Experimental Thermal and Fluid Science 37 (2012) 4656

Contents lists available at SciVerse ScienceDirect

Experimental Thermal and Fluid Science

Experimental investigation of ow and heat transfer in a rectangular channel with 45 angled ribs on one/two walls
Francesca Satta, Daniele Simoni, Giovanni Tanda
DIMSET, University of Genova, via Montallegro 1, I-16145 Genova, Italy

a r t i c l e

i n f o

a b s t r a c t
The ow and heat transfer characteristics in a rectangular channel (aspect ratio AR = 5) with angled rib turbulators, inclined at 45, have been investigated. The angled ribs were attached, with parallel orientations, on one or two surfaces of the channel. The ratio of rib height to hydraulic diameter (e/D) was 0.09, while the rib pitch-to-height ratio (P/e) was 10. Experiments were carried out using air as the convective uid, with the Reynolds number Re = 29,000 and periodic fully developed conditions. The ow patterns were investigated by using laser-Doppler velocimetry (LDV), while the local heat transfer coefcients were provided by liquid crystal (LC) thermography. Inclined ribs are responsible for two-cell secondary ow when installed on two opposite walls of the channel and for one-cell secondary ow when placed on one wall only. The secondary ow brings relatively cold core uid towards the rib leading end regions and warm uid towards the opposite rib trailing end regions, causing marked spanwise variations in the local heat transfer coefcient. 2011 Elsevier Inc. All rights reserved.

Article history: Received 15 April 2011 Received in revised form 27 September 2011 Accepted 27 September 2011 Available online 13 October 2011 Keywords: Flow eld Heat transfer Ribbed channel 45 Angled ribs LDV Liquid crystals

1. Introduction Articial roughness elements such as periodic ribs are often employed in order to enhance forced convective heat transfer between a wall and a uid. The presence of periodic ribs alters the ow characteristics, inducing secondary ow, recirculation and reattachment so as to enhance the heat transfer process. Many parameters affect the performance of rib-roughened channels: the shape and dimensions of the channel, rib characteristics (including shape, size, arrangement, etc.), and ow characteristics (through the Reynolds and Prandtl numbers). Among these parameters, the role played by the rib installation has often been recognized as the most important. Consequently, a considerable amount of literature work has been devoted to evaluating thermal performance of rectangular channels with various forms of rib turbulators (see, for instance, [13]). Although heat transfer and friction results have been extensively documented in the literature, only a limited amount of studies reported ow eld measurements, which are of great support to understand the complicated heat transfer distributions induced by rib turbulators. Rau et al. [4] used laser-Doppler velocimetry (LDV) to measure the ow eld in a rectangular channel with repeated transverse ribs on one or two walls. Flow separation and reattachment downstream of ribs were clearly identied; ow structure was found to

Corresponding author. Tel.: +39 010 3532557; fax: +39 010 3532566.
E-mail address: giovanni.tanda@unige.it (G. Tanda). 0894-1777/$ - see front matter 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.expthermusci.2011.09.020

be three-dimensional with signicant impingement on the smooth side walls. Angling the ribs makes the ow structure more complex, with secondary motions that cause signicant spanwise variation in the mean ow; LDV measurements by Olsson and Sundn [5] and Iacovides et al. [6] showed larger mean velocities in regions where the secondary ow is directed towards the surface. As a result, the separation bubble behind each rib is very intense at one rib end, where high momentum uid is transported by secondary ow through the duct core, while the intensity becomes weaker towards the opposite rib end. Further investigations of ow in angled ribbed channels were provided by Bonhoff et al. [7], Chanteloup et al. [8], and Gao and Sundn [9] by means of particle image velocimetry (PIV); accurate velocity distributions over several planes were obtained with the primary purpose of obtaining a comprehensive understanding of the ow phenomena and a large database for the validation of CFD codes. Despite the different channel and rib geometry, the LDV and PIV experiments documented in the literature showed that the ow eld is typically dominated by the rib-induced secondary ow motion, which leads to marked spanwise velocities. The aim of this study was to experimentally investigate ow and heat transfer characteristics in a rectangular channel, with a large width-to-height ratio and 45 angled ribs arranged parallel on one or two opposite walls. These two rib arrangements are of interest in several engineering applications; ribs on one wall of the channel are used to increase the performance of solar air heaters, while channels roughened on two opposite walls are often encountered in the internal passages of gas turbine blades.

