Upload
others
View
13
Download
0
Embed Size (px)
Citation preview
Hindawi Publishing CorporationJournal of Applied MathematicsVolume 2013 Article ID 596218 10 pageshttpdxdoiorg1011552013596218
Research ArticleComparison Study on the Performances of Finite VolumeMethod and Finite Difference Method
Renwei Liu1 Dongjie Wang1 Xinyu Zhang1 Wang Li12 and Bo Yu1
1 National Engineering Laboratory for Pipeline Safety and Beijing Key Laboratory of Urban Oil and Gas Distribution TechnologyChina University of Petroleum Beijing 102249 China
2 PetroChina Southwest Pipeline Company Chengdu 610041 China
Correspondence should be addressed to Bo Yu yuboboxvip163com
Received 26 March 2013 Revised 9 June 2013 Accepted 9 June 2013
Academic Editor Shuyu Sun
Copyright copy 2013 Renwei Liu et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
Vorticity-stream function method and MAC algorithm are adopted to systemically compare the finite volume method (FVM)and finite difference method (FDM) in this paper Two typical problemsmdashlid-driven flow and natural convection flow in a squarecavitymdashare taken as examples to compare and analyze the calculation performances of FVM and FDMwith variant mesh densitiesdiscrete forms and treatments of boundary condition It is indicated that FVM is superior to FDM from the perspective of accuracystability of convection term robustness and calculation efficiency Particularly when the mesh is coarse and taken as 20 times 20 theresults of FDM suffer severe oscillation and even lose physical meaning
1 Introduction
In the numerical solution of flow and heat transfer problemsthe concepts of conservative and nonconservative equationswere firstly proposed in [1 2] in the 1980s It is noteworthythat from the perspective of differential unit the conservativeand nonconservative equations are equivalent which areboth the mathematical expression of physical conservationlaw Nevertheless the numerical calculation is implementedon the calculation unit of finite size for which the twoforms of equations are of different characteristics Practicalcalculation process shows that the influence of differencesbetween conservative and nonconservative forms on accu-racy stability and efficiency of numerical calculation issignificant
Conservative governing equation and the correspondingdiscrete form show some advantages for example in the con-trol volume with finite size only the conservative equationcan guarantee that the conservation law of the problem stud-ied is satisfied [3ndash5] The result obtained by the conservativeequation is of higher accuracy generally In the calculation offlow problem involving shock wave the obtained flow field isusually smooth and stable employing the conservation formof governing equation while using the non-conservation
equation might lead to unsatisfactory spatial oscillations inthe upstream and downstream regions of the shock wave[6ndash8] Moreover when the conservation equation is usedto a body-fitted coordinate system the conservativeness ofgoverning equation can still be satisfied [9 10] More relatedresearches and applications are presented in the literature[11ndash15] Hence the conservation property of discretizedequations is desirable in engineering calculation Based onthis it is of great significance to investigate the influencefrom conservation property on the calculation performancesand provide guidance for the better implementation of itThe present paper compares the calculation performances byanalyzing the finite volumemethodwhich is conservative andfinite difference method which is nonconservative
Some researchers made comparisons of the twomethodsPatankar [16] and West and Fuller [17] pointed out thatthe results obtained in the situation that grids numbersdecrease and mass conservation is rigorously controlled loseits physical meaning Leonard [18] made a comparison ofthe accuracy of truncation error of the convection term inFVM and FDM By theoretical derivation he indicated thatthe truncation error of FVM is smaller than that of FDMin second-order central difference and second-order upwinddifference Botte et al [19] made comparison of the accuracy
2 Journal of Applied Mathematics
mass conservation and computation time It was validated byexamples that FDM cannot ensure mass conservation whileFVM can Meanwhile the influence of boundary conditionson the accuracy of algorithm was taken into consideration
Previous researchers mainly focused on the comparisonof accuracy There are few systemic studies on the stabilityrobustness and computation efficiencyThis paper comparedthe calculation performances of FVM and FDM based onprevious research Comprehensive comparison of accuracystability of convection term robustness and calculationefficiency was performed The results obtained may enrichthe study on the calculation performances of conservativeand nonconservativemethods which are expected to providesome reference for the practical calculation
2 Physical Problem
Tomake the study possess general significance it is necessaryto analyze the calculation performances of the two methodsbased on variant problems Lid-driven flow and naturalconvection flow in a square cavity which are both typicalproblems in computational fluid mechanics and numericalheat transfer are solved in this paper to make a systemicanalysis on the subject
As Figure 1(a) shows in the lid-driven flow the lid moveshorizontally at a constant velocity 119906top while the other threeboundaries keep still The Reynolds number Re (Reynoldsnumber) adopted in this paper is 1000
Figure 1(b) shows the schematic of natural convectionflow in a square cavity The left and right boundaries are thefirst boundary conditions between which the temperatureon the left boundary is higher The other two boundaries areadiabatic ones In calculation Pr (Prandtl number) is takenas 071 and Ra (Rayleigh number) is taken as 106
3 Numerical Method
Considering that the calculation approach is influential tothe calculation in order to compare the difference of cal-culation performances between FVM and FDM this paperadopts both primitive variable method and nonprimitivevariable method to solve the governing equations For prim-itive variable method MAC algorithm is adopted and fornonprimitive variable method the generally used vorticity-stream function algorithm is adopted The conservationproperty generally refers to the convection term and thusonly the difference of convective term between conservativeand nonconservative forms is compared MAC method ischaracterized by the direct solving of velocity and pressureThe key point of this method is to obtain a nondivergentvelocity field in every time step or iteration step That is thevelocity field of every time layer or iteration layer shouldsatisfy the continuity equation and pressure and the velocityis decoupled by solving Poisson equation which is relevant topressure
31 Governing Equation In primitive variable method pres-sure and velocity are directly used as variables to solve the
Table 1 Parameters of equations for MAC of lid-driven flow
Name of variables Φ Γ120601
119878120601
119883 direction 1198801
Reminus120597119875
120597119883
119884 direction 1198811
Reminus120597119875
120597119884
Table 2 Parameters of equations for MAC of natural convectionflow
Name of variables Φ Γ120601
119878120601
119883 direction 119880 1 minus120597119875
120597119883
119884 direction 119881 1 RaΘPr
minus120597119875
120597119884
Temperature Θ1
Pr0
equations As to the incompressible fluid the dimensionlessunsteady-state conservative and nonconservative equationscan be written as respectively
120597Φ
120597120591+120597 (119880Φ)
120597119883+120597 (119881Φ)
120597119884= Γ120601(1205972Φ
1205971198832+1205972Φ
1205971198842) + 119878120601 (1)
120597Φ
120597120591+ 119880
120597Φ
120597119883+ 119881
120597Φ
120597119884= Γ120601(1205972Φ
1205971198832+1205972Φ
1205971198842) + 119878120601 (2)
where the values of Φ Γ120601 and 119878
120601for lid-driven flow and
natural convection flow are shown in Table 1 and Table 2respectively From (1) and (2) we can see that the convec-tion term of conservative equation is given in the form ofdivergence while that in nonconservative equation is notwhere the nondimensional excessive temperature can bedescribed as Θ = (119879 minus 119879
119897)(119879ℎminus 119879119897) and U V are defined as
nondimensional velocities in separate directions (X directionor Y direction) Governing equations of conservative andnonconservative equations of primitive variable method aregiven before In the calculation process MAC algorithmis adopted to obtain the result in steady state by unsteadycalculation over enough time The result of steady state isanalyzed as follows from the angle of primitive variablemethod Taking the vorticity-stream function method asexample the governing equations of conservative form andnonconservative form can be written as respectively
119886120601(120597 (119880Φ)
120597119883+120597 (119881Φ)
120597119884) = Γ120601(1205972Φ
1205971198832+1205972Φ
1205971198842) + 119878120601 (3)
119886120601(119880
120597Φ
120597119883+ 119881
120597Φ
120597119884) = Γ120601(1205972Φ
1205971198832+1205972Φ
1205971198842) + 119878120601 (4)
As to the vorticity-stream function method only the lid-driven flow is studied in this paper Values ofΦ 119886
120601 Γ120601 and 119878
120601
are shown in Table 3 where ldquo120596rdquo stands for vorticity and ldquo120595rdquostands for stream function
Journal of Applied Mathematics 3
l
l
y
x
119984top
(a) Lid-Driven Flow
l
l
y
x
gTh Tl
Adiabatic
Adiabatic
(b) Natural Convection Flow
Figure 1 Schematic of different physical problem in a square cavity
Table 3 Parameters of equations for vorticity-stream functionalgorithm of lid-driven flow
Name of variables Φ 119886120601
Γ120601
119878120601
Vorticity equation 120596 1 1
Re0
Stream function equation 120595 0 1 minus120596
where ldquo120596rdquo stands for vorticity and ldquo120595rdquo stands for stream function
32 Discretization of Equations Uniform mesh is adopted inthis paper In order to compare the calculation performancesof FVM and FDM with variant discretized schemes centraldifference and second-order upwind scheme are adopted Asto FVM taking e interface on X direction as example theexpressions can be written as
Φ+
119890= Φminus
119890=Φ119875+ Φ119864
2 (5)
Φ+
119890= 15Φ
119875minus 05Φ
119882 119880
119890ge 0 (6a)
Φminus
119890= 15Φ
119864minus 05Φ
119864119864 119880
119890lt 0 (6b)
where ldquoerdquo stands for the east interface between two nodes ldquoErdquostands for east node and ldquoWrdquo means west node and ldquo119864119864rdquo canbe defined as the east node of ldquoErdquo node
As to FDM taking node119875 as example the expressions canbe written as
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
=120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
=Φ119864minus 2Φ119875+ Φ119882
2Δ119883 (7)
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
=3Φ119875minus 4Φ119882+ Φ119882119882
2Δ119883 119880
119875ge 0 (8a)
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
=minus3Φ119875+ 4Φ119864minus Φ119864119864
2Δ119883 119880
119875lt 0 (8b)
xy
ΔY (i 1)
(i 2)
(i 3)
Figure 2 Schematic of the treatment for boundary conditions ofvorticity-stream function method
33 Treatment of the Boundary Condition of Vorticity Sincethe influence from boundary condition of vorticity on thesolution of equations is significant in this section the calcula-tion performances of FVM and FDMwith variant treatmentsof the boundary condition for vorticity are studied
Three boundary conditions of vorticity are comparedand studied In the boundary as shown in Figure 2 theirexpressions Thom equations [20] Woods equation [21] andJenson equation [22] as well as corresponding accuracy areshown in (9)ndash(11)
Thom equation
1205961198941=2 (1205951198942minus 1205951198941)
1205751198842
+ 119874 (Δ119884) (9)
Woods equation
1205961198941=3 (1205951198942minus 1205951198941)
Δ1198842
minus1
21205961198942+ 119874 (Δ119884
2) (10)
Jenson equation
1205961198941=minus71205951198941+ 81205951198942minus 1205951198943
2Δ1198842
+ 119874 (Δ1198842) (11)
4 Journal of Applied Mathematics
34 Deferred Correction When MAC is adopted momen-tum and energy equations are both solved in explicit schemewhile in vorticity-stream function method the convectionterm is treated in implicit scheme by deferred correctionwhich can guarantee the main-diagonal domination of thesolvingmatrix and thus guarantee the calculation stability Asto FVM taking Φ
119890as example the expressions of deferred
correction are as follows
when 119880119890ge 0 Φ
119890= Φ119875+ (Φ+
119890minus Φ119875)lowast
when 119880119890lt 0 Φ
119890= Φ119864+ (Φminus
119890minus Φ119864)lowast
(12)
where ldquolowastrdquo in these equations stands for the former calculationstep
As to FDM similarly to the treatment of Φ119890in FVM the
deferred correction for node 119875 can be written as the sum offirst-order upwind gradient and correction gradient which isshown as follows
when 119880119875ge 0 119880
119875
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
= 119880119875(Φ119875minus Φ119882
Δ119883)
+ [119880119875(120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119882
Δ119883)]
lowast
(13)
when 119880119875lt 0 119880
119875
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
= 119880119875(Φ119864minus Φ119875
Δ119883)
+ [119880119875(120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119864minus Φ119875
Δ119883)]
lowast
(14)
The first terms in the right side of (12)ndash(14) are all first-orderupwind term The second terms are incorporated into thesource term as the form of correction term The discretizedequation is solved by G-S solver with underrelaxation ofwhich the discrete expression is shown as (15) and thediscretized equations of FVM and FDM are shown in (16a)(16b) (17a) and (17b) respectively
119886119875Φ119875= 119886119864Φ119864+ 119886119882Φ119882+ 119886119873Φ119873+ 119886119878Φ119878+ 119887 (15)
119886119864=max (minus119880
119890 0)
Δ119883+
1
ReΔ1198832
119886119882=max (119880
119908 0)
Δ119883+
1
ReΔ1198832
119886119873=max (minus119881
119899 0)
Δ119884+
1
ReΔ1198842
119886119878=max (119881
119904 0)
Δ119884+
1
ReΔ1198842
119886119875=
max (119880119890 0)
Δ119883+max (minus119880
119908 0)
Δ119883+max (119881
119899 0)
Δ119884
+max (minus119881
119904 0)
Δ119884+
2
ReΔ1198832+
2
ReΔ1198842
(16a)
119887 = minusmax (119880119890 0) [
Φ+
119890minus Φ119875
Δ119883] +max (minus119880
119890 0) [
Φminus
119890minus Φ119864
Δ119883]
+max (119880119908 0)[
Φ+
119908minus Φ119882
Δ119883]minusmax (minus119880
119908 0)[
Φminus
119908minus Φ119875
Δ119883]
minusmax (119881119899 0) [
Φ+
119899minus Φ119875
Δ119884] +max (minus119881
119899 0) [
Φminus
119899minus Φ119873
Δ119884]
+max (119881119904 0) [
Φ+
119904minus Φ119878
Δ119884] minusmax (minus119881
119904 0) [
Φminus
119904minus Φ119875
Δ119884]
(16b)
119886119864=max (minus119880
119875 0)
Δ119883+
1
ReΔ1198832
119886119882=max (119880
119875 0)
Δ119883+
1
ReΔ1198832
119886119873=max (minus119881
119875 0)
Δ119884+
1
ReΔ1198842
119886119878=max (119881
119875 0)
Δ119884+
1
ReΔ1198842
119886119875=
max (minus119880119875 0)
Δ119883+max (119880
119875 0)
Δ119883+max (minus119881
119875 0)
Δ119884
+max (119881
119875 0)
Δ119884+
2
ReΔ1198832+
2
ReΔ1198842
(17a)
119887 = minusmax (119880119875 0) [
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119882
Δ119883]
+max (minus119880119875 0) [
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119864minus Φ119875
Δ119883]
minusmax (119881119875 0) [
120597Φ
120597119884
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119878
Δ119884]
+max (minus119881119875 0) [
120597Φ
120597119884
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119873minus Φ119875
Δ119883]
(17b)
where 119886119873 119886119875 119886119864 119886119882 and 119886
119878are coefficients in discretized
equations and 119887 is the source term in discretized equations
4 Results Comparison and Analysis
In this section the accuracy stability of convection termrobustness and calculation efficiency of variant algorithmsschemes and treatments of boundary condition which havebeen discussed previously are systemically compared andanalyzed based on the calculation results of lid-driven flowand natural convection flow in a square cavity
41 Accuracy Firstly the accuracy of FVM and FDM isevaluated by comparing benchmark solution with the results
Journal of Applied Mathematics 5
Table 4 Truncation error of FVM and FDMgoverning equations for vorticity-stream function algorithm of lid driven flow in a square cavity
Convection item Diffusion item
FVM (format)Φ119890
Δ119883minusΦ119908
Δ119883
1
Δ119883
120597Φ
120597119909
10038161003816100381610038161003816100381610038161003816119890
minus1
Δ119883
120597Φ
120597119909
10038161003816100381610038161003816100381610038161003816119908
FVM (TE) minusΔ1198832
8Φ101584010158401015840
119875minusΔ1198834
128Φ(5)
119875minus sdot sdot sdot minus
Δ1198832
24Φ(4)
119875minus 13
Δ1198834
5760Φ(6)
119875minus sdot sdot sdot
FDM (format) 120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816119875
1205972Φ
1205971198832
100381610038161003816100381610038161003816100381610038161003816119875
FDM (TE) minusΔ1198832
6Φ101584010158401015840
119875minusΔ1198834
120Φ(5)
119875minus sdot sdot sdot minus
Δ1198832
12Φ(4)
119875minusΔ1198834
360Φ(6)
119875minus sdot sdot sdot
obtained from variant mesh density discrete schemes andboundary conditions Through precision analysis FVM andFDM are respectively adopted with the Taylor expansioncommenced at midpoint (P node) and their respectivetruncation error (TE) is just as shown Table 4 [19] Thetruncation errors of FVM are obviously less than the FDMrsquos
411 Influence of Mesh Density In [4] it is indicated thatthe conservative property can affect the accuracy while thenonconservative FDM loses its conservative property in thesituation when the grid is too small Tomake further analysisthe calculation results obtained by FVM and FDM withvariant mesh densities are compared The results of thevelocity along the horizontal central axisV and velocity alongthe vertical central axis U obtained by FDM and FVM withcentral difference scheme of MAC algorithm are comparedand illustrated in Figure 3
From Figure 3 we can see that when the mesh densityis taken as 80 times 80 the results obtained by FVM and FDMboth agree well with the benchmark solutionWhen themeshdensity is reduced to 40 times 40 the results obtained by FVMare closer to the benchmark solution and if the mesh densityis further reduced to 20 times 20 the results obtained by FDMdiffer from the benchmark solution greatly in which FVMshows its advantage in accuracyThis phenomenon is in goodagreement with the theoretical analysis in [18] that is theconservative property of FVM better ensures the accuracyof calculation Therefore whether the conservative propertycan be realized is of much significance to the accuracy of theobtained result
412 Influence of Boundary Condition In the followingsection FVM and FDM with variant boundary conditionsare compared from the perspective of accuracy by vorticity-stream function method The results of the velocity alongthe horizontal central axis V and velocity along the verticalcentral axis U obtained by FDM and FVM with centraldifference scheme are compared and illustrated in Figure 4and the influence of boundary condition of vorticity is alsogiven From Figure 4 we can see that no matter whichtreatment of boundary condition is adopted the accuracy ofFVM is higher than that of FDM in vorticity-stream functionmethod
413 Influence of Discrete Scheme The selection of discreteforms is also influential to the accuracy Thus the resultsobtained with variant discrete forms are studied Taking thevorticity-stream function with boundary condition treatedby first-order Thom equation as example the results of thevelocity along the horizontal central axis V and velocityalong the vertical central axis U obtained by FDM and FVMwith variant discrete schemes are compared and illustratedin Figure 5 from which we can see that whether second-order central difference or second-order upwind scheme isadopted the accuracy of FVM is higher than that of FDM thecomparison between (a) and (b) indicates that second-orderupwind difference possesses higher accuracy than second-order central difference
414 Influence of Physical Problem Finally two methods invariant physical problems are compared from the perspectiveof accuracy based on the convection flow in a square cavityTaking the result obtained by MAC algorithm as exampleFigure 6 shows the comparison of results of the velocity alongthe horizontal central axis V and velocity along the verticalcentral axisU obtained by FDMandFVMTemperature fieldsobtained by FVM and FDM are compared with the grid-independent solution (obtained with grid of 256 times 256) asFigure 7 shows
From the comparison in Figure 6 and Figure 7 we cansee that based on the results obtained by MAC of naturalconvection flow in a square cavity the accuracy of FVM isgenerally superior to that of FDM Thus it is indicated bycomparison that no matter which mesh density scheme andtreatment of boundary condition is adopted FVM showsbetter accuracy than FDM
