決ࣜ沲ࣜ汆࣭ࣨࣜ卆ࣜ氪ࣜ柣࣮ࣨࣜ劒ࣜ沫ࣜ笇ࣦ࣯
A CCURACY AND C ONVERGENCE OF THE L OCAL P RECONDITIONING ON THE H IGH A SPECT R ATIO G RIDS
J.E. Lee,
1Y. Kim
2and J.H. Kwon
*3The local preconditioning method has both robust convergence and accurate solutions by using local flow properties for parameters in the preconditioning matrix. Preconditioning methods have been very effective to low speed inviscid flows. In the viscous and turbulent flows, deterioration of convergence should be overcame on the high aspect ratio grids to get better convergence and accuracy. In the present study, the local time stepping and min-CFL/max-VNN definitions are applied to compare the results and we propose the method that switches between two methods. The min-CFL definition is applied for inviscid flow problems and the min-CFL/max-VNN definition is implemented to viscous and turbulent flow problems.
Key Words : (Local Preconditioning) High Aspect Ratio Grid in-CFL in-CFL Definition), mn-CFL/max-VNN in-CFL/max-VNN Definition)
* Corresponding author, E-mail: jhkwon@kaist.ac.kr
1.
(local preconditioning method) (stiffness)
[4].
(stagnation point) (singularity)
. (limiter)
.
. ,
,
.
.
. [1], [12], [13] min-CFL/max-VNN
.
min-CFL /max-VNN «
«
.
‘switching' min-CFL /max-VNN
.
2 Navier-
Park
. 3 MUSCL van Albada
DADI
. k-w WD+
.
10% bump ,
. 2 NACA 4412
.
2.
2.1
Navier-Stokes
.
ġćƒ
ª â ćƖ
â ćƗ
á ćƖ
Ɣ
â ćƗ
Ɣ
(1)
ªá ƙ
Ɯ
ƚ
ƓƎƔ ƛ
Ɲ
ƞ
ì áƙƜ
ƚ
ŇƓŇƓƔŇƓÏâ ƎŇƓ¡ ƛ
Ɲ
ƞ
ì áƙƜ
ƚ
ŇƔŇƓƔŇƔÏâ ƎŇƔ¡ ƛ
Ɲ
ƞ
(2)Ɣáã × ì ňƖƖì ňƗƗì ƓňƖƖâ ƔňƖƗà ƏƖ ä (3)
Ɣáã × ì ňƗƖì ňƗƗì ƓňƗƖâ ƔňƗƗà ƏƗä (4)
(1) ª , ,
, Ɣ Ɩ
(flux) . Choi Merkle
[1] Weiss Smith
(preconditioning matrix)[2] ġ .
(6) Ħ (parameter) Ɓ
. ®Ɛ (reference velocity) (local preconditioning)
(limiter)
. (7) ¤ 0.5 .
ġ á
Ɯ
ƚ
Ħ¡ĦƓĦƔà Î ŇƓ ŇƔ ŇÞŇ× ×Ň à ćà ćà ćƎŇƓŇƔà ć
¡ß
Ɲ
ƞ
(5)zĦ á
Þ
ć®ÎƐÏà ćƁÎÏß
â ć«Î (6)®Ɛá À¼Áƙ
Ɯ
ƚ
Ɓì À´ËÞç®çß줮ƁƓƒì®ƔƇƑì
ö
ććŇ Ɲƞ
(7)®ƁƓƒá À¼ÁÞ®tì Ɓtß ƎƊ ƎƐ
Ɗâ ƎƐßîÏ
®ƁƓƒ
®ƔƇƑ ĢƂ łîŇĢƂ
Ɖ à ƕ ¬ Ɖ ƕ (©)
() . łƊ łƒ
Sutherland łƒá ŇķƔƁłƍÞƉîƕß
[6]. Ɖ à ƕ WD+ [5]
[5,6,7] .
2.2
Roe FDS . ,
[1]. ,
3 MUSCL (Monotone Upstream-
centered Schemes for Conservation Laws) ,
[8].
DADI (Diagonalized Alternate Directional Implicit) [3,9-11] .
Navier-Stokes Ɖ à ƕ
loosely coupled algorithm .
[7] .
3.
3.1
,
[3,7,8].
