Professional Documents
Culture Documents
HYDRODYNAMIC MODEL
ABSTRACT: A mathematical model of three-dimensional (3D) free surface flows has been applied to simulate
the curved channel flows and mass transport. In the horizontal plane, a channel-fitted curvilinear coordinate
system is used, whereas in the vertical plane, the a-transformation is adopted to track the free surface and
variable bed topography. To reduce the numerical diffusion, the second-order upwind scheme of Roe is incor-
porated to discretize the convection terms. The standard k-e model has been modified to take account of some
anisotropic effects appearing in shallow curved channels, i.e., streamline curvature and the damping effects of
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 08/18/18. Copyright ASCE. For personal use only; all rights reserved.
free surface and solid walls. The governing equations are solved in a collocated grid system by a fractional
three-step implicit algorithm. Two test cases from curved flumes have been studied: (1) A 270 channel bend
0
with a sloped outer bank; and (2) a meandering channel with mass transport. The results are compared with the
available data, which shows generally good agreement.
=
where U, V contravariant velocities that are perpendicular
to the 11, ~-curvilinear coordinates, respectively.
where u, v, w = velocity components in the horizontal x-, yo, By the chain rule, the governing equations can be expressed
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 08/18/18. Copyright ASCE. For personal use only; all rights reserved.
and vertical z-directions, respectively; g = gravity force; V.h = in a nonorthogonal curvilinear coordinate system. The Carte-
VIII + v, V. = V tv + v (where V'h, V tv = the turbulent viscosities
V
sian velocity components (u, v, w) are retained as dependent
which reflect the anisotropy effects in the horizontal and ver- variables in the momentum equations. The conservative fonn
tical direction, respectively, v = fluid kinematic viscosity); P of the equations are as follows:
= water kinetic pressure (pressure divided by fluid density); Continuity equation:
and t = time.
The free surface is denoted by H + Zb (where H = water aHJ + aHU + aHV + aHJW = 0
depth and Zb = bed elevation). The total pressure P at any at a~ a11 aa
(8)
position (x, y, z) can be subdivided into (1) the hydrostatic
component, i.e., Po + g(H + Zb - z), where Po = the free Momentum equations:
surface kinetic pressure; and (2) the residual hydrodynamic
component p, i.e., p = P - Po - g(H + Zb - z).
The momentum equations of (2)-(4) can be written in a
general fonn of a balance of convection, diffusion, and source
tenn S: as follows:
a" a SURFACE
a- 1 B02'2'OH
z
~(x,y)
lation from the values at the interior nodes. H + Zb at the Case 1 is a 270 open channel bend with a 2: 1 sloped outer
outlet is given. bank (Hicks 1985); Case 2 is a meandering channel (Chang
In the vicinity of the solid walls and bed, the variations of 1971).
flow properties are steep. To avoid the need for detailed cal-
culations in the near wall region of low-turbulent Reynolds Case 1
number, the wall-function method is employed to simulate the
wall effect and estimate the values of k, E at the near wall Hicks (1985) measured the distribution of velocities and tur-
nodes (Rodi 1984). The wall-function method, which allows bulent fluctuations in a hydraulic smooth, 2700 open channel
the use of a fairly coarse grid in the near wall region, is an bend with a laser Doppler anemometer. The purpose was to
economical way of turbulent flow modeling. Here the refined examine the developing flow near a sloped bank in a smooth
690 I JOURNAL OF HYDRAULIC ENGINEERING I JULY 1998
'~ J
probably extends to the end of the downstream straight reach.
The model simulated the same straight section (2.25B) as the
experiment. The other parameters are n = 0.85 and ~t = 0.6
s. After 1,000 time steps, the simulation terminates when the
~I.
Ll" 865.8
nondimensional overall mass residual for the 3D continuity
equation (8) is less than 6.0 X 10-4 and that for the depth- B=1066.8
averaged form (14) is less than 4.4 X 10- 3 • It is obvious that I· -I
the a-transformation is most suitable in the present case due (b)
to its capacity of fitting both the free surface and the sloped
outer bank. The results of velocity vector fields, distribution FIG. 2. Flume Layout and Dimensions (Unit: mm): (a) Plan
of longitudinal and lateral velocities, and turbulence intensities View; (b) Channel Section with 2:1 Side Slope
.