42 Stability of Convection Term Stability of mathematicalmeaning can only ensure that the oscillation of solution iscontrolled within a range but cannot eliminate the absenceof oscillation The amplification of oscillation may result indivergence Stability of convection term originates from var-ied difference schemes for convective itemand is independentof the introducing of rounding error in the calculation Thisstability is the key point to obtain a solution with physicalmeaning It is concluded that all the unstable schemes willresult in the oscillation of solution Stability condition of
6 Journal of Applied Mathematics
0
04
08
12
0
03
06
U V
X(Y)
0 02 04 06 08 1minus04
minus06
minus03
(a) 80 times 80
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) 40 times 40
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(c) 20 times 20
Figure 3 Velocity on the central axis of lid-driven flow obtained by MAC with variant mesh densities
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(a) Thom
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Woods
Figure 4 Velocity along the central axis obtained by second-order central difference scheme with two treatments of boundary conditionwhen the mesh density is 40 times 40
Journal of Applied Mathematics 7
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12U
minus04
0
03
06
V
minus06
minus03
(a) Second-order central difference scheme
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Second-order upwind difference scheme
Figure 5 Velocity along the central axis obtained by vorticity-stream function method for lid-driven flow with different schemes when themesh density is 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(a) 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(b) 80 times 80
Figure 6 Velocity along the central axis of convection flow in a square cavity with variant gird densities
40 times 40
Grid-independent
(a) 40 times 40
80 times 80
Grid-independent
(b) 80 times 80
Figure 7 Comparison of temperature contours of natural convection flow in a square cavity with variant mesh densities (FVM in the leftcolumn and FDM in the right column)
8 Journal of Applied Mathematics
(a) 40 times 40 (b) 20 times 20
Figure 8 Comparison of stream lines in lid-driven flow with variant mesh densities (FVM in the left column and FDM in the right column)
(a) Thom (b) Woods
(c) Jenson
Figure 9 Stream functions of lid-driven flow under variant boundary treatments with the grid of 20 times 20 (FVM in the left column and FDMin the right column)
the convection item (Mesh Peclet number) is affected bythe following factors such as mesh density and boundarycondition Taking lid-driven flow as example the stabilityof convection term of FVM and FDM with variant meshdensities treatments of boundary condition is compared
421 Influence of Mesh Density Figure 8 shows the compar-ison of stream lines between FVM and FDM with variantmesh densities With the grid of 40 times 40 the two methodsboth can capture the structure of primary vortex and sec-ondary vortex as Figure 8(a) shows It is seen in Figure 8(b)that when the mesh of 20 times 20 is adopted concussionphenomenon appears which ismore serious in FDM and thelocation of the vortex center obviously moves upward
422 Influence of Boundary Condition Similarly to the accu-racy analysis the results obtained by variant treatments for
vorticity of wall boundary are compared as Figure 9 showsThree treatments Thom Woods and Jenson are analyzedFrom Figure 9 we can see that only when Jenson equationis adopted the results obtained by FVM and FDM arerelatively close while when Thom and Woods are adoptedFVM shows better results than FDM The results obtainedby FDM suffer severe oscillation and even lack physicalmeaning It is clear that based on the lid-driven flow stabilityof convection term of FVM is superior to that of FDMThe influence from the same boundary error (truncationerror and rounding error) on difference solving methodsis illustrated in Figure 9 With the treatment of boundarycondition with the same truncation error or rounding errorthe stability of FVM is better than that of FDM especiallyin the situation with boundary treatment of (a) and (b)with relatively low precision truncation error FDM is veryunstable and very easy to slip into oscillation
Journal of Applied Mathematics 9
120572
02 04 06 08 1
Nite
r
80 times 80
40 times 4020 times 20
FVM FDM
5
4
3
2
1
0
times104
(a) Second-order central difference scheme
02 04 06 08 1120572
Nite
r
5
4
3
2
1
0
80 times 80
40 times 4020 times 20
FVM FDM
times104
(b) Second-order upwind scheme
Figure 10 Comparison of the robustness of lid-driven flow with variant schemes
43 Robustness Robustness is a very critical index to evaluatethe numerical approaches In this section the robustness ofFVM and FDM is compared based on the lid-driven flowproblem which is solved by vorticity-stream function Gridsof 20 times 20 40 times 40 and 80 times 80 are adopted respectively inthe calculation Relaxation factor is tentatively taken up from02 to the maximum which can ensure the convergence ofcalculation Iterations are correspondingly recorded as well
It is shown in Figure 10 when the whether the cen-tral difference scheme or second-order upwind scheme isadopted with the refining of calculation grids the maximumrelaxation factor (120572) increases in both FVM and FDM therobustness is enhanced and the iterations (Niter) corre-spondingly decrease as well
As Figure 10(a) shows in the situation that central differ-ence scheme is adopted with the grid of 20 times 20 and 40 times 40the maximum relaxation factor of FVM is bigger than thatof FVM while with the grid of 80 times 80 the relaxation factorcan be taken as 10 in bothmethods Similarly as Figure 10(b)shows generally speaking when second-order upwind dif-ference scheme is adopted the maximum relaxation factor isbigger than that of FDM Based on the result of lid-drivenflow the conservative FVM is superior to the nonconservativeFDM from the perspective of numerical robustness
In the numerical simulation of complex flow field cal-culation efficiency is another critical index on which thispapermakes analysis FromFigure 10we can see that with theincreasing of relaxation factor the iterations of two methodsboth decrease generally What needs to be notified is that theiterations of FVM are smaller than those of FDM
44 Efficiency In the numerical simulation of complex flowfield calculation efficiency is another critical index on whichthis paper makes analysis
As is shown in Table 5 although the calculation timerequired by every iteration step in FVM and FDM is almostthe same with the increasing of mesh numbers calculationtime required by FVM slightly decreases Thus total calcula-tion time of FVM is much smaller than that of FDMThat isFVM is much more efficient that FDM Thus it is validatedthat the efficiency of FVM is better than that of FDM in thesame calculation condition
5 Conclusions
Based on vorticity-stream function and MAC algorithmlid-driven flow and natural convection in a square cavityare calculated The accuracy stability of convective termrobustness of FVM and FDM with variant mesh densitiesdiscrete forms and treatments of boundary condition arecompared The following conclusions are drawn
(1) No matter which algorithm discrete form or treat-ment of boundary condition is adopted the resultsobtained by conservative FVM are closer to bench-mark solution than those obtained by FDM Theaccuracy of FVM is higher than FDM especiallywhen the mesh density is relatively small
(2) The results of lid-driven flow indicate that the resultsobtained by FDM show more serious oscillation andthe results obtained even lose physicalmeaningThusthe stability of convection term of FVM is superiorthan that of FDM
(3) As to the lid-driven flow it is indicated that whenvorticity-stream function method is adopted therelaxation factor of conservative method is biggerthan that of the nonconservative one while theiterations are less which shows that the robustness
10 Journal of Applied Mathematics
Table 5 Computation iterations and time of FVM and FDM in second-order central difference scheme for vorticity-stream functionalgorithm of lid driven flow
Mesh 120572FVM FDM
Iterations Time(s) Time of each iteration Iterations Time(s) Time of each iteration20 02 4950 738 0015 2100 307 001520 01 10500 1544 0015 4300 616 001440 06 3250 1519 0047 4350 1993 004640 05 4450 2035 0046 5800 2629 004540 04 6300 2963 0047 7950 3693 004680 08 4950 8223 0166 5050 8584 017080 06 8500 14134 0166 9000 15395 017180 04 17250 29694 0172 17750 32420 0183
and calculation efficiency of conservative method arebetter than those of the nonconservative method
(4) In the lid-driven flow with vorticity-stream functionmethod the calculation time required by every itera-tion step of FVM is slightly smaller than that of FDMTogether with the smaller total iteration steps FVMis more efficient than FDM
Acknowledgment
The study is supported by the National Science Foundation ofChina (nos 51134006 51176204 and 51276198)
References
[1] J D Anderson Jr Computational Fluid Dynamics The BasicsWith Applications McGraw-Hill 1995
[2] P J Roache Computational Fluid Dynamics Hermosa 1976[3] W Q Tao Heat Transfer Science Press 2001[4] W Q Tao Advances in Computational Heat Transfer Science
Press 2000[5] Y Jaluria and K E Torrance Computational Heat Transfer
Hemisphere Publishing Corporation 1986[6] P D Lax ldquoWeak solutions of nonlinear hyperbolic equations
and their numerical computationrdquo Communications on Pureand Applied Mathematics vol 7 pp 159ndash193 1954
[7] P Lax and B Wendroff ldquoSystems of conservation lawsrdquo Com-munications on Pure and Applied Mathematics vol 13 no 2 pp217ndash237 1960
[8] A Jameson ldquoTransonic potential flow calculations using con-servation formrdquo in Proceedings of the Computational fluiddynamics conference pp 148ndash161 Hartford Conn USA 1975
[9] M Vinokur ldquoConservation equations of gasdynamics in curvi-linear coordinate systemsrdquo Journal of Computational Physicsvol 14 pp 105ndash125 1974
[10] T J Bridges ldquoConservation laws in curvilinear coordinatesa short proof of Vinokurrsquos theorem using differential formsrdquoApplied Mathematics and Computation vol 202 no 2 pp 882ndash885 2008
[11] R J DiPerna ldquoDecay of solutions of hyperbolic systems ofconservation laws with a convex extensionrdquoArchive for RationalMechanics and Analysis vol 64 no 1 pp 1ndash46 1977
[12] A Bressan ldquoHyperbolic systems of conservation lawsrdquo RevistaMatematica Complutense vol 12 no 1 pp 135ndash198 1999
[13] R K S Hankin ldquoThe Euler equations for multiphase com-pressible flow in conservation form-Simulation of shock-bubbleinteractionsrdquo Journal of Computational Physics vol 172 no 2pp 808ndash826 2001
[14] C Lai R A Baltzer andRW Schaffranek ldquoConservation-formequations of unsteady open-channel flowrdquo Journal of HydraulicResearch vol 40 no 5 pp 567ndash578 2002
[15] L M Gao K T Li B Liu and J Su ldquoGeneral integralconservation form of Navier-Stokes equations and numericalapplicationrdquo Chinese Journal of Computational Physics vol 25no 2 pp 172ndash178 2008
[16] S Patankar Numerical Heat Transfer and Fluid Flow Hemi-sphere Publishing 1980
[17] A West and T Fuller ldquoInfluence of rib spacing in proton-exchange membrane electrode assembliesrdquo Journal of AppliedElectrochemistry vol 26 no 6 pp 557ndash565 1996
[18] B P Leonard ldquoComparison of truncation error of finite-difference and finite-volume formulations of convection termsrdquoApplied Mathematical Modelling vol 18 no 1 pp 46ndash50 1994
[19] G G Botte J A Ritter and R E White ldquoComparison of finitedifference and control volume methods for solving differentialequationsrdquoComputers and Chemical Engineering vol 24 no 12pp 2633ndash2654 2000
[20] AThom ldquoThe flow past cylinders at low speedsrdquo Proceedings ofthe Royal Society A vol 141 pp 651ndash659 1993
[21] L C Woods ldquoA note on the numerical solution of fourth orderdifferential equationsrdquo vol 5 pp 176ndash182 1954
[22] V G Jenson ldquoViscous flow round a sphere at low Reynoldsnumbers (Re iexcl 40)rdquo Proceedings of the Royal Society A vol 249pp 346ndash366 1959
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
2 Journal of Applied Mathematics
mass conservation and computation time It was validated byexamples that FDM cannot ensure mass conservation whileFVM can Meanwhile the influence of boundary conditionson the accuracy of algorithm was taken into consideration
Previous researchers mainly focused on the comparisonof accuracy There are few systemic studies on the stabilityrobustness and computation efficiencyThis paper comparedthe calculation performances of FVM and FDM based onprevious research Comprehensive comparison of accuracystability of convection term robustness and calculationefficiency was performed The results obtained may enrichthe study on the calculation performances of conservativeand nonconservativemethods which are expected to providesome reference for the practical calculation
2 Physical Problem
Tomake the study possess general significance it is necessaryto analyze the calculation performances of the two methodsbased on variant problems Lid-driven flow and naturalconvection flow in a square cavity which are both typicalproblems in computational fluid mechanics and numericalheat transfer are solved in this paper to make a systemicanalysis on the subject
As Figure 1(a) shows in the lid-driven flow the lid moveshorizontally at a constant velocity 119906top while the other threeboundaries keep still The Reynolds number Re (Reynoldsnumber) adopted in this paper is 1000
Figure 1(b) shows the schematic of natural convectionflow in a square cavity The left and right boundaries are thefirst boundary conditions between which the temperatureon the left boundary is higher The other two boundaries areadiabatic ones In calculation Pr (Prandtl number) is takenas 071 and Ra (Rayleigh number) is taken as 106
3 Numerical Method
Considering that the calculation approach is influential tothe calculation in order to compare the difference of cal-culation performances between FVM and FDM this paperadopts both primitive variable method and nonprimitivevariable method to solve the governing equations For prim-itive variable method MAC algorithm is adopted and fornonprimitive variable method the generally used vorticity-stream function algorithm is adopted The conservationproperty generally refers to the convection term and thusonly the difference of convective term between conservativeand nonconservative forms is compared MAC method ischaracterized by the direct solving of velocity and pressureThe key point of this method is to obtain a nondivergentvelocity field in every time step or iteration step That is thevelocity field of every time layer or iteration layer shouldsatisfy the continuity equation and pressure and the velocityis decoupled by solving Poisson equation which is relevant topressure
31 Governing Equation In primitive variable method pres-sure and velocity are directly used as variables to solve the
Table 1 Parameters of equations for MAC of lid-driven flow
Name of variables Φ Γ120601
119878120601
119883 direction 1198801
Reminus120597119875
120597119883
119884 direction 1198811
Reminus120597119875
120597119884
Table 2 Parameters of equations for MAC of natural convectionflow
Name of variables Φ Γ120601
119878120601
119883 direction 119880 1 minus120597119875
120597119883
119884 direction 119881 1 RaΘPr
minus120597119875
120597119884
Temperature Θ1
Pr0
equations As to the incompressible fluid the dimensionlessunsteady-state conservative and nonconservative equationscan be written as respectively
120597Φ
120597120591+120597 (119880Φ)
120597119883+120597 (119881Φ)
120597119884= Γ120601(1205972Φ
1205971198832+1205972Φ
1205971198842) + 119878120601 (1)
120597Φ
120597120591+ 119880
120597Φ
120597119883+ 119881
120597Φ
120597119884= Γ120601(1205972Φ
1205971198832+1205972Φ
1205971198842) + 119878120601 (2)
where the values of Φ Γ120601 and 119878
120601for lid-driven flow and
natural convection flow are shown in Table 1 and Table 2respectively From (1) and (2) we can see that the convec-tion term of conservative equation is given in the form ofdivergence while that in nonconservative equation is notwhere the nondimensional excessive temperature can bedescribed as Θ = (119879 minus 119879
119897)(119879ℎminus 119879119897) and U V are defined as
nondimensional velocities in separate directions (X directionor Y direction) Governing equations of conservative andnonconservative equations of primitive variable method aregiven before In the calculation process MAC algorithmis adopted to obtain the result in steady state by unsteadycalculation over enough time The result of steady state isanalyzed as follows from the angle of primitive variablemethod Taking the vorticity-stream function method asexample the governing equations of conservative form andnonconservative form can be written as respectively
119886120601(120597 (119880Φ)
120597119883+120597 (119881Φ)
120597119884) = Γ120601(1205972Φ
1205971198832+1205972Φ
1205971198842) + 119878120601 (3)
119886120601(119880
120597Φ
120597119883+ 119881
120597Φ
120597119884) = Γ120601(1205972Φ
1205971198832+1205972Φ
1205971198842) + 119878120601 (4)
As to the vorticity-stream function method only the lid-driven flow is studied in this paper Values ofΦ 119886
120601 Γ120601 and 119878
120601
are shown in Table 3 where ldquo120596rdquo stands for vorticity and ldquo120595rdquostands for stream function
Journal of Applied Mathematics 3
l
l
y
x
119984top
(a) Lid-Driven Flow
l
l
y
x
gTh Tl
Adiabatic
Adiabatic
(b) Natural Convection Flow
Figure 1 Schematic of different physical problem in a square cavity
Table 3 Parameters of equations for vorticity-stream functionalgorithm of lid-driven flow
Name of variables Φ 119886120601
Γ120601
119878120601
Vorticity equation 120596 1 1
Re0
Stream function equation 120595 0 1 minus120596
where ldquo120596rdquo stands for vorticity and ldquo120595rdquo stands for stream function
32 Discretization of Equations Uniform mesh is adopted inthis paper In order to compare the calculation performancesof FVM and FDM with variant discretized schemes centraldifference and second-order upwind scheme are adopted Asto FVM taking e interface on X direction as example theexpressions can be written as
Φ+
119890= Φminus
119890=Φ119875+ Φ119864
2 (5)
Φ+
119890= 15Φ
119875minus 05Φ
119882 119880
119890ge 0 (6a)
Φminus
119890= 15Φ
119864minus 05Φ
119864119864 119880
119890lt 0 (6b)
where ldquoerdquo stands for the east interface between two nodes ldquoErdquostands for east node and ldquoWrdquo means west node and ldquo119864119864rdquo canbe defined as the east node of ldquoErdquo node
As to FDM taking node119875 as example the expressions canbe written as
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
=120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
=Φ119864minus 2Φ119875+ Φ119882
2Δ119883 (7)
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
=3Φ119875minus 4Φ119882+ Φ119882119882
2Δ119883 119880
119875ge 0 (8a)
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
=minus3Φ119875+ 4Φ119864minus Φ119864119864
2Δ119883 119880
119875lt 0 (8b)
xy
ΔY (i 1)
(i 2)
(i 3)
Figure 2 Schematic of the treatment for boundary conditions ofvorticity-stream function method
33 Treatment of the Boundary Condition of Vorticity Sincethe influence from boundary condition of vorticity on thesolution of equations is significant in this section the calcula-tion performances of FVM and FDMwith variant treatmentsof the boundary condition for vorticity are studied
Three boundary conditions of vorticity are comparedand studied In the boundary as shown in Figure 2 theirexpressions Thom equations [20] Woods equation [21] andJenson equation [22] as well as corresponding accuracy areshown in (9)ndash(11)
Thom equation
1205961198941=2 (1205951198942minus 1205951198941)
1205751198842
+ 119874 (Δ119884) (9)
Woods equation
1205961198941=3 (1205951198942minus 1205951198941)
Δ1198842
minus1
21205961198942+ 119874 (Δ119884
2) (10)
Jenson equation
1205961198941=minus71205951198941+ 81205951198942minus 1205951198943
2Δ1198842
+ 119874 (Δ1198842) (11)
4 Journal of Applied Mathematics
34 Deferred Correction When MAC is adopted momen-tum and energy equations are both solved in explicit schemewhile in vorticity-stream function method the convectionterm is treated in implicit scheme by deferred correctionwhich can guarantee the main-diagonal domination of thesolvingmatrix and thus guarantee the calculation stability Asto FVM taking Φ
119890as example the expressions of deferred
correction are as follows
when 119880119890ge 0 Φ
119890= Φ119875+ (Φ+
119890minus Φ119875)lowast
when 119880119890lt 0 Φ
119890= Φ119864+ (Φminus
119890minus Φ119864)lowast
(12)
where ldquolowastrdquo in these equations stands for the former calculationstep
As to FDM similarly to the treatment of Φ119890in FVM the
deferred correction for node 119875 can be written as the sum offirst-order upwind gradient and correction gradient which isshown as follows
when 119880119875ge 0 119880
119875
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
= 119880119875(Φ119875minus Φ119882
Δ119883)
+ [119880119875(120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119882
Δ119883)]
lowast
(13)