F. Satta et al. / Experimental Thermal and Fluid Science 37 (2012) 4656

47

Nomenclature AR D e k H h L m0 Nu P Pnet qconv Re TLC Tair channel aspect ratio, W/H hydraulic diameter of the rectangular channel, 2 WH/ (W + H) rib height thermal conductivity of air channel height convective heat transfer coefcient length of heated surface air mass ow rate Nusselt number rib pitch p inter-rib distance, P e 2 convective heat ux Reynolds number, UmD/m surface temperature (by liquid crystals) air bulk temperature U Um V W x x0 X y Y z Z velocity component along x coordinate mean air velocity velocity component along y coordinate channel width streamwise coordinate auxiliary streamwise coordinate starting from the rib downstream side dimensionless x0 coordinate spanwise coordinate dimensionless y coordinate transverse coordinate dimensionless z coordinate

Greek symbols q density of air m kinematic viscosity of air

For both rib congurations, attention was focused on the velocity eld, measured by means of the laser-Doppler velocimetry. Flow measurements have been integrated by local heat transfer coefcient measurements, using the steady-state liquid crystal technique. The convective uid used in the experiments was air, with the Reynolds number xed at 29,000. The investigated domain was set between two consecutive ribs sufciently far from the test section entrance, where periodic ow and heat transfer conditions in the streamwise direction are deemed to have been reached. The experimental results here obtained are useful to improve the understanding of the heat and uid ow process and to provide reference data for the validation of numerical ow solvers.

relative to a smooth channel under the constant pumping power constraint. Two different rib arrangements were considered: (i) the 45 angled, parallel ribs were xed only on the heated plate, the opposite adiabatic plate being at, (ii) the 45 angled, parallel ribs, were xed symmetrically (in-line) on the heated plate and on the opposite adiabatic plate. The rst case is referred as one-ribbed wall channel (1RW), while the second one is termed two-ribbed wall channel (2RW). Flow and thermal measurements were restricted to a region, between a pair of consecutive ribs (namely between the seventh and the eighth ribs), sufciently far from the test section entrance (more than 3.6D), as shown in Fig. 1. This choice is consistent with previous investigations on ribbed channels with repeated inclined ribs [6,11], where ow and thermal measurements were carried out for the 7th and the 5th rib modules, respectively. 2.2. Experimental setup and procedure A sketch of the experimental setup is shown in Fig. 2. An opencircuit wind tunnel was used for this study. Air, drawn at room temperature, passed through a lter and entered a rectangular channel, where the test section was placed. The rectangular channel, as wide and high as the test section, included a hydrodynamic development duct longer than 60D, and a 30D long exit duct, preceding and following the test section respectively. A Venturi owmeter was used for the measurement of mass ow rate. Fluid ow and heat transfer measurements were carried out separately, by holding the Reynolds number xed (Re = 29,000). The Reynolds number is determined from

2. Experimental arrangement 2.1. Test section The test section is schematically illustrated in Fig. 1. It consisted of a rectangular channel with the following geometric characteristics: width W = 0.1 m, height H = 0.02 m, aspect ratio AR = W/H = 5, hydraulic diameter D = 2 WH/(W + H) = 0.033 m. This channel was delimited by the heated plate (width W = 0.1 m, length L = 0.28 m) and lateral and frontal, 5-mm-thick, unheated, Plexiglas walls. Plexiglas was used to provide optical access for both velocity and temperature measurements. The heated plate was made of 0.5 mm thick stainless steel to which a plane heater had been attached to provide a controllable uniform heat ux. It was used for thermal experiments only (measurement of local heat transfer coefcients), while LDV measurements were performed under isothermal conditions (i.e., no power delivered to the heater). Ribs, made of balsa wood, had a square section (side e is equal to 3 mm) and a blockage ratio e/H = 0.15 (e/D = 0.09). Owing to their low thermal conductivity, ribs were considered to be virtually adiabatic and their function was to generate turbulence in the airow in order to increase the inter-rib heat transfer. They were attached, by double-sided adhesive tape, at periodic streamwise stations with an inclination (relative to the main ow direction) of 45 and a rib pitch P of 30 mm (P/e = 10). This pitch value was chosen since a parallel study [10] showed that a pitch-to-height value P/e from 10 to 13.3 yields, for a rectangular channel (AR = 5) with 45 angled ribs, the best thermal performance