(a)
i •i •
o 0.2 mil
\ ttl
FIG. 3. Velocity Vector Fields at: (a) Surface; (b) Cross Section 3
o 1
Sec. 1
ru
Sec.3
~~.,J] .J .J
Sec.4
~c]
FIG. 4. Comparlaon of Longitudinal Velocity Component (uIUo) Distribution
0 0.4
~
sij
1 'i\ \\
sec~
~ **::+;: :s;: 'h:
S8CjJ
~ .*~ 'b '5b:
s{j
~ '4;: 'b: s: FIG. 5. Comparlaon of Lateral Velocity Component (vlUo) Distribution
are shown in Figs. 3-8 at the four typical locations along the nomenon and is in satisfactory agreement with the experi-
channel and compared with the data of Hicks (1985) and Hicks mental data for the longitudinal and lateral velocities as shown,
et aI. (1990). Because the experiments were taken perpendic- respectively, in Figs. 4 and 5. A notable disparity appears at
ular to the slope on the outer bank, while the vertical a-co- Section 1 at the entrance of the bend. The data of Hicks (1985)
ordinate is used in modeling, some discrepancies are expected indicated that the secondary current starts upstream of the
along the sloped bank. curve, whereas the model shows that the secondary current
Figs. 3(a) and 3(b) give the surface velocity vector field and starts to develop just at the entrance of the curve; both the
a typical secondary current at Section 3, respectively. The sim- model and the experimental data show a higher initial strength
ulation shows that the velocity increases continuously along near the inner bank. Fig. 5 also indicates that the simulated
the sloped outer bank in the bend. The secondary currents that secondary current initially developed near the bottom (Section
move toward the inner bank at the bottom and toward the outer 1) and moved toward the inner bank because of the super-
bank at the top, bring fluid with relatively low longitudinal elevation extended upstream, and the net lateral (depth-aver-
momentum toward the center portion of the channel causing aged) flow is relatively small. This is consistent with other
a lower velocity at the surface and the maximum velocity to computations [e.g., Leschziner and Rodi (1979), Demuren and
lie below the water surface. The model reproduces this phe- Rodi (1986) with two-equation turbulence models] and some
692/ JOURNAL OF HYDRAULIC ENGINEERING / JULY 1998
~_'*S;:_'*:g;:_~*:+;: d
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 08/18/18. Copyright ASCE. For personal use only; all rights reserved.
0 0.2
Sec. 1
~t 1\& IT [I
Sec.2
rn
~C 11. 1\. 1\.. lIi
Sec.3
~l.. FIG. 7.
I\,~ [\M. 1\.
Comparison of Turbulence Intensity y;;'i/Uo Distribution
[j
experiments [e.g., Chang (1971); Steffler (1984)]. The mea- Case 2
surement, however, illustrated that the secondary motion at the
entrance is strong near the free surface and away from the A series of experiments in meander channels was conducted
inner bank, which could indicate that there was some distur- by Chang (1971) in which both the flow and a neutrally buoy-
bance upstream of the entrance in the experiments. Despite the ant pollutant transport were measured. The channels with hy-
aforementioned discrepancy, the model predicted the overall draulic smooth bed and walls and rectangular cross sections
behavior of the velocity fields once the spiral motion devel- had unifonn 90° bends in alternating directions interconnected
oped, Le., at Sections 2-4. by straight reaches, One of the single meander channels is
Fig. 6 shows the velocity distributions at Section 3 obtained illustrated in Fig. 9 with the water depth h = 0.115 m and bulk
by the standard k-e model and the proposed modification. The velocity Uo = 0.366 mis, The measurements were carried out
standard k-e model underpredicted the pressure-driven second- along the second bend of the channel from Sections 1 to 13.
ary currents, which results in the discrepancy in the longitu- Demuren and Rodi (1986) employed a 3D mathematical
dinal velocity distribution compared with the measured data. model with cylindrical coordinates to calculate this meander-
Fig~ and ~ompare the distribution of turbulence inten- ing flow. Some simplifications were made, e.g., the "rigid lid"
sities U,2 and V,2, respectively, with the measurements. The approximation was used to deal with the free surface problem;
agreement is generally very close, except in the vicinity of the all diffusion tenns in longitudinal direction were neglected.
bed where-!Qe simulations are not able to capture the peak Comparison with the measurements showed generally good
values of U,2 due in part to insufficient grid density in that agreement, but the 3D concentration field which is more sen-
area. Another discrepancy appears at Sections 3 and 4 in the sitive to turbulent diffusivity was not reported, and the cylin-
sloped bank where the distributions of U,2 have been smoothed drical coordinates have limited applicability.
by the simulations but the depth-averaged values are similar As indicated in Fig. 9, the straight exit section after the
to the experiment, as illustrated in Fig, 7, second bend was relatively short in the experiments (-0.918).
