This action might not be possible to undo. Are you sure you want to continue?

BooksAudiobooksComicsSheet Music### Categories

### Categories

### Categories

Editors' Picks Books

Hand-picked favorites from

our editors

our editors

Editors' Picks Audiobooks

Hand-picked favorites from

our editors

our editors

Editors' Picks Comics

Hand-picked favorites from

our editors

our editors

Editors' Picks Sheet Music

Hand-picked favorites from

our editors

our editors

Top Books

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Audiobooks

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Comics

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Sheet Music

What's trending, bestsellers,

award-winners & more

award-winners & more

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

www.elsevier.com/locate/apm

**Finite element analysis of laminar and turbulent ﬂows using LES and subgrid-scale models
**

T.L. Popiolek

a

a,*

, A.M. Awruch b, P.R.F. Teixeira

a

´ lia km 8, 96200-000 Rio Grande, Department of Mathematics, Federal University Foundation of Rio Grande, Av. Ita RS, Brazil b Graduate Program in Civil Engineering, Federal University of Rio Grande do Sul, Av. Osvaldo Aranha 99 -3° andar, 90035-190, Porto Alegre, RS, Brazil Received 1 November 2003; accepted 18 March 2005 Available online 17 May 2005

Abstract Numerical simulations of laminar and turbulent ﬂows in a lid driven cavity and over a backward-facing step are presented in this work. The main objectives of this research are to know more about the structure of turbulent ﬂows, to identify their three-dimensional characteristic and to study physical eﬀects due to heat transfer. The ﬁltered Navier–Stokes equations are used to simulate large scales, however they are supplemented by subgrid-scale (SGS) models to simulate the energy transfer from large scales toward subgridscales, where this energy will be dissipated by molecular viscosity. Two SGS models are applied: the classical SmagorinskyÕs model and the Dynamic model for large eddy simulation (LES). Both models are implemented in a three-dimensional ﬁnite element code using linear tetrahedral elements. Qualitative and quantitative aspects of two and three-dimensional ﬂows in a lid-driven cavity and over a backward-facing step, using LES, are analyzed comparing numerical and experimental results obtained by other authors. Ó 2005 Elsevier Inc. All rights reserved.

Keywords: Laminar and turbulent ﬂows; Large eddy simulation; Finite elements; Subgrid-scale model

*

Corresponding author. Fax: +55 5323 15382. E-mail address: dmttales@furg.br (T.L. Popiolek).

0307-904X/$ - see front matter Ó 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.apm.2005.03.019

178

T.L. Popiolek et al. / Applied Mathematical Modelling 30 (2006) 177–199

1. Introduction Turbulent ﬂows are of great practical interest in several engineering ﬁelds and may be deﬁned as a three-dimensional ﬂow with highly disordered, intermittent and rotational ﬂuid motion and with diﬀusive and dissipative characteristics. Although formulation of mathematical models to simulate numerically such complex ﬂows is a challenging task, many researches have been developed and some reliable results have been obtained. Flows with high Reynolds numbers, where the inﬂuence of the diﬀerent turbulence scales must be taken into account, cannot be solved by direct numerical simulation (DNS) due to the large amount of data and unknowns involved in the computational solution (and the corresponding requirements in terms of CPU time and computer memory). Turbulent ﬂows may be simulated using the Reynolds Averaged Navier–Stokes (RANS) equations. This approach is based in the separation of the instantaneous value of a speciﬁc ﬂow variable in its mean value and ﬂuctuations with respect to this mean value. The well-known Reynolds stress components are originated substituting mean values and ﬂuctuations of the variables in the conservation equations. Details about this subject can be found in traditional texts such as Hinze [1], Schlichting [2] and Tennekes and Lumley [3], among others. The RANS equations have more unknowns than equations, and for this reason it is necessary to use closure models to deﬁne the Reynolds stress components. Several models have been employed by diﬀerent authors in the last three decades, and most of these models are described by the state of the art reviews presented by Launder and Spalding [4,5], Rodi [6] and Markatos [7], among others. In recent years several authors have implemented diﬀerent modiﬁcations to the original equations in order to get some improvements of the numerical models behaviour. Alternatively, large eddy simulation (LES) may be used to analyze turbulent ﬂows. This methodology was initially proposed by Smagorinsky [8], and it consists in the separation of the large eddies and subgrid-scales using a grid ﬁlter. Large eddies are associated to the low ﬂow frequencies and they are originated by the domain geometry and the boundaries. Subgrid-scales (SGS) are associated to high frequencies and they have an isotropic and homogeneous behaviour, maintaining their independence with respect to the main stream. As in RANS equations, in LES is also necessary to use closure models, and due to the characteristics of the SGS (homogeneity, isotropy and no signiﬁcant variations for diﬀerent ﬂows), they are more appropriated to be represented by mathematical models. Then, in LES the Large Eddies are simulated directly, whereas SGS are simulated using closure models. Although RANS equations and LES seem to be similar, in the ﬁrst one the closure models simulate the momentum and energy transfer from the mean ﬂow to the ﬂuctuating part, while in LES the closure models simulate the momentum and energy transfer from the large eddies to the small turbulence scales (or subgrid-scales). Comparisons, advances and trends of turbulence models applied to bluﬀ bodies were presented by Ferzinger [9], Leschziner [10] and Murakami [11]. In this work, studies to simulate turbulent ﬂows using LES, with the classical SmagorinskyÕs model and the dynamic subgrid-scale model are presented. The three-dimensional ﬂow in a lid-driven cavity is simulated, and statistical studies with respect to the velocity mean value, turbulence intensity and Reynolds stresses are performed. Two and three-dimensional ﬂows in a backward-facing step are also analyzed in order to verify the behaviour of the two models comparing results of this work with those obtained numerically and experimentally by other authors.

