Renewable Energy 31 (2006) 317331 www.elsevier.

com/locate/renene

Analysis of heat transfer augmentation and ow characteristics due to rib roughness over absorber plate of a solar air heater
Alok Chaube, P.K. Sahoo, S.C. Solanki*
Department of Mechanical and Industrial Engineering, Indian Institute of Technology Roorkee, India Received 20 October 2004; accepted 16 January 2005 Available online 12 September 2005

Abstract A computational analysis of heat transfer augmentation and ow characteristics due to articial roughness in the form of ribs on a broad, heated wall of a rectangular duct for turbulent ow (Reynolds number range 300020,000, which is relevant in solar air heater) has been carried out. Shear stress transport kKu turbulence model is selected by comparing the predictions of different turbulence models with experimental results available in the literature. A detailed analysis of heat transfer variation within inter rib region is done by using the selected turbulence model. The analysis shows that peak in local heat transfer coefcient occurs at the point of reattachment of the separated ow as observed experimentally. The results predict a signicant enhancement of heat transfer in comparison to that for a smooth surface. There is a good matching between the predictions by SST kKu and experimental results. In this work, nine different shapes of rib are examined using SST kKu model and compared on the basis of heat transfer enhancement, friction characteristics and performance index considering heat transfer enhancement with the same pumping power. q 2005 Elsevier Ltd. All rights reserved.
Keywords: Solar air heater; Turbulence; Heat transfer; Friction; Ribbed roughened duct.

* Corresponding author. Tel.: C91 1332 285697 (O), C91 1332 285126 (R). E-mail address: scsolfme@iitr.ernet.in (S.C. Solanki).

0960-1481/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.renene.2005.01.012

318

A. Chaube et al. / Renewable Energy 31 (2006) 317331

Nomenclature Dh hydraulic diameterZ4 area/perimeter friction factor Gk turbulent kinetic energy due to mean velocity gradient Gb Turbulent kinetic energy due to Buoyancy h heat transfer coefcientZQ/DtA 02 02 k turbulent kinetic energyZ1=2u C v C w02 Z 1=2ui i u kair thermal conductivity of air Nu Nusselt numberZhDh/kair p/e relative roughness pitch Pr Prandtl numberZmCp/kair  Re Reynolds numberZruDh =m St Stanton numberZNu/Re Pr Sk, S3 source terms u0 uctuating axial velocity component Gk, Gu effective diffusivity of k and 3 r density of air 3 dissipation rate u specic dissipation rate d boundary layer thickness m dynamic viscosity yC dimensionless distance from the wall (Zruty/m) Subscript s smooth duct

1. Introduction Articial roughness up to laminar sub-layer to enhance heat transfer coefcient is used in various applications like gas turbine blade cooling channels, heat exchangers, nuclear reactors and solar air heaters. A number of experimental studies [1,2] in this area have been carried out but very few attempts of numerical investigation have been made so far due to complexity of ow pattern and computational limitations. In this work, an attempt is made to predict numerically the details of both the velocity and temperature elds responsible for heat transfer enhancement. The presence of rib may enhance heat transfer because of interruption of the viscous sub layer, which yields ow turbulence, separation and reattachment leading to a higher heat transfer coefcient. The enhancement of heat transfer by ow separation and reattachment caused by ribs is signicantly higher compared to that by the increased heat transfer area due to ribs (n-effect) [3]. The heat transfer measurements results for two different rib spacings, p/eZ14 and 8, indicate the importance of roughness geometry [4]. Liou et al. [5] have performed both the numerical analysis and experimental study to investigate the heat transfer and uid ow behavior in a rectangular channel ow with stream wise periodic ribs mounted on one of the principal walls. They have concluded that the ow acceleration and the turbulence intensity are two