JOURNAL OF HYDRAULIC ENGINEERING / JULY 1998/693
0 0.2
Sec. 1
~ [j 11 [) [jj
Sec.2
~\ [J a 1\ [j
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 08/18/18. Copyright ASCE. For personal use only; all rights reserved.
sec~
~ 1\ 1\ 1\
~
"'I~'C
Sec~
1\ 1\
FIG. 8. Comparl8on of Turbulencelnten81ty Y;:;,Uo DI8tribution
1\
To test the sensitivity of the exit length, a longer straight sec- almost the same (the relative difference is within 1%) because
tion (-2.5B) is used in simulations. Initially, a same grid den- of the parabolic nature of the flow at the exit in this case.
sity was applied to these two different exit section cases, i.e., Comparison of the longitudinal and lateral velocity profiles
the meshes of 132 X 26 X 12 (for the longer section, denoted at the B1 grid with the measurements at selected locations are
as AI) and 122 X 26 X 12 (for the shorter section, denoted shown in Figs. 10 and 11, respectively. The results at the Al
as A2) in ~, 11, a-directions, respectively. The sizes of the grid are also presented in comparison of grid independence of
control volumes in the physical plane are between 0.26 and the solutions. It shows that the solutions are quite close. The
0.36 m and 0.065 and 0.105 m in longitudinal and lateral di- maximum relative difference of the simulated velocities at Al
rections, respectively. Uniform distribution of a (Aa = 0.0833) and Bl grids is about 4.5%. The experiments showed that
is set in vertical. To check grid independence of the solutions, there exists another smaller counterrotating eddy near the sur-
a fine grid (denoted as Bl) is adopted for the longer exit sec- face at the concave bank of the second bend (Chang 1971), at
tion with the 166 X 44 X 16 nonuniformly distributed seg- Sections 7 and 13 in Fig. 11. It also is notable that there exists
ments in ~, 11, a-directions, respectively. The sizes of the con- a small counterrotating eddy near the surface at the convex
trol volumes in the physical plane are between 0.21 and 0.32 bank in the single bend experiments of Hicks (1985), as shown
m and 0.021 and 0.065 m in longitudinal and lateral directions, at Section 3 in Fig. 5. This phenomenon may be due to a
respectively. A uniform distribution of a (Aa = 0.0625) is hydrodynamic instability similar to the Gortler vortices present
used. In the lateral 11-direction, grids are concentrated near the in boundary layers along concave walls, as stated by De
side wall regions for all the cases. The other parameters are Vriend (1981). The present model could only predict the main
n = 0.85 and At = 0.5 s. After 1,000-1,400 time steps, the helical motion of secondary currents even with the fine B 1
solutions are regarded as steady state when the mass residual grid. The possible reasons are that the grid should be further
for the 3D continuity equation (8) is less than 7.0 X 10-4 and refined in the longitudinal direction and the complex higher-
that for the depth-averaged form (10) is less than 1.1 X 10- 4 • order turbulence models would be used. Secondary currents,
The computations are performed at A 1 and A2 grids to study especially, are difficult to measure precisely, since they are a
the influence of the length of the straight exit section. Simu- fraction of the main flow velocities. Though some discrepancy
lations show that the results of the velocity distributions are exists, the simulations at the B 1 grid represent the general
894/ JOURNAL OF HYDRAULIC ENGINEERING / JULY 1998
0.5
behavior of the secondary currents and the resultant main icantly in the meandering channel. This 3D effect cannot be
flows with reasonable accuracy. Therefore, the flow field sim- represented properly by two-dimensional models (Demuren
ulated at the B1 grid is used to calculate the mass transport in and Rodi 1986; Ye and McCorquodale 1997). As the experi-
the meandering channel. ments indicated (Chang 1971), the mixing rate could be up to
Performance of the standard k-e model in the A1 grid is also an order of eight times higher in a meandering channel com-
carried out and the results at Sections 3 and 7 are presented pared to an equivalent straight channel. The prediction of the
in Figs. lO(a), lO(c), II(a), and II(c). The standard k-e model pollutant concentration distributions is compared with the
underpredicts the secondary currents and consequently gen- measurements of Chang (1971). The case considered was a
erates relatively larger error in the longitudinal velocity distri- dye injection point at the middepth of the centerline and the
bution. entrance to the first bend, as shown in Fig. 9. The development
Because of the centrifugal force effect, the helical motion of the ensuing concentration field was shown and measured
carries the flow near free surface toward the concave bank and for the second bend.