T. but in the ﬁnite element context D may be taken as the cubic root of the element volume.1 and 0. ð8Þ si/ ¼ ÀDt oxi r where r is the Prandtl number for heat transfer problems. si/ ¼ qð ui / ui /0 À u0i / ui / i i i ð4Þ ð5Þ In Eqs. k is the volumetric viscosity. ðq/Þ þ ot oxi oxi oxj ð1Þ ð2Þ ð3Þ is a ﬁltered scalar p are the ﬁltered velocity components and pressure. momentum and energy equations. while (Æ) 0 in Eqs. si/ is also assumed as a non-linear functions of the strain rate.2 and D is the grid ﬁlter width (representing a length scale). þk ot oxj oxj oxj oxj oxi oxk o o o o / Þ À ðq ui / Dij À si/ À S / ¼ 0. . S / is a ﬁltered source. q is assumed to be constant and represent the speciﬁc mass. Popiolek et al. (4) and (5) represents components of the small turbulence scales or ui subgrid-scales. where the ﬁltered strain rate component S ij and the eddy viscosity mt are given by qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 1 o ui o uj 2 S ij ¼ þ . 2 oxj oxi ð7Þ In Eq. respectively. They are given by ui ui uj À uj À ui u0j À uj u0i À u0i u0j Þ ¼ qðLij À C ij À u0i u0j Þ ﬃ Àqu0i u0j . the following ﬁltered expressions for a Newtonian slightly compressible ﬂuid are obtained: i 1 op ou ¼ 0. Dij are the components of the molecular diﬀusion coeﬃcient tensor. / Applied Mathematical Modelling 30 (2006) 177–199 179 2. (7) Cs is the SmagorinskyÕs constant. in three-dimensional ﬂows D ¼ ðDxDy DzÞ1=3 .). The eddy viscosity model is frequently used to represent the eﬀects of SGS in LES. pollutant concentration. þq 2 c ot oxi ! o o op o o ui o uj o uk ðq ui Þ þ uj Þ þ ðq ui dij À qm þ dij À sij ¼ 0. Usually. respectively. etc. c is the sij and si/ are the components of the SGS stress tensound speed and dij is the Kroenecker delta. mt ¼ ðC s DÞ 2S ij S ij ¼ ðC s DÞ2 jS j. which has values varying between 0. and may be written as follows: o/ qmt with Dt ¼ . the overbar indicates ﬁltered quantities and represent components of the large turbulence scales. Crossed terms C ij ¼ ui u0j þ u0i uj as well as Leonard terms Lij ¼ ui uj À uj have not much inﬂuence and they may be omitted. (1)–(5). / where ui and ﬁeld (which may be the temperature. and it may be written as follows: ð6Þ sij ¼ À2qmt S ij . representing the inﬂuence of the SGS in the ﬂow structure. sor and of the SGS ﬂux vector. m is the molecular kinematic viscosity.L. sij ¼ qð À À À u0 /0 Þ ¼ qðLi/ À C i/ À u0 /0 Þ ﬃ Àqu0 /0 . respectively. Smagorinsky’s subgrid-scale model (SSGS) Applying a grid ﬁlter to the continuity. sij is assumed as a non-linear function of the strain rate.

180 T. In the dynamic subgrid-scale model. Values of Cs depend on the userÕs criterion and of the diﬀerent applications. a test ﬁlter is applied. These values may be calculated in a systematic way. þq c 2 ot oxi ! o o oh pi o oh ui i oh uj i oh uk i þk ðqh ui ih uj iÞ þ dij À qm þ dij À T ij ¼ 0. 2 ð12Þ ð13Þ ð14Þ ð15Þ ð16Þ ð17Þ Lilly [13] solved the system of equations given by Eq. uj i À h This expression was also presented by Germano et al. T ij ¼ qðh uj i À h i À hu0 /0 iÞ. (1)–(3).L. and the corresponding equation are given by i 1 ohp oh ui i ¼ 0. Dynamic subgrid-scale model (DSGS) In SmagorinskyÕs model Cs has a constant value over the whole domain. with their components given by ui ui ih uj i À hu0i u0j iÞ. (16). i À h ui / ui ih/ T i/ ¼ qðh i j Taking into account Eqs. and Tij and Ti/ are component of a stress tensor and a ﬂux vector. (14) and (15) the following system of equations is obtained: sij i ¼ À2CM ij . (6). values of Cs have variations in space and time. obtaining the following expression for C: . ðqh/iÞ þ ot oxi oxi oxj ð9Þ ð10Þ ð11Þ where hÆi indicates the application of the second ﬁltering process using a test ﬁlter. the ﬁltered expressions are given by Eqs. The dynamic subgrid-scale model is characterized by two ﬁltering processes: In the ﬁrst one. [12] and modiﬁed after by Lilly [13]. From Eqs. (4) and (12). Using Eqs. Lij ¼ T ij À h where M ij ¼ qhDi2 jhS ijhS ij i À qhD jS jS ij i. / Applied Mathematical Modelling 30 (2006) 177–199 3. (6) and (7). (7). without any interference of the user. In the second ﬁltering process. where the SGS Reynolds stress and SGS ﬂux vector components were included. [12]. it is obtained: Lij ¼ T ij À h sij i ¼ qðh ui ui ih uj iÞ. where hDi ¼ 2D. Popiolek et al. using the grid ﬁlter. Tij may be expressed as follows: T ij ¼ À2qC hDi2 jhS ijhS ij i. formulated ﬁrst by Germano et al. ðqh ui iÞ þ ot oxj oxj oxj oxj oxi oxk i o o o oh/ ðqh ui ih/iÞ À Dij À T i/ À hS / i ¼ 0. and this value remains constant during the time marching process.