A. Chaube et al. / Renewable Energy 31 (2006) 317331

319

major factors inuencing the heat transfer coefcient. The combined effect is found to be optimum for the pitch to rib height ratio equal to 10, which results in the maximum value of average heat transfer coefcient. Rau et al. [6] experimentally found optimum pitch to rib height ratio to be equal to 9. Hence, these investigations reveal that not only the rib geometry but also its geometrical arrangement play a vital role in enhancing the heat transfer coefcient. Karwa [7] has reported an experimental investigation for the same conguration for the Reynolds number range of 400016,000. Tanda [8] has carried out experimental investigation of heat transfer in a rectangular channel with transverse and V-shaped broken ribs using liquid crystal thermography. He concluded that features of the inter-rib distributions of the heat transfer coefcient are strongly related to rib shape and geometry; a relative maximum is typically attained down stream of each rib for continuous transverse ribs (due to ow reattachment). The main aim of the present analysis is to investigate the ow and heat transfer characteristics of a two-dimensional rib roughened rectangular duct with only one principal (broad) wall subjected to uniform heat ux by making use of computer simulation. The ribs are provided only on the heated wall. The other three walls are smooth (without ribs) and insulated. Such a case is encountered in solar air heaters with articially roughened absorber plate. The following section presents the selection and validation of model and detailed analysis using selected model followed by results and discussion.

2. Solution domain The solution domain shown in Fig. 1(a) has been selected as per the experimental details given by Karwa [7]. A rectangular duct with height (H) of 40 mm, rib height (e) of 3.4 mm, rib width of 5.8 mm and pitch (p) of 34 mm has been taken for analysis. In the experimental details [7], the thickness of the heated plate is only 1 mm, which is very small in comparison to the surface area normal to the heat ow. Hence the Biot number is also very small (less than 10K3), which allows us to neglect the internal resistance in comparison to convective resistance. Therefore, the uniform heat ux of 4 kW/m2 is given on ribbed surface, neglecting the conduction resistance within the plate. A 2-D analysis of heat transfer and uid ow through a high aspect ratio (7.5) rectangular duct with

Fig. 1. (a) Ribbed duct geometry and the solution domain. (b) Rectangular mesh with grid adaption for yCZ1 at wall.

320

A. Chaube et al. / Renewable Energy 31 (2006) 317331

transverse ribs provided on a broad, heated wall and other walls smooth and insulated, is carried out using commercially available CFD software, FLUENT 6.1. A non-uniform rectangular mesh with grid adoption for yCZ1 at adjacent wall region is used to resolve the laminar sub-layer (yC!5) and is shown in Fig. 1(b). For grid independence test, the number of cells is varied from 27,650 to 66,524 in various steps. It is found that after 60,910 cells, further increase in cells has less than 1% variation in Nusselt number value which is taken as criterion for grid independence. Similar analysis is carried out for a smooth duct of same dimensions for same range of Reynolds number 300020,000 to nd out the ratio of Stanton number for ribbed duct and smooth duct as experimentally determined by Karwa [7]. The inlet velocity ranging from 0.68 to 3.4 m/s, outlet pressure equals to atmospheric pressure, and no slip wall boundary conditions are used for the analysis

3. Selection and validation of model The selection of model is carried out by comparing the predictions by different low Reynolds number models with experimental results available in the literature [7]. The selected model is further validated, by comparing the heat transfer predictions within the inter-rib regions with the experimental results of Tanda [8]. Low-Reynolds number models are used for the near wall regions because high-Reynolds number models do not perform well in these regions [9,10]. For examples, standard kK3 model and Reynolds stress model (RSM) do not work well near wall region where k&3 approach to zero. Large numerical problems appear in the 3-equation, as k becomes zero. The destruction term in 3-equation includes 32/k, and this causes problem as k/0 even if 3 also goes to zero; they must go to zero at an appropriate rate to avoid problem and this is often not the case. Similarly large eddy simulation (LES) model does not catch the small eddies near the wall [11]. Taking above difculties into consideration low Reynolds number models have been developed. The kKu model replaces dissipation rate (3) term by a specic dissipation rate (u) term which transfers the k from the denominator to the numerator in the specic dissipation rate equation to avoid numerical difculties. The renormalizationgroup (RNG) kK3 model is developed using renormalization theory, to modify the kK3 model for near wall region, by including additional term in 3-equation [10]. The realizable kK3 model contains a new formulation for turbulent viscosity and new transport equations for 3, which is derived from an exact equation for the transport of the mean-square vorticity uctuations. For the ow situations where core and wall bounded regions both are to be modeled with the same accuracies, the blending of both types of models can give satisfactory performance. The SST kKu model is developed using blending function between kK3 and kKu models [9]. It is developed to effectively blend the robust and accurate formulation of kKu model in the near wall region with the free stream independence of kK3 model in the far eld. Both the standard kKu model and kK3 model are multiplied by a blending function and then added together. The blending function is designed to be one in the near wall region, which activates the standard kKu model and zero away from the surface, which activates the kK3 model.