that near the bottom toward the convex bank. This process Considering the initial dilution of the dye injection into the
increases mixing and influences lateral mass transport signif- flume, the 3D model of mass transport has been applied to
JOURNAL OF HYDRAULIC ENGINEERING / JULY 1998/895
0.0 '-----'---'-->---'-_-I-_--l.-_---l._-L-L_-L_...J
Sec. 3
Sec.9
1.0 ,-----,c---,---,--,------r-.,--,...--,---r--n
0.6
Sec. 3
::WJJ } /) 1{ (~\
Sec.11
J \ I
0.0
Sec.9
CONCLUSIONS
Sec.11
A 3D hydrodynamic model of free surface turbulent flows
~O[/
:.:
/ 7j
Ii-Ill
JI I I
has been developed to simulate currents and mass transport in
curved channels. The proposed model accounts for the major
3D flow features and their consequent effects on mass trans-
port. The model has the capability to handle an unprescribed
free surface and nonrectangular, nonprismatic practical chan-
nel geometries owing to the applicability of a-transformation.
Algebraic formulations for the horizontal and vertical eddy
Sec. 13 viscosities have been used to take account of the streamline
FIG. 13. Measured Concentration (ClCo) Profiles at Various curvature and damping effects of the free surface and solid
Sections boundaries; this, partly, overcomes the drawback of the stan-
dard isotropic k-e model. This modification proved useful in
obtain the concentration distribution based on the simulated the present test cases of shallow curved open channel flows;
velocity field. Fig. 12 shows the results of the depth-averaged however, further study is needed to define the general appli-
concentration distribution, whereas the 3D concentration pro- cability of this conclusion.
files of both the measured and calculated fields are presented The model has been tested by applying it to two typical
in Figs. 13 and 14, respectively, where Co = cross-channel curved open channel flows: (1) A single 2700 channel bend
averaged concentration. The trend of the nonuniform distri- with a sloped outer bank; and (2) a m( andering channel with
bution in vertical is related to the direction of the spiral mo- pollutant transport. The model results have been compared
tion. The agreement with the experimental data is generally with the available data, and the agreement is generally good.
good with the following two exceptions: The simulations show that the strength of secondary currents
are much stronger in a single bend than that in meanders due
to the opposite spiral motion generated by the alternate be.Js
1. There was difficulty in simulating correctly the actual of the meandering channels. The lateral mass transport due to
situation near the discharge port, which contributes some secondary currents in curved channel can be considerable. Fur-
of the discrepancy; this problem also appeared in the ther studies should be carried out for hydraulically rough beds,
work of Demuren and Rodi (1986). different width-depth ratio channels, and natural rivers.
2. The experiments showed evidence of a secondary vortex
near the surface at the outer bank of the second bend, as APPENDIX!.
illustrated in Figs. 13(b) and 13(c), while the simulations
only represent the effect of the main secondary current In (9), the metric tensor components lX, 13, 'Y, etc. are ex-
patterns on the mass transport, as indicated in Fig. 14. pressed as
896/ JOURNAL OF HYDRAULIC ENGINEERING / JULY 1998
at
(Hut+
= - llH'
+1 (dP' dP')
- Y - - Y
d~" dTJ (
q33 =>"(Y(z" - y.,zi + >"(XM - X.,z()2 + 12;
z(~, TJ, cr, t) = H + Zb - crH (24a-i) (30a)
The diffusion coefficients r~, r ol>v and source terms Sol> of 1 213
+1 (dP'
J (Hvt+ (Hvt+ dP')
-
u, v, ware = - llH' - x - - x
at dTJ ( d~"
<I> =u: f =V. h =Voj>h th + v, f ol>v =v., =v", + v (25a)
dp'
- II dcr (XM - X.,z() (30b)
S. = -H {y., .i- [g(H + Zb) + p] - Y( .!- [g(H + + P]}
Downloaded from ascelibrary.org by Universidad Nacional De Ingenieria on 08/18/18. Copyright ASCE. For personal use only; all rights reserved.