.T.2 may be adopted. at a speciﬁc node i. If the dynamic subgrid-scale model (DSGS) is used. C = (Cs)2 remains constant. þk uj Þ þ ot oxj oxj oxj oxj oxi oxk o o Þ À o ðDij þ Dt Þ o/ À S / ¼ 0. was obtained calculating the mean value of the same variable with values computed in the ﬁrst ﬁltering process belonging to all nodes connected to node i. / Applied Mathematical Modelling 30 (2006) 177–199 181 C¼À 1 Lij M ij . because it was veriﬁed that in regions where the energy is transferred from small to large scales. simulating the energy transfer from small to large scales (the backscatter phenomenon). numerical instabilities may occur. 2 M ij M ij ð18Þ This model has some important characteristics: (a) The eddy viscosity is equal to zero in laminar ﬂows. using Eqs. ð23Þ When classical SmagorinskyÕs model (SSGS) is used. In this work. 2 c ot oxi ! o o op o o ui o uj o uk ðq ui Þ þ ðq ui dij À qðm þ mt Þ þ dij ¼ 0. (b) The eddy viscosity may take negative values.1 to 0. In the dynamics model. (15)–(17). r ð22Þ . (c) The model has an appropriated asymptotic behaviour near the solid boundaries.L. the system of equations to simulate turbulent ﬂows of slightly compressible ﬂuids are given by i 1 op ou þq ¼ 0. Popiolek et al. the box ﬁlter with its length scale deﬁned by the ﬁnite element mesh) are submitted to a second ﬁltering process using a test ﬁlter. and commonly values varying from 0. a variable resulting from the second ﬁltering process. obtained by the addition of the molecular and the eddy viscosities. were not allowed. and it is obtained with Eq. C has not a constant value. Negative values of the total viscosity. ðq/Þ þ ðq ui / ot oxi oxi oxj where mt ¼ C D jS j and Dt ¼ With jS j ¼ ð2S ij S ij Þ 1=2 2 ð19Þ ð20Þ ð21Þ qmt . S ij ¼ þ 2 oxj oxi D ¼ ðElement volumeÞ1=3 . for example. i ou j 1 ou . the ﬁeld variables obtained with a ﬁrst ﬁltering process (using. 4. The model to simulate turbulent ﬂows with LES Taking into account the two models described in the previous sections. (18).

may be expressed by the following equation: oq 1 op oU i . by the following algorithm: (1) Determine e ðnþ1Þ=2 with Eq. i Applying the classical Galerkin method for space discretization. (28). Using n ofij osn op n ij e ðnþ1Þ=2 ¼ U n À Dt U . 2 X X oxj X oxj X oxi ð30Þ . The ﬁnite element algorithm The mass conservation for slightly compressible ﬂuids. ¼ ¼À ot c2 ot oxi ð24Þ where c is the sound speed and Ui = qui. (4) Determine U n with Eq. with Dp = pn+1 À pn. respectively Z Z Z Z ! n ðnþ1Þ=2 n Dt oN oN oN n ^ b e XE U jE ¼ N dX U i À dX f ij À dX ^ sij þ dX ^ pn .182 T. (25) is given by the following expression: Ui ðnþ1Þ=2 ð26Þ e ðnþ1Þ=2 À Dt oDp . Expanding the momentum conservation equations in Taylor series the following expression is obtained for the ﬁrst time step (at time t + Dtn/2): n osn Dt oU n Dt ofij opn 1 oDp ij ðnþ1Þ=2 n i Ui . assuming constant entropy. U i ¼ U i þ Dt À þ ¼ U i À Dt ot oxj oxj oxi ð28Þ ð29Þ Then the ﬂow is analyzed. (24) in time and using Eq. Popiolek et al. (26). it is obtained: " ðnþ1Þ=2 # ðnþ1Þ=2 e 1 oU i oU D t o o D p i . À þ i i 2 oxj oxj oxi Eq. after space discretization. ¼U i 4 oxi ð27Þ In Eq. (26). (29). (28) and calculate pn+1 = pn + Dp. the following matrix expressions are obtained for Eqs. ¼ ÀD t À Dq ¼ 2 Dp ¼ ÀDt c 4 oxi oxi oxi oxi The second time step is given by the following expression (at time t + Dtn): ! ðnþ1Þ=2 ðnþ1Þ=2 ðnþ1Þ=2 ofij osij oU i opðnþ1Þ=2 nþ 1 n n .L. / Applied Mathematical Modelling 30 (2006) 177–199 5. (2) Determine Dp with Eq. Discretizing Eq. (27). (26). (27) and (29). ð25Þ ¼ Un þ À À þ þ ¼ U i i 2 ot 2 oxj oxj oxi 2 oxi where fij = ujUi and p(n+1)/2 = pn + Dp/2. (3) Determine U i ðnþ1Þ=2 þ1 Ui with Eq. sij represents viscous terms (including molecular and turbulence eﬀects). (27).