Re U m D=m

where m is the air kinematic viscosity and Um is the mean air velocity, calculated from mass ow rate m0 divided by density q and the maximum cross-sectional area (WH). Properties of air have been evaluated either at the uid temperature (velocity measurements) or at the lm temperature (heat transfer measurements). 2.2.1. The measurement domain Details of the coordinate system and of the measurement plane locations are indicated in Fig. 3. A convenient coordinate system has been introduced, as shown in the three-dimensional view of the measurement domain plotted in Fig. 3a (1RW case) and Fig. 3b (2RW case). For the 1RW conguration, the wall at z = 0 is

48

F. Satta et al. / Experimental Thermal and Fluid Science 37 (2012) 4656

(a)

HEATER

air flow B

Heated ribbed surface

H=0.02 m
Adiabatic ribbed (or smooth) surface

W= 0.1 m

(b)

air flow

Section B-B P=
30 mm

e=3 mm

= 45 L= 0.28 m

Fig. 1. Sketch of the test section: (a) side view, (b) frontal view (section BB). The heated plate is covered with a LC sheet on the air side. The wall opposed to the heated plate is unheated and smooth (1RW) or ribbed (2RW). It is made of Plexiglas in order to allow the optical access for LDV and LC measurements. The shaded area represents the measurement eld.

VENTURIMETER

OUT OF THE ROOM

SEEDING UNIT
LAMPS

Compressed air
CCD VIDEOCAMERA

PC with IEEE 1396 interface

Fig. 2. Schematic layout of the experimental setup.

unribbed, while ribs are located on the wall at z = H. For the 2RW conguration, both walls at z = 0 and z = H are ribbed. For both 1RW and 2RW congurations, only the wall at z = H is heated during the heat transfer experiments. A view of the heated wall is provided in Fig. 3c, where the leading end of ribs denotes the rib end closest to the ow inlet, while the other rib end is called the trailing end. Similarly, the side wall (at y = 0) conning with the leading ends of ribs is called leading side-wall and the opposite side (at y = W) is called trailing side-wall. To facilitate the presentation of velocity and heat transfer results, an auxiliary coordinate x0 in the streamwise direction has been used. Coordinate x0 originates, for any y location, from the downstream side of rst rib and coincides with x only along the leading side-wall. The trio x0 yz has been normalized as follows: X = x0 /Pnet, Y = y/W and Z = z/H, where Pnet is the inter-rib distance, p equal to (P e 2). Based on the previous denitions, the normalized coordinates X, Y, and Z range from 0 to 1 within the measurement domain. Five XZ planes, identied by Y = 0.1, 0.3, 0.5, 0.7 and 0.9, normal to the ribbed walls, have been selected for the velocity measurements, as shown in Fig. 3a and b. For each plane, a coarse

grid of 13 (along Z) 6 (along X) points was used. The grid has more measurement points along the Z-direction (where the highest velocity gradients are expected) and a ner mesh close to the walls (Z = 0 and Z = 1), with rst and last measurement points 1.0 mm away from the walls. With the adopted grid of 13 6 points, the ow structure was expected to be captured with reasonable resolution. 2.2.2. LDV measurements A four beams, two colors, laser-Doppler velocimeter (Dantec Fiber Flow), in the backward scattered conguration, was employed for the present investigation. The light sources were two diode-pumped solid state lasers with a power of 200 mW each. The two pairs of green and blue beams, employed to investigate the two velocity components, had wavelengths of 532 nm and 488 nm, respectively. The frequency of one of each pair of beams was shifted by 40 MHz with a Bragg cell. The probe consisted of an optical transducer head of 60 mm diameter, with a focal length of 300 mm, connected to the emitting optics and to the photomultipliers by means of optic bers. The fringe separation is 4.1 lm for the green pair of beams and 3.8 lm for the blue one. The probe

