥𝑦𝑖,𝑗 + 𝑠𝑥𝑦𝑖,𝑗−1 2ℎ𝑦 ℎ𝑥 ℎ𝑦2 𝑞7 + 𝑠 + 𝑠𝑥𝑦𝑖−1,𝑗−1 − 𝑠𝑥𝑦𝑖−1,𝑗+1 − 𝑠𝑥𝑦𝑖+1,𝑗−1 4ℎ𝑥 ℎ𝑦 𝑥𝑦𝑖+1,𝑗+1 𝑖,𝑗−1 38 Discretised vorticity – upwind dif. at dominated convective transport Kontroloval: xxxxxx 𝜕2 𝜕2 𝜕2 𝜕 𝜕 µ 𝑔 + 𝑔 + 𝑔 + 𝑣 + 𝑣 𝜕𝜉 2 𝑥𝑥 𝜕2 𝑦𝑦 𝜕 𝑥𝑦 𝜕 𝑥 𝜕 𝑦 𝜕 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝜕𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝑠𝑥𝑦 = 𝑞1 + 𝑞2 + 𝑞3 + 𝑞 + 𝑞 + 𝑞 + 𝑞 − 𝑞 𝜕2 𝜕 𝜕 𝜕 4 𝜕2 5 𝜕2 6 𝜕 7 𝜕2 8 Interpretation of these velocity components is not immediately apparent (may be negative) 𝜕µ 𝜕 𝑣𝑥 = (𝑢 − 2 ) 𝜕𝑥 𝜕𝑥 𝜕2 𝜕µ 𝜕 𝜕µ 𝜕 𝑣𝑦 = µ 2 + (𝑣 − 2 ) + (𝑢 − 2 ) 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑥 𝑔𝑦𝑦 𝑔𝑥𝑦 𝑔𝑥𝑥 − 2 + + − 2 + + 𝑖+1,𝑗+1 + 𝑖−1,𝑗−1 − 𝑖−1,𝑗+1 − 𝑖+1,𝑗−1 𝑖−1,𝑗 𝑖,𝑗 𝑖+1,𝑗 𝑖,𝑗−1 𝑖,𝑗 𝑖,𝑗+1 2 2 4ℎ ℎ ℎ𝑥 ℎ𝑦 𝑥 𝑦 µ + 𝑣𝑦 + 𝑣𝑦 𝑣𝑦 − 𝑣𝑦 𝑣𝑥 + 𝑣𝑥 𝑣𝑥 − 𝑣𝑥 (𝑖,𝑗 − 𝑖−1,𝑗 ) + (𝑖+1,𝑗 − 𝑖,𝑗 ) + (𝑖,𝑗 − 𝑖,𝑗−1 ) + (𝑖,𝑗+1 − 𝑖,𝑗 ) = 𝐺𝑖𝑗 2ℎ𝑥 2ℎ𝑥 2ℎ𝑦 2ℎ𝑦 Comparing the size of the convective and diffusive member 𝜕 2 𝜕 2 𝜕 𝜕 2 𝜕 ( ) + ⋯ + µ + 𝑢 +⋯ 𝜕𝜉 2 𝜕𝑥 𝜕 𝜕𝑥 2 𝜕𝑥 39 Boundary conditions – vorticity Kontroloval: xxxxxx The boundary conditions for vorticity resulting from the requirement of zero velocity on the wall u = v = 0 =0, 1 𝜕 𝜕 𝜕 𝜕 1 𝜕 2𝐸2 (1 − )(1 − 𝜅) 𝑢=𝑣=0 = =0 = + =0 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕 𝐿𝑘 𝜕 (1 − (1 − )𝐸) 𝐿𝑘 𝜕 𝜕 1 = =0 𝜕𝑦 𝜕 (1 − (1 − )𝐸) 𝐻 It follows from these equations that the first partial derivative must be zero in physical and computational plane. In the computational plane complete quadratic polynomial will replace the stream function around a node on the wall 2 + 𝑎 2 + 𝑎 = 𝑎 + 𝑎 + 𝑎 + 𝑎 0 1 2 3 4 5 i,j=Ny =1 Six polynomial coefficients should ensure 4 values in hy nodes and two conditions of zero derivatives at point i,Ny-1 i,Ny. Coefficients of coordinates are zero. =0 i,j=1 =0 =2 𝑖,𝑁𝑦−1 − 𝑤 () = 𝑤 + (1 − )2 =1 Gaussian wall 2 ℎ𝑦 𝑖,2 2 () = =0 plain wall 2 ℎ𝑦 40 Boundary conditions – vorticity Kontroloval: xxxxxx Substituting the definition of a polynomial we get a strong vorticity boundary condition at Gaussian wall geometry 𝜕2𝜓 𝑤 = − 2 𝜕 𝜕 𝜕𝑥 2 2 𝜕 + 𝜕𝑦 𝑖,𝑁𝑦−1 − 𝑤 2𝐸2 1 − 1 − 𝜅 =2 ( 1 − 1 − 𝐸 𝐿𝑘 ℎ𝑦2 2 + 1 Approximate profiles u,v, , for power law fluid 2) 1− 1− 𝐸 𝐻 iteration Elastic stress sxx, syy, sxy (hyperbolic MW) Strong boundary condition at the bottom wall 𝜕2𝜓 𝑤 = − 2 𝜕 𝜕 𝜕𝑥 2 2 𝜕 + 𝜕𝑦 𝑖,2 =2 2 ℎ𝑦 Poisson equation for Velocities, µ , BC for 1 1− 1− 𝐸 𝐻 Poisson equation for 2 Boundary conditions for , at input and output expect stabilized velocity profile Poisson equation for pressure p -3 6 x 10 𝐻 ∗ = 𝐻(1 − 1 − 𝑒 − ) 2 5 4 3 5 =1 y 2 1 0 0 0.05 0.1 x 0.15 0.2 x 10 ) -3 4.5 =0 0.25 0.3 0.35 0.4 where um is the maximum velocity in the flow channel axis (width H) corresponding to a power law fluid flow index n 4 