(30)–(33) N is a vector containing the shape functions. The ﬁnite element mesh with 40.1. . In the second case. 1. as the ﬂow is essentially laminar. For Re = 3200. a time interval. Popiolek et al. The symbol ^ indicates vectors with nodal values of the corresponding variables. was identiﬁed and statistical studies in this period were carried out. Eq. oxi ð32Þ Z Z ðnþ1Þ=2 oNT ^ ^ nþ1 ¼ ^ n þ Dt NT N dX U NT N dX U d X f ijE i X X X oxj Z Z oNT oNT n sij þ Dt À Dt N dX ^ N dX ð^ p n þ D^ p=2Þ X oxj X oxj Z Z ðnþ1Þ=2 T T ^ À Dt N nj dC f ijE þ Dt N Nnj dC ^ sn ij C C Z T À Dt N Nni dC ð^ p þ D^ p=2Þ. (31) is solved using the conjugate gradient method with diagonal pre-conditioning. The statistical analysis of the turbulence quantities (intensity and Reynolds stress components) were performed taking velocity ﬁeld data corresponding to central horizontal and vertical lines belonging to the plane of symmetry (z = 0). Three-dimensional ﬂow in a lid-driven cavity The three-dimensional ﬂow in a cavity. / Applied Mathematical Modelling 30 (2006) 177–199 183 Z 1 Dt 2 N 2 N dX þ c 4 X T Z X Z oNT oN o NT e ðnþ1Þ=2 dX D^ p ¼ Dt dX U iE oxi oxi oxi Z X ! ðnþ1Þ=2 T e N ni dC U iE . the index E indicates that the corresponding variables are taken with a constant value over the element domain X. C ð33Þ In Eqs. going from the left to the right side. The 12 edges have the same dimension (1. although according Zang et al. À C ð31Þ XE UiE Z ðnþ1Þ=2 e ¼ XE U iE ðnþ1Þ=2 À Dt 4 Z X oN dX D^ p. for a three-dimensional ﬂow. NT is the transpose of vector N and n is a unit vector normal to the boundary C. is analyzed in this section. a steady state may be obtained. In the ﬁrst case. Numerical applications 6. with boundaries C. periodic oscillations occur [14] and the steady state cannot be attained. where oscillations of the velocity ﬁeld occur. [15].817 nodes and 184.320 linear tetrahedral elements is presented in Fig.T. for this Reynolds number. Flows with Re = 1000 and 3200 were considered. 6.L.00 m) and non-slip boundary conditions were applied on the cavity walls. the ﬂow is still a laminar ﬂow. due to a prescribed velocity U0 applied to its top surface.

Dimensionless mean velocity components on the two lines deﬁned previously and located at the plane of symmetry were calculated with the following expression: Ui ¼ ui U0 ði ¼ 1. / Applied Mathematical Modelling 30 (2006) 177–199 Y X Z Fig. 2Þ.184 T. 2Þ. U2 0 ð37Þ . and U0 is the prescribed velocity at the top surface. ð35Þ where u0i is the ﬂuctuating part.L. and it may be written as follows: ui ¼ ui þ u0i ði ¼ 1. The turbulence intensity may be deﬁned by the following dimensionless expression: Uirms ¼ 10ðu0i2 Þ U0 1=2 ði ¼ 1. ð36Þ The dimensionless Reynolds stress components are given by U 1U 2 ¼ 500ðu01 u02 Þ . respectively. 1. Finite element mesh. 2Þ. the instantaneous velocity can be separated in its mean value and a ﬂuctuating part. ð34Þ where u1 and u2 are mean velocity components in x and y directions. Taking into account Reynolds hypothesis. Popiolek et al.

Figs. respectively. streamlines.00 Fig.1 0 0. (36) and (37). In Fig.T. 5 the path of a particle located initially near the opposite wall to the plane of symmetry (x = 0. 2–4. [16]. In Fig. pressure distribution and dimensionless mean velocity components at the central horizontal and vertical lines belonging to the plane of symmetry.7 0.5 0. the particle follows a secondary vortex located at the right side of the cavity. In Figs.63.005) toward the frontal wall is present. z = 0.50.4 0.60. In this case.25 0. is presented. located on the right side. Popiolek et al. The particle displacement takes place with a spiral motion (characterized by a growing radius) toward the plane of symmetry.8 0. respectively. 6 the particle motion from a position near the symmetry plane (x = 0. A statistical study was carried out analyzing the three-dimensional ﬂow in a cavity with Re = 3200. / Applied Mathematical Modelling 30 (2006) 177–199 185 Constants 10 and 500. z = 0. located at the left side of the cavity.9 0. The ﬁnite element algorithm described in the previous section and LES with the dynamic SGS model were used. Streamlines for Re = 1000. returning after to the frontal wall. In Fig.6 y 0.3 0. 7 the velocity ﬁeld in a secondary vortex.L. leaving the plane of symmetry region following a secondary vortex.2 0. 5 and 6 show the complex characteristics of a three-dimensional ﬂow in a lid-driven cavity. 2. at the bottom of the plane of symmetry. y = 0.75 1.50 x 0. Instants with weak and strong oscillations are shown 1 0. in Eqs. are presented for Re = 1000 and good agreement have been obtained with respect to the result reported by Tang et al. . y = 0.46) is presented.60. were used to amplify values of turbulence intensity and Reynolds stress in order to get a suitable graphical representation.00 0.

9 0.2 0.0 Tang et al.9 0. (1995) 0. (1995) U2 -0.0 Y 0. Pressure distribution for Re = 1000.0 -1.0 (a) X (b) U1 Fig.75 1.25 0.4 0.8 0.1 0.4 0.0 0.0 0.2 Present work 1.8 0.5 0. Popiolek et al.0 -0.7 0.5 0.8 0.6 0.4 0.0 Present work Tang et al.6 -0. Mean velocity components proﬁles at central lines belonging to the plane of symmetry: (a) horizontal line and (b) vertical line.2 0.5 0.7 0.0 0.2 -0.3 0. 3. .4 0. 4. / Applied Mathematical Modelling 30 (2006) 177–199 1 0.8 1.6 0.1 0 0.186 T.3 0.6 y 0.4 -0.L.8 -1.5 1. 1.0 0.2 0.00 Fig.00 0.6 0.50 x 0.