A. Chaube et al. / Renewable Energy 31 (2006) 317331

321

1.0

Experimental Results Shear Stress Transport k Standard k Renormalization k Realizable k

1.0

St/Sts
1.0 1.0 2000

4000

6000

8000

10000

12000

14000

16000

Re
Fig. 2. Comparison between heat transfer enhancement predictions of models with experimental results.

The experimental results of heat transfer enhancement in the form of graph between Stanton number ratio of ribbed and smooth duct versus Reynolds number are compared with the computational results of different models (Fig. 2). After comparing the performance of different selected models with the experimental results, the SST kKu model has been found to yield results closer to the experimental results as compared to other models. The transport equations for SST kKu are:   v v v vk rk C rkui Z Gk (1) C Gk K Yk C Sk vt vxi vxj vxj   v v v vu ru C ruui Z Gu C Gu K Yu C Su vt vxi vxj vxj where Gk represents the generation of turbulent kinetic energy due to mean velocity gradient; Gu represents the generation of u; Gk represents the effective diffusivity of k; Gu represents the effective diffusivity of u; Yk represents the dissipation of k due to turbulence; Yu represents the dissipation of u due to turbulence; Sk and Su represent user-dened source terms. The SST kKu model is further validated by comparing its heat transfer predictions for inter-rib regions with the experimental results presented by Tanda [8]. To carryout this

(2)

322

A. Chaube et al. / Renewable Energy 31 (2006) 317331

study the domain of analysis is taken as per the experimental details given by Tanda [8], and are as follows: Duct height (H)Z20 mm Rib height (e)Z3 mm (square rib) p/eZ13.3 Inlet lengthZ245 mm Uniform heat at bottom surfaceZ1100 W/m2 (the surface below a rib is considered insulated) Aspect ratio (AR)Z5 Pitch pZ40 mm Length of test sectionZ280 mm Outlet lengthZ115 mm Width of ductZ100 mm Reynolds numberZ8900. Fig. 3(a) shows the domain of analysis and Fig. 3(b) shows the non-uniform mesh with grid adaption for yCZ1. The inter-rib region is provided with high-density mesh to observe the heat transfer variation as well as to catch reattachment point. The mean inlet velocity equals to 4.3 m/s, atmospheric pressure as outlet pressure, and no-slip wall boundary conditions are taken for the purpose of analysis. A uniform heat ux of 1100 W/m2 is given at bottom broad wall keeping all other walls smooth and insulated. The ribs provided on the heated wall are also insulated from the wall, thus no heat is given to the ribs as done by Tanda [8]. Both the 3D and 2D analyses are carried out and compared with experimental results. The 3D analysis is carried out

Fig. 3. (a) Schematic diagram of the duct without top cover. (b) Non-uniform mesh with high density in inter-rib region.

A. Chaube et al. / Renewable Energy 31 (2006) 317331

323

Fig. 4. (a) Distribution of center line Nusselt number experimental result. (b) Distribution of center line Nusselt number with similar conguration 3D analysis with 1,203,000 hexahedral cells. (c) Distribution of Nusselt number 2D analysis with 113,006 quadrilateral cells.