3.5 3 y [m] H Velocity profile at inlet (for y<H*/2) 2𝑦 𝑢 𝑦 = 𝑢𝑚 (1 − 1 − ∗ 𝐻 𝑛+1 𝑛 2.5 2 1.5 1 0.5 0 0 0.02 0.04 0.06 u 41 0.08 0.1 Boundary conditions – stream function and vorticity 𝐻 ∗ = 𝐻(1 − 1 − 𝑒 − ) 2 Boundary conditions for , at inlet and outlet with gap thickness 𝐻 2𝑦 𝑛 2𝑦 2𝑦 2𝑛+1 𝑦 = 𝑢𝑚 ( ∗ − (1 − 𝑠𝑖𝑔𝑛(1 − ∗ )|1 − ∗ | 𝑛 )) 2 𝐻 2𝑛 + 1 𝐻 𝐻 Stream function at inlet Overall flowrate 𝑛+1 𝑤 = 𝑢 𝐻 2𝑛 + 1 𝑚 -3 5 4.5 4.5 4 4 3.5 3.5 3 3 y [m] y [m] x 10 is the boundary condition for at upper wall (w=0 at bottom wall) 1 𝜕𝑢 𝑛 + 1 2𝑢𝑚 2𝑦 𝑛 =− =− ( − 1) 𝜕𝑦 𝑛 𝐻∗ 𝐻∗ Vorticity at inlet and outlet 5 2.5 2 1.5 1.5 1 1 0.5 0.5 0 1 2 3 psi 4 5 6 x 10 -4 x 10 Approximation valid only for fully developed velocity profile -3 2.5 2 0 Kontroloval: xxxxxx 0 -140 -120 -100 -80 -60 omega -40 -20 0 42 Iterative solution of OMG.m geometry Approximate profiles u,v, , for power law fluid iteration Elastic stress sxx, syy, sxy (hyperbolic MW) OMG Poisson equation for Velocities, µ , BC for Nodal velocities and boundary conditions for () i=2..Nx-1 Poisson equation for Poisson equation for pressure p Solution of tridiagonal systém for in vertical direction i. (upwind) 43 Iterative solution of (OMG.m) Previously derived transport by pseudovelocities and implementation of upwind 𝜕µ 𝜕 𝜕2 𝜕µ 𝜕 𝜕µ 𝜕 𝑣𝑥 = (𝑢 − 2 ) 𝑣𝑦 = µ 2 + (𝑣 − 2 ) + (𝑢 − 2 ) 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑥 𝑣𝑥 + 𝑣𝑥 𝑣𝑥 − 𝑣𝑥 𝑣𝑦 + 𝑣𝑦 𝑣𝑦 − 𝑣𝑦 𝑉𝑥𝑝 = 𝑉𝑥𝑚 = 𝑉𝑦𝑝 = 𝑉𝑦𝑚 = 2µℎ𝑥 2µℎ𝑥 2µℎ𝑦 2µℎ𝑦 In a similar way like solution of stream function we separate discretized algebraic equations for vorticity to the time marching solution consisting in horizontal sweep (time step dt1) and vertical sweep (time step dt2). 2𝑖,𝑗 1𝑖𝑗 𝑔𝑥𝑥 𝑔𝑦𝑦 𝑔 𝑔 1 𝑦𝑦 𝑦𝑦 + 2 2 − 𝑉𝑦𝑝 + 𝑉𝑦𝑚 − 2𝑖,𝑗−1 − 𝑉𝑦𝑝 − 2𝑖,𝑗+1 + 𝑉𝑦𝑚 = + 𝑖−1,𝑗 − 2𝑖,𝑗 + 𝑖+1,𝑗 2 𝑡2 𝑡2 ℎ𝑥2 ℎ𝑦 ℎ𝑦 ℎ𝑦2 𝑔𝑥𝑦 𝐺𝑖𝑗 +𝑉𝑥𝑝 𝑖,𝑗 − 𝑖−1,𝑗 + 𝑉𝑥𝑚 𝑖+1,𝑗 − 𝑖,𝑗 + + 𝑖−1,𝑗−1 − 𝑖−1,𝑗+1 − 𝑖+1,𝑗−1 − 4ℎ𝑥 ℎ𝑦 𝑖+1,𝑗+1 𝜕 𝜕 𝑞1 = 𝜕𝑦 𝜕𝑥 𝜕2 𝑞2 = 𝜕𝑥𝑦 𝜕 𝜕 𝑞3 = 𝜕𝑦 𝜕𝑥 𝜕2 𝑞4 = 2 𝜕𝑥 𝑞5 = ( 𝜕 2 ) 𝜕𝑥 𝑞6 = ( 𝜕 2 ) 𝜕𝑥 𝑞7 = 2 𝜕 𝜕 𝜕𝑥 𝜕𝑥 𝑞8 = ( 𝐺𝑖𝑗 𝑞1 𝑞2 = 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 − 2 𝑠𝑦𝑦 − 𝑠𝑥𝑥 + 𝑠𝑦𝑦 − 𝑠𝑥𝑥 + 𝑠𝑦𝑦 − 𝑠𝑥𝑥 − 𝑠𝑦𝑦 − 𝑠𝑥𝑥 𝑖,𝑗+1 𝑖,𝑗 𝑖,𝑗−1 𝑖,𝑗+1 𝑖,𝑗−1 2ℎ𝑦 ℎ𝑦 𝑞3 𝑞4 + 𝑠𝑦𝑦 − 𝑠𝑥𝑥 + 𝑠𝑦𝑦 − 𝑠𝑥𝑥 − 𝑠𝑦𝑦 − 𝑠𝑥𝑥 − 𝑠𝑦𝑦 − 𝑠𝑥𝑥 + 𝑠 − 𝑠𝑥𝑦𝑖,𝑗−1 𝑖+1,𝑗+1 𝑖−1,𝑗−1 𝑖+1,𝑗−1 𝑖−1,𝑗+1 4ℎ𝑥 ℎ𝑦 2ℎ𝑦 𝑥𝑦𝑖,𝑗+1 𝑞5 𝑞6 − 𝑞8 𝑞7 + 2 𝑠𝑥𝑦𝑖+1,𝑗 − 2𝑠𝑥𝑦𝑖,𝑗 + 𝑠𝑥𝑦𝑖−1,𝑗 + 𝑠 − 2𝑠 + 𝑠 + 𝑠 + 𝑠𝑥𝑦𝑖−1,𝑗−1 − 𝑠𝑥𝑦𝑖−1,𝑗+1 − 𝑠𝑥𝑦𝑖+1,𝑗−1 𝑥𝑦𝑖,𝑗+1 𝑥𝑦𝑖,𝑗 𝑥𝑦𝑖,𝑗−1 4ℎ𝑥 ℎ𝑦 𝑥𝑦𝑖+1,𝑗+1 ℎ𝑥 ℎ𝑦2 44 𝜕 2 ) 𝜕𝑦 Iterative solution of OMG.m %vorticity omegaold=omega; for iter=1:20 %vertical sweep for i=2:nx-1 aa(1)=0;bb(1)=1;cc(1)=0;dd(1)=omega(i,1); aa(ny)=0;bb(ny)=1;cc(ny)=0;dd(ny)=omega(i,ny); for j=2:ny-1 