Path of the particle starting from the point (0. 0. were used to perform the statistical study. . for Re = 1000. 5.000 time steps.63. the same characteristics are preserved. close to the frontal wall. Popiolek et al. reattachment and recirculation. and they are representative of a periodic phenomenon.46).2. 8–10. Two.and three-dimensional ﬂows in a backward-facing step are analyzed in this section. velocity ﬁeld data corresponding to a period of approximately 12 s or 40.and three-dimensional ﬂow over a backward-facing step Two. specially in the peak values. Although some diﬀerences can be observed. 0. turbulence intensity and Reynolds stress components are presented in Figs.T. They have characteristics of a very complex ﬂow with layers separation. / Applied Mathematical Modelling 30 (2006) 177–199 187 Fig.L. where strong oscillations occur.60. respectively. 6. In this work. Comparisons of numerical and experimental results for the mean velocity.

At the channel entrance the velocity proﬁle corresponds to a fully developed ﬂow. / Applied Mathematical Modelling 30 (2006) 177–199 Fig. Path of the particle starting from the point (0. One of the more important characteristics is the relation between Reynolds number and the reattachment length. h = 1.94 m.60. s = 0. because it allows to compare results obtained in this work with numerical and experimental results obtained previously by other authors. close to the plane of symmetry. At the channel exit only pressure was prescribed (p = 0). perpendicular to the entrance plane. The inﬂow velocity proﬁle is given by . Popiolek et al. xe = 1. while in the other face there is a plane of symmetry. at the bottom and at the frontal face of the channel. There are solid walls at the top.0 m and xt = 30.005).0 m. 6. The computational domain is depicted in Fig. for Re = 1000. Non-slip boundary conditions were applied at the solid walls. were w = 5. 0. while in the plane of symmetry the perpendicular velocity was taken equal to zero.0 m.0 m.50.188 T. 11.L. 0.

1.90 1. Popiolek et al.0 0.2 Present work Zang et al.6 0.0 0.1 0 0 0.1 0.0 0.2 0.5 0. .0 (a) X (b) U1 Fig.3 0. 2 ! 2 y À h=2 . / Applied Mathematical Modelling 30 (2006) 177–199 0. 8.8 1.0 Y -0.05 m2/s.0 °C and at the solid boundaries the temperature remains constant and equal to zero.4 0. 7.80 0.1 0. Mean velocity at central lines belonging to the plane of symmetry: (a) horizontal line and (b) vertical line.4 189 0.0 Present work Zang et al.7 0. U ðy Þ ¼ U max 1 À 3 h=2 ð38Þ where Umax is the maximum velocity.T.3 0.0 -1.80 0.2 -0. For the thermal problem. The value adopted for the thermal diﬀusion coeﬃcient was 0.4 0.8 0.00 0.3 0.0 0.4 0.6 0.8 -1.00 (a) X (b) X Fig.5 1.6 -0.70 0.2 y 0.4 -0. the temperature at the entrance was taken equal to 1.6 0.) U2 0.8 0.70 0.) 1.9 0.0 0.0 -0. (Exp.5 0.2 0.90 1.4 0.L. (Exp. Velocity ﬁeld in a secondary vortex: (a) instant with weak oscillations and (b) instant with strong oscillations.2 y 0.

(Exp.0 0.0 0.2 0.6 -0.6 0.0 0.7 0.6 0.255 tetrahedral elements was adopted.0 0.4 Present work Zang et al.2 0.0 Y 0.1 0.0 -1.549 nodes and 291.6 0. 12.600 elements was adopted.190 1. a mesh with 7594 nodes and 18. 1. using only one layer of elements in the perpendicular direction to the ﬂow.0 0.0 -0.L.5 0.5 1.0 -1. (Exp.2 -0. This mesh is shown in Fig. For the three-dimensional ﬂow.0 0.4 0.) U1U2 0.0 -0.9 0.9 0. / Applied Mathematical Modelling 30 (2006) 177–199 1.2 0. .8 0.5 0. 9.8 1. ð39Þ ¼ m 3 m m For two-dimensional ﬂows.4 0. Popiolek et al.4 0.6 -0.8 0.) U2rms 0.8 0. Here.8 -1.0 -0.0 Present work Zang et al.6 0.5 1.2 0.8 0.4 -0.0 0.0 Present work Zang et al. 10.3 0. The Reynolds number for this problem is deﬁned as [17] À2 Á U max ð2hÞ 4 U max h UD 3 Re ¼ ¼ . (Exp.8 1.) Present work 0. Reynolds stress components at central lines belonging to the plane of symmetry: (a) horizontal line and (b) vertical line. in the two lateral surfaces the perpendicular velocity remains constant and equal to zero.) 1.0 Y 0.4 -0.4 0.6 0.8 -1.2 0.1 0.0 0.5 0. a mesh with 64.6 0.4 T.3 0.0 (a) X (b) U1U2 Fig. Turbulence intensity at central lines belonging to the plane of symmetry: (a) horizontal line and (b) vertical line.2 -0.5 0.0 Zang et al.2 -0.0 (a) X (b) U1rms Fig.7 0. (Exp.

the diﬀerences with the experimental results grow as Reynolds number increases.T. 14. simulating two-dimensional ﬂows with several Reynolds numbers (until Re = 1000). 11. Popiolek et al.L. In Fig. Although numerical results present good agreement. [17] and the numerical results given by Williams and Baker [19] are compared to the results obtained in the present work for three-dimensional ﬂows. Good agreement is observed between numerical and experimental simulations. Values of the reattachment lengths obtained with the classical SmagorinskyÕs (SSGS) model is . 13. [17] and numerical results obtained by Kim and Moin [18] for the reattachment length are compared with the results of the present work. Computational domain. Y X Z Fig. the experimental work of Armaly et al. In Fig. experimental results obtained by Armaly et al. 12. The ﬁnite element mesh for three-dimensional ﬂows. / Applied Mathematical Modelling 30 (2006) 177–199 Plane of symmetry Channel entrance 191 w Solid wall i Ex t h Solid wall y z s x xe xr Solid wall xt Fig.