with 1,203,000 cells and 2D analysis is carried out with 113,006 cells. The results are shown in Fig. 4 and it is observed that even less than 1/10th number of cells in 2D analysis gives comparable results. It is due to the fact that the secondary ow effect is negligible for continuous transverse ribs [9]. Hence, further analysis is carried out with 2D mesh, which not only saves computer memory but also lot of computational time. For grid independence test, the number of cells is varied from 60,681 to 126,346 in various steps. It is found that after 113,006 cells, further increase in cells has negligible effect on the results. The inter-rib Nusselt number variation between 4th and 5th ribs is plotted, because it is observed that periodic inter-rib Nu-distribution is reached after 34 ribs [8]. The predicted result is compared with the experimental results of Tanda [8], and is shown in Fig. 5. It shows that the prediction by the SST kKu model is comparable in the middle high heat transfer region, but it under predicts near rib surface. In the experimental

324

A. Chaube et al. / Renewable Energy 31 (2006) 317331

Fig. 5. Nusselt number distribution in inter-rib region.

case the whole surface is heated with uniform ux and plexiglas ribs are glued over the heated surface, but in the computational model the surface under rib is considered insulated, which may be a possible cause of this under prediction.

4. Detailed analysis To understand the mechanism of heat transfer in the inter-rib region, it is required to analyze the uid ow characteristics, which are responsible for this variation. The ow pattern and turbulence intensity are two major factors inuencing the local heat transfer coefcient [6]. Many investigators [36,10] have observed that the peak in local heat transfer coefcient occurs at reattachment point. Hence detailed information of ow pattern including point of separation, reattachment point, dead zones, etc. and distribution of turbulence intensity within the domain is desired for the analysis, which are obtained by using FLUENT 6.1 software. After selection and validation of the SST kKu model, detailed uid ow and heat transfer analysis is carried out using different rib shapes to nd out the shape that gives the best performance.

5. Results and discussion Fig. 6(a) depicts the ow pattern in the inter-rib region. It clearly shows the dead regions (eddy formation) adjacent to the rib, both on upstream and downstream-sides. The downstream has larger eddies. These are regimes of low heat transfer and are shown in Fig. 4. Fig. 6(b) shows the x-velocity (u) vector distribution around the ribs. The negative x-velocity around the ribs indicates the existence of separation and reattachment of ow over heated surface. Fig. 7 is x-velocity gradient plot, which gives exact position of reattachment point and length of reattachment. At the point of separation du/dy becomes zero, moving from positive to negative value. Again it becomes zero at point of reattachment, moving from negative to positive value. The reattachment lengths (Lr) for successive inter-rib regions are shown in Fig. 8. It is found that reattachment occurs further

A. Chaube et al. / Renewable Energy 31 (2006) 317331

325

Fig. 6. (a) Stream line pattern in inter-rib region. (b) X-velocity vector plot.

Fig. 7. X-velocity (u) gradient distribution on heated surface over the length of the duct.

6 5 Lr/e 4 3 2 0 10 20 30 40 50 60 70 80 90 Locations of reattachment point X/e

Fig. 8. Reattachment lengths for the successive inter-ribs.

Series1

326

A. Chaube et al. / Renewable Energy 31 (2006) 317331

Fig. 9. Variation of local heat transfer coefcient over the length of the duct.

down stream for the rst inter-rib region compared with the others and the reattachment length decreases rapidly to reach an asymptotic value. The reattachment length decreases to a value after 34 ribs, which remains the same for consequent ribs. Figs. 7 and 9 clearly show that peak of local heat transfer coefcient occurs at the reattachment point. Fig. 10 shows the contours of turbulence intensity inside the duct. The peaks of turbulence intensity are found downstream in the vicinity of the ribs. The peaks of local surface heat transfer coefcient are also found at similar downstream locations, which describes the strong inuence of turbulence intensity on heat transfer enhancement. Using the SST kKu model nine different rib-shapes under following categories have been analyzed for similar duct parameters and roughness geometry: (a) (b) (c) (d) (e) Rectangular rib2!3 mm, 4!3 mm, and 5!3 mm (w!e) Square rib3!3 mm (w!e) Chamfered ribChamfer angle FZ11, 13, and 158 Semicircular ribradius rZ3 mm Circular ribdiameter dZ3 mm.

Reynolds number range is selected from 3000 to 20,000 in which solar air heaters normally operate. The results are shown in Figs. 11 and 12, which depict their behavior through which their performance may be compared. The results show that all the roughness

Fig. 10. Contours of turbulence intensity (%).