when 119880119875lt 0 119880
119875
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
= 119880119875(Φ119864minus Φ119875
Δ119883)
+ [119880119875(120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119864minus Φ119875
Δ119883)]
lowast
(14)
The first terms in the right side of (12)ndash(14) are all first-orderupwind term The second terms are incorporated into thesource term as the form of correction term The discretizedequation is solved by G-S solver with underrelaxation ofwhich the discrete expression is shown as (15) and thediscretized equations of FVM and FDM are shown in (16a)(16b) (17a) and (17b) respectively
119886119875Φ119875= 119886119864Φ119864+ 119886119882Φ119882+ 119886119873Φ119873+ 119886119878Φ119878+ 119887 (15)
119886119864=max (minus119880
119890 0)
Δ119883+
1
ReΔ1198832
119886119882=max (119880
119908 0)
Δ119883+
1
ReΔ1198832
119886119873=max (minus119881
119899 0)
Δ119884+
1
ReΔ1198842
119886119878=max (119881
119904 0)
Δ119884+
1
ReΔ1198842
119886119875=
max (119880119890 0)
Δ119883+max (minus119880
119908 0)
Δ119883+max (119881
119899 0)
Δ119884
+max (minus119881
119904 0)
Δ119884+
2
ReΔ1198832+
2
ReΔ1198842
(16a)
119887 = minusmax (119880119890 0) [
Φ+
119890minus Φ119875
Δ119883] +max (minus119880
119890 0) [
Φminus
119890minus Φ119864
Δ119883]
+max (119880119908 0)[
Φ+
119908minus Φ119882
Δ119883]minusmax (minus119880
119908 0)[
Φminus
119908minus Φ119875
Δ119883]
minusmax (119881119899 0) [
Φ+
119899minus Φ119875
Δ119884] +max (minus119881
119899 0) [
Φminus
119899minus Φ119873
Δ119884]
+max (119881119904 0) [
Φ+
119904minus Φ119878
Δ119884] minusmax (minus119881
119904 0) [
Φminus
119904minus Φ119875
Δ119884]
(16b)
119886119864=max (minus119880
119875 0)
Δ119883+
1
ReΔ1198832
119886119882=max (119880
119875 0)
Δ119883+
1
ReΔ1198832
119886119873=max (minus119881
119875 0)
Δ119884+
1
ReΔ1198842
119886119878=max (119881
119875 0)
Δ119884+
1
ReΔ1198842
119886119875=
max (minus119880119875 0)
Δ119883+max (119880
119875 0)
Δ119883+max (minus119881
119875 0)
Δ119884
+max (119881
119875 0)
Δ119884+
2
ReΔ1198832+
2
ReΔ1198842
(17a)
119887 = minusmax (119880119875 0) [
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119882
Δ119883]
+max (minus119880119875 0) [
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119864minus Φ119875
Δ119883]
minusmax (119881119875 0) [
120597Φ
120597119884
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119878
Δ119884]
+max (minus119881119875 0) [
120597Φ
120597119884
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119873minus Φ119875
Δ119883]
(17b)
where 119886119873 119886119875 119886119864 119886119882 and 119886
119878are coefficients in discretized
equations and 119887 is the source term in discretized equations
4 Results Comparison and Analysis
In this section the accuracy stability of convection termrobustness and calculation efficiency of variant algorithmsschemes and treatments of boundary condition which havebeen discussed previously are systemically compared andanalyzed based on the calculation results of lid-driven flowand natural convection flow in a square cavity
41 Accuracy Firstly the accuracy of FVM and FDM isevaluated by comparing benchmark solution with the results
Journal of Applied Mathematics 5
Table 4 Truncation error of FVM and FDMgoverning equations for vorticity-stream function algorithm of lid driven flow in a square cavity
Convection item Diffusion item
FVM (format)Φ119890
Δ119883minusΦ119908
Δ119883
1
Δ119883
120597Φ
120597119909
10038161003816100381610038161003816100381610038161003816119890
minus1
Δ119883
120597Φ
120597119909
10038161003816100381610038161003816100381610038161003816119908
FVM (TE) minusΔ1198832
8Φ101584010158401015840
119875minusΔ1198834
128Φ(5)
119875minus sdot sdot sdot minus
Δ1198832
24Φ(4)
119875minus 13
Δ1198834
5760Φ(6)
119875minus sdot sdot sdot
FDM (format) 120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816119875
1205972Φ
1205971198832
100381610038161003816100381610038161003816100381610038161003816119875
FDM (TE) minusΔ1198832
6Φ101584010158401015840
119875minusΔ1198834
120Φ(5)
119875minus sdot sdot sdot minus
Δ1198832
12Φ(4)
119875minusΔ1198834
360Φ(6)
119875minus sdot sdot sdot
obtained from variant mesh density discrete schemes andboundary conditions Through precision analysis FVM andFDM are respectively adopted with the Taylor expansioncommenced at midpoint (P node) and their respectivetruncation error (TE) is just as shown Table 4 [19] Thetruncation errors of FVM are obviously less than the FDMrsquos
411 Influence of Mesh Density In [4] it is indicated thatthe conservative property can affect the accuracy while thenonconservative FDM loses its conservative property in thesituation when the grid is too small Tomake further analysisthe calculation results obtained by FVM and FDM withvariant mesh densities are compared The results of thevelocity along the horizontal central axisV and velocity alongthe vertical central axis U obtained by FDM and FVM withcentral difference scheme of MAC algorithm are comparedand illustrated in Figure 3
From Figure 3 we can see that when the mesh densityis taken as 80 times 80 the results obtained by FVM and FDMboth agree well with the benchmark solutionWhen themeshdensity is reduced to 40 times 40 the results obtained by FVMare closer to the benchmark solution and if the mesh densityis further reduced to 20 times 20 the results obtained by FDMdiffer from the benchmark solution greatly in which FVMshows its advantage in accuracyThis phenomenon is in goodagreement with the theoretical analysis in [18] that is theconservative property of FVM better ensures the accuracyof calculation Therefore whether the conservative propertycan be realized is of much significance to the accuracy of theobtained result
412 Influence of Boundary Condition In the followingsection FVM and FDM with variant boundary conditionsare compared from the perspective of accuracy by vorticity-stream function method The results of the velocity alongthe horizontal central axis V and velocity along the verticalcentral axis U obtained by FDM and FVM with centraldifference scheme are compared and illustrated in Figure 4and the influence of boundary condition of vorticity is alsogiven From Figure 4 we can see that no matter whichtreatment of boundary condition is adopted the accuracy ofFVM is higher than that of FDM in vorticity-stream functionmethod
413 Influence of Discrete Scheme The selection of discreteforms is also influential to the accuracy Thus the resultsobtained with variant discrete forms are studied Taking thevorticity-stream function with boundary condition treatedby first-order Thom equation as example the results of thevelocity along the horizontal central axis V and velocityalong the vertical central axis U obtained by FDM and FVMwith variant discrete schemes are compared and illustratedin Figure 5 from which we can see that whether second-order central difference or second-order upwind scheme isadopted the accuracy of FVM is higher than that of FDM thecomparison between (a) and (b) indicates that second-orderupwind difference possesses higher accuracy than second-order central difference
414 Influence of Physical Problem Finally two methods invariant physical problems are compared from the perspectiveof accuracy based on the convection flow in a square cavityTaking the result obtained by MAC algorithm as exampleFigure 6 shows the comparison of results of the velocity alongthe horizontal central axis V and velocity along the verticalcentral axisU obtained by FDMandFVMTemperature fieldsobtained by FVM and FDM are compared with the grid-independent solution (obtained with grid of 256 times 256) asFigure 7 shows
From the comparison in Figure 6 and Figure 7 we cansee that based on the results obtained by MAC of naturalconvection flow in a square cavity the accuracy of FVM isgenerally superior to that of FDM Thus it is indicated bycomparison that no matter which mesh density scheme andtreatment of boundary condition is adopted FVM showsbetter accuracy than FDM
42 Stability of Convection Term Stability of mathematicalmeaning can only ensure that the oscillation of solution iscontrolled within a range but cannot eliminate the absenceof oscillation The amplification of oscillation may result indivergence Stability of convection term originates from var-ied difference schemes for convective itemand is independentof the introducing of rounding error in the calculation Thisstability is the key point to obtain a solution with physicalmeaning It is concluded that all the unstable schemes willresult in the oscillation of solution Stability condition of
6 Journal of Applied Mathematics
0
04
08
12
0
03
06
U V
X(Y)
0 02 04 06 08 1minus04
minus06
minus03
(a) 80 times 80
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) 40 times 40
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(c) 20 times 20
Figure 3 Velocity on the central axis of lid-driven flow obtained by MAC with variant mesh densities
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(a) Thom
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Woods
Figure 4 Velocity along the central axis obtained by second-order central difference scheme with two treatments of boundary conditionwhen the mesh density is 40 times 40
Journal of Applied Mathematics 7
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12U
minus04
0
03
06
V
minus06
minus03
(a) Second-order central difference scheme
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Second-order upwind difference scheme
Figure 5 Velocity along the central axis obtained by vorticity-stream function method for lid-driven flow with different schemes when themesh density is 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(a) 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(b) 80 times 80
Figure 6 Velocity along the central axis of convection flow in a square cavity with variant gird densities
40 times 40
Grid-independent
(a) 40 times 40
80 times 80
Grid-independent
(b) 80 times 80
Figure 7 Comparison of temperature contours of natural convection flow in a square cavity with variant mesh densities (FVM in the leftcolumn and FDM in the right column)
8 Journal of Applied Mathematics
(a) 40 times 40 (b) 20 times 20
Figure 8 Comparison of stream lines in lid-driven flow with variant mesh densities (FVM in the left column and FDM in the right column)
(a) Thom (b) Woods
(c) Jenson
Figure 9 Stream functions of lid-driven flow under variant boundary treatments with the grid of 20 times 20 (FVM in the left column and FDMin the right column)
the convection item (Mesh Peclet number) is affected bythe following factors such as mesh density and boundarycondition Taking lid-driven flow as example the stabilityof convection term of FVM and FDM with variant meshdensities treatments of boundary condition is compared
421 Influence of Mesh Density Figure 8 shows the compar-ison of stream lines between FVM and FDM with variantmesh densities With the grid of 40 times 40 the two methodsboth can capture the structure of primary vortex and sec-ondary vortex as Figure 8(a) shows It is seen in Figure 8(b)that when the mesh of 20 times 20 is adopted concussionphenomenon appears which ismore serious in FDM and thelocation of the vortex center obviously moves upward
422 Influence of Boundary Condition Similarly to the accu-racy analysis the results obtained by variant treatments for
vorticity of wall boundary are compared as Figure 9 showsThree treatments Thom Woods and Jenson are analyzedFrom Figure 9 we can see that only when Jenson equationis adopted the results obtained by FVM and FDM arerelatively close while when Thom and Woods are adoptedFVM shows better results than FDM The results obtainedby FDM suffer severe oscillation and even lack physicalmeaning It is clear that based on the lid-driven flow stabilityof convection term of FVM is superior to that of FDMThe influence from the same boundary error (truncationerror and rounding error) on difference solving methodsis illustrated in Figure 9 With the treatment of boundarycondition with the same truncation error or rounding errorthe stability of FVM is better than that of FDM especiallyin the situation with boundary treatment of (a) and (b)with relatively low precision truncation error FDM is veryunstable and very easy to slip into oscillation
Journal of Applied Mathematics 9
120572
02 04 06 08 1
Nite
r
80 times 80
40 times 4020 times 20
FVM FDM
5
4
3
2
1
0
times104
(a) Second-order central difference scheme
02 04 06 08 1120572
Nite
r
5
4
3
2
1
0
80 times 80
40 times 4020 times 20
FVM FDM
times104
(b) Second-order upwind scheme
Figure 10 Comparison of the robustness of lid-driven flow with variant schemes
43 Robustness Robustness is a very critical index to evaluatethe numerical approaches In this section the robustness ofFVM and FDM is compared based on the lid-driven flowproblem which is solved by vorticity-stream function Gridsof 20 times 20 40 times 40 and 80 times 80 are adopted respectively inthe calculation Relaxation factor is tentatively taken up from02 to the maximum which can ensure the convergence ofcalculation Iterations are correspondingly recorded as well
It is shown in Figure 10 when the whether the cen-tral difference scheme or second-order upwind scheme isadopted with the refining of calculation grids the maximumrelaxation factor (120572) increases in both FVM and FDM therobustness is enhanced and the iterations (Niter) corre-spondingly decrease as well
As Figure 10(a) shows in the situation that central differ-ence scheme is adopted with the grid of 20 times 20 and 40 times 40the maximum relaxation factor of FVM is bigger than thatof FVM while with the grid of 80 times 80 the relaxation factorcan be taken as 10 in bothmethods Similarly as Figure 10(b)shows generally speaking when second-order upwind dif-ference scheme is adopted the maximum relaxation factor isbigger than that of FDM Based on the result of lid-drivenflow the conservative FVM is superior to the nonconservativeFDM from the perspective of numerical robustness
In the numerical simulation of complex flow field cal-culation efficiency is another critical index on which thispapermakes analysis FromFigure 10we can see that with theincreasing of relaxation factor the iterations of two methodsboth decrease generally What needs to be notified is that theiterations of FVM are smaller than those of FDM
44 Efficiency In the numerical simulation of complex flowfield calculation efficiency is another critical index on whichthis paper makes analysis
As is shown in Table 5 although the calculation timerequired by every iteration step in FVM and FDM is almostthe same with the increasing of mesh numbers calculationtime required by FVM slightly decreases Thus total calcula-tion time of FVM is much smaller than that of FDMThat isFVM is much more efficient that FDM Thus it is validatedthat the efficiency of FVM is better than that of FDM in thesame calculation condition
5 Conclusions
Based on vorticity-stream function and MAC algorithmlid-driven flow and natural convection in a square cavityare calculated The accuracy stability of convective termrobustness of FVM and FDM with variant mesh densitiesdiscrete forms and treatments of boundary condition arecompared The following conclusions are drawn
(1) No matter which algorithm discrete form or treat-ment of boundary condition is adopted the resultsobtained by conservative FVM are closer to bench-mark solution than those obtained by FDM Theaccuracy of FVM is higher than FDM especiallywhen the mesh density is relatively small
(2) The results of lid-driven flow indicate that the resultsobtained by FDM show more serious oscillation andthe results obtained even lose physicalmeaningThusthe stability of convection term of FVM is superiorthan that of FDM
(3) As to the lid-driven flow it is indicated that whenvorticity-stream function method is adopted therelaxation factor of conservative method is biggerthan that of the nonconservative one while theiterations are less which shows that the robustness
10 Journal of Applied Mathematics
Table 5 Computation iterations and time of FVM and FDM in second-order central difference scheme for vorticity-stream functionalgorithm of lid driven flow
Mesh 120572FVM FDM
Iterations Time(s) Time of each iteration Iterations Time(s) Time of each iteration20 02 4950 738 0015 2100 307 001520 01 10500 1544 0015 4300 616 001440 06 3250 1519 0047 4350 1993 004640 05 4450 2035 0046 5800 2629 004540 04 6300 2963 0047 7950 3693 004680 08 4950 8223 0166 5050 8584 017080 06 8500 14134 0166 9000 15395 017180 04 17250 29694 0172 17750 32420 0183
and calculation efficiency of conservative method arebetter than those of the nonconservative method
(4) In the lid-driven flow with vorticity-stream functionmethod the calculation time required by every itera-tion step of FVM is slightly smaller than that of FDMTogether with the smaller total iteration steps FVMis more efficient than FDM
Acknowledgment
The study is supported by the National Science Foundation ofChina (nos 51134006 51176204 and 51276198)
References
[1] J D Anderson Jr Computational Fluid Dynamics The BasicsWith Applications McGraw-Hill 1995
[2] P J Roache Computational Fluid Dynamics Hermosa 1976[3] W Q Tao Heat Transfer Science Press 2001[4] W Q Tao Advances in Computational Heat Transfer Science
Press 2000[5] Y Jaluria and K E Torrance Computational Heat Transfer
Hemisphere Publishing Corporation 1986[6] P D Lax ldquoWeak solutions of nonlinear hyperbolic equations
and their numerical computationrdquo Communications on Pureand Applied Mathematics vol 7 pp 159ndash193 1954
[7] P Lax and B Wendroff ldquoSystems of conservation lawsrdquo Com-munications on Pure and Applied Mathematics vol 13 no 2 pp217ndash237 1960
[8] A Jameson ldquoTransonic potential flow calculations using con-servation formrdquo in Proceedings of the Computational fluiddynamics conference pp 148ndash161 Hartford Conn USA 1975
[9] M Vinokur ldquoConservation equations of gasdynamics in curvi-linear coordinate systemsrdquo Journal of Computational Physicsvol 14 pp 105ndash125 1974
[10] T J Bridges ldquoConservation laws in curvilinear coordinatesa short proof of Vinokurrsquos theorem using differential formsrdquoApplied Mathematics and Computation vol 202 no 2 pp 882ndash885 2008
[11] R J DiPerna ldquoDecay of solutions of hyperbolic systems ofconservation laws with a convex extensionrdquoArchive for RationalMechanics and Analysis vol 64 no 1 pp 1ndash46 1977
[12] A Bressan ldquoHyperbolic systems of conservation lawsrdquo RevistaMatematica Complutense vol 12 no 1 pp 135ndash198 1999
[13] R K S Hankin ldquoThe Euler equations for multiphase com-pressible flow in conservation form-Simulation of shock-bubbleinteractionsrdquo Journal of Computational Physics vol 172 no 2pp 808ndash826 2001
[14] C Lai R A Baltzer andRW Schaffranek ldquoConservation-formequations of unsteady open-channel flowrdquo Journal of HydraulicResearch vol 40 no 5 pp 567ndash578 2002
[15] L M Gao K T Li B Liu and J Su ldquoGeneral integralconservation form of Navier-Stokes equations and numericalapplicationrdquo Chinese Journal of Computational Physics vol 25no 2 pp 172ndash178 2008
[16] S Patankar Numerical Heat Transfer and Fluid Flow Hemi-sphere Publishing 1980
[17] A West and T Fuller ldquoInfluence of rib spacing in proton-exchange membrane electrode assembliesrdquo Journal of AppliedElectrochemistry vol 26 no 6 pp 557ndash565 1996
[18] B P Leonard ldquoComparison of truncation error of finite-difference and finite-volume formulations of convection termsrdquoApplied Mathematical Modelling vol 18 no 1 pp 46ndash50 1994
[19] G G Botte J A Ritter and R E White ldquoComparison of finitedifference and control volume methods for solving differentialequationsrdquoComputers and Chemical Engineering vol 24 no 12pp 2633ndash2654 2000
[20] AThom ldquoThe flow past cylinders at low speedsrdquo Proceedings ofthe Royal Society A vol 141 pp 651ndash659 1993
[21] L C Woods ldquoA note on the numerical solution of fourth orderdifferential equationsrdquo vol 5 pp 176ndash182 1954
[22] V G Jenson ldquoViscous flow round a sphere at low Reynoldsnumbers (Re iexcl 40)rdquo Proceedings of the Royal Society A vol 249pp 346ndash366 1959
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 3
l
l
y
x
119984top
(a) Lid-Driven Flow
l
l
y
x
gTh Tl
Adiabatic
Adiabatic
(b) Natural Convection Flow
Figure 1 Schematic of different physical problem in a square cavity
Table 3 Parameters of equations for vorticity-stream functionalgorithm of lid-driven flow
Name of variables Φ 119886120601
Γ120601
119878120601
Vorticity equation 120596 1 1
Re0
Stream function equation 120595 0 1 minus120596
where ldquo120596rdquo stands for vorticity and ldquo120595rdquo stands for stream function
32 Discretization of Equations Uniform mesh is adopted inthis paper In order to compare the calculation performancesof FVM and FDM with variant discretized schemes centraldifference and second-order upwind scheme are adopted Asto FVM taking e interface on X direction as example theexpressions can be written as
Φ+
119890= Φminus
119890=Φ119875+ Φ119864
2 (5)
Φ+
119890= 15Φ
119875minus 05Φ
119882 119880
119890ge 0 (6a)
Φminus
119890= 15Φ
119864minus 05Φ
119864119864 119880
119890lt 0 (6b)