vx=(rho*u(i,j)-2*dmjudx(i,j))*dksidx(i,j); vy=(rho*v(i,j)-2*dmjudy(i,j))*detady(i,j)+mju(i,j)*detadxdx(i,j); Vxp=(vx+abs(vx))/(2*mju(i,j)*hx); Vxn=(vx-abs(vx))/(2*mju(i,j)*hx); Vyp=(vy+abs(vy))/(2*mju(i,j)*hy); Vym=(vy-abs(vy))/(2*mju(i,j)*hy); G1=q1(i,j)/hy^2*(syy(i,j+1)-sxx(i,j+1)-2*(syy(i,j)-sxx(i,j))+syy(i,j-1)-sxx(i,j-1)); G2=q2(i,j)/(2*hy)*(syy(i,j+1)-sxx(i,j+1)-syy(i,j-1)+sxx(i,j-1)); G3=q3(i,j)/(4*hx*hy)*(syy(i+1,j+1)-sxx(i+1,j+1)+syy(i-1,j-1)-sxx(i-1,j-1)-syy(i+1,j-1)+sxx(i+1,j-1)-syy(i-1,j+1)+syy(i-1,j+1)); G4=q4(i,j)/(2*hy)*(sxy(i,j+1)-sxy(i,j-1)); G5=q5(i,j)/hx^2*(sxy(i+1,j)-2*sxy(i,j)+sxy(i-1,j)); G6=(q6(i,j)-q8(i,j))/hy^2*(sxy(i,j+1)-sxy(i,j)+sxy(i,j-1)); G7=q7(i,j)/(4*hx*hy)*(sxy(i+1,j+1)+sxy(i-1,j-1)-sxy(i-1,j+1)-sxy(i+1,jk-1)); Gij=G1+G2+G3+G4+G5+G6+G7; aa(j)=-gyy(i,j)/hy^2+Vyp; cc(j)=-gyy(i,j)/hy^2-Vym; bb(j)=1/dt2+2*gyy(i,j)/hy^2-Vyp+Vym; dd1=omega(i,j)/dt2+gxx(i,j)/hx^2*(omega(i-1,j)-2*omega(i,j)+omega(i+1,j)); dd2=Vxp*(omega(i,j)-omega(i-2,j))+Vxm*(omega(i+1,j)-omega(i,j)); dd3=gxy(i,j)/(4*hx*hy)*(omega(i+1,j+1)+omega(i-1,j-1)-omega(i+1,j-1)-omega(i-1,j+1)); dd(j)=dd1+dd2+dd3-Gij/mju(i,j); end omega(i,:)=tridag(aa,bb,cc,dd,ny); end sum=0; for i=1:nx for j=1:ny sum=sum+abs(omega(i,j)-omegaold(i,j)); end end sdev=sum/(nx*ny)/psiw; disp(sprintf(' iter=%d relsum=%.5g',iter,sdev)) if sdev<0.00001 break end end 45 Constitutive equations White Metzner. Geometry Approximate profiles u,v, , for power law fluid iteration Elastic stress sxx, syy, sxy (hyperbolické MW) When you skip these three sections (iterative refinement of the stream function and vorticity) will result Pavlovec approximation of the solution, but for a smooth Gaussian gap profile. Poisson equation for Nodal velocities and viscosity. BC for Poisson equation for Poisson equation for pressure p 46 Constitutive equation White Metzner. Kontroloval: xxxxxx 2 𝑠𝑥𝑥 2𝜇 𝜕𝑢 𝜇 𝜕𝑠𝑥𝑥 𝜕𝑠𝑥𝑥 2𝜇 𝜕𝑢 2𝜇2 𝜕 2 𝑢 𝜕2𝑣 𝜕𝑢 1− + 𝑢 +𝑣 = 𝑠𝑥𝑦 − (𝑢 2 − 𝑣 2 − 2 𝐺 𝜕𝑥 𝐺 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝑦 𝐺 𝜕𝑥 𝜕𝑦 𝜕𝑥 2 𝑠𝑦𝑦 𝜕𝑠𝑦𝑦 𝜕𝑠𝑦𝑦 2𝜇 𝜕𝑢 𝜇 2𝜇 𝜕𝑣 2𝜇2 𝜕 2 𝑢 𝜕2𝑣 𝜕𝑣 1+ + 𝑢 +𝑣 = 𝑠𝑥𝑦 + (𝑢 2 − 𝑣 2 + 2 𝐺 𝜕𝑥 𝐺 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝑥 𝐺 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑢 − 𝜕𝑦 2 𝜕𝑣 + 𝜕𝑥 2 − 𝜕𝑢 𝜕𝑣 ) 𝜕𝑦 𝜕𝑥 + 𝜕𝑢 𝜕𝑣 ) 𝜕𝑦 𝜕𝑥 𝜕𝑠𝑥𝑦 𝜕𝑠𝑥𝑦 𝜇 𝜇 𝜕𝑣 𝜕𝑢 𝜇2 𝜕2𝑣 𝜕2𝑣 𝜕2𝑢 𝜕2𝑢 𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕𝑣 𝑠𝑥𝑦 + 𝑢 +𝑣 = 𝑠 + 𝑠 − (𝑢 − +𝑣 − −2 −2 ) 𝐺 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝑥 𝑥𝑥 𝜕𝑦 𝑦𝑦 𝐺 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑦 2 𝜕𝑥 2 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑣 𝜕𝑣 𝜕𝑢 = −2 ( − ) 𝜕𝑦 𝜕𝑥 𝜕𝑦 47 Kontroloval: xxxxxx Constitutive equation White Metzner. Boundary conditions (BC) at wall. For upper curved wall 𝑠𝑥𝑥 𝑠𝑦𝑦 𝑠𝑥𝑦 2𝜇 𝜕𝑢 2𝜇 𝜕𝑢 2𝜇2 𝜕𝑢 1− = 𝑠𝑥𝑦 + (2 𝐺 𝜕𝑥 𝐺 𝜕𝑦 𝐺 𝜕𝑥 2 2𝜇 𝜕𝑢 2𝜇 𝜕𝑣 2𝜇2 𝜕𝑣 1+ = 𝑠 + (2 𝐺 𝜕𝑥 𝐺 