Ģƒ á Ɣß â ƐÞƔß
¥ (8)
ƐÞƔß á Ï
Þ
ć©ÅƊłƊ â ć©Åƒ
łƒ
ß
ćŇĹćö
ćÞńƖÏ£â ńƗÏß (9)(8)
(9) Jacobian .
DADI .
(8) .
max-CFL .
3.2 Min-CFL
.
max-CFL . 1
.
, 1
.
(10) min-CFL
[12,13].
Ģƒ á À´Ëƙ
Ɯ
ƚ
ćŁƖ
¥ĢƖ
ì ćŁƗ
¥ĢƗ ƛ
Ɲ
ƞ
(10)ŁƖ Ɨ (11)
(10) (8)
.
.
3.3 Min-CFL/max-VNN
. min-CFL
, min-CFL
.
®ƔƇƑ
[1]. Choi Merkle
[12,13] .
[12,13] (12) .
®ƔƇƑÏ á À´Ëƙ
Ɯ
ƚ
ćÞķ à Îß â ƁÏîƓÏ ķÞķ à Îß
ì ćÞĸ à Îß â ƁÏîƔÏ ĸÞĸ à Îß ƛ
Ɲ
ƞ
ƁÏ (12)ķ á ć¯§§
¥ ć«ƃĢƖ
¬ ì ĸ á ć¯§§
¥ ć¬ Îć«ƃĢƗ
Î (13)
«á ć ĢƗ
ĢƖ ì «ƃĢƖá ćŃ
ƓĢƖ ì «ƃĢƗá ćŃ
ƔĢƗ (14)
(12) « 1 ,
. (13)
¯§§ Von Neumann Number .
(14)
Ń . ¬ Î[1], «[13], «Ï[14]
Î . ¯§§
Î××
. CFL VNN
. (10) (12)
, «
CFL
VNN . (12)
min-CFL
(12) min-CFL/max-VNN .
3.4 Switching
10% bump min-CFL/max-VNN
Fig. 2 Pressure isolines of the original definition when M=0.1 : AR=5, 50, 500, 5000 (from the top).
(a) AR = 5
Fig. 1 (b) AR = 500
Iteration L2normof∆Q
1000 2000 3000
10-14 10-12 10-10 10-8 10-6 10-4 10-2
Original∆t, AR = 5 Original∆t, AR = 50 Original∆t, AR = 500 Original∆t, AR = 5000
∞
AR = 5000
AR = 500
AR = 50
AR = 5
Fig. 3 Convergence histories of the original definition at AR from 5 to 5000, M=0.1
[1,13] ,
[12]
min-CFL/max-VNN switching . ,
min-CFL/max-VNN .
(12) switching
10%
.
4.
10% bump
(original), min-CFL (Min-CFL),
switching (present)
. ,
2 NACA 4412
(original), min-CFL/max-VNN (Buelow), switching (present)
.
, bump
.
4.1 10% bump
10% bumpFig. 1 . 65×71 .
CFL 5 bump
4 . 0.1
0.001 « 5, 50, 500, 5000
. «
« 5000 bump 2200
. 0.001 0.1
0.1 .
Fig. 2 «
,
. «
5 5000
. ,
.
Fig. 3 5 « ,
Fig. 5 Convergence histories of the switching definition at AR from 5 to 5000, M=0.1
(a) AR = 10
Fig. 6 (b) AR = 105 Iteration
L2normof∆Q
1000 2000 3000
10-14 10-12 10-10 10-8 10-6 10-4 10-2 100
Min-CFL∆t, AR = 5 Min-CFL∆t, AR = 50 Min-CFL∆t, AR = 500 Min-CFL∆t, AR = 5000 M∞= 0.1
AR = 5000
AR = 500 50
5
Fig. 4 Convergence histories of the min-CFL definition at AR from 5 to 5000, M=0.1
min-CFL , switching
. « 5 50 Fig. 3
Fig. 4 min-CFL
. « 500 5000
min-CFL . Fig. 3
Fig. 4 «
min-CFL «
[1,12,13].
switching
« min-CFL
Fig. 3 Fig. 4 Fig. 5 .
4.2
Fig. 6 .
97×65 Fig. 7
Ɩ á Ï .