where ldquoerdquo stands for the east interface between two nodes ldquoErdquostands for east node and ldquoWrdquo means west node and ldquo119864119864rdquo canbe defined as the east node of ldquoErdquo node
As to FDM taking node119875 as example the expressions canbe written as
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
=120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
=Φ119864minus 2Φ119875+ Φ119882
2Δ119883 (7)
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
=3Φ119875minus 4Φ119882+ Φ119882119882
2Δ119883 119880
119875ge 0 (8a)
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
=minus3Φ119875+ 4Φ119864minus Φ119864119864
2Δ119883 119880
119875lt 0 (8b)
xy
ΔY (i 1)
(i 2)
(i 3)
Figure 2 Schematic of the treatment for boundary conditions ofvorticity-stream function method
33 Treatment of the Boundary Condition of Vorticity Sincethe influence from boundary condition of vorticity on thesolution of equations is significant in this section the calcula-tion performances of FVM and FDMwith variant treatmentsof the boundary condition for vorticity are studied
Three boundary conditions of vorticity are comparedand studied In the boundary as shown in Figure 2 theirexpressions Thom equations [20] Woods equation [21] andJenson equation [22] as well as corresponding accuracy areshown in (9)ndash(11)
Thom equation
1205961198941=2 (1205951198942minus 1205951198941)
1205751198842
+ 119874 (Δ119884) (9)
Woods equation
1205961198941=3 (1205951198942minus 1205951198941)
Δ1198842
minus1
21205961198942+ 119874 (Δ119884
2) (10)
Jenson equation
1205961198941=minus71205951198941+ 81205951198942minus 1205951198943
2Δ1198842
+ 119874 (Δ1198842) (11)
4 Journal of Applied Mathematics
34 Deferred Correction When MAC is adopted momen-tum and energy equations are both solved in explicit schemewhile in vorticity-stream function method the convectionterm is treated in implicit scheme by deferred correctionwhich can guarantee the main-diagonal domination of thesolvingmatrix and thus guarantee the calculation stability Asto FVM taking Φ
119890as example the expressions of deferred
correction are as follows
when 119880119890ge 0 Φ
119890= Φ119875+ (Φ+
119890minus Φ119875)lowast
when 119880119890lt 0 Φ
119890= Φ119864+ (Φminus
119890minus Φ119864)lowast
(12)
where ldquolowastrdquo in these equations stands for the former calculationstep
As to FDM similarly to the treatment of Φ119890in FVM the
deferred correction for node 119875 can be written as the sum offirst-order upwind gradient and correction gradient which isshown as follows
when 119880119875ge 0 119880
119875
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
= 119880119875(Φ119875minus Φ119882
Δ119883)
+ [119880119875(120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119882
Δ119883)]
lowast
(13)
when 119880119875lt 0 119880
119875
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
= 119880119875(Φ119864minus Φ119875
Δ119883)
+ [119880119875(120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119864minus Φ119875
Δ119883)]
lowast
(14)
The first terms in the right side of (12)ndash(14) are all first-orderupwind term The second terms are incorporated into thesource term as the form of correction term The discretizedequation is solved by G-S solver with underrelaxation ofwhich the discrete expression is shown as (15) and thediscretized equations of FVM and FDM are shown in (16a)(16b) (17a) and (17b) respectively
119886119875Φ119875= 119886119864Φ119864+ 119886119882Φ119882+ 119886119873Φ119873+ 119886119878Φ119878+ 119887 (15)
119886119864=max (minus119880
119890 0)
Δ119883+
1
ReΔ1198832
119886119882=max (119880
119908 0)
Δ119883+
1
ReΔ1198832
119886119873=max (minus119881
119899 0)
Δ119884+
1
ReΔ1198842
119886119878=max (119881
119904 0)
Δ119884+
1
ReΔ1198842
119886119875=
max (119880119890 0)
Δ119883+max (minus119880
119908 0)
Δ119883+max (119881
119899 0)
Δ119884
+max (minus119881
119904 0)
Δ119884+
2
ReΔ1198832+
2
ReΔ1198842
(16a)
119887 = minusmax (119880119890 0) [
Φ+
119890minus Φ119875
Δ119883] +max (minus119880
119890 0) [
Φminus
119890minus Φ119864
Δ119883]
+max (119880119908 0)[
Φ+
119908minus Φ119882
Δ119883]minusmax (minus119880
119908 0)[
Φminus
119908minus Φ119875
Δ119883]
minusmax (119881119899 0) [
Φ+
119899minus Φ119875
Δ119884] +max (minus119881
119899 0) [
Φminus
119899minus Φ119873
Δ119884]
+max (119881119904 0) [
Φ+
119904minus Φ119878
Δ119884] minusmax (minus119881
119904 0) [
Φminus
119904minus Φ119875
Δ119884]
(16b)
119886119864=max (minus119880
119875 0)
Δ119883+
1
ReΔ1198832
119886119882=max (119880
119875 0)
Δ119883+
1
ReΔ1198832
119886119873=max (minus119881
119875 0)
Δ119884+
1
ReΔ1198842
119886119878=max (119881
119875 0)
Δ119884+
1
ReΔ1198842
119886119875=
max (minus119880119875 0)
Δ119883+max (119880
119875 0)
Δ119883+max (minus119881
119875 0)
Δ119884
+max (119881
119875 0)
Δ119884+
2
ReΔ1198832+
2
ReΔ1198842
(17a)
119887 = minusmax (119880119875 0) [
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119882
Δ119883]
+max (minus119880119875 0) [
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119864minus Φ119875
Δ119883]
minusmax (119881119875 0) [
120597Φ
120597119884
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119878
Δ119884]
+max (minus119881119875 0) [
120597Φ
120597119884
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119873minus Φ119875
Δ119883]
(17b)
where 119886119873 119886119875 119886119864 119886119882 and 119886
119878are coefficients in discretized
equations and 119887 is the source term in discretized equations
4 Results Comparison and Analysis
In this section the accuracy stability of convection termrobustness and calculation efficiency of variant algorithmsschemes and treatments of boundary condition which havebeen discussed previously are systemically compared andanalyzed based on the calculation results of lid-driven flowand natural convection flow in a square cavity
41 Accuracy Firstly the accuracy of FVM and FDM isevaluated by comparing benchmark solution with the results
Journal of Applied Mathematics 5
Table 4 Truncation error of FVM and FDMgoverning equations for vorticity-stream function algorithm of lid driven flow in a square cavity
Convection item Diffusion item
FVM (format)Φ119890
Δ119883minusΦ119908
Δ119883
1
Δ119883
120597Φ
120597119909
10038161003816100381610038161003816100381610038161003816119890
minus1
Δ119883
120597Φ
120597119909
10038161003816100381610038161003816100381610038161003816119908
FVM (TE) minusΔ1198832
8Φ101584010158401015840
119875minusΔ1198834
128Φ(5)
119875minus sdot sdot sdot minus
Δ1198832
24Φ(4)
119875minus 13
Δ1198834
5760Φ(6)
119875minus sdot sdot sdot
FDM (format) 120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816119875
1205972Φ
1205971198832
100381610038161003816100381610038161003816100381610038161003816119875
FDM (TE) minusΔ1198832
6Φ101584010158401015840
119875minusΔ1198834
120Φ(5)
119875minus sdot sdot sdot minus
Δ1198832
12Φ(4)
119875minusΔ1198834
360Φ(6)
119875minus sdot sdot sdot
obtained from variant mesh density discrete schemes andboundary conditions Through precision analysis FVM andFDM are respectively adopted with the Taylor expansioncommenced at midpoint (P node) and their respectivetruncation error (TE) is just as shown Table 4 [19] Thetruncation errors of FVM are obviously less than the FDMrsquos
411 Influence of Mesh Density In [4] it is indicated thatthe conservative property can affect the accuracy while thenonconservative FDM loses its conservative property in thesituation when the grid is too small Tomake further analysisthe calculation results obtained by FVM and FDM withvariant mesh densities are compared The results of thevelocity along the horizontal central axisV and velocity alongthe vertical central axis U obtained by FDM and FVM withcentral difference scheme of MAC algorithm are comparedand illustrated in Figure 3
From Figure 3 we can see that when the mesh densityis taken as 80 times 80 the results obtained by FVM and FDMboth agree well with the benchmark solutionWhen themeshdensity is reduced to 40 times 40 the results obtained by FVMare closer to the benchmark solution and if the mesh densityis further reduced to 20 times 20 the results obtained by FDMdiffer from the benchmark solution greatly in which FVMshows its advantage in accuracyThis phenomenon is in goodagreement with the theoretical analysis in [18] that is theconservative property of FVM better ensures the accuracyof calculation Therefore whether the conservative propertycan be realized is of much significance to the accuracy of theobtained result
412 Influence of Boundary Condition In the followingsection FVM and FDM with variant boundary conditionsare compared from the perspective of accuracy by vorticity-stream function method The results of the velocity alongthe horizontal central axis V and velocity along the verticalcentral axis U obtained by FDM and FVM with centraldifference scheme are compared and illustrated in Figure 4and the influence of boundary condition of vorticity is alsogiven From Figure 4 we can see that no matter whichtreatment of boundary condition is adopted the accuracy ofFVM is higher than that of FDM in vorticity-stream functionmethod
413 Influence of Discrete Scheme The selection of discreteforms is also influential to the accuracy Thus the resultsobtained with variant discrete forms are studied Taking thevorticity-stream function with boundary condition treatedby first-order Thom equation as example the results of thevelocity along the horizontal central axis V and velocityalong the vertical central axis U obtained by FDM and FVMwith variant discrete schemes are compared and illustratedin Figure 5 from which we can see that whether second-order central difference or second-order upwind scheme isadopted the accuracy of FVM is higher than that of FDM thecomparison between (a) and (b) indicates that second-orderupwind difference possesses higher accuracy than second-order central difference
414 Influence of Physical Problem Finally two methods invariant physical problems are compared from the perspectiveof accuracy based on the convection flow in a square cavityTaking the result obtained by MAC algorithm as exampleFigure 6 shows the comparison of results of the velocity alongthe horizontal central axis V and velocity along the verticalcentral axisU obtained by FDMandFVMTemperature fieldsobtained by FVM and FDM are compared with the grid-independent solution (obtained with grid of 256 times 256) asFigure 7 shows
From the comparison in Figure 6 and Figure 7 we cansee that based on the results obtained by MAC of naturalconvection flow in a square cavity the accuracy of FVM isgenerally superior to that of FDM Thus it is indicated bycomparison that no matter which mesh density scheme andtreatment of boundary condition is adopted FVM showsbetter accuracy than FDM
42 Stability of Convection Term Stability of mathematicalmeaning can only ensure that the oscillation of solution iscontrolled within a range but cannot eliminate the absenceof oscillation The amplification of oscillation may result indivergence Stability of convection term originates from var-ied difference schemes for convective itemand is independentof the introducing of rounding error in the calculation Thisstability is the key point to obtain a solution with physicalmeaning It is concluded that all the unstable schemes willresult in the oscillation of solution Stability condition of
6 Journal of Applied Mathematics
0
04
08
12
0
03
06
U V
X(Y)
0 02 04 06 08 1minus04
minus06
minus03
(a) 80 times 80
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) 40 times 40
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(c) 20 times 20
Figure 3 Velocity on the central axis of lid-driven flow obtained by MAC with variant mesh densities
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(a) Thom
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Woods
Figure 4 Velocity along the central axis obtained by second-order central difference scheme with two treatments of boundary conditionwhen the mesh density is 40 times 40
Journal of Applied Mathematics 7
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12U
minus04
0
03
06
V
minus06
minus03
(a) Second-order central difference scheme
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Second-order upwind difference scheme
Figure 5 Velocity along the central axis obtained by vorticity-stream function method for lid-driven flow with different schemes when themesh density is 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(a) 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(b) 80 times 80
Figure 6 Velocity along the central axis of convection flow in a square cavity with variant gird densities
40 times 40
Grid-independent
(a) 40 times 40
80 times 80
Grid-independent
(b) 80 times 80
Figure 7 Comparison of temperature contours of natural convection flow in a square cavity with variant mesh densities (FVM in the leftcolumn and FDM in the right column)
8 Journal of Applied Mathematics
(a) 40 times 40 (b) 20 times 20
Figure 8 Comparison of stream lines in lid-driven flow with variant mesh densities (FVM in the left column and FDM in the right column)
(a) Thom (b) Woods
(c) Jenson
Figure 9 Stream functions of lid-driven flow under variant boundary treatments with the grid of 20 times 20 (FVM in the left column and FDMin the right column)
the convection item (Mesh Peclet number) is affected bythe following factors such as mesh density and boundarycondition Taking lid-driven flow as example the stabilityof convection term of FVM and FDM with variant meshdensities treatments of boundary condition is compared
421 Influence of Mesh Density Figure 8 shows the compar-ison of stream lines between FVM and FDM with variantmesh densities With the grid of 40 times 40 the two methodsboth can capture the structure of primary vortex and sec-ondary vortex as Figure 8(a) shows It is seen in Figure 8(b)that when the mesh of 20 times 20 is adopted concussionphenomenon appears which ismore serious in FDM and thelocation of the vortex center obviously moves upward
422 Influence of Boundary Condition Similarly to the accu-racy analysis the results obtained by variant treatments for
vorticity of wall boundary are compared as Figure 9 showsThree treatments Thom Woods and Jenson are analyzedFrom Figure 9 we can see that only when Jenson equationis adopted the results obtained by FVM and FDM arerelatively close while when Thom and Woods are adoptedFVM shows better results than FDM The results obtainedby FDM suffer severe oscillation and even lack physicalmeaning It is clear that based on the lid-driven flow stabilityof convection term of FVM is superior to that of FDMThe influence from the same boundary error (truncationerror and rounding error) on difference solving methodsis illustrated in Figure 9 With the treatment of boundarycondition with the same truncation error or rounding errorthe stability of FVM is better than that of FDM especiallyin the situation with boundary treatment of (a) and (b)with relatively low precision truncation error FDM is veryunstable and very easy to slip into oscillation
Journal of Applied Mathematics 9
120572
02 04 06 08 1
Nite
r
80 times 80
40 times 4020 times 20
FVM FDM
5
4
3
2
1
0
times104
(a) Second-order central difference scheme
02 04 06 08 1120572
Nite
r
5
4
3
2
1
0
80 times 80
40 times 4020 times 20
FVM FDM
times104
(b) Second-order upwind scheme
Figure 10 Comparison of the robustness of lid-driven flow with variant schemes
43 Robustness Robustness is a very critical index to evaluatethe numerical approaches In this section the robustness ofFVM and FDM is compared based on the lid-driven flowproblem which is solved by vorticity-stream function Gridsof 20 times 20 40 times 40 and 80 times 80 are adopted respectively inthe calculation Relaxation factor is tentatively taken up from02 to the maximum which can ensure the convergence ofcalculation Iterations are correspondingly recorded as well
It is shown in Figure 10 when the whether the cen-tral difference scheme or second-order upwind scheme isadopted with the refining of calculation grids the maximumrelaxation factor (120572) increases in both FVM and FDM therobustness is enhanced and the iterations (Niter) corre-spondingly decrease as well
As Figure 10(a) shows in the situation that central differ-ence scheme is adopted with the grid of 20 times 20 and 40 times 40the maximum relaxation factor of FVM is bigger than thatof FVM while with the grid of 80 times 80 the relaxation factorcan be taken as 10 in bothmethods Similarly as Figure 10(b)shows generally speaking when second-order upwind dif-ference scheme is adopted the maximum relaxation factor isbigger than that of FDM Based on the result of lid-drivenflow the conservative FVM is superior to the nonconservativeFDM from the perspective of numerical robustness
In the numerical simulation of complex flow field cal-culation efficiency is another critical index on which thispapermakes analysis FromFigure 10we can see that with theincreasing of relaxation factor the iterations of two methodsboth decrease generally What needs to be notified is that theiterations of FVM are smaller than those of FDM
44 Efficiency In the numerical simulation of complex flowfield calculation efficiency is another critical index on whichthis paper makes analysis
As is shown in Table 5 although the calculation timerequired by every iteration step in FVM and FDM is almostthe same with the increasing of mesh numbers calculationtime required by FVM slightly decreases Thus total calcula-tion time of FVM is much smaller than that of FDMThat isFVM is much more efficient that FDM Thus it is validatedthat the efficiency of FVM is better than that of FDM in thesame calculation condition
5 Conclusions
Based on vorticity-stream function and MAC algorithmlid-driven flow and natural convection in a square cavityare calculated The accuracy stability of convective termrobustness of FVM and FDM with variant mesh densitiesdiscrete forms and treatments of boundary condition arecompared The following conclusions are drawn
(1) No matter which algorithm discrete form or treat-ment of boundary condition is adopted the resultsobtained by conservative FVM are closer to bench-mark solution than those obtained by FDM Theaccuracy of FVM is higher than FDM especiallywhen the mesh density is relatively small
(2) The results of lid-driven flow indicate that the resultsobtained by FDM show more serious oscillation andthe results obtained even lose physicalmeaningThusthe stability of convection term of FVM is superiorthan that of FDM
(3) As to the lid-driven flow it is indicated that whenvorticity-stream function method is adopted therelaxation factor of conservative method is biggerthan that of the nonconservative one while theiterations are less which shows that the robustness
10 Journal of Applied Mathematics
Table 5 Computation iterations and time of FVM and FDM in second-order central difference scheme for vorticity-stream functionalgorithm of lid driven flow
Mesh 120572FVM FDM
Iterations Time(s) Time of each iteration Iterations Time(s) Time of each iteration20 02 4950 738 0015 2100 307 001520 01 10500 1544 0015 4300 616 001440 06 3250 1519 0047 4350 1993 004640 05 4450 2035 0046 5800 2629 004540 04 6300 2963 0047 7950 3693 004680 08 4950 8223 0166 5050 8584 017080 06 8500 14134 0166 9000 15395 017180 04 17250 29694 0172 17750 32420 0183
and calculation efficiency of conservative method arebetter than those of the nonconservative method
(4) In the lid-driven flow with vorticity-stream functionmethod the calculation time required by every itera-tion step of FVM is slightly smaller than that of FDMTogether with the smaller total iteration steps FVMis more efficient than FDM
Acknowledgment
The study is supported by the National Science Foundation ofChina (nos 51134006 51176204 and 51276198)
References
[1] J D Anderson Jr Computational Fluid Dynamics The BasicsWith Applications McGraw-Hill 1995
[2] P J Roache Computational Fluid Dynamics Hermosa 1976[3] W Q Tao Heat Transfer Science Press 2001[4] W Q Tao Advances in Computational Heat Transfer Science
Press 2000[5] Y Jaluria and K E Torrance Computational Heat Transfer
Hemisphere Publishing Corporation 1986[6] P D Lax ldquoWeak solutions of nonlinear hyperbolic equations
and their numerical computationrdquo Communications on Pureand Applied Mathematics vol 7 pp 159ndash193 1954
[7] P Lax and B Wendroff ldquoSystems of conservation lawsrdquo Com-munications on Pure and Applied Mathematics vol 13 no 2 pp217ndash237 1960
[8] A Jameson ldquoTransonic potential flow calculations using con-servation formrdquo in Proceedings of the Computational fluiddynamics conference pp 148ndash161 Hartford Conn USA 1975
[9] M Vinokur ldquoConservation equations of gasdynamics in curvi-linear coordinate systemsrdquo Journal of Computational Physicsvol 14 pp 105ndash125 1974
[10] T J Bridges ldquoConservation laws in curvilinear coordinatesa short proof of Vinokurrsquos theorem using differential formsrdquoApplied Mathematics and Computation vol 202 no 2 pp 882ndash885 2008
[11] R J DiPerna ldquoDecay of solutions of hyperbolic systems ofconservation laws with a convex extensionrdquoArchive for RationalMechanics and Analysis vol 64 no 1 pp 1ndash46 1977
[12] A Bressan ldquoHyperbolic systems of conservation lawsrdquo RevistaMatematica Complutense vol 12 no 1 pp 135ndash198 1999