0 Xr/S Armaly et al.0 15. / Applied Mathematical Modelling 30 (2006) 177–199 17.0 Xr/S Armaly et al. smaller with respect to those obtained with the dynamic (DSGS) model.0 11. (b) As numerical and experimental results are coincident only for three-dimensional ﬂows when Re > 600.0 5.0 13.0 13. the three-dimensional characteristic of the turbulence phenomenon is also evident.0 7.0 15.0 3. . 13.192 T. using SSGS and DSGS models are shown.0 0 200 400 Re 600 800 1000 Fig. and three facts are conﬁrmed (observing the corresponding results for Re > 600): (a) Reattachment lengths with SSGS model are smaller with respect to those simulated by the DSGS models.0 7.) present work (DSGS model) present work (SSGS model) Kim and Moin 9. which evidences the dissipative character of SSGS model. One of the reasons of this fact is the transversal ﬂow generated by the non-slip boundary conditions prescribed in one of the side walls (on the other side. 17.0 1.and three-dimensional ﬂows in the present work. Popiolek et al. (Exp. showing that SSGS model is more dissipative than the DSGS model. Reattachment length for two-dimensional ﬂows.) present work (DSGS model) present work (SSGS model) Williams and Baker 9.L. results obtained for two. (Exp.0 11. Reattachment length for three-dimensional ﬂows.0 1.0 5.0 0 200 400 Re 600 800 1000 Fig. (c) Reattachment lengths in three-dimensional ﬂows are greater than those of two-dimensional ﬂows. the plane of symmetry is located).0 3. 14. 15. In Fig.

0 5. 10. streamlines corresponding to Reynolds numbers 400. 2-D 193 Xr/S 9.0 3. 800 and 1000. In Fig.40. 3-D SSGS model.T.0 DSGS model.L.0 0 200 400 Re 600 800 1000 Fig.0 11.0 13. Streamlines for Re = 400.0 1. 600. 800 and 1000. [17]. respectively. 3-D DSGS model. for two-dimensional ﬂow. 600. 16. 15. 800 and 1000. Dimensionless separation lengths of the secondary vortices obtained in the present work are 5. separation lengths equal to 5. A secondary vortex near to the top of the channel can be observed for Reynolds numbers greater or equal to 600. 9. Reattachment length for two. Popiolek et al. for two-dimensional ﬂow are shown.17 and 10. respectively. However two aspects may be considered: (a) 2 1 0 0 5 10 15 20 25 30 y x y 2 1 0 0 5 10 15 20 25 30 x y 2 1 0 0 5 10 15 20 25 30 x y 2 1 0 0 5 10 15 20 25 30 x Fig.87. .0 7. [17] found (through an experimental work) for the same Reynolds numbers.11 and 9. Armaly et al. respectively.0 15. 2-D SSGS model. / Applied Mathematical Modelling 30 (2006) 177–199 17. It can be observed that results of the present work for two-dimensional ﬂow are similar to those obtained experimentally by Armaly et al.52. 16. respectively.25 for Reynolds numbers 600.and three-dimensional ﬂows.

0 mm). 800 and 1000. 800 and 1000. are presented. the secondary vortex is very thin. are present for a two-dimensional ﬂow. and by Sohn [20]. for Re = 600 and 800. Pressure distribution for Re = 400. In the present work it seems that the transition regime begins in the interval of Reynolds numbers between 800 and 1000. with very small elements in this region. In Fig. for Re = 1000. but it is still very weak. 800 and 1000. respectively. [17] reported that the secondary vortex is very thin (about 0. Other numerical results for twodimensional such as those present by Kim and Moin [18]. 19. In Fig.4 mm). It will be seen after that when three-dimensional ﬂows are simulated. (b) Armaly et al. but between the top and the bottom boundaries the eﬀect of diﬀusion is more important. In Fig. For Re = 800 the eﬀect of advection between the top and bottom boundaries increases.L. 18. are shown. For Re = 600. The secondary vortex is a consequence of adverse pressure gradients. the streamlines on the plane of symmetry of a three-dimensional ﬂow. for Re = 800.194 T.0 mm. dominant heat transfer along the channel is due to advection eﬀect. During the interval corresponding to the transition regime (Reynolds number between 1200 and 6600) the separation length of the secondary vortices decreases. 800 and 1000. advection eﬀects are dominant in both directions. 17 pressure contours for Reynolds numbers 400. disappear. for Reynolds numbers 600. temperature contours for Reynolds numbers 600. Finally. respectively. In the two-dimensional simulations of the present work thickness greater than 0. 17. It can be observed that the secondary vortex on the top of the channel. 2 y 1 0 0 5 10 15 20 25 30 x 2 y 1 0 0 2 5 10 15 20 25 30 x y 1 0 0 2 5 10 15 20 25 30 x y 1 0 0 5 10 15 20 25 30 x Fig. . 600. respectively. Secondary vortices was not captured in the present work because the elements size in this region is approximately equal to 26. Popiolek et al. which begins when Re ﬃ 1200. have the same characteristics. / Applied Mathematical Modelling 30 (2006) 177–199 Armaly et al. Chiang and Sheu [21] presented similar results for Re = 1000 and Williams and Baker [19] expressed that the separation region is very thin for Re = 800 (they reported a value equal to 1. respectively. 600. and to capture this vortex it is necessary to use a reﬁned mesh.4 mm were obtained. [17] aﬃrmed that the separation length of the secondary vortex increases with Reynolds number until the transition regime.