Zb)
o~ dTJ
(HW)"+I - (Hwt+ 213 dp'
-"--~--"---'-- = II - (30e)
at dcr
H'+I - H'+213 o(HUt+ 213 d(HV)n+213 d(HW)"+213
J + + + 1 --'----''----
at O~ dTJ dcr
<I> = v: f oj>h = V.h = Vth + v, fol>v = v.v = v", + v (200)
dHU" dHV" dHW")
+ll ( - - + - - + J - - =0
S. = -H {X( d~ [g(H + Zb) + p] - x., :~ [g(H + Zb) + P]} O~ OTJ dcr (30d)
dP
H'+1 = H'+213 (30e)
- dcr (x M - x.,z() (26b) It is easy to prove that the working equations (29) and (30)
of the present fractional step algorithm are consistent to the
original governing equation set (8)-(10) and (25)-(27) by fol-
lowing the procedure of Yanenko (1971), Le., combining (28)
(when <I> = u), (29a) and (30a) results in (9) (when <I> = u),
Sw =1dP
-
dcr
(27b) etc. The velocity correction (u;, U;), water depth correction
(H'), and the velocity correction (u7, Un, hydrodynamic pres-
sure correction (P') equations can be obtained directly from
APPENDIX II. WORKING EQUATIONS (29) of Step 2 and (30) of Step 3, respectively. This is one of
the main advantages of the present algorithm over the SIM-
The working equations of the fractional algorithm are as PLEC algorithm of Van Doormaal et al. (1984).
follows.
ACKNOWLEDGMENTS
Step 1. Convection-diffusion process
The writers appreciate Dr. F. E. Hicks, University of Alberta, for pro-
I13 viding the original experimental data. This research was supported by the
1 (H<I>r+ - (H<I»" + [(Hu)"<I>n+ll3] + [(Hv)"<I>n+ll3]
at (., Great Lakes Institute for Environmental Research, University of Windsor,
and the Natural Science and Engineering Research Council of Canada.
+ [¥ ('Y<I>~+113 - 13<1>~ + i; <1>:) 1 Chang, H. H. (1992). Fluvial processes in river engineering. Krieger
Publishing Co., Malabar, Fla.
Chang, Y. C. (1971). "Lateral mixing in meandering channels," PhD
thesis, Univ. of Iowa, Iowa City, Iowa.
Cheng, G. C., and Farokhi, S. (1992). "On turbulent flows dominated by
+ [~(>..q
1 31 <1>"( + >..q32.,
<1>" + H' <l>n+II3)] q33
a a
+ S ol> (28)
curvature effects." J. Fluid Engrg., ASME, 114(1), 52-57.
Cokljat, D., and Younis, B. A. (1995). "Second-order closure study of
open-channel flows." J. Hydr. Engrg., ASCE, 121(2),94-105.
where 8"+113 = H n ; and the expressions of Sol> for momentum Demuren, A. O. (1991). "Calculation of turbulence-driven secondary mo-
equations can be found in (25)-(27). tion in ducts with arbitrary cross section." AIAA J., 29(4), 531-537.
Demuren, A. O. (1993). "A numerical model for flow in meandering
Step 2. Hydrostatic propagation process channels with natural bed topography." Water Resour. Res., 29(4),
1269-1277.
(Hut+ 213 - (Hut+l 13 _ _ gH' (dH' _ dH' ) Demuren. A. 0., and Rodi, W. (1986). "Calculation of flow and pollutant
at - II J o~ Y., dTJ Y(
(29a) dispersion in meandering channels." J. Fluid Mech., 172(11),65-92.
De Vriend, H. J. (1981). "Velocity redistribution in curved rectangle
l13 channels." J. Fluid Mech., 107(6),423-439.
(Hvt+
213
- (Hvt+ = -ll gHn (OH' x _ dH' x) (29b)
Dou, G. (1986). "General laws of turbulent flows." Proc., 3rd Int. Symp.
at 1 dTJ ( d~" on River Sedimentation, 47-57.