𝜕𝑥 𝑥𝑦 𝐺 𝜕𝑦 2 𝜕𝑢 + 𝜕𝑦 2 𝜕𝑣 + 𝜕𝑥 2 + 𝜕𝑢 𝜕𝑣 ) 𝜕𝑦 𝜕𝑥 System of 3 linear algebraic equations for 3 unknowns 𝜕𝑢 𝜕𝑣 + ) 𝜕𝑦 𝜕𝑥 2𝜇2 𝜕𝑢 𝜇 𝜕𝑣 𝜕𝑢 𝜕𝑣 𝜕𝑢 = 𝑠𝑥𝑥 + 𝑠𝑦𝑦 + ( − ) 𝐺 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝑥 𝜕𝑥 𝜕𝑦 2𝜇 𝜕𝑢 1− 𝐺 𝜕𝑥 0 − 𝜇 𝜕𝑣 𝐺 𝜕𝑥 0 2𝜇 𝜕𝑢 1+ 𝐺 𝜕𝑥 𝜇 𝜕𝑢 − 𝐺 𝜕𝑦 2𝜇 𝜕𝑢 − 𝐺 𝜕𝑦 2𝜇 𝜕𝑣 − 𝐺 𝜕𝑥 1 𝑠𝑥𝑥 𝑠𝑦𝑦 = 𝑠𝑥𝑦 2𝜇 2 𝜕𝑢 (2 𝐺 𝜕𝑥 2 2 𝜕𝑢 + 𝜕𝑦 2 + 𝜕𝑢 𝜕𝑣 ) 𝜕𝑦 𝜕𝑥 2 2𝜇 2 𝜕𝑣 𝜕𝑣 𝜕𝑢 𝜕𝑣 (2 + + ) 𝐺 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 2𝜇 2 𝜕𝑢 𝜕𝑣 𝜕𝑢 ( − ) 𝐺 𝜕𝑥 𝜕𝑥 𝜕𝑦 For bottom (planar) wall 𝑠𝑥𝑥 2𝜇2 𝜕𝑢 = 𝐺 𝜕𝑦 𝑠𝑦𝑦 = 0 2 At wall with transducers is syy=0 and yy =0. It offers the possibility to integrate WM hyperbolic equations from left to right and bottom to top 𝑠𝑥𝑦 = 0 48 Transformation of constitutive equations White Metzner. 𝑠𝑥𝑥 1 − Kontroloval: xxxxxx 2𝜇 𝜕𝑢 𝜕𝜉 𝜕𝑢 𝜕𝜂 𝜇 𝜕𝑠𝑥𝑥 𝜕𝜉 𝜕𝑠𝑥𝑥 𝜕𝜂 𝜕𝑠𝑥𝑥 𝜕𝜂 2𝜇 𝜕𝑢 𝜕𝜂 ( + ) + 𝑢( + ) +𝑣 = 𝑠 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 2 2 2𝜇2 𝜕𝑢 𝜕 2 𝜂 𝜕 2 𝑢 𝜕𝜉 𝜕 2 𝑢 𝜕𝜂 𝜕 2 𝑢 𝜕𝜂 𝜕𝜉 𝜕 2 𝑣 𝜕𝜂 − (𝑢( + + 2 +2 )−𝑣 2 𝐺 𝜕𝜂 𝜕𝑥 2 𝜕𝜉 2 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝜉𝜕𝜂 𝜕𝑥 𝜕𝑥 𝜕𝜂 𝜕𝑦 2 2 𝜕𝑢 𝜕𝜉 𝜕𝑢 𝜕𝜂 𝜕𝑢 𝜕𝜂 𝜕𝑢 𝜕𝜂 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 −2 + − − ( + )) 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝜂 𝜕𝑦 𝜕𝜂 𝜕𝑦 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝑠𝑦𝑦 1 + 2 𝜕𝑠𝑦𝑦 𝜕𝜉 𝜕𝑠𝑦𝑦 𝜕𝜂 𝜕𝑠𝑦𝑦 𝜕𝜂 2𝜇 𝜕𝑢 𝜕𝜉 𝜕𝑢 𝜕𝜂 𝜇 2𝜇 𝜕𝑢 𝜕𝜂 ( + ) + 𝑢( + ) +𝑣 = 𝑠 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 2 2 2𝜇2 𝜕𝑢 𝜕 2 𝜂 𝜕 2 𝑢 𝜕𝜉 𝜕 2 𝑢 𝜕𝜂 𝜕 2 𝑢 𝜕𝜂 𝜕𝜉 𝜕 2 𝑣 𝜕𝜂 + (𝑢( + + 2 +2 )−𝑣 2 𝐺 𝜕𝜂 𝜕𝑥 2 𝜕𝜉 2 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝜉𝜕𝜂 𝜕𝑥 𝜕𝑥 𝜕𝜂 𝜕𝑦 2 2 𝜕𝑢 𝜕𝜉 𝜕𝑢 𝜕𝜂 𝜕𝑢 𝜕𝜉 𝜕𝑢 𝜕𝜂 𝜕𝑢 𝜕𝜂 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 +2 + + + + ( + )) 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝜂 𝜕𝑦 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 2 𝜕𝑠𝑥𝑦 𝜕𝜉 𝜕𝑠𝑥𝑦 𝜕 𝜇 𝜕𝑠𝑥𝑦 𝜕𝜂 𝜇 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 𝜕𝑢 𝜕𝜂 𝑠𝑥𝑦 + 𝑢( + )+𝑣 = ( + )𝑠 + 𝑠 𝐺 𝜕 𝜕𝑥 𝜕 𝜕𝑥 𝜕𝜂 𝜕𝑦 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝑥𝑥 𝜕𝜂 𝜕𝑦 𝑦𝑦 𝜇2 𝜕𝑣 𝜕 2 𝜂 𝜕 2 𝑣 𝜕𝜉 − (𝑢 + 𝐺 𝜕𝜂 𝜕𝑥 2 𝜕𝜉 2 𝜕𝑥 𝜕 2 𝑢 𝜕𝜂 2 𝜕𝑢 𝜕 2 𝜂 𝜕 2 𝑢 𝜕𝜉 +𝑣 ( ) − − 𝜕𝜂2 𝜕𝑦 𝜕𝜂 𝜕𝑥 2 𝜕𝜉 2 𝜕𝑥 2 2 𝜕 2 𝑣 𝜕𝜂 + 2 𝜕𝜂 𝜕𝑥 𝜕 2 𝑢 𝜕𝜂 − 2 𝜕𝜂 𝜕𝑥 2 2 𝜕 2 𝑣 𝜕𝜂 𝜕𝜉 𝜕 2 𝑣 𝜕𝜂 +2 − 𝜕𝜉𝜕𝜂 𝜕𝑥 𝜕𝑥 𝜕𝜂2 𝜕𝑦 2 𝜕 2 𝑢 𝜕𝜂 𝜕𝜉 𝜕𝑣 𝜕𝜂 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 𝜕𝑢 𝜕𝜂 −2 −2 ( + − )) 𝜕𝜉𝜕𝜂 𝜕𝑥 𝜕𝑥 𝜕𝜂 𝜕𝑦 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝜂 𝜕𝑦 49 Transformation of constitutive equations White Metzner (auxilliary variables) 𝑠𝑥𝑥 𝑠𝑦𝑦 Kontroloval: xxxxxx 2𝜇 𝜕𝑢 𝜕𝜉 𝜕𝑢 𝜕𝜂 𝐴= ( + ) 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜇 𝜕𝑠𝑥𝑥 𝜕𝜉 𝜕𝑠𝑥𝑥 𝜕𝜂 𝜕𝜂 2𝜇 𝜕𝑢 𝜕𝜂 1−𝐴 + 𝑢 + (𝑢 +𝑣 ) = 𝑠 − 𝑄𝑥𝑥 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 Interpretation as material derivative 𝜇 𝜕𝑠𝑦𝑦 𝜕𝜉 𝜕𝑠𝑦𝑦 𝜕𝜂 𝜕𝜂 2𝜇 𝜕𝑢 