F. Satta et al. / Experimental Thermal and Fluid Science 37 (2012) 4656

49

(a)
W

P - e2

air flow

y x

(b)
W

air flow

y x

P - e2

(c)
air flow

trailing end of ribs

trailing side-wall

P - e2
x y
leading end of ribs

W e

x
leading side-wall

Fig. 3. Denition of coordinate system and measurement domain: (a) channel with one ribbed wall (1RW), (b) channel with two ribbed walls (2RW), (c) heated wall (at z = H) for both 1RW and 2RW congurations.

volume was 0.09 mm 0.09 mm 1.4 mm. The ow was seeded with mineral oil droplets with a mean diameter of 1.5 lm. The seeding material led to the contamination of the Plexiglas surfaces which progressively lowered the data rate; to solve this problem, one lateral wall of the test section was equipped with a 10-mm diameter optical window to periodically permit the cleaning of the surfaces affected by the oil deposition. To process bursts, two Dantec Enhanced Burst Spectrum Analysers were employed. For each measurement point 30,000 samples were collected with a maximum record length in time of 120 s. To allow a high movement precision between two successive measurement points, the probe has been traversed using a three-axis computer controlled traversing mechanism with a minimum linear translation step of 8 lm. At each measurement point, streamwise and spanwise velocity components, U and V respectively, have been recorded. According to the notation of Fig. 3, spanwise velocity is positive when directed from the leading side-wall to the trailing side-wall. The velocity components U and V have been scaled by the bulk mean velocity Um; at the selected Reynolds number of 29,000, Um was equal to 13.5 m/s.

Fine-gauge thermocouples were placed inside the rectangular channel directly exposed to the airow and in several places inside the channel wall material. These sensors were used to measure the air temperature at the test section inlet, estimate conduction heat losses to the surroundings, and control the attainment of the steady-state conditions. Thermosensitive, cholesteric liquid crystals were used to measure temperature distributions on the heated surface. The prepackaged LC sheet (0.15 mm thick) consisted of a thermochromic liquid crystal layer on a black background applied onto a mylar lm and backed with a pressure-sensitive adhesive. The color distribution of the liquid crystals was observed by a CCD videocamera through the Plexiglas wall opposite to the heated surface. Four 100-W lamps were used to illuminate the test section during image storing. The lamps were switched on, and the frontal insulation layer was removed, only for the time strictly required for image acquisition. Only a reduced region of the test surface (shaded area in Fig. 1), encompassing the eld of interest (region between the seventh and the eighth ribs), was framed by properly adjusting the magnication of the videocamera. After digitizing and ltering processes the resolution was about 0.37 mm/pixel. The relationship between the color and the temperature of the thermosensitive liquid crystals was found by a separate calibration experiment. The calibration test was carried out by gradually heating a 5 mm thick aluminum plate (calibration plate) covered by a LC lm identical to that used in the convective experiments. The calibration plate was equipped with ten ne-gauge thermocouples, calibrated to 0.1 K, housed in small holes drilled in the material at different positions as close to the surface as possible. The same videocamera, lighting conditions and frontal Plexiglas wall as for the convective experiments were used in the calibration runs. For a given wall temperature, the corresponding color image displayed by the LCs was digitized and processed in order to obtain, pixel by pixel, the HSI (hue, saturation, intensity) contents from the RGB (red, green, blue) domain. Among the new parameters (H, S, I) only hue was retained since it was found to be the only one correlated with the surface temperature. LCs here employed had a red start temperature of 30 C with a bandwidth of 4 C. The hue-temperature correlation was found to be fairly linear and repeatable in the range from 31.5 to 32.5 C (hue in the 70140 range): this eld of hue was therefore selected for the quantitative analysis of each LC image. The corresponding calibration line had an uncertainty band (at the 95% condence interval) of 0.2 K. Convective experiments were performed according to the following procedure. After the rib array had been set and the airow had been adjusted to the prescribed velocity, the DC current was supplied to the heater. The surface temperature was maintained within the thermosensitivity region of the liquid crystals by controlling the input power delivered to the heater. Once steady-state conditions were reached, input power, thermocouple readings were recorded and the LC image was captured by the CCD videocamera and stored in a PC. At a given mass ow rate, 715 LC images were taken for different values of the input power so as to move the green color throughout the entire optical eld. The images of the colored pattern of LCs were then processed in order to extract the local heat transfer coefcient according to the following relationship:

2.2.3. Heat transfer measurements A thin liquid crystal (LC) sheet was applied to the air side of the heated plate to measure local wall temperature. The outer sides of the test section walls were thermally insulated in order to convey as much as possible of the electric power dissipated by the heater to the convective air ow. Power was supplied by an adjustable DC source and measured by a voltmeter and an amperometer.

h qconv =T LC T air

where qconv is the net convective heat ux, assumed to be uniformly distributed over the heated plate, TLC is the surface temperature detected by the LCs, and Tair is the local bulk temperature of the air along the streamwise direction, calculated by an energy balance. The difference between surface and bulk uid temperature was typically in the 1114 K range. The net convective heat ux was

50

F. Satta et al. / Experimental Thermal and Fluid Science 37 (2012) 4656

calculated as the product of the measured voltage and current supplied to the heater (minus the calculated heat losses due to radiation and conduction escaping from the test section) divided by the surface area exposed to the air ow. Radiative and conductive losses were calculated by considering an enclosure of gray surfaces and a simple 1-D conduction model, respectively. At the selected Re number, heat transfer rate leaving the test section by radiation and conduction accounted for about 6% of the input power to the heater. Additional details on the heat transfer measurements based on liquid crystal thermography are given in [10,12,13]. Heat transfer data are presented in dimensionless form by introducing the Nusselt number Nu:

The uncertainty (at the 95% condence level) in local h (or Nu) values was estimated to be 6% [12]. This value takes into account the effects of (i) errors in the measurement of the electrical power, uid temperature (thermocouple readings) and wall temperature (liquidcrystal thermographic reading), (ii) errors in the estimation of heat losses, and (iii) errors due to thermal conduction, along transverse and longitudinal directions, inside the stainless steel, heated plate. The Reynolds number had a calculated uncertainty of 4%.

3. Results and discussion 3.1. Channel with ribs on one wall First, the results for the channel with angled ribs on one side (1RW case) are presented and discussed. Figs. 48 show the normalized streamwise and spanwise velocities in the XZ planes, respectively identied by the coordinates Y = 0.1, 0.3, 0.5, 0.7, and 0.9. Figs. 9 and 10 refer to the streamwise Nusselt number distributions, again at different Y coordinates (from 0.1 to 0.9) over the heated ribbed surface (Z = 1) and to the contours of lines at constant Nusselt number, respectively. A general view of the ow eld reveals asymmetric distributions of the normalized U/Um component along the channel height due to the presence of ribs on the wall at Z = 1. At the plane Y = 0.1 (Fig. 4), the uid, near the ribbed side (Z = 0.90.95), has a streamwise velocity close to zero just downstream of the rib (X up to 0.28). Here a separation zone is likely to exist, as found in similar experimental circumstances by Gao and Sundn [9], even though measurement points close enough to the wall to detect negative U/Um values were unfortunately impossible to reach. For larger X values, U/Um, close to the ribbed wall, is increased up to about unity, due to the ow reattachment. Sufciently far from the ribbed wall (Z = 0.70.8), the U/Um proles have a maximum, whose magnitude (U/Um = 1.251.4) is only slightly dependent on streamwise X coordinate. For lower values of Z, the streamwise velocity component does not seem to be affected by the X value. The plot of the normalized spanwise velocity V/Um shows that, in the channel region facing the ribbed side

Nu hD=k

where k is the air thermal conductivity, evaluated at the lm temperature. With reference to the sketch in Fig. 3, the map of h (and Nu) values has been obtained in the xy plane at z = H (i.e., Z = 1), identifying the surface of the heated plate in contact with the convective uid. 2.3. Experimental uncertainties Statistical uncertainty in mean velocities depends on several elements: the number of independent samples, the turbulence intensity based on the local velocity, and the condence interval. To increase the number of statistically independent samples, the maximum data rate value was limited according to George [14] and Satta et al. [15]. Moreover, statistical moments were weightaveraged with transit time to avoid statistical bias. Thanks to the large number of independent samples, the statistical uncertainty on the averaged velocity due to the nite number of samples was estimated to be better than 4% for a probability of 95% and a local turbulence intensity of 100% (rms of the velocity equal to mean velocity) in the near-wall region, where turbulence intensities based on local mean velocity as large as 100% can occur. Taking into account also the experimental uncertainty associated with the LDV technique (as evaluated by Modarress et al. [16]), the overall mean velocity uncertainty is still conned within 4%.