[13] R K S Hankin ldquoThe Euler equations for multiphase com-pressible flow in conservation form-Simulation of shock-bubbleinteractionsrdquo Journal of Computational Physics vol 172 no 2pp 808ndash826 2001
[14] C Lai R A Baltzer andRW Schaffranek ldquoConservation-formequations of unsteady open-channel flowrdquo Journal of HydraulicResearch vol 40 no 5 pp 567ndash578 2002
[15] L M Gao K T Li B Liu and J Su ldquoGeneral integralconservation form of Navier-Stokes equations and numericalapplicationrdquo Chinese Journal of Computational Physics vol 25no 2 pp 172ndash178 2008
[16] S Patankar Numerical Heat Transfer and Fluid Flow Hemi-sphere Publishing 1980
[17] A West and T Fuller ldquoInfluence of rib spacing in proton-exchange membrane electrode assembliesrdquo Journal of AppliedElectrochemistry vol 26 no 6 pp 557ndash565 1996
[18] B P Leonard ldquoComparison of truncation error of finite-difference and finite-volume formulations of convection termsrdquoApplied Mathematical Modelling vol 18 no 1 pp 46ndash50 1994
[19] G G Botte J A Ritter and R E White ldquoComparison of finitedifference and control volume methods for solving differentialequationsrdquoComputers and Chemical Engineering vol 24 no 12pp 2633ndash2654 2000
[20] AThom ldquoThe flow past cylinders at low speedsrdquo Proceedings ofthe Royal Society A vol 141 pp 651ndash659 1993
[21] L C Woods ldquoA note on the numerical solution of fourth orderdifferential equationsrdquo vol 5 pp 176ndash182 1954
[22] V G Jenson ldquoViscous flow round a sphere at low Reynoldsnumbers (Re iexcl 40)rdquo Proceedings of the Royal Society A vol 249pp 346ndash366 1959
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
4 Journal of Applied Mathematics
34 Deferred Correction When MAC is adopted momen-tum and energy equations are both solved in explicit schemewhile in vorticity-stream function method the convectionterm is treated in implicit scheme by deferred correctionwhich can guarantee the main-diagonal domination of thesolvingmatrix and thus guarantee the calculation stability Asto FVM taking Φ
119890as example the expressions of deferred
correction are as follows
when 119880119890ge 0 Φ
119890= Φ119875+ (Φ+
119890minus Φ119875)lowast
when 119880119890lt 0 Φ
119890= Φ119864+ (Φminus
119890minus Φ119864)lowast
(12)
where ldquolowastrdquo in these equations stands for the former calculationstep
As to FDM similarly to the treatment of Φ119890in FVM the
deferred correction for node 119875 can be written as the sum offirst-order upwind gradient and correction gradient which isshown as follows
when 119880119875ge 0 119880
119875
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
= 119880119875(Φ119875minus Φ119882
Δ119883)
+ [119880119875(120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119882
Δ119883)]
lowast
(13)
when 119880119875lt 0 119880
119875
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
= 119880119875(Φ119864minus Φ119875
Δ119883)
+ [119880119875(120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119864minus Φ119875
Δ119883)]
lowast
(14)
The first terms in the right side of (12)ndash(14) are all first-orderupwind term The second terms are incorporated into thesource term as the form of correction term The discretizedequation is solved by G-S solver with underrelaxation ofwhich the discrete expression is shown as (15) and thediscretized equations of FVM and FDM are shown in (16a)(16b) (17a) and (17b) respectively
119886119875Φ119875= 119886119864Φ119864+ 119886119882Φ119882+ 119886119873Φ119873+ 119886119878Φ119878+ 119887 (15)
119886119864=max (minus119880
119890 0)
Δ119883+
1
ReΔ1198832
119886119882=max (119880
119908 0)
Δ119883+
1
ReΔ1198832
119886119873=max (minus119881
119899 0)
Δ119884+
1
ReΔ1198842
119886119878=max (119881
119904 0)
Δ119884+
1
ReΔ1198842
119886119875=
max (119880119890 0)
Δ119883+max (minus119880
119908 0)
Δ119883+max (119881
119899 0)
Δ119884
+max (minus119881
119904 0)
Δ119884+
2
ReΔ1198832+
2
ReΔ1198842
(16a)
119887 = minusmax (119880119890 0) [
Φ+
119890minus Φ119875
Δ119883] +max (minus119880
119890 0) [
Φminus
119890minus Φ119864
Δ119883]
+max (119880119908 0)[
Φ+
119908minus Φ119882
Δ119883]minusmax (minus119880
119908 0)[
Φminus
119908minus Φ119875
Δ119883]
minusmax (119881119899 0) [
Φ+
119899minus Φ119875
Δ119884] +max (minus119881
119899 0) [
Φminus
119899minus Φ119873
Δ119884]
+max (119881119904 0) [
Φ+
119904minus Φ119878
Δ119884] minusmax (minus119881
119904 0) [
Φminus
119904minus Φ119875
Δ119884]
(16b)
119886119864=max (minus119880
119875 0)
Δ119883+
1
ReΔ1198832
119886119882=max (119880
119875 0)
Δ119883+
1
ReΔ1198832
119886119873=max (minus119881
119875 0)
Δ119884+
1
ReΔ1198842
119886119878=max (119881
119875 0)
Δ119884+
1
ReΔ1198842
119886119875=
max (minus119880119875 0)
Δ119883+max (119880
119875 0)
Δ119883+max (minus119881
119875 0)
Δ119884
+max (119881
119875 0)
Δ119884+
2
ReΔ1198832+
2
ReΔ1198842
(17a)
119887 = minusmax (119880119875 0) [
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119882
Δ119883]
+max (minus119880119875 0) [
120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119864minus Φ119875
Δ119883]
minusmax (119881119875 0) [
120597Φ
120597119884
10038161003816100381610038161003816100381610038161003816
+
119875
minusΦ119875minus Φ119878
Δ119884]
+max (minus119881119875 0) [
120597Φ
120597119884
10038161003816100381610038161003816100381610038161003816
minus
119875
minusΦ119873minus Φ119875
Δ119883]
(17b)
where 119886119873 119886119875 119886119864 119886119882 and 119886
119878are coefficients in discretized
equations and 119887 is the source term in discretized equations
4 Results Comparison and Analysis
In this section the accuracy stability of convection termrobustness and calculation efficiency of variant algorithmsschemes and treatments of boundary condition which havebeen discussed previously are systemically compared andanalyzed based on the calculation results of lid-driven flowand natural convection flow in a square cavity
41 Accuracy Firstly the accuracy of FVM and FDM isevaluated by comparing benchmark solution with the results
Journal of Applied Mathematics 5
Table 4 Truncation error of FVM and FDMgoverning equations for vorticity-stream function algorithm of lid driven flow in a square cavity
Convection item Diffusion item
FVM (format)Φ119890
Δ119883minusΦ119908
Δ119883
1
Δ119883
120597Φ
120597119909
10038161003816100381610038161003816100381610038161003816119890
minus1
Δ119883
120597Φ
120597119909
10038161003816100381610038161003816100381610038161003816119908
FVM (TE) minusΔ1198832
8Φ101584010158401015840
119875minusΔ1198834
128Φ(5)
119875minus sdot sdot sdot minus
Δ1198832
24Φ(4)
119875minus 13
Δ1198834
5760Φ(6)
119875minus sdot sdot sdot
FDM (format) 120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816119875
1205972Φ
1205971198832
100381610038161003816100381610038161003816100381610038161003816119875
FDM (TE) minusΔ1198832
6Φ101584010158401015840
119875minusΔ1198834
120Φ(5)
119875minus sdot sdot sdot minus
Δ1198832
12Φ(4)
119875minusΔ1198834
360Φ(6)
119875minus sdot sdot sdot
obtained from variant mesh density discrete schemes andboundary conditions Through precision analysis FVM andFDM are respectively adopted with the Taylor expansioncommenced at midpoint (P node) and their respectivetruncation error (TE) is just as shown Table 4 [19] Thetruncation errors of FVM are obviously less than the FDMrsquos
411 Influence of Mesh Density In [4] it is indicated thatthe conservative property can affect the accuracy while thenonconservative FDM loses its conservative property in thesituation when the grid is too small Tomake further analysisthe calculation results obtained by FVM and FDM withvariant mesh densities are compared The results of thevelocity along the horizontal central axisV and velocity alongthe vertical central axis U obtained by FDM and FVM withcentral difference scheme of MAC algorithm are comparedand illustrated in Figure 3
From Figure 3 we can see that when the mesh densityis taken as 80 times 80 the results obtained by FVM and FDMboth agree well with the benchmark solutionWhen themeshdensity is reduced to 40 times 40 the results obtained by FVMare closer to the benchmark solution and if the mesh densityis further reduced to 20 times 20 the results obtained by FDMdiffer from the benchmark solution greatly in which FVMshows its advantage in accuracyThis phenomenon is in goodagreement with the theoretical analysis in [18] that is theconservative property of FVM better ensures the accuracyof calculation Therefore whether the conservative propertycan be realized is of much significance to the accuracy of theobtained result
412 Influence of Boundary Condition In the followingsection FVM and FDM with variant boundary conditionsare compared from the perspective of accuracy by vorticity-stream function method The results of the velocity alongthe horizontal central axis V and velocity along the verticalcentral axis U obtained by FDM and FVM with centraldifference scheme are compared and illustrated in Figure 4and the influence of boundary condition of vorticity is alsogiven From Figure 4 we can see that no matter whichtreatment of boundary condition is adopted the accuracy ofFVM is higher than that of FDM in vorticity-stream functionmethod
413 Influence of Discrete Scheme The selection of discreteforms is also influential to the accuracy Thus the resultsobtained with variant discrete forms are studied Taking thevorticity-stream function with boundary condition treatedby first-order Thom equation as example the results of thevelocity along the horizontal central axis V and velocityalong the vertical central axis U obtained by FDM and FVMwith variant discrete schemes are compared and illustratedin Figure 5 from which we can see that whether second-order central difference or second-order upwind scheme isadopted the accuracy of FVM is higher than that of FDM thecomparison between (a) and (b) indicates that second-orderupwind difference possesses higher accuracy than second-order central difference
414 Influence of Physical Problem Finally two methods invariant physical problems are compared from the perspectiveof accuracy based on the convection flow in a square cavityTaking the result obtained by MAC algorithm as exampleFigure 6 shows the comparison of results of the velocity alongthe horizontal central axis V and velocity along the verticalcentral axisU obtained by FDMandFVMTemperature fieldsobtained by FVM and FDM are compared with the grid-independent solution (obtained with grid of 256 times 256) asFigure 7 shows
From the comparison in Figure 6 and Figure 7 we cansee that based on the results obtained by MAC of naturalconvection flow in a square cavity the accuracy of FVM isgenerally superior to that of FDM Thus it is indicated bycomparison that no matter which mesh density scheme andtreatment of boundary condition is adopted FVM showsbetter accuracy than FDM
42 Stability of Convection Term Stability of mathematicalmeaning can only ensure that the oscillation of solution iscontrolled within a range but cannot eliminate the absenceof oscillation The amplification of oscillation may result indivergence Stability of convection term originates from var-ied difference schemes for convective itemand is independentof the introducing of rounding error in the calculation Thisstability is the key point to obtain a solution with physicalmeaning It is concluded that all the unstable schemes willresult in the oscillation of solution Stability condition of
6 Journal of Applied Mathematics
0
04
08
12
0
03
06
U V
X(Y)
0 02 04 06 08 1minus04
minus06
minus03
(a) 80 times 80
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) 40 times 40
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(c) 20 times 20
Figure 3 Velocity on the central axis of lid-driven flow obtained by MAC with variant mesh densities
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(a) Thom
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Woods
Figure 4 Velocity along the central axis obtained by second-order central difference scheme with two treatments of boundary conditionwhen the mesh density is 40 times 40
Journal of Applied Mathematics 7
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12U
minus04
0
03
06
V
minus06
minus03
(a) Second-order central difference scheme
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Second-order upwind difference scheme
Figure 5 Velocity along the central axis obtained by vorticity-stream function method for lid-driven flow with different schemes when themesh density is 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(a) 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(b) 80 times 80
Figure 6 Velocity along the central axis of convection flow in a square cavity with variant gird densities
40 times 40
Grid-independent
(a) 40 times 40
80 times 80
Grid-independent
(b) 80 times 80
Figure 7 Comparison of temperature contours of natural convection flow in a square cavity with variant mesh densities (FVM in the leftcolumn and FDM in the right column)
8 Journal of Applied Mathematics
(a) 40 times 40 (b) 20 times 20
Figure 8 Comparison of stream lines in lid-driven flow with variant mesh densities (FVM in the left column and FDM in the right column)
(a) Thom (b) Woods
(c) Jenson
Figure 9 Stream functions of lid-driven flow under variant boundary treatments with the grid of 20 times 20 (FVM in the left column and FDMin the right column)
the convection item (Mesh Peclet number) is affected bythe following factors such as mesh density and boundarycondition Taking lid-driven flow as example the stabilityof convection term of FVM and FDM with variant meshdensities treatments of boundary condition is compared
421 Influence of Mesh Density Figure 8 shows the compar-ison of stream lines between FVM and FDM with variantmesh densities With the grid of 40 times 40 the two methodsboth can capture the structure of primary vortex and sec-ondary vortex as Figure 8(a) shows It is seen in Figure 8(b)that when the mesh of 20 times 20 is adopted concussionphenomenon appears which ismore serious in FDM and thelocation of the vortex center obviously moves upward
422 Influence of Boundary Condition Similarly to the accu-racy analysis the results obtained by variant treatments for
vorticity of wall boundary are compared as Figure 9 showsThree treatments Thom Woods and Jenson are analyzedFrom Figure 9 we can see that only when Jenson equationis adopted the results obtained by FVM and FDM arerelatively close while when Thom and Woods are adoptedFVM shows better results than FDM The results obtainedby FDM suffer severe oscillation and even lack physicalmeaning It is clear that based on the lid-driven flow stabilityof convection term of FVM is superior to that of FDMThe influence from the same boundary error (truncationerror and rounding error) on difference solving methodsis illustrated in Figure 9 With the treatment of boundarycondition with the same truncation error or rounding errorthe stability of FVM is better than that of FDM especiallyin the situation with boundary treatment of (a) and (b)with relatively low precision truncation error FDM is veryunstable and very easy to slip into oscillation
Journal of Applied Mathematics 9
120572
02 04 06 08 1
Nite
r
80 times 80
40 times 4020 times 20
FVM FDM
5
4
3
2
1
0
times104
(a) Second-order central difference scheme
02 04 06 08 1120572
Nite
r
5
4
3
2
1
0
80 times 80
40 times 4020 times 20
FVM FDM
times104
(b) Second-order upwind scheme
Figure 10 Comparison of the robustness of lid-driven flow with variant schemes
43 Robustness Robustness is a very critical index to evaluatethe numerical approaches In this section the robustness ofFVM and FDM is compared based on the lid-driven flowproblem which is solved by vorticity-stream function Gridsof 20 times 20 40 times 40 and 80 times 80 are adopted respectively inthe calculation Relaxation factor is tentatively taken up from02 to the maximum which can ensure the convergence ofcalculation Iterations are correspondingly recorded as well
It is shown in Figure 10 when the whether the cen-tral difference scheme or second-order upwind scheme isadopted with the refining of calculation grids the maximumrelaxation factor (120572) increases in both FVM and FDM therobustness is enhanced and the iterations (Niter) corre-spondingly decrease as well
As Figure 10(a) shows in the situation that central differ-ence scheme is adopted with the grid of 20 times 20 and 40 times 40the maximum relaxation factor of FVM is bigger than thatof FVM while with the grid of 80 times 80 the relaxation factorcan be taken as 10 in bothmethods Similarly as Figure 10(b)shows generally speaking when second-order upwind dif-ference scheme is adopted the maximum relaxation factor isbigger than that of FDM Based on the result of lid-drivenflow the conservative FVM is superior to the nonconservativeFDM from the perspective of numerical robustness
In the numerical simulation of complex flow field cal-culation efficiency is another critical index on which thispapermakes analysis FromFigure 10we can see that with theincreasing of relaxation factor the iterations of two methodsboth decrease generally What needs to be notified is that theiterations of FVM are smaller than those of FDM
44 Efficiency In the numerical simulation of complex flowfield calculation efficiency is another critical index on whichthis paper makes analysis
As is shown in Table 5 although the calculation timerequired by every iteration step in FVM and FDM is almostthe same with the increasing of mesh numbers calculationtime required by FVM slightly decreases Thus total calcula-tion time of FVM is much smaller than that of FDMThat isFVM is much more efficient that FDM Thus it is validatedthat the efficiency of FVM is better than that of FDM in thesame calculation condition
5 Conclusions
Based on vorticity-stream function and MAC algorithmlid-driven flow and natural convection in a square cavityare calculated The accuracy stability of convective termrobustness of FVM and FDM with variant mesh densitiesdiscrete forms and treatments of boundary condition arecompared The following conclusions are drawn
(1) No matter which algorithm discrete form or treat-ment of boundary condition is adopted the resultsobtained by conservative FVM are closer to bench-mark solution than those obtained by FDM Theaccuracy of FVM is higher than FDM especiallywhen the mesh density is relatively small
(2) The results of lid-driven flow indicate that the resultsobtained by FDM show more serious oscillation andthe results obtained even lose physicalmeaningThusthe stability of convection term of FVM is superiorthan that of FDM
(3) As to the lid-driven flow it is indicated that whenvorticity-stream function method is adopted therelaxation factor of conservative method is biggerthan that of the nonconservative one while theiterations are less which shows that the robustness
10 Journal of Applied Mathematics
Table 5 Computation iterations and time of FVM and FDM in second-order central difference scheme for vorticity-stream functionalgorithm of lid driven flow
Mesh 120572FVM FDM
Iterations Time(s) Time of each iteration Iterations Time(s) Time of each iteration20 02 4950 738 0015 2100 307 001520 01 10500 1544 0015 4300 616 001440 06 3250 1519 0047 4350 1993 004640 05 4450 2035 0046 5800 2629 004540 04 6300 2963 0047 7950 3693 004680 08 4950 8223 0166 5050 8584 017080 06 8500 14134 0166 9000 15395 017180 04 17250 29694 0172 17750 32420 0183
and calculation efficiency of conservative method arebetter than those of the nonconservative method
(4) In the lid-driven flow with vorticity-stream functionmethod the calculation time required by every itera-tion step of FVM is slightly smaller than that of FDMTogether with the smaller total iteration steps FVMis more efficient than FDM
Acknowledgment
The study is supported by the National Science Foundation ofChina (nos 51134006 51176204 and 51276198)
References
[1] J D Anderson Jr Computational Fluid Dynamics The BasicsWith Applications McGraw-Hill 1995
[2] P J Roache Computational Fluid Dynamics Hermosa 1976[3] W Q Tao Heat Transfer Science Press 2001[4] W Q Tao Advances in Computational Heat Transfer Science
Press 2000[5] Y Jaluria and K E Torrance Computational Heat Transfer
Hemisphere Publishing Corporation 1986[6] P D Lax ldquoWeak solutions of nonlinear hyperbolic equations
and their numerical computationrdquo Communications on Pureand Applied Mathematics vol 7 pp 159ndash193 1954
[7] P Lax and B Wendroff ldquoSystems of conservation lawsrdquo Com-munications on Pure and Applied Mathematics vol 13 no 2 pp217ndash237 1960
[8] A Jameson ldquoTransonic potential flow calculations using con-servation formrdquo in Proceedings of the Computational fluiddynamics conference pp 148ndash161 Hartford Conn USA 1975
[9] M Vinokur ldquoConservation equations of gasdynamics in curvi-linear coordinate systemsrdquo Journal of Computational Physicsvol 14 pp 105ndash125 1974
[10] T J Bridges ldquoConservation laws in curvilinear coordinatesa short proof of Vinokurrsquos theorem using differential formsrdquoApplied Mathematics and Computation vol 202 no 2 pp 882ndash885 2008
[11] R J DiPerna ldquoDecay of solutions of hyperbolic systems ofconservation laws with a convex extensionrdquoArchive for RationalMechanics and Analysis vol 64 no 1 pp 1ndash46 1977