𝜕𝜂 1+𝐴 + 𝑢 + (𝑢 +𝑣 ) = 𝑠 + 𝑄𝑦𝑦 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 𝐷 𝜕𝜂 𝜕𝜂 =𝑢 +𝑣 𝐷𝑡 𝜕𝑥 𝜕𝑦 (corresponding to motion of fluid particle along =konst) 𝜇 𝜕𝑠𝑥𝑦 𝜕𝜉 𝜕𝑠𝑥𝑦 𝜕 𝜕𝜂 𝜇 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 𝜕𝑢 𝜕𝜂 𝑠𝑥𝑦 + 𝑢 + (𝑢 +𝑣 ) = ( + )𝑠𝑥𝑥 + 𝑠 − 𝑄𝑥𝑦 𝐺 𝜕 𝜕𝑥 𝜕 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝜂 𝜕𝑦 𝑦𝑦 2𝜇2 𝜕𝑢 𝜕 2 𝜂 𝜕 2 𝑢 𝜕𝜉 = (𝑢( + 𝐺 𝜕𝜂 𝜕𝑥 2 𝜕𝜉 2 𝜕𝑥 2 2𝜇2 𝜕𝑢 𝜕 2 𝜂 𝜕 2 𝑢 𝜕𝜉 𝑄𝑦𝑦 = (𝑢( + 𝐺 𝜕𝜂 𝜕𝑥 2 𝜕𝜉 2 𝜕𝑥 2 𝜇2 𝜕𝑣 𝜕 2 𝜂 𝜕 2 𝑣 𝜕𝜉 𝑄𝑥𝑦 = (𝑢 + 𝐺 𝜕𝜂 𝜕𝑥 2 𝜕𝜉 2 𝜕𝑥 2 𝑄𝑥𝑥 𝜕 2 𝑢 𝜕𝜂 + 2 𝜕𝜂 𝜕𝑥 2 𝜕 2 𝑢 𝜕𝜂 + 2 𝜕𝜂 𝜕𝑥 2 𝜕 2 𝑣 𝜕𝜂 + 2 𝜕𝜂 𝜕𝑥 2 𝜕 2 𝑢 𝜕𝜂 𝜕𝜉 𝜕 2 𝑣 𝜕𝜂 +2 )−𝑣 2 𝜕𝜉𝜕𝜂 𝜕𝑥 𝜕𝑥 𝜕𝜂 𝜕𝑦 2 𝜕 2 𝑢 𝜕𝜂 𝜕𝜉 𝜕 2 𝑣 𝜕𝜂 +2 )−𝑣 2 𝜕𝜉𝜕𝜂 𝜕𝑥 𝜕𝑥 𝜕𝜂 𝜕𝑦 2 2 𝜕𝑢 𝜕𝜉 𝜕𝑢 𝜕𝜂 −2 + 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 2 𝜕𝑢 𝜕𝜉 𝜕𝑢 𝜕𝜂 +2 + 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 2 𝜕 2 𝑣 𝜕𝜂 𝜕𝜉 𝜕 2 𝑣 𝜕𝜂 𝜕 2 𝑢 𝜕𝜂 +2 − +𝑣 𝜕𝜉𝜕𝜂 𝜕𝑥 𝜕𝑥 𝜕𝜂2 𝜕𝑦 𝜕𝜂2 𝜕𝑦 𝜕𝑣 𝜕𝜂 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 𝜕𝑢 𝜕𝜂 −2 ( + − )) 𝜕𝜂 𝜕𝑦 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝜂 𝜕𝑦 2 𝜕𝑢 𝜕𝜂 − 𝜕𝜂 𝜕𝑦 2 − 𝜕𝑢 𝜕𝜂 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 ( + )) 𝜕𝜂 𝜕𝑦 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝑢 𝜕𝜉 𝜕𝑢 𝜕𝜂 + + 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕𝑢 𝜕 2 𝜂 𝜕 2 𝑢 𝜕𝜉 − − 𝜕𝜂 𝜕𝑥 2 𝜕𝜉 2 𝜕𝑥 2 2 + 𝜕𝑢 𝜕𝜂 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 ( + )) 𝜕𝜂 𝜕𝑦 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝜕 2 𝑢 𝜕𝜂 − 2 𝜕𝜂 𝜕𝑥 2 𝜕 2 𝑢 𝜕𝜂 𝜕𝜉 −2 𝜕𝜉𝜕𝜂 𝜕𝑥 𝜕𝑥 50 Discretized constitutive equation White Metzner (UP direction) geometry =1 =0 Kontroloval: xxxxxx i-1,j i,j 𝑠𝑥𝑥 𝑖,1 = hy i,j-1 =0 i,j=1 =2 2𝜇 𝐺 2 𝜕𝑢 𝜕𝑦 2 Approximate profiles u,v, , for power law fluid iteration Elastic stress sxx, syy, sxy (hyperbolic MW) 𝑠𝑦𝑦 𝑖,1 = 0 Poisson equation for Velocities, µ , BC for 𝑠𝑥𝑦 𝑖,1 = 0 Poisson equation for Poisson equation for pressure p 𝑠𝑥𝑥 𝑖,𝑗 𝜇 𝑠𝑥𝑥 𝑖,𝑗 − 𝑠𝑥𝑥 𝑖−1,𝑗 𝜕𝜉 𝑠𝑥𝑥 𝑖,𝑗 − 𝑠𝑥𝑥 𝑖,𝑗−1 𝜕𝜂 𝜕𝜂 2𝜇 𝜕𝑢 𝜕𝜂 1−𝐴 + 𝑢 + (𝑢 +𝑣 ) = 𝑠 − 𝑄𝑥𝑥 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 𝑖,𝑗 𝑠𝑦𝑦 𝑖, 𝑗 𝜇 𝑠𝑦𝑦 𝑖,𝑗 − 𝑠𝑦𝑦 𝑖−1,𝑗 𝜕𝜉 𝑠𝑦𝑦 𝑖,𝑗 − 𝑠𝑦𝑦 𝑖,𝑗−1 𝜕𝜂 𝜕𝜂 2𝜇 𝜕𝑢 𝜕𝜂 1+𝐴 + 𝑢 + (𝑢 +𝑣 ) = 𝑠 + 𝑄𝑦𝑦 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 𝑖,𝑗 𝜇 𝑠𝑥𝑦 𝑖,𝑗 − 𝑠𝑥𝑦 𝑖−1,𝑗 𝜕𝜉 𝑠𝑥𝑦 𝑖,𝑗 − 𝑠𝑥𝑦 𝑖,𝑗−1 𝜕 𝜕𝜂 𝜇 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 𝜕𝑢 𝜕𝜂 𝑠𝑥𝑦 𝑖,𝑗 + 𝑢 + (𝑢 +𝑣 ) = ( + )𝑠 + 𝑠 − 𝑄𝑥𝑦 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝑥𝑥 𝑖,𝑗 𝜕𝜂 𝜕𝑦 𝑦𝑦 𝑖,𝑗 51 Discretized constitutive equation White Metzner (UP direction). Kontroloval: xxxxxx Resulting system of algebraic equations for three stresses at point i,j evaluated from stresses at left and from bottom 𝜇 𝑢 𝜕 𝑢 𝜕 𝑣 𝜕 1−𝐴+ ( + + ) 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 ℎ𝑦 𝜕𝑦 0 = 𝜇 𝐺 𝜇 𝐺 𝜇 𝐺 𝜇 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 − ( + ) 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝑠𝑥𝑥 𝑖−1,𝑗 𝜕𝜉 𝑠𝑥𝑥 𝑖,𝑗−1 𝜕𝜂 𝑢 + (𝑢 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 𝑠𝑦𝑦 𝑖−1,𝑗 𝜕𝜉 𝑠𝑦𝑦 𝑖,𝑗−1 𝜕𝜂 𝑢 + (𝑢 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 𝑠𝑥𝑦 𝑖−1,𝑗 𝜕𝜉 𝑠𝑥𝑦 𝑖,𝑗−1 𝜕𝜂 𝑢 + (𝑢 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 2𝜇 𝜕𝑢 𝜕𝜂 0 − 𝐺 𝜕𝜂 𝜕𝑦 𝜇 𝑢 𝜕 𝑢 𝜕 𝑣 𝜕 2𝜇 𝜕𝑢 𝜕𝜂 1+𝐴+ ( + + ) − 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 ℎ𝑦 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝜇 𝜕𝑢 𝜕𝜂 𝜇 𝑢 𝜕 𝑢 𝜕 𝑣 𝜕 − 1+ ( + + ) 𝐺 𝜕𝜂 𝜕𝑦 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 ℎ𝑦 𝜕𝑦 𝜕𝜂 + 𝑣 ) − 𝑄𝑥𝑥 𝜕𝑦 𝜕𝜂 + 𝑣 ) − 𝑄𝑦𝑦 𝜕𝑦 𝜕𝜂 + 𝑣 ) − 𝑄𝑥𝑦 𝜕𝑦 𝑠𝑥𝑥 , 𝑖𝑗 𝑠𝑦𝑦 , 𝑖𝑗 𝑠𝑥𝑦 , 𝑖𝑗 52 Discretized constitutive equation White Metzner (DOWN direction) geometry =1 i-1,j =0 Kontroloval: xxxxxx =0 i,j+1 i,j i,j=1 𝑠𝑥𝑥 𝑖,1 = hy =2 2𝜇 𝐺 2 𝜕𝑢 𝜕𝑦 2 Approximate profiles u,v, , for power law fluid iteration Elastic stress sxx, syy, sxy (hyperbolic MW) 𝑠𝑦𝑦 𝑖,1 = 0 Poisson equation for Velocities, µ , BC for 𝑠𝑥𝑦 𝑖,1 = 0 Poisson equation for Poisson equation for pressure p 𝑠𝑥𝑥 𝑖,𝑗 𝜇 𝑠𝑥𝑥 𝑖,𝑗 − 𝑠𝑥𝑥 𝑖−1,𝑗 𝜕𝜉 𝑠𝑥𝑥 𝑖,𝑗+1 − 𝑠𝑥𝑥 𝑖,𝑗 𝜕𝜂 𝜕𝜂 2𝜇 𝜕𝑢 𝜕𝜂 1−𝐴 + 𝑢 + (𝑢 +𝑣 ) = 𝑠 − 𝑄𝑥𝑥 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 𝑖,𝑗 𝑠𝑦𝑦 𝑖, 𝑗 𝜇 𝑠𝑦𝑦 𝑖,𝑗 − 𝑠𝑦𝑦 𝑖−1,𝑗 𝜕𝜉 𝑠𝑦𝑦 𝑖,𝑗+1 − 𝑠𝑦𝑦 𝑖,𝑗 𝜕𝜂 𝜕𝜂 2𝜇 𝜕𝑢 𝜕𝜂 1+𝐴 + 𝑢 + (𝑢 +𝑣 ) = 𝑠 + 𝑄𝑦𝑦 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 𝑖,𝑗 𝜇 𝑠𝑥𝑦 𝑖,𝑗 − 𝑠𝑥𝑦 𝑖−1,𝑗 𝜕𝜉 𝑠𝑥𝑦 𝑖,𝑗+1 − 𝑠𝑥𝑦 𝑖,𝑗 𝜕 𝜕𝜂 𝜇 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 𝜕𝑢 𝜕𝜂 𝑠𝑥𝑦 𝑖,𝑗 + 𝑢 + (𝑢 +𝑣 ) = ( + )𝑠 + 𝑠 − 𝑄𝑥𝑦 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝑥𝑥 𝑖,𝑗 𝜕𝜂 𝜕𝑦 𝑦𝑦 𝑖,𝑗 53 Discretized constitutive equation White Metzner (DOWN direction). Kontroloval: xxxxxx Resulting system of algebraic equations for three stresses at point i,j evaluated from stresses at left and top 1−𝐴+ 𝜇 𝑢 𝜕 𝑢 𝜕 𝑣 𝜕 ( − − ) 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 ℎ𝑦 𝜕𝑦 0 = 𝜇 𝐺 𝜇 𝐺 𝜇 𝐺 𝜇 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 − ( + ) 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝑠𝑥𝑥 𝑖−1,𝑗 𝜕𝜉 𝑠𝑥𝑥 𝑖,𝑗+1 𝜕𝜂 𝑢 − (𝑢 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 𝑠𝑦𝑦 𝑖−1,𝑗 𝜕𝜉 𝑠𝑦𝑦 𝑖,𝑗+1 𝜕𝜂 𝑢 − (𝑢 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 𝑠𝑥𝑦 𝑖−1,𝑗 𝜕𝜉 𝑠𝑥𝑦 𝑖,𝑗+1 𝜕𝜂 𝑢 − (𝑢 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 2𝜇 𝜕𝑢 𝜕𝜂 𝐺 𝜕𝜂 𝜕𝑦 𝜇 𝑢 𝜕 𝑢 𝜕 𝑣 𝜕 2𝜇 𝜕𝑢 𝜕𝜂 1+𝐴+ ( − − ) − 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 ℎ𝑦 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝜇 𝜕𝑢 𝜕𝜂 𝜇 𝑢 𝜕 𝑢 𝜕 𝑣 𝜕 − 1+ ( − − ) 𝐺 𝜕𝜂 𝜕𝑦 𝐺 ℎ𝑥 𝜕𝑥 ℎ𝑦 𝜕𝑥 ℎ𝑦 𝜕𝑦 𝜕𝜂 + 𝑣 ) − 𝑄𝑥𝑥 𝜕𝑦 𝜕𝜂 + 𝑣 ) − 𝑄𝑦𝑦 𝜕𝑦 𝜕𝜂 + 𝑣 ) − 𝑄𝑥𝑦 𝜕𝑦 0 − 𝑠𝑥𝑥 , 𝑖𝑗 𝑠𝑦𝑦 , 𝑖𝑗 𝑠𝑥𝑦 , 𝑖𝑗 54 Kontroloval: xxxxxx Discretized constitutive equation White Metzner (iterative threediagonal) 2𝜇 𝜕𝑢 1− 𝐺 𝜕𝑥 0 =1 i-1,j =0 =0 𝑠𝑥𝑥 