18. It is necessary to stand out that results of a three-dimensional ﬂow is greatly inﬂuenced by the dimension adopted for the transversal direction (which is the width of the channel). 800 and 1000. respectively.78 0.45 195 y 1 0 0 0.33 0. simulations with Re = 104 and 4 · 104 were performed.64 0. 19. In Fig.78 0.45 0.89 0.73 0.09 0.0.89 0.11 0.18 0. instantaneous values for t = 150 s and Re = 104 are shown. Streamlines for three-dimensional ﬂows with Re = 600.22 5 15 20 25 30 x 2 0.18 36 0.46 0. A strong coherence may be veriﬁed between the velocity ﬁeld and the other variables (streamlines.91 0. / Applied Mathematical Modelling 30 (2006) 177–199 2 0.L. because.33 5 10 0.56 0.50.56 0. In order to study real turbulent ﬂows.18 5 10 15 20 25 30 x Fig. using DSGS model.2 2 0. It can be observed that separation bubbles are distributed near the walls at the top and the bottom of the channel.36 0.27 0. and the corresponding consequence is that the loss of temperature is faster in turbulent ﬂows than in laminar ﬂows.45 20 0.67 10 0.33 25 30 0 15 x 2 1 0 0 y 0. It is necessary to emphasize that results for Re = 800 and 1000 were plotted using time average values to identify with accuracy the reattachment length. Temperature contours for Re = 600. the diﬀerence between a two.22 0.27 0. pressure and temperature).T. 20.64 8 0. respectively.11 0. for these Reynolds numbers. 2 y 1 0 0 5 10 15 20 25 30 x 2 y 1 0 0 5 10 15 20 25 30 x 2 y 1 0 0 5 10 15 20 25 30 Fig.1 5 0. .22 0.11 y 1 0 0.33 0. vortex shedding. Popiolek et al.27 0. Advective eﬀects are dominant in both directions. If this dimension is not large enough to weaken the transversal ﬂow.67 0. where advective eﬀects are only important in the longitudinal direction.82 0. separation and small perturbations along the walls occur. 800 and 1000.and threedimensional ﬂow will be more intense.

Time average values for velocity. 21. 21. 20. 2 y 1 0 0 2 5 10 15 20 25 30 x y 1 0 0 5 10 15 20 25 30 x 2 y 1 0 0 5 10 15 20 25 30 x Fig.08 15 20 0. it is possible to observe the more intense oscillations and faster loss of temperature occur when Re = 4 · 104.17 0.196 2 T. pressure and temperature ﬁelds.08 7 0 . . but comparing Figs.42 0.50 0. Comparing instantaneous velocity ﬁelds and streamlines. 21 time average values of the velocity and pressure ﬁelds as well as the corresponding streamlines are shown. presented in Fig. respectively.33 0. This comparison shows also that vortices near the exit section oscillate and change continuously their positions. In Figs.59 0. respectively. Turbulent ﬂow for Re = 104 in t = 150 s. Velocity. In Fig.8 4 0. presented in Fig. 20 and 22 where instantaneous values are presented.25 10 15 20 25 30 x Fig. 22 and 23 instantaneous (at t = 150 s) and time average values of ﬁeld variables for a turbulent ﬂow with Re = 4 · 104 are shown.92 0 . 20 with the average values.1 25 30 x y 1 0 0 5 0. Popiolek et al.L. Results are similar to those obtained for Re = 104. streamlines. / Applied Mathematical Modelling 30 (2006) 177–199 y 1 0 0 5 10 15 20 25 30 x 2 y 1 0 0 5 10 15 20 25 30 x 2 y 1 0 0 2 5 10 0.08 0. streamlines and pressure ﬁelds. Turbulent ﬂow for Re = 104. it is possible to conclude that vortices located near the step oscillate. but in terms of mean values they remain always in the same positions.

22. Time average values for velocity.L. Turbulent ﬂow for Re = 4 · 104. respectively.40. For Re = 104 a dimensionless value equal to 8. 23. the dimensionless separation length obtained in the present work was 6.85 1 0 0 0.0 obtained experimentally by Kim et al. . [17].08 25 30 0. 7. streamlines and pressure ﬁelds. Using Figs. / Applied Mathematical Modelling 30 (2006) 177–199 2 197 y 1 0 0 5 10 15 20 25 30 2 x y 1 0 0 5 10 15 20 25 30 x 2 y 1 0 0 5 10 15 20 25 30 x 2 y 0.08 0. separation length of the main vortices can be measured. Velocity.23 0.0 ± 1.62 0. This work was used as a reference by Sohn [20]. respectively.T.15 0.08 10 15 20 5 x Fig. pressure and temperature ﬁelds.05 was obtained. Conclusions SmagorinskyÕs classical model (SSGS model) and the dynamic model (DSGS model) were analyzed to simulate laminar and turbulent ﬂows in a lid-driven cavity and a backward-facing step. which is very close to the value obtained in a experimental work by Armaly et al. For Reynolds 4 · 104. 21 and 23. streamlines. 2 1 0 0 2 1 0 0 5 10 15 20 25 30 y 5 10 15 20 25 30 x y x y 2 1 0 0 5 10 15 20 25 30 x Fig.15 0.93 0.00. Turbulent ﬂow for Re = 4 · 104 in t = 150 s. which is inside the interval 7.77 0. [22]. These authors reported a separation length equal to 8.62 0. Popiolek et al.