[12] A Bressan ldquoHyperbolic systems of conservation lawsrdquo RevistaMatematica Complutense vol 12 no 1 pp 135ndash198 1999
[13] R K S Hankin ldquoThe Euler equations for multiphase com-pressible flow in conservation form-Simulation of shock-bubbleinteractionsrdquo Journal of Computational Physics vol 172 no 2pp 808ndash826 2001
[14] C Lai R A Baltzer andRW Schaffranek ldquoConservation-formequations of unsteady open-channel flowrdquo Journal of HydraulicResearch vol 40 no 5 pp 567ndash578 2002
[15] L M Gao K T Li B Liu and J Su ldquoGeneral integralconservation form of Navier-Stokes equations and numericalapplicationrdquo Chinese Journal of Computational Physics vol 25no 2 pp 172ndash178 2008
[16] S Patankar Numerical Heat Transfer and Fluid Flow Hemi-sphere Publishing 1980
[17] A West and T Fuller ldquoInfluence of rib spacing in proton-exchange membrane electrode assembliesrdquo Journal of AppliedElectrochemistry vol 26 no 6 pp 557ndash565 1996
[18] B P Leonard ldquoComparison of truncation error of finite-difference and finite-volume formulations of convection termsrdquoApplied Mathematical Modelling vol 18 no 1 pp 46ndash50 1994
[19] G G Botte J A Ritter and R E White ldquoComparison of finitedifference and control volume methods for solving differentialequationsrdquoComputers and Chemical Engineering vol 24 no 12pp 2633ndash2654 2000
[20] AThom ldquoThe flow past cylinders at low speedsrdquo Proceedings ofthe Royal Society A vol 141 pp 651ndash659 1993
[21] L C Woods ldquoA note on the numerical solution of fourth orderdifferential equationsrdquo vol 5 pp 176ndash182 1954
[22] V G Jenson ldquoViscous flow round a sphere at low Reynoldsnumbers (Re iexcl 40)rdquo Proceedings of the Royal Society A vol 249pp 346ndash366 1959
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 5
Table 4 Truncation error of FVM and FDMgoverning equations for vorticity-stream function algorithm of lid driven flow in a square cavity
Convection item Diffusion item
FVM (format)Φ119890
Δ119883minusΦ119908
Δ119883
1
Δ119883
120597Φ
120597119909
10038161003816100381610038161003816100381610038161003816119890
minus1
Δ119883
120597Φ
120597119909
10038161003816100381610038161003816100381610038161003816119908
FVM (TE) minusΔ1198832
8Φ101584010158401015840
119875minusΔ1198834
128Φ(5)
119875minus sdot sdot sdot minus
Δ1198832
24Φ(4)
119875minus 13
Δ1198834
5760Φ(6)
119875minus sdot sdot sdot
FDM (format) 120597Φ
120597119883
10038161003816100381610038161003816100381610038161003816119875
1205972Φ
1205971198832
100381610038161003816100381610038161003816100381610038161003816119875
FDM (TE) minusΔ1198832
6Φ101584010158401015840
119875minusΔ1198834
120Φ(5)
119875minus sdot sdot sdot minus
Δ1198832
12Φ(4)
119875minusΔ1198834
360Φ(6)
119875minus sdot sdot sdot
obtained from variant mesh density discrete schemes andboundary conditions Through precision analysis FVM andFDM are respectively adopted with the Taylor expansioncommenced at midpoint (P node) and their respectivetruncation error (TE) is just as shown Table 4 [19] Thetruncation errors of FVM are obviously less than the FDMrsquos
411 Influence of Mesh Density In [4] it is indicated thatthe conservative property can affect the accuracy while thenonconservative FDM loses its conservative property in thesituation when the grid is too small Tomake further analysisthe calculation results obtained by FVM and FDM withvariant mesh densities are compared The results of thevelocity along the horizontal central axisV and velocity alongthe vertical central axis U obtained by FDM and FVM withcentral difference scheme of MAC algorithm are comparedand illustrated in Figure 3
From Figure 3 we can see that when the mesh densityis taken as 80 times 80 the results obtained by FVM and FDMboth agree well with the benchmark solutionWhen themeshdensity is reduced to 40 times 40 the results obtained by FVMare closer to the benchmark solution and if the mesh densityis further reduced to 20 times 20 the results obtained by FDMdiffer from the benchmark solution greatly in which FVMshows its advantage in accuracyThis phenomenon is in goodagreement with the theoretical analysis in [18] that is theconservative property of FVM better ensures the accuracyof calculation Therefore whether the conservative propertycan be realized is of much significance to the accuracy of theobtained result
412 Influence of Boundary Condition In the followingsection FVM and FDM with variant boundary conditionsare compared from the perspective of accuracy by vorticity-stream function method The results of the velocity alongthe horizontal central axis V and velocity along the verticalcentral axis U obtained by FDM and FVM with centraldifference scheme are compared and illustrated in Figure 4and the influence of boundary condition of vorticity is alsogiven From Figure 4 we can see that no matter whichtreatment of boundary condition is adopted the accuracy ofFVM is higher than that of FDM in vorticity-stream functionmethod
413 Influence of Discrete Scheme The selection of discreteforms is also influential to the accuracy Thus the resultsobtained with variant discrete forms are studied Taking thevorticity-stream function with boundary condition treatedby first-order Thom equation as example the results of thevelocity along the horizontal central axis V and velocityalong the vertical central axis U obtained by FDM and FVMwith variant discrete schemes are compared and illustratedin Figure 5 from which we can see that whether second-order central difference or second-order upwind scheme isadopted the accuracy of FVM is higher than that of FDM thecomparison between (a) and (b) indicates that second-orderupwind difference possesses higher accuracy than second-order central difference
414 Influence of Physical Problem Finally two methods invariant physical problems are compared from the perspectiveof accuracy based on the convection flow in a square cavityTaking the result obtained by MAC algorithm as exampleFigure 6 shows the comparison of results of the velocity alongthe horizontal central axis V and velocity along the verticalcentral axisU obtained by FDMandFVMTemperature fieldsobtained by FVM and FDM are compared with the grid-independent solution (obtained with grid of 256 times 256) asFigure 7 shows
From the comparison in Figure 6 and Figure 7 we cansee that based on the results obtained by MAC of naturalconvection flow in a square cavity the accuracy of FVM isgenerally superior to that of FDM Thus it is indicated bycomparison that no matter which mesh density scheme andtreatment of boundary condition is adopted FVM showsbetter accuracy than FDM
42 Stability of Convection Term Stability of mathematicalmeaning can only ensure that the oscillation of solution iscontrolled within a range but cannot eliminate the absenceof oscillation The amplification of oscillation may result indivergence Stability of convection term originates from var-ied difference schemes for convective itemand is independentof the introducing of rounding error in the calculation Thisstability is the key point to obtain a solution with physicalmeaning It is concluded that all the unstable schemes willresult in the oscillation of solution Stability condition of
6 Journal of Applied Mathematics
0
04
08
12
0
03
06
U V
X(Y)
0 02 04 06 08 1minus04
minus06
minus03
(a) 80 times 80
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) 40 times 40
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(c) 20 times 20
Figure 3 Velocity on the central axis of lid-driven flow obtained by MAC with variant mesh densities
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(a) Thom
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Woods
Figure 4 Velocity along the central axis obtained by second-order central difference scheme with two treatments of boundary conditionwhen the mesh density is 40 times 40
Journal of Applied Mathematics 7
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12U
minus04
0
03
06
V
minus06
minus03
(a) Second-order central difference scheme
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Second-order upwind difference scheme
Figure 5 Velocity along the central axis obtained by vorticity-stream function method for lid-driven flow with different schemes when themesh density is 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(a) 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(b) 80 times 80
Figure 6 Velocity along the central axis of convection flow in a square cavity with variant gird densities
40 times 40
Grid-independent
(a) 40 times 40
80 times 80
Grid-independent
(b) 80 times 80
Figure 7 Comparison of temperature contours of natural convection flow in a square cavity with variant mesh densities (FVM in the leftcolumn and FDM in the right column)
8 Journal of Applied Mathematics
(a) 40 times 40 (b) 20 times 20
Figure 8 Comparison of stream lines in lid-driven flow with variant mesh densities (FVM in the left column and FDM in the right column)
(a) Thom (b) Woods
(c) Jenson
Figure 9 Stream functions of lid-driven flow under variant boundary treatments with the grid of 20 times 20 (FVM in the left column and FDMin the right column)
the convection item (Mesh Peclet number) is affected bythe following factors such as mesh density and boundarycondition Taking lid-driven flow as example the stabilityof convection term of FVM and FDM with variant meshdensities treatments of boundary condition is compared
421 Influence of Mesh Density Figure 8 shows the compar-ison of stream lines between FVM and FDM with variantmesh densities With the grid of 40 times 40 the two methodsboth can capture the structure of primary vortex and sec-ondary vortex as Figure 8(a) shows It is seen in Figure 8(b)that when the mesh of 20 times 20 is adopted concussionphenomenon appears which ismore serious in FDM and thelocation of the vortex center obviously moves upward
422 Influence of Boundary Condition Similarly to the accu-racy analysis the results obtained by variant treatments for
vorticity of wall boundary are compared as Figure 9 showsThree treatments Thom Woods and Jenson are analyzedFrom Figure 9 we can see that only when Jenson equationis adopted the results obtained by FVM and FDM arerelatively close while when Thom and Woods are adoptedFVM shows better results than FDM The results obtainedby FDM suffer severe oscillation and even lack physicalmeaning It is clear that based on the lid-driven flow stabilityof convection term of FVM is superior to that of FDMThe influence from the same boundary error (truncationerror and rounding error) on difference solving methodsis illustrated in Figure 9 With the treatment of boundarycondition with the same truncation error or rounding errorthe stability of FVM is better than that of FDM especiallyin the situation with boundary treatment of (a) and (b)with relatively low precision truncation error FDM is veryunstable and very easy to slip into oscillation
Journal of Applied Mathematics 9
120572
02 04 06 08 1
Nite
r
80 times 80
40 times 4020 times 20
FVM FDM
5
4
3
2
1
0
times104
(a) Second-order central difference scheme
02 04 06 08 1120572
Nite
r
5
4
3
2
1
0
80 times 80
40 times 4020 times 20
FVM FDM
times104
(b) Second-order upwind scheme
Figure 10 Comparison of the robustness of lid-driven flow with variant schemes
43 Robustness Robustness is a very critical index to evaluatethe numerical approaches In this section the robustness ofFVM and FDM is compared based on the lid-driven flowproblem which is solved by vorticity-stream function Gridsof 20 times 20 40 times 40 and 80 times 80 are adopted respectively inthe calculation Relaxation factor is tentatively taken up from02 to the maximum which can ensure the convergence ofcalculation Iterations are correspondingly recorded as well
It is shown in Figure 10 when the whether the cen-tral difference scheme or second-order upwind scheme isadopted with the refining of calculation grids the maximumrelaxation factor (120572) increases in both FVM and FDM therobustness is enhanced and the iterations (Niter) corre-spondingly decrease as well
As Figure 10(a) shows in the situation that central differ-ence scheme is adopted with the grid of 20 times 20 and 40 times 40the maximum relaxation factor of FVM is bigger than thatof FVM while with the grid of 80 times 80 the relaxation factorcan be taken as 10 in bothmethods Similarly as Figure 10(b)shows generally speaking when second-order upwind dif-ference scheme is adopted the maximum relaxation factor isbigger than that of FDM Based on the result of lid-drivenflow the conservative FVM is superior to the nonconservativeFDM from the perspective of numerical robustness
In the numerical simulation of complex flow field cal-culation efficiency is another critical index on which thispapermakes analysis FromFigure 10we can see that with theincreasing of relaxation factor the iterations of two methodsboth decrease generally What needs to be notified is that theiterations of FVM are smaller than those of FDM
44 Efficiency In the numerical simulation of complex flowfield calculation efficiency is another critical index on whichthis paper makes analysis
As is shown in Table 5 although the calculation timerequired by every iteration step in FVM and FDM is almostthe same with the increasing of mesh numbers calculationtime required by FVM slightly decreases Thus total calcula-tion time of FVM is much smaller than that of FDMThat isFVM is much more efficient that FDM Thus it is validatedthat the efficiency of FVM is better than that of FDM in thesame calculation condition
5 Conclusions
Based on vorticity-stream function and MAC algorithmlid-driven flow and natural convection in a square cavityare calculated The accuracy stability of convective termrobustness of FVM and FDM with variant mesh densitiesdiscrete forms and treatments of boundary condition arecompared The following conclusions are drawn
(1) No matter which algorithm discrete form or treat-ment of boundary condition is adopted the resultsobtained by conservative FVM are closer to bench-mark solution than those obtained by FDM Theaccuracy of FVM is higher than FDM especiallywhen the mesh density is relatively small
(2) The results of lid-driven flow indicate that the resultsobtained by FDM show more serious oscillation andthe results obtained even lose physicalmeaningThusthe stability of convection term of FVM is superiorthan that of FDM
(3) As to the lid-driven flow it is indicated that whenvorticity-stream function method is adopted therelaxation factor of conservative method is biggerthan that of the nonconservative one while theiterations are less which shows that the robustness
10 Journal of Applied Mathematics
Table 5 Computation iterations and time of FVM and FDM in second-order central difference scheme for vorticity-stream functionalgorithm of lid driven flow
Mesh 120572FVM FDM
Iterations Time(s) Time of each iteration Iterations Time(s) Time of each iteration20 02 4950 738 0015 2100 307 001520 01 10500 1544 0015 4300 616 001440 06 3250 1519 0047 4350 1993 004640 05 4450 2035 0046 5800 2629 004540 04 6300 2963 0047 7950 3693 004680 08 4950 8223 0166 5050 8584 017080 06 8500 14134 0166 9000 15395 017180 04 17250 29694 0172 17750 32420 0183
and calculation efficiency of conservative method arebetter than those of the nonconservative method
(4) In the lid-driven flow with vorticity-stream functionmethod the calculation time required by every itera-tion step of FVM is slightly smaller than that of FDMTogether with the smaller total iteration steps FVMis more efficient than FDM
Acknowledgment
The study is supported by the National Science Foundation ofChina (nos 51134006 51176204 and 51276198)
References
[1] J D Anderson Jr Computational Fluid Dynamics The BasicsWith Applications McGraw-Hill 1995
[2] P J Roache Computational Fluid Dynamics Hermosa 1976[3] W Q Tao Heat Transfer Science Press 2001[4] W Q Tao Advances in Computational Heat Transfer Science
Press 2000[5] Y Jaluria and K E Torrance Computational Heat Transfer
Hemisphere Publishing Corporation 1986[6] P D Lax ldquoWeak solutions of nonlinear hyperbolic equations
and their numerical computationrdquo Communications on Pureand Applied Mathematics vol 7 pp 159ndash193 1954
[7] P Lax and B Wendroff ldquoSystems of conservation lawsrdquo Com-munications on Pure and Applied Mathematics vol 13 no 2 pp217ndash237 1960
[8] A Jameson ldquoTransonic potential flow calculations using con-servation formrdquo in Proceedings of the Computational fluiddynamics conference pp 148ndash161 Hartford Conn USA 1975
[9] M Vinokur ldquoConservation equations of gasdynamics in curvi-linear coordinate systemsrdquo Journal of Computational Physicsvol 14 pp 105ndash125 1974
[10] T J Bridges ldquoConservation laws in curvilinear coordinatesa short proof of Vinokurrsquos theorem using differential formsrdquoApplied Mathematics and Computation vol 202 no 2 pp 882ndash885 2008
[11] R J DiPerna ldquoDecay of solutions of hyperbolic systems ofconservation laws with a convex extensionrdquoArchive for RationalMechanics and Analysis vol 64 no 1 pp 1ndash46 1977
[12] A Bressan ldquoHyperbolic systems of conservation lawsrdquo RevistaMatematica Complutense vol 12 no 1 pp 135ndash198 1999
[13] R K S Hankin ldquoThe Euler equations for multiphase com-pressible flow in conservation form-Simulation of shock-bubbleinteractionsrdquo Journal of Computational Physics vol 172 no 2pp 808ndash826 2001
[14] C Lai R A Baltzer andRW Schaffranek ldquoConservation-formequations of unsteady open-channel flowrdquo Journal of HydraulicResearch vol 40 no 5 pp 567ndash578 2002
[15] L M Gao K T Li B Liu and J Su ldquoGeneral integralconservation form of Navier-Stokes equations and numericalapplicationrdquo Chinese Journal of Computational Physics vol 25no 2 pp 172ndash178 2008
[16] S Patankar Numerical Heat Transfer and Fluid Flow Hemi-sphere Publishing 1980
[17] A West and T Fuller ldquoInfluence of rib spacing in proton-exchange membrane electrode assembliesrdquo Journal of AppliedElectrochemistry vol 26 no 6 pp 557ndash565 1996
[18] B P Leonard ldquoComparison of truncation error of finite-difference and finite-volume formulations of convection termsrdquoApplied Mathematical Modelling vol 18 no 1 pp 46ndash50 1994
[19] G G Botte J A Ritter and R E White ldquoComparison of finitedifference and control volume methods for solving differentialequationsrdquoComputers and Chemical Engineering vol 24 no 12pp 2633ndash2654 2000
[20] AThom ldquoThe flow past cylinders at low speedsrdquo Proceedings ofthe Royal Society A vol 141 pp 651ndash659 1993
[21] L C Woods ldquoA note on the numerical solution of fourth orderdifferential equationsrdquo vol 5 pp 176ndash182 1954
[22] V G Jenson ldquoViscous flow round a sphere at low Reynoldsnumbers (Re iexcl 40)rdquo Proceedings of the Royal Society A vol 249pp 346ndash366 1959
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
6 Journal of Applied Mathematics
0
04
08
12
0
03
06
U V
X(Y)
0 02 04 06 08 1minus04
minus06
minus03
(a) 80 times 80
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) 40 times 40
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(c) 20 times 20
Figure 3 Velocity on the central axis of lid-driven flow obtained by MAC with variant mesh densities
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(a) Thom
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Woods
Figure 4 Velocity along the central axis obtained by second-order central difference scheme with two treatments of boundary conditionwhen the mesh density is 40 times 40
Journal of Applied Mathematics 7
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12U
minus04
0
03
06
V
minus06
minus03
(a) Second-order central difference scheme
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Second-order upwind difference scheme
Figure 5 Velocity along the central axis obtained by vorticity-stream function method for lid-driven flow with different schemes when themesh density is 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(a) 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(b) 80 times 80
Figure 6 Velocity along the central axis of convection flow in a square cavity with variant gird densities
40 times 40
Grid-independent
(a) 40 times 40
80 times 80
Grid-independent
(b) 80 times 80
Figure 7 Comparison of temperature contours of natural convection flow in a square cavity with variant mesh densities (FVM in the leftcolumn and FDM in the right column)
8 Journal of Applied Mathematics
(a) 40 times 40 (b) 20 times 20
Figure 8 Comparison of stream lines in lid-driven flow with variant mesh densities (FVM in the left column and FDM in the right column)
(a) Thom (b) Woods
(c) Jenson
Figure 9 Stream functions of lid-driven flow under variant boundary treatments with the grid of 20 times 20 (FVM in the left column and FDMin the right column)
the convection item (Mesh Peclet number) is affected bythe following factors such as mesh density and boundarycondition Taking lid-driven flow as example the stabilityof convection term of FVM and FDM with variant meshdensities treatments of boundary condition is compared
421 Influence of Mesh Density Figure 8 shows the compar-ison of stream lines between FVM and FDM with variantmesh densities With the grid of 40 times 40 the two methodsboth can capture the structure of primary vortex and sec-ondary vortex as Figure 8(a) shows It is seen in Figure 8(b)that when the mesh of 20 times 20 is adopted concussionphenomenon appears which ismore serious in FDM and thelocation of the vortex center obviously moves upward