CFL 5 VNN 100 . 4
0.1, 4×103, 4×104, 4×105, 4×106, 4×107 .
«á10, 100, 1000, 104, 105 ,
min-CFL/max-VNN , switching .
Fig. 7
« 4×104(100), 4×105(100), 4×105(1000) .
.
Fig. 8 « 4×103(10),
4×104(10), 4×105(10)
min-CFL/max-VNN .
«
.
Fig. 9 « 4×104(100),
Fig. 7 Accuracy of the original definition near the wall, Re = 4×104(100), 4×105(100), 4×105(1000)
Fig. 8 Convergence comparisons between the original and min- CFL /max-VNN definitions, Re = 4×103(10), 4×104(10), 4×105(10)
Fig. 10 Accuracy of the original definition near the wall, Re = 4×106(105), 4×107(105).
Fig. 9 Convergence comparisons among the original, min-CFL/
max-VNN, present definitions, Re = 4×104(100), 4×105(100)
4×105(100) min-CFL/max-VNN switching
. «
switching
.
«
.
4.3
k-w WD+
Fig. 10 .
.
.
Fig. 11 « 4×106(104),
4×107(104) min-CFL/max-VNN .
« min-CFL/max-VNN
.
Fig. 11 Convergence comparisons between the min-CFL/max -VNN and present definitions, Re = 4×106(104), 4×107(104).
Fig. 12 Convergence comparisons between the min-CFL/max- VNN and present definitions, Re = 4×106(105), 4×107(105).
switching min-CFL/max-VNN
.
Fig. 12 « 4×106(105),
4×107(105) min-CFL/max-VNN
. «
min-CFL/max-VNN .
5.
.
, min-CFL ,
min-CFL/max-VNN , switching
,
. switching
« , «
min-CFL/max-VNN .
«
. NACA 4412
.
[1] 1993, Choi, Y.-H. and Merkle, C.L., "The Application of Preconditioning in Viscous Flows," Journal of Computational Physics, Vol.105, No.1, pp.207-223.
[2] 1995, Weiss, J.M. and Smith, W.A., "Preconditioning Applied to Variable and Constant Density Flows," AIAA Journal, Vol.33, No.11, pp.2050-2057.
[3] 2006, Park, S.H., Lee, J.E. and Kwon, J.H., "Preconditioning HLLE Method for Flows at All Mach Numbers," AIAA Journal, Vol.44, No.11, pp.2645-2653.
[4] 2004, Yildrim, B. and Cinnella, P., "On the Validation of a Global Preconditioner for the Euler Equations," AIAA Paper 2004-740.
[5] 2002, Thivet, F., "Lessons Learned from RANS Simulations of Shock-Wave/Boundary-Layer Interactions," AIAA Paper 2002-0583.
[6] 1988, Wilcox, D.C., "Reassessment of the Scale-Determining Equation for Advanced Trubulence Models," AIAA Journal, Vol.26, No.11, pp.1299-13130.
[7] 2004, Park, S.H. and Kwon, J.H., "Implementation of k-w Turbulence Models in an Implicit Multigrid Method," AIAA Journal, Vol.42, No.7, pp.1348-1357.
[8] 1986, Anderson, W.K., Tomas, J.L. and van Leer, B.,
"Comparison of Finite Volume Flux Vector Splittings for the Euler Equations," AIAA Journal, Vol.24, No.9, pp.1453-1460.
[9] 1981, Pulliam, T.H. and Chaussee, D.S., "A Diagonal Form of and Implicit Approximate-Facotrization Algorithm," J.
Comp. Phys., Vol.39, pp.347-363.
Euler Equations," AIAA J., Vol.2, No.11, pp.1565-1571.
[11] 1997, Jespersen, D., Pulliam, T. and Burning, P., "Recent Enhancements to OVERFLOW," AIAA Paper 97-0644.
[12] 2001, Buelow, P.E.O., Venkateswaran, S. and Merkle, C.L.,
Schemes," Computers & Fluids, Vol.30, pp.961-988.
[13] 1994, Buelow P.E.O., Venkateswaran, S., Merkle, C.L., "The Effect of Grid Aspect Ratio on Convergence," AIAA Journal, Vol.32, No.12, pp.2401-2408.