𝑖,𝑗 1 − 𝐴 + 𝑠𝑦𝑦 𝑖, 𝑗 i,j+1 i,j i,j-1 i,j=1 hy − 𝜇 𝜕𝑣 𝐺 𝜕𝑥 0 2𝜇 𝜕𝑢 1+ 𝐺 𝜕𝑥 𝜇 𝜕𝑢 − 𝐺 𝜕𝑦 =2 𝑠𝑥𝑥 𝑖,1 2𝜇 𝜕𝑢 − 𝐺 𝜕𝑦 2𝜇 𝜕𝑣 − 𝐺 𝜕𝑥 𝑠𝑥𝑥𝑖,𝑛𝑦 𝑠𝑦𝑦 𝑖, 𝑛𝑦 = 𝑠𝑥𝑦𝑖,𝑛𝑦 1 2𝜇2 𝜕𝑢 = 𝐺 𝜕𝑦 2𝜇 2 𝜕𝑢 (2 𝐺 𝜕𝑥 2 2 𝜕𝑢 + 𝜕𝑦 2 + 𝜕𝑢 𝜕𝑣 ) 𝜕𝑦 𝜕𝑥 2 2𝜇 2 𝜕𝑣 𝜕𝑣 𝜕𝑢 𝜕𝑣 (2 + + ) 𝐺 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 2𝜇 2 𝜕𝑢 𝜕𝑣 𝜕𝑢 ( − ) 𝐺 𝜕𝑥 𝜕𝑥 𝜕𝑦 2 𝑠𝑦𝑦 𝑖,1 = 0 𝑠𝑥𝑦 𝑖,1 = 0 𝜇 𝑠𝑥𝑥 𝑖,𝑗 − 𝑠𝑥𝑥 𝑖−1,𝑗 𝜕𝜉 𝑠𝑥𝑥 𝑖,𝑗+1 − 𝑠𝑥𝑥 𝑖,𝑗−1 𝜕𝜂 𝜕𝜂 2𝜇 𝜕𝑢 𝜕𝜂 𝑢 + (𝑢 +𝑣 ) = 𝑠 − 𝑄𝑥𝑥 𝐺 ℎ𝑥 𝜕𝑥 2ℎ𝑦 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 𝑖,𝑗 𝜇 𝑠𝑦𝑦 𝑖,𝑗 − 𝑠𝑦𝑦 𝑖−1,𝑗 𝜕𝜉 𝑠𝑦𝑦 𝑖,𝑗+1 − 𝑠𝑦𝑦 𝑖,𝑗−1 𝜕𝜂 𝜕𝜂 2𝜇 𝜕𝑢 𝜕𝜂 1+𝐴 + 𝑢 + (𝑢 +𝑣 ) = 𝑠 + 𝑄𝑦𝑦 𝐺 ℎ𝑥 𝜕𝑥 2ℎ𝑦 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 𝑖,𝑗 𝜇 𝑠𝑥𝑦 𝑖,𝑗 − 𝑠𝑥𝑦 𝑖−1,𝑗 𝜕𝜉 𝑠𝑥𝑦 𝑖,𝑗+1 − 𝑠𝑥𝑦 𝑖,𝑗−1 𝜕 𝜕𝜂 𝜇 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 𝜕𝑢 𝜕𝜂 𝑠𝑥𝑦 𝑖,𝑗 + 𝑢 + (𝑢 +𝑣 ) = ( + )𝑠 + 𝑠 − 𝑄𝑥𝑦 𝐺 ℎ𝑥 𝜕𝑥 2ℎ𝑦 𝜕𝑥 𝜕𝑦 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝑥𝑥 𝑖,𝑗 𝜕𝜂 𝜕𝑦 𝑦𝑦 𝑖,𝑗 55 Kontroloval: xxxxxx Discretized constitutive equation White Metzner (iterative threediagonal) Diagonal (vector bb) WM BC for sxx, syy, sxy i=2,3,…,nx Low diagonal (vector aa) Upper diagonal (vector cc) 𝜇 𝑢 𝜕𝜉 𝜇 𝜕𝜂 𝜕𝜂 𝜇 𝜕𝜂 𝜕𝜂 − 𝑠𝑥𝑥 𝑖,𝑗−1 𝑢 +𝑣 + 𝑠𝑥𝑥 𝑖,𝑗+1 𝑢 +𝑣 𝐺 ℎ𝑥 𝜕𝑥 2𝐺ℎ𝑦 𝜕𝑥 𝜕𝑦 2𝐺ℎ𝑦 𝜕𝑥 𝜕𝑦 𝜇 𝜕𝜉 2𝜇 𝜕𝑢 𝜕𝜂 = 𝑠𝑥𝑥 𝑖−1,𝑗 𝑢 + 𝑠 − 𝑄𝑥𝑥 𝐺ℎ𝑥 𝜕𝑥 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 𝑖,𝑗 𝑠𝑥𝑥 𝑖,𝑗 1 − 𝐴 + iteration stress sxx (3diagonal solver) stress syy (3diagonal solver) stress sxy (3diagonal solver) 𝜇 𝑢 𝜕𝜉 𝜇 𝜕𝜂 𝜕𝜂 𝜇 𝜕𝜂 𝜕𝜂 − 𝑠𝑦𝑦 𝑖,𝑗−1 𝑢 +𝑣 + 𝑠𝑦𝑦 𝑖,𝑗+1 𝑢 +𝑣 𝐺 ℎ𝑥 𝜕𝑥 2𝐺ℎ𝑦 𝜕𝑥 𝜕𝑦 2𝐺ℎ𝑦 𝜕𝑥 𝜕𝑦 𝜇 𝜕𝜉 2𝜇 𝜕𝑢 𝜕𝜂 = 𝑠 𝑢 + 𝑠 − 𝑄𝑦𝑦 𝐺ℎ𝑥 𝑦𝑦 𝑖−1,𝑗 𝜕𝑥 𝐺 𝜕𝜂 𝜕𝑦 𝑥𝑦 𝑖,𝑗 𝑠𝑦𝑦 𝑖,𝑗 1 + 𝐴 + I section 𝜇 𝑢 𝜕𝜉 𝜇 𝜕𝜂 𝜕𝜂 𝜇 𝜕𝜂 𝜕𝜂 𝑠𝑥𝑦 𝑖,𝑗 1 + − 𝑠𝑥𝑦 𝑖,𝑗−1 𝑢 +𝑣 + 𝑠𝑥𝑦 𝑖,𝑗+1 𝑢 +𝑣 𝐺 ℎ𝑥 𝜕𝑥 2𝐺ℎ𝑦 𝜕𝑥 𝜕𝑦 2𝐺ℎ𝑦 𝜕𝑥 𝜕𝑦 𝜇 𝜕𝜉 𝜇 𝜕𝑣 𝜕𝜉 𝜕𝑣 𝜕𝜂 𝜕𝑢 𝜕𝜂 = 𝑠 𝑢 + ( + )𝑠 + 𝑠 − 𝑄𝑥𝑦 𝐺ℎ𝑥 𝑥𝑦 𝑖−1,𝑗 𝜕𝑥 𝐺 𝜕𝜉 𝜕𝑥 𝜕𝜂 𝜕𝑥 𝑥𝑥 𝑖,𝑗 𝜕𝜂 𝜕𝑦 𝑦𝑦 𝑖,𝑗 56 Discretized constitutive equation White Metzner MW.m (Matlab) Kontroloval: xxxxxx Resulting system of algebraic equations for three stresses at point i,j evaluated from stresses at left and from bottom 57 Poisson equation for pressure (𝑢 𝜕𝑢 𝜕𝑢 𝜕𝑝 𝜕𝜏𝑥𝑥 𝜕𝜏𝑥𝑦 +𝑣 )=− + + 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑥 𝜕𝑦 Kontroloval: xxxxxx geometry Approximate profiles u,v, , for power law fluid iteration 𝜕𝑣 𝜕𝑣 𝜕𝑝 𝜕𝜏𝑥𝑦 𝜕𝜏𝑦𝑦 (𝑢 +𝑣 )= − + + 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕2𝑝 𝜕2𝑝 𝜕𝑢 𝜕𝑢 𝜕2𝑢 𝜕2𝑣 𝜕𝑣 𝜕𝑢 𝜕2𝑢 𝜕 2 𝑣 𝜕𝑣 𝜕𝑣 + + +𝑢 2+𝑣 2+2 +𝑣 +𝑢 + 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥𝑦 𝜕𝑥𝑦 𝜕𝑦 𝜕𝑦 𝜕 2 𝑥𝑦 𝜕2𝑝 𝜕2𝑝 𝜕𝑣 𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕 2 𝑥𝑥 𝜕 2 𝑦𝑦 = 2 + 2 + 2 − = + +2 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥𝜕𝑦 Elastic stress sxx, syy, sxy (hyperbolic MW) Poisson equation for Velocities, µ , BC for Poisson equation for Poisson equation for pressure p