422 Influence of Boundary Condition Similarly to the accu-racy analysis the results obtained by variant treatments for
vorticity of wall boundary are compared as Figure 9 showsThree treatments Thom Woods and Jenson are analyzedFrom Figure 9 we can see that only when Jenson equationis adopted the results obtained by FVM and FDM arerelatively close while when Thom and Woods are adoptedFVM shows better results than FDM The results obtainedby FDM suffer severe oscillation and even lack physicalmeaning It is clear that based on the lid-driven flow stabilityof convection term of FVM is superior to that of FDMThe influence from the same boundary error (truncationerror and rounding error) on difference solving methodsis illustrated in Figure 9 With the treatment of boundarycondition with the same truncation error or rounding errorthe stability of FVM is better than that of FDM especiallyin the situation with boundary treatment of (a) and (b)with relatively low precision truncation error FDM is veryunstable and very easy to slip into oscillation
Journal of Applied Mathematics 9
120572
02 04 06 08 1
Nite
r
80 times 80
40 times 4020 times 20
FVM FDM
5
4
3
2
1
0
times104
(a) Second-order central difference scheme
02 04 06 08 1120572
Nite
r
5
4
3
2
1
0
80 times 80
40 times 4020 times 20
FVM FDM
times104
(b) Second-order upwind scheme
Figure 10 Comparison of the robustness of lid-driven flow with variant schemes
43 Robustness Robustness is a very critical index to evaluatethe numerical approaches In this section the robustness ofFVM and FDM is compared based on the lid-driven flowproblem which is solved by vorticity-stream function Gridsof 20 times 20 40 times 40 and 80 times 80 are adopted respectively inthe calculation Relaxation factor is tentatively taken up from02 to the maximum which can ensure the convergence ofcalculation Iterations are correspondingly recorded as well
It is shown in Figure 10 when the whether the cen-tral difference scheme or second-order upwind scheme isadopted with the refining of calculation grids the maximumrelaxation factor (120572) increases in both FVM and FDM therobustness is enhanced and the iterations (Niter) corre-spondingly decrease as well
As Figure 10(a) shows in the situation that central differ-ence scheme is adopted with the grid of 20 times 20 and 40 times 40the maximum relaxation factor of FVM is bigger than thatof FVM while with the grid of 80 times 80 the relaxation factorcan be taken as 10 in bothmethods Similarly as Figure 10(b)shows generally speaking when second-order upwind dif-ference scheme is adopted the maximum relaxation factor isbigger than that of FDM Based on the result of lid-drivenflow the conservative FVM is superior to the nonconservativeFDM from the perspective of numerical robustness
In the numerical simulation of complex flow field cal-culation efficiency is another critical index on which thispapermakes analysis FromFigure 10we can see that with theincreasing of relaxation factor the iterations of two methodsboth decrease generally What needs to be notified is that theiterations of FVM are smaller than those of FDM
44 Efficiency In the numerical simulation of complex flowfield calculation efficiency is another critical index on whichthis paper makes analysis
As is shown in Table 5 although the calculation timerequired by every iteration step in FVM and FDM is almostthe same with the increasing of mesh numbers calculationtime required by FVM slightly decreases Thus total calcula-tion time of FVM is much smaller than that of FDMThat isFVM is much more efficient that FDM Thus it is validatedthat the efficiency of FVM is better than that of FDM in thesame calculation condition
5 Conclusions
Based on vorticity-stream function and MAC algorithmlid-driven flow and natural convection in a square cavityare calculated The accuracy stability of convective termrobustness of FVM and FDM with variant mesh densitiesdiscrete forms and treatments of boundary condition arecompared The following conclusions are drawn
(1) No matter which algorithm discrete form or treat-ment of boundary condition is adopted the resultsobtained by conservative FVM are closer to bench-mark solution than those obtained by FDM Theaccuracy of FVM is higher than FDM especiallywhen the mesh density is relatively small
(2) The results of lid-driven flow indicate that the resultsobtained by FDM show more serious oscillation andthe results obtained even lose physicalmeaningThusthe stability of convection term of FVM is superiorthan that of FDM
(3) As to the lid-driven flow it is indicated that whenvorticity-stream function method is adopted therelaxation factor of conservative method is biggerthan that of the nonconservative one while theiterations are less which shows that the robustness
10 Journal of Applied Mathematics
Table 5 Computation iterations and time of FVM and FDM in second-order central difference scheme for vorticity-stream functionalgorithm of lid driven flow
Mesh 120572FVM FDM
Iterations Time(s) Time of each iteration Iterations Time(s) Time of each iteration20 02 4950 738 0015 2100 307 001520 01 10500 1544 0015 4300 616 001440 06 3250 1519 0047 4350 1993 004640 05 4450 2035 0046 5800 2629 004540 04 6300 2963 0047 7950 3693 004680 08 4950 8223 0166 5050 8584 017080 06 8500 14134 0166 9000 15395 017180 04 17250 29694 0172 17750 32420 0183
and calculation efficiency of conservative method arebetter than those of the nonconservative method
(4) In the lid-driven flow with vorticity-stream functionmethod the calculation time required by every itera-tion step of FVM is slightly smaller than that of FDMTogether with the smaller total iteration steps FVMis more efficient than FDM
Acknowledgment
The study is supported by the National Science Foundation ofChina (nos 51134006 51176204 and 51276198)
References
[1] J D Anderson Jr Computational Fluid Dynamics The BasicsWith Applications McGraw-Hill 1995
[2] P J Roache Computational Fluid Dynamics Hermosa 1976[3] W Q Tao Heat Transfer Science Press 2001[4] W Q Tao Advances in Computational Heat Transfer Science
Press 2000[5] Y Jaluria and K E Torrance Computational Heat Transfer
Hemisphere Publishing Corporation 1986[6] P D Lax ldquoWeak solutions of nonlinear hyperbolic equations
and their numerical computationrdquo Communications on Pureand Applied Mathematics vol 7 pp 159ndash193 1954
[7] P Lax and B Wendroff ldquoSystems of conservation lawsrdquo Com-munications on Pure and Applied Mathematics vol 13 no 2 pp217ndash237 1960
[8] A Jameson ldquoTransonic potential flow calculations using con-servation formrdquo in Proceedings of the Computational fluiddynamics conference pp 148ndash161 Hartford Conn USA 1975
[9] M Vinokur ldquoConservation equations of gasdynamics in curvi-linear coordinate systemsrdquo Journal of Computational Physicsvol 14 pp 105ndash125 1974
[10] T J Bridges ldquoConservation laws in curvilinear coordinatesa short proof of Vinokurrsquos theorem using differential formsrdquoApplied Mathematics and Computation vol 202 no 2 pp 882ndash885 2008
[11] R J DiPerna ldquoDecay of solutions of hyperbolic systems ofconservation laws with a convex extensionrdquoArchive for RationalMechanics and Analysis vol 64 no 1 pp 1ndash46 1977
[12] A Bressan ldquoHyperbolic systems of conservation lawsrdquo RevistaMatematica Complutense vol 12 no 1 pp 135ndash198 1999
[13] R K S Hankin ldquoThe Euler equations for multiphase com-pressible flow in conservation form-Simulation of shock-bubbleinteractionsrdquo Journal of Computational Physics vol 172 no 2pp 808ndash826 2001
[14] C Lai R A Baltzer andRW Schaffranek ldquoConservation-formequations of unsteady open-channel flowrdquo Journal of HydraulicResearch vol 40 no 5 pp 567ndash578 2002
[15] L M Gao K T Li B Liu and J Su ldquoGeneral integralconservation form of Navier-Stokes equations and numericalapplicationrdquo Chinese Journal of Computational Physics vol 25no 2 pp 172ndash178 2008
[16] S Patankar Numerical Heat Transfer and Fluid Flow Hemi-sphere Publishing 1980
[17] A West and T Fuller ldquoInfluence of rib spacing in proton-exchange membrane electrode assembliesrdquo Journal of AppliedElectrochemistry vol 26 no 6 pp 557ndash565 1996
[18] B P Leonard ldquoComparison of truncation error of finite-difference and finite-volume formulations of convection termsrdquoApplied Mathematical Modelling vol 18 no 1 pp 46ndash50 1994
[19] G G Botte J A Ritter and R E White ldquoComparison of finitedifference and control volume methods for solving differentialequationsrdquoComputers and Chemical Engineering vol 24 no 12pp 2633ndash2654 2000
[20] AThom ldquoThe flow past cylinders at low speedsrdquo Proceedings ofthe Royal Society A vol 141 pp 651ndash659 1993
[21] L C Woods ldquoA note on the numerical solution of fourth orderdifferential equationsrdquo vol 5 pp 176ndash182 1954
[22] V G Jenson ldquoViscous flow round a sphere at low Reynoldsnumbers (Re iexcl 40)rdquo Proceedings of the Royal Society A vol 249pp 346ndash366 1959
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 7
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12U
minus04
0
03
06
V
minus06
minus03
(a) Second-order central difference scheme
FDMFVM (U) benchmark solution
(V) benchmark solution
X(Y)
0 02 04 06 08 1
0
04
08
12
U
minus04
0
03
06
V
minus06
minus03
(b) Second-order upwind difference scheme
Figure 5 Velocity along the central axis obtained by vorticity-stream function method for lid-driven flow with different schemes when themesh density is 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(a) 40 times 40
FDMFVM (U) grid-independent
(V) grid-independent
X(Y)
0 02 04 06 08 1
0
60
120
U
minus120
minus60
0
200
400
V
minus400
minus200
(b) 80 times 80
Figure 6 Velocity along the central axis of convection flow in a square cavity with variant gird densities
40 times 40
Grid-independent
(a) 40 times 40
80 times 80
Grid-independent
(b) 80 times 80
Figure 7 Comparison of temperature contours of natural convection flow in a square cavity with variant mesh densities (FVM in the leftcolumn and FDM in the right column)
8 Journal of Applied Mathematics
(a) 40 times 40 (b) 20 times 20
Figure 8 Comparison of stream lines in lid-driven flow with variant mesh densities (FVM in the left column and FDM in the right column)
(a) Thom (b) Woods
(c) Jenson
Figure 9 Stream functions of lid-driven flow under variant boundary treatments with the grid of 20 times 20 (FVM in the left column and FDMin the right column)
the convection item (Mesh Peclet number) is affected bythe following factors such as mesh density and boundarycondition Taking lid-driven flow as example the stabilityof convection term of FVM and FDM with variant meshdensities treatments of boundary condition is compared
421 Influence of Mesh Density Figure 8 shows the compar-ison of stream lines between FVM and FDM with variantmesh densities With the grid of 40 times 40 the two methodsboth can capture the structure of primary vortex and sec-ondary vortex as Figure 8(a) shows It is seen in Figure 8(b)that when the mesh of 20 times 20 is adopted concussionphenomenon appears which ismore serious in FDM and thelocation of the vortex center obviously moves upward
422 Influence of Boundary Condition Similarly to the accu-racy analysis the results obtained by variant treatments for
vorticity of wall boundary are compared as Figure 9 showsThree treatments Thom Woods and Jenson are analyzedFrom Figure 9 we can see that only when Jenson equationis adopted the results obtained by FVM and FDM arerelatively close while when Thom and Woods are adoptedFVM shows better results than FDM The results obtainedby FDM suffer severe oscillation and even lack physicalmeaning It is clear that based on the lid-driven flow stabilityof convection term of FVM is superior to that of FDMThe influence from the same boundary error (truncationerror and rounding error) on difference solving methodsis illustrated in Figure 9 With the treatment of boundarycondition with the same truncation error or rounding errorthe stability of FVM is better than that of FDM especiallyin the situation with boundary treatment of (a) and (b)with relatively low precision truncation error FDM is veryunstable and very easy to slip into oscillation
Journal of Applied Mathematics 9
120572
02 04 06 08 1
Nite
r
80 times 80
40 times 4020 times 20
FVM FDM
5
4
3
2
1
0
times104
(a) Second-order central difference scheme
02 04 06 08 1120572
Nite
r
5
4
3
2
1
0
80 times 80
40 times 4020 times 20
FVM FDM
times104
(b) Second-order upwind scheme
Figure 10 Comparison of the robustness of lid-driven flow with variant schemes
43 Robustness Robustness is a very critical index to evaluatethe numerical approaches In this section the robustness ofFVM and FDM is compared based on the lid-driven flowproblem which is solved by vorticity-stream function Gridsof 20 times 20 40 times 40 and 80 times 80 are adopted respectively inthe calculation Relaxation factor is tentatively taken up from02 to the maximum which can ensure the convergence ofcalculation Iterations are correspondingly recorded as well
It is shown in Figure 10 when the whether the cen-tral difference scheme or second-order upwind scheme isadopted with the refining of calculation grids the maximumrelaxation factor (120572) increases in both FVM and FDM therobustness is enhanced and the iterations (Niter) corre-spondingly decrease as well
As Figure 10(a) shows in the situation that central differ-ence scheme is adopted with the grid of 20 times 20 and 40 times 40the maximum relaxation factor of FVM is bigger than thatof FVM while with the grid of 80 times 80 the relaxation factorcan be taken as 10 in bothmethods Similarly as Figure 10(b)shows generally speaking when second-order upwind dif-ference scheme is adopted the maximum relaxation factor isbigger than that of FDM Based on the result of lid-drivenflow the conservative FVM is superior to the nonconservativeFDM from the perspective of numerical robustness
In the numerical simulation of complex flow field cal-culation efficiency is another critical index on which thispapermakes analysis FromFigure 10we can see that with theincreasing of relaxation factor the iterations of two methodsboth decrease generally What needs to be notified is that theiterations of FVM are smaller than those of FDM
44 Efficiency In the numerical simulation of complex flowfield calculation efficiency is another critical index on whichthis paper makes analysis
As is shown in Table 5 although the calculation timerequired by every iteration step in FVM and FDM is almostthe same with the increasing of mesh numbers calculationtime required by FVM slightly decreases Thus total calcula-tion time of FVM is much smaller than that of FDMThat isFVM is much more efficient that FDM Thus it is validatedthat the efficiency of FVM is better than that of FDM in thesame calculation condition
5 Conclusions
Based on vorticity-stream function and MAC algorithmlid-driven flow and natural convection in a square cavityare calculated The accuracy stability of convective termrobustness of FVM and FDM with variant mesh densitiesdiscrete forms and treatments of boundary condition arecompared The following conclusions are drawn
(1) No matter which algorithm discrete form or treat-ment of boundary condition is adopted the resultsobtained by conservative FVM are closer to bench-mark solution than those obtained by FDM Theaccuracy of FVM is higher than FDM especiallywhen the mesh density is relatively small
(2) The results of lid-driven flow indicate that the resultsobtained by FDM show more serious oscillation andthe results obtained even lose physicalmeaningThusthe stability of convection term of FVM is superiorthan that of FDM
(3) As to the lid-driven flow it is indicated that whenvorticity-stream function method is adopted therelaxation factor of conservative method is biggerthan that of the nonconservative one while theiterations are less which shows that the robustness
10 Journal of Applied Mathematics
Table 5 Computation iterations and time of FVM and FDM in second-order central difference scheme for vorticity-stream functionalgorithm of lid driven flow
Mesh 120572FVM FDM
Iterations Time(s) Time of each iteration Iterations Time(s) Time of each iteration20 02 4950 738 0015 2100 307 001520 01 10500 1544 0015 4300 616 001440 06 3250 1519 0047 4350 1993 004640 05 4450 2035 0046 5800 2629 004540 04 6300 2963 0047 7950 3693 004680 08 4950 8223 0166 5050 8584 017080 06 8500 14134 0166 9000 15395 017180 04 17250 29694 0172 17750 32420 0183
and calculation efficiency of conservative method arebetter than those of the nonconservative method
(4) In the lid-driven flow with vorticity-stream functionmethod the calculation time required by every itera-tion step of FVM is slightly smaller than that of FDMTogether with the smaller total iteration steps FVMis more efficient than FDM
Acknowledgment
The study is supported by the National Science Foundation ofChina (nos 51134006 51176204 and 51276198)
References
[1] J D Anderson Jr Computational Fluid Dynamics The BasicsWith Applications McGraw-Hill 1995
[2] P J Roache Computational Fluid Dynamics Hermosa 1976[3] W Q Tao Heat Transfer Science Press 2001[4] W Q Tao Advances in Computational Heat Transfer Science
Press 2000[5] Y Jaluria and K E Torrance Computational Heat Transfer
Hemisphere Publishing Corporation 1986[6] P D Lax ldquoWeak solutions of nonlinear hyperbolic equations
and their numerical computationrdquo Communications on Pureand Applied Mathematics vol 7 pp 159ndash193 1954
[7] P Lax and B Wendroff ldquoSystems of conservation lawsrdquo Com-munications on Pure and Applied Mathematics vol 13 no 2 pp217ndash237 1960
[8] A Jameson ldquoTransonic potential flow calculations using con-servation formrdquo in Proceedings of the Computational fluiddynamics conference pp 148ndash161 Hartford Conn USA 1975
[9] M Vinokur ldquoConservation equations of gasdynamics in curvi-linear coordinate systemsrdquo Journal of Computational Physicsvol 14 pp 105ndash125 1974
[10] T J Bridges ldquoConservation laws in curvilinear coordinatesa short proof of Vinokurrsquos theorem using differential formsrdquoApplied Mathematics and Computation vol 202 no 2 pp 882ndash885 2008
[11] R J DiPerna ldquoDecay of solutions of hyperbolic systems ofconservation laws with a convex extensionrdquoArchive for RationalMechanics and Analysis vol 64 no 1 pp 1ndash46 1977
[12] A Bressan ldquoHyperbolic systems of conservation lawsrdquo RevistaMatematica Complutense vol 12 no 1 pp 135ndash198 1999
[13] R K S Hankin ldquoThe Euler equations for multiphase com-pressible flow in conservation form-Simulation of shock-bubbleinteractionsrdquo Journal of Computational Physics vol 172 no 2pp 808ndash826 2001
[14] C Lai R A Baltzer andRW Schaffranek ldquoConservation-formequations of unsteady open-channel flowrdquo Journal of HydraulicResearch vol 40 no 5 pp 567ndash578 2002
[15] L M Gao K T Li B Liu and J Su ldquoGeneral integralconservation form of Navier-Stokes equations and numericalapplicationrdquo Chinese Journal of Computational Physics vol 25no 2 pp 172ndash178 2008
[16] S Patankar Numerical Heat Transfer and Fluid Flow Hemi-sphere Publishing 1980
[17] A West and T Fuller ldquoInfluence of rib spacing in proton-exchange membrane electrode assembliesrdquo Journal of AppliedElectrochemistry vol 26 no 6 pp 557ndash565 1996
[18] B P Leonard ldquoComparison of truncation error of finite-difference and finite-volume formulations of convection termsrdquoApplied Mathematical Modelling vol 18 no 1 pp 46ndash50 1994
[19] G G Botte J A Ritter and R E White ldquoComparison of finitedifference and control volume methods for solving differentialequationsrdquoComputers and Chemical Engineering vol 24 no 12pp 2633ndash2654 2000
[20] AThom ldquoThe flow past cylinders at low speedsrdquo Proceedings ofthe Royal Society A vol 141 pp 651ndash659 1993
[21] L C Woods ldquoA note on the numerical solution of fourth orderdifferential equationsrdquo vol 5 pp 176ndash182 1954
[22] V G Jenson ldquoViscous flow round a sphere at low Reynoldsnumbers (Re iexcl 40)rdquo Proceedings of the Royal Society A vol 249pp 346ndash366 1959
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
8 Journal of Applied Mathematics
(a) 40 times 40 (b) 20 times 20
Figure 8 Comparison of stream lines in lid-driven flow with variant mesh densities (FVM in the left column and FDM in the right column)
(a) Thom (b) Woods
(c) Jenson
Figure 9 Stream functions of lid-driven flow under variant boundary treatments with the grid of 20 times 20 (FVM in the left column and FDMin the right column)
the convection item (Mesh Peclet number) is affected bythe following factors such as mesh density and boundarycondition Taking lid-driven flow as example the stabilityof convection term of FVM and FDM with variant meshdensities treatments of boundary condition is compared
421 Influence of Mesh Density Figure 8 shows the compar-ison of stream lines between FVM and FDM with variantmesh densities With the grid of 40 times 40 the two methodsboth can capture the structure of primary vortex and sec-ondary vortex as Figure 8(a) shows It is seen in Figure 8(b)that when the mesh of 20 times 20 is adopted concussionphenomenon appears which ismore serious in FDM and thelocation of the vortex center obviously moves upward