1.0 0.9 0.8 0.7 0.6

1.0 0.9 0.8 0.7

0.5 0.4 0.3 0.2 0.1

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76 X = 0.88

0.6 0.5 0.4 0.3 0.2 0.1

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76 X = 0.88

0.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

0.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

U / Um

V / Um

Fig. 4. Proles of normalized streamwise U/Um and spanwise V/Um velocity components, at selected streamwise X location, in the XZ plane at Y = 0.1. Channel with one ribbed wall (1RW).

F. Satta et al. / Experimental Thermal and Fluid Science 37 (2012) 4656

51

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76 X = 0.88

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76 X = 0.88

U / Um

V / Um

Fig. 5. Proles of normalized streamwise U/Um and spanwise V/Um velocity components, at selected streamwise X location, in the XZ plane at Y = 0.3. Channel with one ribbed wall (1RW).

(0.5 < Z < 1), the rib presence gives rise to a spanwise velocity component, which is close to zero or slightly negative just downstream of the rib (smaller X values) while it grows to positive values as the X coordinate increases. On the contrary, for Z < 0.5, no spanwise velocity components can be recognized in this plane (Y = 0.1). As the Y-coordinate is increased to 0.3 (Fig. 5), the magnitude of U/Um is generally reduced in the channel region close to the ribbed wall (Z > 0.5), while it increases close to the smooth wall (Z < 0.5). The U/Um proles have the maximum (U/Um = 1.4) located at Z = 0.3, regardless of the X value. Streamwise velocity proles recorded at the midplane (Y = 0.5, Fig. 6) are qualitatively similar to those obtained over the previous plane. The comparison of the spanwise velocity distributions measured for 0.1 < Y < 0.5 (Figs. 46) clearly shows that, as Y coordinate

increases, the U/Um proles associated with the different streamwise positions tend to collapse together in the region close to the ribbed surface, where all the distributions show positive spanwise velocity component values. As the smooth surface is approached, the ow returns towards the opposite side wall with a negative spanwise velocity. For further increases of Y-coordinate to 0.7 (Fig. 7) and 0.9 (Fig. 8), values of U/Um close to the ribbed wall remain nearly unchanged, whereas reductions in U/Um are observed rst in the region close to the smooth wall (passing from Y = 0.50.7) and then in the central region (from Y = 0.7 to 0.9). This streamwise velocity reduction seems to be directly linked to the progressive increase in the spanwise velocity magnitude taking place as the Y coordinate increases.

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76 X = 0.88

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76 X = 0.88

U / Um

V / Um

Fig. 6. Proles of normalized streamwise U/Um and spanwise V/Um velocity components, at selected streamwise X location, in the XZ plane at Y = 0.5. Channel with one ribbed wall (1RW).

52
1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

F. Satta et al. / Experimental Thermal and Fluid Science 37 (2012) 4656

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76 X = 0.88

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76 X = 0.88

0.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

U / Um

V / Um

Fig. 7. Proles of normalized streamwise U/Um and spanwise V/Um velocity components, at selected streamwise X location, in the XZ plane at Y = 0.7. Channel with one ribbed wall (1RW).

What was observed in the U/Um and V/Um plots associated with the ow in the whole channel can be ascribed to one-cell secondary ow, bringing uid from the ribbed wall (Z = 1) to the smooth wall (Z = 0). More precisely, secondary ow moves along the rib prole from the leading side-wall to the trailing side-wall; when the trailing side-wall is approached, secondary ow turns and moves along the smooth wall. The secondary ow clearly affects the heat transfer characteristics, since it brings core uid at relatively low temperature towards the region close to the leading side-wall, thus promoting high levels of heat transfer coefcient. Conversely, the warmer uid moving towards the opposite trailing side-wall is expected to lead to a progressive reduction of the heat transfer performance in the spanwise direction.