Decomposition to elastic and viscous components 𝜕 2 𝑥𝑥 𝜕 2 𝑠𝑥𝑥 𝜕3𝑢 𝜕µ 𝜕 2 𝑢 𝜕 2 µ 𝜕𝑢 = + 2µ 3 + 4 +2 2 𝜕𝑥 2 𝜕𝑥 2 𝜕𝑥 𝜕𝑥 𝜕𝑥 2 𝜕𝑥 𝜕𝑥 𝜕 2 𝑦𝑦 𝜕 2 𝑠𝑦𝑦 𝜕3𝑣 𝜕µ 𝜕 2 𝑣 𝜕 2 µ 𝜕𝑣 = + 2µ 3 + 4 +2 2 𝜕𝑦 2 𝜕𝑦 2 𝜕𝑦 𝜕𝑦 𝜕𝑦 2 𝜕𝑦 𝜕𝑦 𝜕 2 𝑥𝑦 𝜕 2 𝑠𝑥𝑦 𝜕3𝑢 𝜕3𝑣 𝜕µ 𝜕 2 𝑢 𝜕 2 𝑣 𝜕µ 𝜕 2 𝑣 𝜕2𝑢 𝜕 2 µ 𝜕𝑢 𝜕𝑣 = +µ + + + + + + + 𝜕𝑥𝜕𝑦 𝜕𝑥𝜕𝑦 𝜕𝑥𝜕𝑦 2 𝜕𝑦𝜕𝑥 2 𝜕𝑦 𝜕𝑥𝜕𝑦 𝜕𝑥 2 𝜕𝑥 𝜕𝑥𝜕𝑦 𝜕𝑦 2 𝜕𝑥𝜕𝑦 𝜕𝑦 𝜕𝑥 58 Poisson equation for pressure Kontroloval: xxxxxx For constant viscosity 𝜕 2 𝑥𝑦 𝜕 2 𝑠𝑥𝑥 𝜕 2 𝑠𝑦𝑦 𝜕 2 𝑠𝑥𝑦 𝜕 2 𝑥𝑥 𝜕 2 𝑦𝑦 𝜕3𝑢 𝜕3𝑣 𝜕3𝑢 𝜕3𝑣 + +2 = + +2 + µ(2 3 + 2 3 + 2 +2 ) 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥𝜕𝑦 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥𝜕𝑦 2 𝜕𝑦𝜕𝑥 2 A sum of all the members in the viscosity coefficient is zero, therefore, the viscosity does not directly affect the relationship between the pressure and the elastic tension 𝜕3𝑢 𝜕3𝑣 𝜕3𝑢 𝜕3𝑣 2 3+2 3+2 +2 =0 𝜕𝑥 𝜕𝑦 𝜕𝑥𝜕𝑦 2 𝜕𝑦𝜕𝑥 2 𝜕 2 𝑠𝑥𝑦 𝜕2𝑝 𝜕2𝑝 𝜕𝑣 𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕 2 𝑠𝑥𝑥 𝜕 2 𝑠𝑦𝑦 + + 2 − = + +2 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥𝜕𝑦 For variable viscosity the relationship between the pressure and the elastic tension appear 𝜕 2 𝑠𝑥𝑦 𝜕2𝑝 𝜕2𝑝 𝜕𝑣 𝜕𝑢 𝜕𝑣 𝜕𝑢 𝜕 2 𝑠𝑥𝑥 𝜕 2 𝑠𝑦𝑦 + + 2 − = + +2 + 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑥 2 𝜕𝑦 2 𝜕𝑥𝜕𝑦 𝜕 𝜕 𝜕 𝜕 𝜕 2 µ 𝜕𝑢 𝜕𝑣 𝜕 2 µ 𝜕𝑢 𝜕 2 µ 𝜕𝑣 2( − + + + 2 + ) 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑥𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑥 𝜕𝑥 𝜕𝑦 2 𝜕𝑦 59 Poisson equation for pressure. Boundary conditions Kontroloval: xxxxxx The equation for pressure distribution is elliptic equation of second order, and in each point of the border is necessary to prescribe exactly one boundary condition: Either directly pressure (at least in one point, because the pressure field in an incompressible liquid is defined up to an arbitrary constant) or the normal derivative of the pressure (Neuman boundary condition of the second kind). At the inlet, where we assume the absence of elastic effects can be fixed e.g. zero pressure over the cross section (p = 0). The solid wall (u = v = 0) is required to enter a pressure gradient, starting from the Cauchy equilibrium equations 𝜕𝑝 𝜕𝜏𝑥𝑥 𝜕𝜏𝑥𝑦 𝜕𝑠𝑥𝑥 𝜕𝑠𝑥𝑦 𝜕2𝑢 𝜕2𝑢 𝜕2𝑣 𝜕 𝜕𝑢 𝜕 𝜕𝑢 𝜕𝑣 = + = + + (2 2 + 2 + )+2 + ( + ) 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥𝜕𝑦 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑝 = 𝜕𝑥 𝜕𝑠𝑥𝑥 𝜕𝑠𝑥𝑦 𝜕 𝜕 𝜕𝑢 𝜕 𝜕𝑢 𝜕𝑣 + − +2 + ( + ) 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑝 𝜕𝜏𝑦𝑦 𝜕𝜏𝑥𝑦 𝜕𝑠𝑦𝑦 𝜕𝑠𝑥𝑦 𝜕2𝑣 𝜕2𝑣 𝜕2𝑢 𝜕 𝜕𝑣 𝜕 𝜕𝑢 𝜕𝑣 = + = + + 2 2 + 2+ +2 + ( + ) 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑥𝜕𝑦 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 𝜕𝑝 = 𝜕𝑦 𝜕𝑠𝑦𝑦 𝜕𝑠𝑥𝑦 𝜕 𝜕 𝜕𝑣 𝜕 𝜕𝑢 𝜕𝑣 + − +2 + ( + ) 𝜕𝑦 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥 60