A. Chaube et al. / Renewable Energy 31 (2006) 317331

Rectangular_3X2 Square_3X3Rib Rectangular_3X4Rib Rectangular_ 3x5Rib Angle Rib_11(5) Angle Rib_13(6) Angle Rib_15(7) SemiCircular Rib Circular Rib Smooth Duct

327

220 200 180 160 140

Nu

120 100 80 60 40 20 0 2000

4000

6000

8000

10000 12000 14000 16000 18000 20000 22000

Re
Fig. 11. Plot between Nusselt number and Reynolds number for different rib-shapes.

element give enhancement in heat transfer as well as friction factor. The angle ribs give highest heat transfer enhancement but at the cost of very high friction factor specically at low Reynolds numbers. The rectangular and square ribs give high heat transfer with moderate pressure penalty. The circular and semi-circular ribs show good performance from friction point of view but give low heat transfer enhancement in comparison to other rib-shapes. Higher Nusselt number is found in chamfered ribs than rectangular ribs. William and Watts [12] had put forward an explanation for this through the observation of ow patterns. They observed that in case of the square or rectangular section rib (FZ08), at relative roughness pitch of 3, a standing vortex was formed between the ribs lling approximately two-third of the cavity and the energy interchange with the main ow was found to be only sufcient to provide vortex shedding occasionally. Thus the mean level of ow disturbance in the main ow above the ribs was small. But with the chamfered ribs, they observed that the vortex between the ribs was more vigorous, growing to full cavity range and shedding more frequently, thus causing a much larger area of disturbance above the ribs. This is also observed in present study as shown in Fig. 13(c) and (d). Thus, increase in Nusselt number and friction factor with the increase in the chamfer angle can be attributed to more frequent shedding of vortices and hence causing greater heat removal from the surface as well as more friction loss. In almost all the cases the friction factor is found high at low Reynolds numbers. At lower Reynolds number, the thickness of laminar sub-layer submerges the roughness elements and viscous forces are sufciently large to damp out the disturbances caused by ribs. At high Reynolds numbers, the laminar sub-layer is thinner, and so ribs are of the height equal to the laminar sub-layer thickness or slightly more. When rib height is such that it could just penetrate the laminar sub-layer in to the region beyond it, enhancement of

328

A. Chaube et al. / Renewable Energy 31 (2006) 317331

Rect_Rib3X2 Square_Rib3X3 Rect_Rib3X4 Rect_Rib3X5 Angle_Rib11 Angle_Rib13 Angle_Rib15 Circular_Rib SemiCir_Rib Smooth-Duct_Mod.Blasius Eq.

0.16 0.14 0.12

Friction-factor

0.10 0.08 0.06 0.04 0.02 0.00 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 22000

Reynolds Number
Fig. 12. Plot between friction-factor and Reynolds number for different rib shapes.

heat transfer is optimum [2,13]. An observation of stream function contours for lowest and highest Reynolds number in Fig. 13(a)(d) also reveals that vortex formation at top of the rib surface provides rolling action to the ow and hence reduces the friction at high Reynolds numbers.

Fig. 13. Contours of stream function at (a) Reynolds numberZ20,000 with rectangular rib; (b) Reynolds numberZ3000 with rectangular rib; (c) Reynolds numberZ20,000 with 138 Chamfered rib; (d) Reynolds numberZ3000 with 138 Chamfered rib.

A. Chaube et al. / Renewable Energy 31 (2006) 317331

329

The comparison of performance by different rib geometries of equal pumping power would be more useful for engineering applications. For which the performance index (h) can be expressed as follows: hZ St=Sts f =fs 1=3 (3)

The Dittus and Boelter correlation as given below in Eq. (4) is used to nd out heat transfer in smooth duct: Nus Z 0:024Re0:8 Pr 0:4 (4)

The modied Blasius equation is used to calculate the friction factor in smooth duct, which is as follows: fs Z 0:085ReK0:25 (5)

Fig.14 gives performance index of different rib shapes for Reynolds number range from 3000 to 20,000. The results predict that rectangular rib of size 3!5 gives best performance over most of the range of Reynolds number.
Rect_Rib3X2 Square_Rib3X3

2.2 2.1 2.0

Rect_Rib3X4 Rect_Rib3X5 SemiCir_Rib Circular_Rib Angle_Rib11 Angle_Rib13 Angle_Rib15

Performance factor

1.9 1.8 1.7 1.6 1.5 1.4 1.3 1.2 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 22000

Reynolds Number
Fig. 14. Plot between performance index and Reynolds number for different rib shapes.