an Introduction to Its Mechanism and Theory. J. Hinze. among other reasons) occurs for a smaller Reynolds number than that reported by the experimental work of Armaly et al. it was shown that for high Reynolds numbers (in the present work Re = 104 and 4 · 104 were simulated). vortices oscillations and loss of temperature due to dominant advective eﬀect in all directions increase together with growing values of Reynolds numbers. In the second case. McGraw-Hill Book Company. In the ﬁrst case the results of this work were compared with those obtained numerically by Tang et al.L. Schlichting.O. However. turbulence intensity and Reynolds stress components were presented and compared with those obtained experimentally by Zang et al. but. Future works will include applications to more practical engineering problems. New York. numerical results for the reattachment length are smaller than those obtained with experimental works.B. several studies were accomplished such as the inﬂuence of Reynolds number. D.L. For two-dimensional simulations the separation length of the secondary vortex is similar to values given in the technical literature. good agreements are obtained between numerical and experimental results. Acknowledgments The authors wish to thank CNPq and CAPES due to their ﬁnancial support and the Supercomputing Center of the Federal University of Rio Grande do Sul. 1968. [16] and good agreements were observed. and a reﬁned mesh with very small elements would be necessary. in this work the transient regime (characterized by a decrease of the separation length. Launder. Lectures in Mathematic Models in Turbulence. for three-dimensional ﬂows. For Reynolds number between 600 and 1000 and two-dimensional ﬂows. 1972. sixth ed. 1972. Cambridge. Finally. [4] B. Academic Press.198 T. For three-dimensional simulations it was not possible to capture the secondary vortex because it is very thin. In three-dimensional ﬂows a transversal ﬂow going from the lateral wall to the plane of symmetry occurs. again.and three-dimensional ﬂows in a backward-facing step. it was conﬁrmed that the SSGS model is more dissipative than the DSGS model and that turbulence is a three-dimensional phenomenon. New York. . comparisons between the classical SmagorinskyÕs model and the Dynamic subgrid-scale model and diﬀerences between two.and three-dimensional ﬂows. McGraw-Hill Book Company.. Massachusetts and London. Tennekes. [17]. When Re < 600 good results for the separation length of the main vortex were obtained in twoand three-dimensional ﬂows. References [1] J. For this interval of Reynolds numbers the SSGS model simulates smaller values of the separation length than those obtained with the DSGS model. [15] and. With respect to the two. / Applied Mathematical Modelling 30 (2006) 177–199 The three-dimensional ﬂows in a lid-driven cavity for Re = 1000 and 3200 were analyzed. A First Course in Turbulence. Popiolek et al.. 1959. London. Boundary-layer Theory. 15th ed. [3] H. [2] H. Turbulence. Lumley. good agreements were obtained. a statistical study were carried out and results for the mean velocity.E. Spalding. From the previous commentaries. and values of SSGS model decrease when Reynolds number increases.

H. A dynamic subgrid-scale eddy viscosity model. [16] L. Zang. Ferzinger. J. reality and prospects. Kim. Lilly. [12] M. [9] J. Scho ¨ nung. J. Meth. Koseﬀ.H.L. Piomelli. [11] S. Weather Rev. Cheng. U. Fluid Eng. Fluids 31 (1999) 721–745. Leschziner. Moin. Numer. [13] D. Math. T. [7] N. . Numer. J. Model. J. Mech. [20] J. Baker. [15] Y. A proposed modiﬁcation of the Germano subgrid-scale closure method. / Applied Mathematical Modelling 30 (2006) 177–199 199 [5] B. Int. The numerical computation of turbulent ﬂows. Month. 102 (1980) 302–308. 91 (1963) 99–164. Tang. Numer. P.E. 67&68 (1997) 3–34. 59 (1985) 308–323. 10 (1986) 190–220. Netherlands. The basic experiment. Moin. Rodi.Q.H. Sohn. Investigation of a reattachment turbulent shear layer: ﬂow over a backwardfacing step. Johnston. Int. Fluids 23 (1996) 125–142. Phys. Markatos.C. Ind. Current status and future trends in computational wind engineering. Meth. J. Spalding. Time evolution of laminar ﬂow over a three-dimensional backward-facing step.T. Application of a fractional-step method to incompressible Navier–Stokes equations. Kim.R. J. Wind Eng. T. A.B. Kline. T. Computational modelling of complex turbulent ﬂows—expectations.F. D. Institute Association of Hydraulics Research (IAHR). Numerical simulations of laminar ﬂow over a 3D backward-facing step. Aerodyn. Ind.W. Numer. Fluids 21 (1995) 413–432.C. J. The mathematical modelling of turbulent ﬂows. Street. Fluid Mech. Turbulence Models and Their Application in Hydraulics—a State of the Art Review. [21] T. [8] J. Meth. F. Fluids 8 (1988) 1469–1490. J. [19] P. 46&47 (1993) 195–212.M. Chiang. Towards a new model-free simulation of high-Reynolds-ﬂows: local average direct numerical simulation. Meth.T. Meth. R. A dynamic mixed subgrid-scale model and its application to turbulent recirculating ﬂows. General circulation experiments with the primitive equations: I.F.H. J. ASME Trans. Cabot. Delft. Durst. 1984. Phys. W. Armaly. J. Aerodyn. Sheu. Comput.. 46–47 (1993) 37–51. Wind Eng. Popiolek et al. Appl.K. Phys. Murakami. Aerodyn. [18] J. Int. Appl. [17] B. Fluids A 5 (1993) 3186–3196.L.J.T. J. Fluids A 3 (1991) 1760–1765. S. Smagorinsky. P. Eng. Simulation of complex turbulent ﬂows: recent advances and prospects in wind engineering.L. [10] M. Wind Eng.P. Tsang. J. Denaro. Experimental and theoretical investigation of backwardfacing step ﬂow. Fluids 24 (1997) 1159–1183. J. Germano. Ind. 127 (1983) 473–496. [22] J. Meth. Pereira.A. 3 (1974) 269–289. Comput. Williams. Int. Launder. Evaluation of FIDAP on some classical laminar and turbulent benchmarks. B.J. [14] F. [6] W. Numer. J. Fluids A 4 (1992) 633– 635. Transient solutions for three-dimensional lid-driven cavity ﬂows by a leastsquares ﬁnite element method.P. Int. Phys.

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue listening from where you left off, or restart the preview.

scribd