422 Influence of Boundary Condition Similarly to the accu-racy analysis the results obtained by variant treatments for
vorticity of wall boundary are compared as Figure 9 showsThree treatments Thom Woods and Jenson are analyzedFrom Figure 9 we can see that only when Jenson equationis adopted the results obtained by FVM and FDM arerelatively close while when Thom and Woods are adoptedFVM shows better results than FDM The results obtainedby FDM suffer severe oscillation and even lack physicalmeaning It is clear that based on the lid-driven flow stabilityof convection term of FVM is superior to that of FDMThe influence from the same boundary error (truncationerror and rounding error) on difference solving methodsis illustrated in Figure 9 With the treatment of boundarycondition with the same truncation error or rounding errorthe stability of FVM is better than that of FDM especiallyin the situation with boundary treatment of (a) and (b)with relatively low precision truncation error FDM is veryunstable and very easy to slip into oscillation
Journal of Applied Mathematics 9
120572
02 04 06 08 1
Nite
r
80 times 80
40 times 4020 times 20
FVM FDM
5
4
3
2
1
0
times104
(a) Second-order central difference scheme
02 04 06 08 1120572
Nite
r
5
4
3
2
1
0
80 times 80
40 times 4020 times 20
FVM FDM
times104
(b) Second-order upwind scheme
Figure 10 Comparison of the robustness of lid-driven flow with variant schemes
43 Robustness Robustness is a very critical index to evaluatethe numerical approaches In this section the robustness ofFVM and FDM is compared based on the lid-driven flowproblem which is solved by vorticity-stream function Gridsof 20 times 20 40 times 40 and 80 times 80 are adopted respectively inthe calculation Relaxation factor is tentatively taken up from02 to the maximum which can ensure the convergence ofcalculation Iterations are correspondingly recorded as well
It is shown in Figure 10 when the whether the cen-tral difference scheme or second-order upwind scheme isadopted with the refining of calculation grids the maximumrelaxation factor (120572) increases in both FVM and FDM therobustness is enhanced and the iterations (Niter) corre-spondingly decrease as well
As Figure 10(a) shows in the situation that central differ-ence scheme is adopted with the grid of 20 times 20 and 40 times 40the maximum relaxation factor of FVM is bigger than thatof FVM while with the grid of 80 times 80 the relaxation factorcan be taken as 10 in bothmethods Similarly as Figure 10(b)shows generally speaking when second-order upwind dif-ference scheme is adopted the maximum relaxation factor isbigger than that of FDM Based on the result of lid-drivenflow the conservative FVM is superior to the nonconservativeFDM from the perspective of numerical robustness
In the numerical simulation of complex flow field cal-culation efficiency is another critical index on which thispapermakes analysis FromFigure 10we can see that with theincreasing of relaxation factor the iterations of two methodsboth decrease generally What needs to be notified is that theiterations of FVM are smaller than those of FDM
44 Efficiency In the numerical simulation of complex flowfield calculation efficiency is another critical index on whichthis paper makes analysis
As is shown in Table 5 although the calculation timerequired by every iteration step in FVM and FDM is almostthe same with the increasing of mesh numbers calculationtime required by FVM slightly decreases Thus total calcula-tion time of FVM is much smaller than that of FDMThat isFVM is much more efficient that FDM Thus it is validatedthat the efficiency of FVM is better than that of FDM in thesame calculation condition
5 Conclusions
Based on vorticity-stream function and MAC algorithmlid-driven flow and natural convection in a square cavityare calculated The accuracy stability of convective termrobustness of FVM and FDM with variant mesh densitiesdiscrete forms and treatments of boundary condition arecompared The following conclusions are drawn
(1) No matter which algorithm discrete form or treat-ment of boundary condition is adopted the resultsobtained by conservative FVM are closer to bench-mark solution than those obtained by FDM Theaccuracy of FVM is higher than FDM especiallywhen the mesh density is relatively small
(2) The results of lid-driven flow indicate that the resultsobtained by FDM show more serious oscillation andthe results obtained even lose physicalmeaningThusthe stability of convection term of FVM is superiorthan that of FDM
(3) As to the lid-driven flow it is indicated that whenvorticity-stream function method is adopted therelaxation factor of conservative method is biggerthan that of the nonconservative one while theiterations are less which shows that the robustness
10 Journal of Applied Mathematics
Table 5 Computation iterations and time of FVM and FDM in second-order central difference scheme for vorticity-stream functionalgorithm of lid driven flow
Mesh 120572FVM FDM
Iterations Time(s) Time of each iteration Iterations Time(s) Time of each iteration20 02 4950 738 0015 2100 307 001520 01 10500 1544 0015 4300 616 001440 06 3250 1519 0047 4350 1993 004640 05 4450 2035 0046 5800 2629 004540 04 6300 2963 0047 7950 3693 004680 08 4950 8223 0166 5050 8584 017080 06 8500 14134 0166 9000 15395 017180 04 17250 29694 0172 17750 32420 0183
and calculation efficiency of conservative method arebetter than those of the nonconservative method
(4) In the lid-driven flow with vorticity-stream functionmethod the calculation time required by every itera-tion step of FVM is slightly smaller than that of FDMTogether with the smaller total iteration steps FVMis more efficient than FDM
Acknowledgment
The study is supported by the National Science Foundation ofChina (nos 51134006 51176204 and 51276198)
References
[1] J D Anderson Jr Computational Fluid Dynamics The BasicsWith Applications McGraw-Hill 1995
[2] P J Roache Computational Fluid Dynamics Hermosa 1976[3] W Q Tao Heat Transfer Science Press 2001[4] W Q Tao Advances in Computational Heat Transfer Science
Press 2000[5] Y Jaluria and K E Torrance Computational Heat Transfer
Hemisphere Publishing Corporation 1986[6] P D Lax ldquoWeak solutions of nonlinear hyperbolic equations
and their numerical computationrdquo Communications on Pureand Applied Mathematics vol 7 pp 159ndash193 1954
[7] P Lax and B Wendroff ldquoSystems of conservation lawsrdquo Com-munications on Pure and Applied Mathematics vol 13 no 2 pp217ndash237 1960
[8] A Jameson ldquoTransonic potential flow calculations using con-servation formrdquo in Proceedings of the Computational fluiddynamics conference pp 148ndash161 Hartford Conn USA 1975
[9] M Vinokur ldquoConservation equations of gasdynamics in curvi-linear coordinate systemsrdquo Journal of Computational Physicsvol 14 pp 105ndash125 1974
[10] T J Bridges ldquoConservation laws in curvilinear coordinatesa short proof of Vinokurrsquos theorem using differential formsrdquoApplied Mathematics and Computation vol 202 no 2 pp 882ndash885 2008
[11] R J DiPerna ldquoDecay of solutions of hyperbolic systems ofconservation laws with a convex extensionrdquoArchive for RationalMechanics and Analysis vol 64 no 1 pp 1ndash46 1977
[12] A Bressan ldquoHyperbolic systems of conservation lawsrdquo RevistaMatematica Complutense vol 12 no 1 pp 135ndash198 1999
[13] R K S Hankin ldquoThe Euler equations for multiphase com-pressible flow in conservation form-Simulation of shock-bubbleinteractionsrdquo Journal of Computational Physics vol 172 no 2pp 808ndash826 2001
[14] C Lai R A Baltzer andRW Schaffranek ldquoConservation-formequations of unsteady open-channel flowrdquo Journal of HydraulicResearch vol 40 no 5 pp 567ndash578 2002
[15] L M Gao K T Li B Liu and J Su ldquoGeneral integralconservation form of Navier-Stokes equations and numericalapplicationrdquo Chinese Journal of Computational Physics vol 25no 2 pp 172ndash178 2008
[16] S Patankar Numerical Heat Transfer and Fluid Flow Hemi-sphere Publishing 1980
[17] A West and T Fuller ldquoInfluence of rib spacing in proton-exchange membrane electrode assembliesrdquo Journal of AppliedElectrochemistry vol 26 no 6 pp 557ndash565 1996
[18] B P Leonard ldquoComparison of truncation error of finite-difference and finite-volume formulations of convection termsrdquoApplied Mathematical Modelling vol 18 no 1 pp 46ndash50 1994
[19] G G Botte J A Ritter and R E White ldquoComparison of finitedifference and control volume methods for solving differentialequationsrdquoComputers and Chemical Engineering vol 24 no 12pp 2633ndash2654 2000
[20] AThom ldquoThe flow past cylinders at low speedsrdquo Proceedings ofthe Royal Society A vol 141 pp 651ndash659 1993
[21] L C Woods ldquoA note on the numerical solution of fourth orderdifferential equationsrdquo vol 5 pp 176ndash182 1954
[22] V G Jenson ldquoViscous flow round a sphere at low Reynoldsnumbers (Re iexcl 40)rdquo Proceedings of the Royal Society A vol 249pp 346ndash366 1959
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 9
120572
02 04 06 08 1
Nite
r
80 times 80
40 times 4020 times 20
FVM FDM
5
4
3
2
1
0
times104
(a) Second-order central difference scheme
02 04 06 08 1120572
Nite
r
5
4
3
2
1
0
80 times 80
40 times 4020 times 20
FVM FDM
times104
(b) Second-order upwind scheme
Figure 10 Comparison of the robustness of lid-driven flow with variant schemes
43 Robustness Robustness is a very critical index to evaluatethe numerical approaches In this section the robustness ofFVM and FDM is compared based on the lid-driven flowproblem which is solved by vorticity-stream function Gridsof 20 times 20 40 times 40 and 80 times 80 are adopted respectively inthe calculation Relaxation factor is tentatively taken up from02 to the maximum which can ensure the convergence ofcalculation Iterations are correspondingly recorded as well
It is shown in Figure 10 when the whether the cen-tral difference scheme or second-order upwind scheme isadopted with the refining of calculation grids the maximumrelaxation factor (120572) increases in both FVM and FDM therobustness is enhanced and the iterations (Niter) corre-spondingly decrease as well
As Figure 10(a) shows in the situation that central differ-ence scheme is adopted with the grid of 20 times 20 and 40 times 40the maximum relaxation factor of FVM is bigger than thatof FVM while with the grid of 80 times 80 the relaxation factorcan be taken as 10 in bothmethods Similarly as Figure 10(b)shows generally speaking when second-order upwind dif-ference scheme is adopted the maximum relaxation factor isbigger than that of FDM Based on the result of lid-drivenflow the conservative FVM is superior to the nonconservativeFDM from the perspective of numerical robustness
In the numerical simulation of complex flow field cal-culation efficiency is another critical index on which thispapermakes analysis FromFigure 10we can see that with theincreasing of relaxation factor the iterations of two methodsboth decrease generally What needs to be notified is that theiterations of FVM are smaller than those of FDM
44 Efficiency In the numerical simulation of complex flowfield calculation efficiency is another critical index on whichthis paper makes analysis
As is shown in Table 5 although the calculation timerequired by every iteration step in FVM and FDM is almostthe same with the increasing of mesh numbers calculationtime required by FVM slightly decreases Thus total calcula-tion time of FVM is much smaller than that of FDMThat isFVM is much more efficient that FDM Thus it is validatedthat the efficiency of FVM is better than that of FDM in thesame calculation condition
5 Conclusions
Based on vorticity-stream function and MAC algorithmlid-driven flow and natural convection in a square cavityare calculated The accuracy stability of convective termrobustness of FVM and FDM with variant mesh densitiesdiscrete forms and treatments of boundary condition arecompared The following conclusions are drawn
(1) No matter which algorithm discrete form or treat-ment of boundary condition is adopted the resultsobtained by conservative FVM are closer to bench-mark solution than those obtained by FDM Theaccuracy of FVM is higher than FDM especiallywhen the mesh density is relatively small
(2) The results of lid-driven flow indicate that the resultsobtained by FDM show more serious oscillation andthe results obtained even lose physicalmeaningThusthe stability of convection term of FVM is superiorthan that of FDM
(3) As to the lid-driven flow it is indicated that whenvorticity-stream function method is adopted therelaxation factor of conservative method is biggerthan that of the nonconservative one while theiterations are less which shows that the robustness
10 Journal of Applied Mathematics
Table 5 Computation iterations and time of FVM and FDM in second-order central difference scheme for vorticity-stream functionalgorithm of lid driven flow
Mesh 120572FVM FDM
Iterations Time(s) Time of each iteration Iterations Time(s) Time of each iteration20 02 4950 738 0015 2100 307 001520 01 10500 1544 0015 4300 616 001440 06 3250 1519 0047 4350 1993 004640 05 4450 2035 0046 5800 2629 004540 04 6300 2963 0047 7950 3693 004680 08 4950 8223 0166 5050 8584 017080 06 8500 14134 0166 9000 15395 017180 04 17250 29694 0172 17750 32420 0183
and calculation efficiency of conservative method arebetter than those of the nonconservative method
(4) In the lid-driven flow with vorticity-stream functionmethod the calculation time required by every itera-tion step of FVM is slightly smaller than that of FDMTogether with the smaller total iteration steps FVMis more efficient than FDM
Acknowledgment
The study is supported by the National Science Foundation ofChina (nos 51134006 51176204 and 51276198)
References
[1] J D Anderson Jr Computational Fluid Dynamics The BasicsWith Applications McGraw-Hill 1995
[2] P J Roache Computational Fluid Dynamics Hermosa 1976[3] W Q Tao Heat Transfer Science Press 2001[4] W Q Tao Advances in Computational Heat Transfer Science
Press 2000[5] Y Jaluria and K E Torrance Computational Heat Transfer
Hemisphere Publishing Corporation 1986[6] P D Lax ldquoWeak solutions of nonlinear hyperbolic equations
and their numerical computationrdquo Communications on Pureand Applied Mathematics vol 7 pp 159ndash193 1954
[7] P Lax and B Wendroff ldquoSystems of conservation lawsrdquo Com-munications on Pure and Applied Mathematics vol 13 no 2 pp217ndash237 1960
[8] A Jameson ldquoTransonic potential flow calculations using con-servation formrdquo in Proceedings of the Computational fluiddynamics conference pp 148ndash161 Hartford Conn USA 1975
[9] M Vinokur ldquoConservation equations of gasdynamics in curvi-linear coordinate systemsrdquo Journal of Computational Physicsvol 14 pp 105ndash125 1974
[10] T J Bridges ldquoConservation laws in curvilinear coordinatesa short proof of Vinokurrsquos theorem using differential formsrdquoApplied Mathematics and Computation vol 202 no 2 pp 882ndash885 2008
[11] R J DiPerna ldquoDecay of solutions of hyperbolic systems ofconservation laws with a convex extensionrdquoArchive for RationalMechanics and Analysis vol 64 no 1 pp 1ndash46 1977
[12] A Bressan ldquoHyperbolic systems of conservation lawsrdquo RevistaMatematica Complutense vol 12 no 1 pp 135ndash198 1999
[13] R K S Hankin ldquoThe Euler equations for multiphase com-pressible flow in conservation form-Simulation of shock-bubbleinteractionsrdquo Journal of Computational Physics vol 172 no 2pp 808ndash826 2001
[14] C Lai R A Baltzer andRW Schaffranek ldquoConservation-formequations of unsteady open-channel flowrdquo Journal of HydraulicResearch vol 40 no 5 pp 567ndash578 2002
[15] L M Gao K T Li B Liu and J Su ldquoGeneral integralconservation form of Navier-Stokes equations and numericalapplicationrdquo Chinese Journal of Computational Physics vol 25no 2 pp 172ndash178 2008
[16] S Patankar Numerical Heat Transfer and Fluid Flow Hemi-sphere Publishing 1980
[17] A West and T Fuller ldquoInfluence of rib spacing in proton-exchange membrane electrode assembliesrdquo Journal of AppliedElectrochemistry vol 26 no 6 pp 557ndash565 1996
[18] B P Leonard ldquoComparison of truncation error of finite-difference and finite-volume formulations of convection termsrdquoApplied Mathematical Modelling vol 18 no 1 pp 46ndash50 1994
[19] G G Botte J A Ritter and R E White ldquoComparison of finitedifference and control volume methods for solving differentialequationsrdquoComputers and Chemical Engineering vol 24 no 12pp 2633ndash2654 2000
[20] AThom ldquoThe flow past cylinders at low speedsrdquo Proceedings ofthe Royal Society A vol 141 pp 651ndash659 1993
[21] L C Woods ldquoA note on the numerical solution of fourth orderdifferential equationsrdquo vol 5 pp 176ndash182 1954
[22] V G Jenson ldquoViscous flow round a sphere at low Reynoldsnumbers (Re iexcl 40)rdquo Proceedings of the Royal Society A vol 249pp 346ndash366 1959
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
10 Journal of Applied Mathematics
Table 5 Computation iterations and time of FVM and FDM in second-order central difference scheme for vorticity-stream functionalgorithm of lid driven flow
Mesh 120572FVM FDM
Iterations Time(s) Time of each iteration Iterations Time(s) Time of each iteration20 02 4950 738 0015 2100 307 001520 01 10500 1544 0015 4300 616 001440 06 3250 1519 0047 4350 1993 004640 05 4450 2035 0046 5800 2629 004540 04 6300 2963 0047 7950 3693 004680 08 4950 8223 0166 5050 8584 017080 06 8500 14134 0166 9000 15395 017180 04 17250 29694 0172 17750 32420 0183
and calculation efficiency of conservative method arebetter than those of the nonconservative method
(4) In the lid-driven flow with vorticity-stream functionmethod the calculation time required by every itera-tion step of FVM is slightly smaller than that of FDMTogether with the smaller total iteration steps FVMis more efficient than FDM
Acknowledgment
The study is supported by the National Science Foundation ofChina (nos 51134006 51176204 and 51276198)
References
[1] J D Anderson Jr Computational Fluid Dynamics The BasicsWith Applications McGraw-Hill 1995
[2] P J Roache Computational Fluid Dynamics Hermosa 1976[3] W Q Tao Heat Transfer Science Press 2001[4] W Q Tao Advances in Computational Heat Transfer Science
Press 2000[5] Y Jaluria and K E Torrance Computational Heat Transfer
Hemisphere Publishing Corporation 1986[6] P D Lax ldquoWeak solutions of nonlinear hyperbolic equations
and their numerical computationrdquo Communications on Pureand Applied Mathematics vol 7 pp 159ndash193 1954
[7] P Lax and B Wendroff ldquoSystems of conservation lawsrdquo Com-munications on Pure and Applied Mathematics vol 13 no 2 pp217ndash237 1960
[8] A Jameson ldquoTransonic potential flow calculations using con-servation formrdquo in Proceedings of the Computational fluiddynamics conference pp 148ndash161 Hartford Conn USA 1975
[9] M Vinokur ldquoConservation equations of gasdynamics in curvi-linear coordinate systemsrdquo Journal of Computational Physicsvol 14 pp 105ndash125 1974
[10] T J Bridges ldquoConservation laws in curvilinear coordinatesa short proof of Vinokurrsquos theorem using differential formsrdquoApplied Mathematics and Computation vol 202 no 2 pp 882ndash885 2008
[11] R J DiPerna ldquoDecay of solutions of hyperbolic systems ofconservation laws with a convex extensionrdquoArchive for RationalMechanics and Analysis vol 64 no 1 pp 1ndash46 1977
[12] A Bressan ldquoHyperbolic systems of conservation lawsrdquo RevistaMatematica Complutense vol 12 no 1 pp 135ndash198 1999
[13] R K S Hankin ldquoThe Euler equations for multiphase com-pressible flow in conservation form-Simulation of shock-bubbleinteractionsrdquo Journal of Computational Physics vol 172 no 2pp 808ndash826 2001
[14] C Lai R A Baltzer andRW Schaffranek ldquoConservation-formequations of unsteady open-channel flowrdquo Journal of HydraulicResearch vol 40 no 5 pp 567ndash578 2002
[15] L M Gao K T Li B Liu and J Su ldquoGeneral integralconservation form of Navier-Stokes equations and numericalapplicationrdquo Chinese Journal of Computational Physics vol 25no 2 pp 172ndash178 2008
[16] S Patankar Numerical Heat Transfer and Fluid Flow Hemi-sphere Publishing 1980
[17] A West and T Fuller ldquoInfluence of rib spacing in proton-exchange membrane electrode assembliesrdquo Journal of AppliedElectrochemistry vol 26 no 6 pp 557ndash565 1996
[18] B P Leonard ldquoComparison of truncation error of finite-difference and finite-volume formulations of convection termsrdquoApplied Mathematical Modelling vol 18 no 1 pp 46ndash50 1994
[19] G G Botte J A Ritter and R E White ldquoComparison of finitedifference and control volume methods for solving differentialequationsrdquoComputers and Chemical Engineering vol 24 no 12pp 2633ndash2654 2000
[20] AThom ldquoThe flow past cylinders at low speedsrdquo Proceedings ofthe Royal Society A vol 141 pp 651ndash659 1993
[21] L C Woods ldquoA note on the numerical solution of fourth orderdifferential equationsrdquo vol 5 pp 176ndash182 1954
[22] V G Jenson ldquoViscous flow round a sphere at low Reynoldsnumbers (Re iexcl 40)rdquo Proceedings of the Royal Society A vol 249pp 346ndash366 1959
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of