330

A. Chaube et al. / Renewable Energy 31 (2006) 317331

6. Conclusions The present analysis demonstrates that the SST kKu model gives good results for the prediction of heat transfer and friction characteristics in high aspect ratio rib roughened rectangular duct. In the view of present analysis the following conclusions are drawn: (1) It is observed that the 2D analysis model itself yields results, which are closer to the experimental ones as compared to 3D models. The 3D model requires much higher memory and computational time compared to 2D ones. It may be because of negligible effect of secondary ow in transverse ribbed duct surface. Hence, it is sufcient to employ a simpler 2D model which being more economical with the memory and computational time requirement. (2) In the inter-rib region, the model predicts well near the central high heat transfer area but it under predicts around ribs. (3) The peaks of local heat transfer are found at the reattachment points. Experiments also conrm it. (4) The turbulence intensity is found maximum at peak of the local heat transfer coefcient in the inter-rib regions. (5) The peaks of the local heat transfer coefcient and the reattachment lengths decreases stream wise in the successive inter-rib regions up to 34 ribs and then the distribution becomes periodic with peak values and the reattachment lengths remained unchanged. (6) The highest heat transfer is achieved with chamfered ribs but the best performance index is found with rectangular rib of size 3!5 mm.

References
[1] Muluwork KB, Solanki SC, Saini JS. Study of heat transfer and friction in solar air heaters roughened with staggered discrete ribs. Proceedings of the fourth ISHMT-ASME heat and mass transfer conference, Pune, India; 2000. p. 3918. [2] Karwa R, Solanki SC, Saini JS. Thermo-hydraulic performance of solar air-heaters having integral chamfered rib roughness on absorber plates. Energy 2001;26:16176. [3] Lee CK, Abdel-Moneim SA. Computational analysis of heat transfer in turbulent ow past a horizontal surface with 2-D ribs. Int Commun Heat Mass Transfer 2001;28(2):16170. [4] Slanciauskas A. Two friendly rules for the turbulent heat transfer enhancement. Int J Heat Mass Transfer 2001;44:215561. [5] Liou T-M, Hwang J-J, Chen S-H. Simulation and measurement of enhanced turbulent heat transfer in a channel with periodic ribs on one principal wall. Int J Heat Mass Transfer. 1993;36:50717. [6] Rau G, Cakan M, Moeller D, Arts T. The effect of periodic ribs on the local aerodynamic and heat transfer performance of a straight cooling channel. ASME 1998;120:36875. [7] Karwa R. Experimental studies of augmented heat transfer and friction in asymmetrically heated rectangular ducts with ribs on the heated wall in transverse, inclined, V-continuous and V-discrete pattern. Int Commun Heat Mass Transfer 2003;30(2):24150. [8] Tanda G. Heat transfer in rectangular channels with transverse and V-shaped broken ribs. Int J Heat Mass Transfer 2004;47:22943.

A. Chaube et al. / Renewable Energy 31 (2006) 317331 [9] [10] [11] [12]

331

Fluent 6.0 Users Guide, vol. 2; 2001. Davidson L. An introduction to turbulence models. Chalmers; 1997. Versteeg HK, Malalasekera W. An introduction to computational uid dynamics. Longman; 1995. Williams F, Watts J. The development of rough surfaces with improved heat transfer performance and a study of mechanisms involved. Proceedings of the fourth international heat transfer conference, Paris, vol. 2, FC 5.5; 1970. p. 111. [13] Kakac S, Shah RK, Aung W. Handbook of single phase convective heat transfer. Dordrecht: Wiley; 1987.