Fisher, H. B., List, E. J., Koh, R. C. Y., Imberger, J., and Brooks, N. H.
H'+213 - H'+1I3 d(HfJ)n+113 d(HV)"+ll3 (1979). Mixing in inlands and coastal waters. Academic Press, New
1 + + --'----''--- York, N.Y.
at d~ dTJ Galmes, J. M., and Lakshminarayana, B. (1984). "Turbulence modeling
for three-dimensional shear flows over curved rotating bodies." AIAA
dHO'
+ II ( - - + dHV')
-- =0 J., 22(10), 1420-1428.
d~ dTJ (29c) Hicks, F. E. (1985). "Shear and velocity near a sloped bank in a curved
channel," MS thesis, Univ. of Alberta, Edmonton, Alta., Canada.
Hicks, F. E., Jin, Y. C., and Steffier, P. M. (1990). "Flow near sloped
(29d) bank in curved channel." J. Hydr. Engrg., ASCE, 116(1), 55-70.
Leschziner, A., and Rodi, W. (1981). "Calculation of annular and twin Yen, B. C. (1965). "Characteristics of subcritical flow in a meandering
parallel jets using various discretisation schemes and turbulence-model channel." Rep., Inst. of Hydr. Res., Univ. of Iowa, Iowa City, Iowa.
variations." J. Fluids Engrg., ASME, 103(2), 352-360.
Majumdar, M. (1988). "Role of underrelaxation in momentum interpo- APPENDIX IV. NOTATION
lation for calculation of flow with non-staggered grids." Numer. Heat
Transfer, 13(1), 125-132. The following symbols are used in this paper:
Naot, D., and Rodi, W. (1982). "Numerical simulation of secondary cur-
rents in a channel flow." J. Hydr. Div., ASCE, 108(8),948-968. C = neutral concentration;
Odgaard, A. J. (1986). "Meander flow model. I: Development." J. Hydr. g = gravitational constant;
Engrg., ASCE, 112(12), 1117 -1136. H = water depth;
Patankar, S. V. (1980). Numerical heat transfer and fluid flow. Hemi- J = Jacobian;
sphere Publishing Corp., Bristol, Pa.
Rodi, W. (1984). Turbulence models and their application in hydraulics-
P = water kinetic pressure;
A state of art review, 2nd Ed., Int. Assoc. for Hydr. Res., Delft, The q3h q32' q13' q23' a, 13, 'Y = metric tensor components;
Netherlands. t = time;
Roe, P. L. (1981). "Approximate Riemann solvers, parameter vectors, u, v, w = velocity components in x, y, z-direc-
and different schemes." J. Compo Phys., 43(2),357-372. tions;
Shimizu, Y., Yamaguchi, Y., and Itakura, T. (1990). "Three-dimensional U, V = contravariant velocities in ~, TJ-direc-
computation of flow and bed deformation." J. Hydr. Engrg., ASCE, tions;
116(9),1090-1107. W = velocity defined in a-coordinate;
Steffler, P. M. (1984). "Turbulent flow in a curved rectangular channel," x, y, Z = Cartesian coordinates;
PhD thesis, Univ. of Alberta, Edmonton, Alta., Canada. Zb = bed elevation;
Stelling, G., and Van Kester, J. A. Th. M. (1994). "On the approximation
of horizontal gradients in SIGMA co-ordinates for bathymetry with
v = kinematic viscosity;
steep bottom slopes." Int. J. Numer. Methods in Fluids, 18(10), 915- Vth' V,. = turbulent viscosities;
~, TJ, a = curvilinear coordinates;
935.
Tamamidis, P., and Assanis, D. N. (1993). "Evaluation of various high- n = implicitness coefficient; and
order-accuracy schemes with and without flux limiters." Int. J. Numer. ac = turbulent Schmidt number for C.
Methods Fluids, 16(10),931-948.
Van Doorrnaal, J. P., and Raithby, G. D. (1984). "Enhancements of the Superscripts
SIMPLE method for predicting incompressible fluid flows." Numer.
Heat Transfer, 7(2),147-163. n, n + I = time level;
Yanenko, N. N. (1971). The method offractional steps. Springer-Verlag n + 1/3, n + 213 = intermediate value between steps; and
KG, Berlin, Germany. ,, " = variation between steps.