The distributions of the Nusselt number along the streamwise direction X are plotted in Fig. 9. As previously mentioned, the highest values of Nu are found close to the leading side-wall, at the lowest Y value (Y = 0.1). Nu values have a relative minimum as the X coordinate is inside the ow separation region or the streamwise velocity is weak (X % 0.1); for increasing values of X, Nu increases up to a relative maximum, due to the ow reattachment and local uid acceleration, then slightly decreases due to the re-development of ow, and nally, after a second relative minimum, increases again, probably due to an intense ow recirculation induced by the downstream rib. As Y coordinate is increased up to 0.5, the streamwise velocities close to the ribbed wall become lower than those recorded at Y = 0.1, as inferred from the comparative inspection of Figs. 46; as a consequence, the

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76 X = 0.88

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76 X = 0.88

0.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

U / Um

V / Um

Fig. 8. Proles of normalized streamwise U/Um and spanwise V/Um velocity components, at selected streamwise X location, in the XZ plane at Y = 0.9. Channel with one ribbed wall (1RW).

F. Satta et al. / Experimental Thermal and Fluid Science 37 (2012) 4656

53

240 220 200 180

trailing side-wall

110
Y=0.1 Y=0.3 Y=0.5 Y=0.7 Y=0.9

115

Nu

160 140 120 100 80 0.0

120 125 135 145 160 180 200 220 240

0.2

0.4

0.6

0.8

1.0
Fig. 10. Contours of lines at equal Nusselt number for the channel with one ribbed wall (1RW).

X
Fig. 9. Nusselt number distributions along the streamwise X direction at various spanwise Y locations. Channel with one ribbed wall (1RW).

magnitude of Nu values is progressively reduced. It is conjectured that, owing to the ow mixing induced by the secondary ow, the separation zone beyond rst rib becomes less distinguishable from the boundary layer development; as a result, the Nu proles, sufciently far from the leading side-wall, are less X-dependent. Finally, Nu distributions recorded for higher values of Y (up to 0.9) are similar to that for Y = 0.5, probably due to the fact that the ow structure, deduced from Fig. 7 (Y = 0.7) and Fig. 8 (Y = 0.9), is not altered in the vicinity of the ribbed wall as the trailing side-wall is approached. Inspection of Fig. 10, showing the contours of lines at constant Nu number, reveals that the region of the heated plate conned within the rib leading end and the leading side-wall is characterized by high heat transfer coefcients, due to the impingement of the core uid carried by the secondary ow. Conversely, low Nu number values are observed close to the opposite region of
1.0 0.9 0.8 0.7 0.6

the heated surface, downstream of the trailing edge of the rib. This region is washed by the uid carried by the secondary ow, preheated by the thermal exchange with the part of heated surface close to the leading side-wall and featured by relatively low streamwise velocities, resulting in the lowest heat transfer zone. 3.2. Channel with ribs on two walls Results for the channel with angled ribs on two opposite sides (2RW case) are reported in Figs. 1113 (streamwise and spanwise normalized velocity components) and in Figs. 14 and 15 (streamwise Nusselt number distributions and iso-Nu contour maps). Velocity components are presented only at Y = 0.1, 0.5 and 0.9, due to space limitation. As expected, symmetric distributions of U/Um and V/Um are encountered, due to the symmetry of the rib arrangement, but the features of the velocity proles are markedly affected by the Y coordinate.
1.0 0.9 0.8 0.7 0.6

X = 0.12 X = 0.28 X = 0.44

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76

0.5 0.4 0.3 0.2 0.1

X = 0.60 X = 0.76 X = 0.88

0.5 0.4 0.3 0.2 0.1

0.0 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

0.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

U / Um

V / Um

Fig. 11. Proles of normalized streamwise U/Um and spanwise V/Um velocity components, at selected streamwise X location, in the XZ plane at Y = 0.1. Channel with two ribbed walls (2RW).

54
1.0 0.9 0.8 0.7

F. Satta et al. / Experimental Thermal and Fluid Science 37 (2012) 4656

X = 0.12
0.6

X = 0.12
0.6

X = 0.28

X = 0.28 X = 0.44 X = 0.60 X = 0.76

0.5 0.4 0.3 0.2 0.1

X = 0.44 X = 0.60 X = 0.76 X = 0.88

0.5 0.4 0.3 0.2 0.1

0.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

0.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

U / Um

V / Um

Fig. 12. Proles of normalized streamwise U/Um and spanwise V/Um velocity components, at selected streamwise X location, in the XZ plane at Y = 0.5. Channel with two ribbed walls (2RW).

At Y = 0.1 (Fig. 11), the uid regions just downstream of the ribs (X 6 0.28) and within a distance of one rib height from the ribbed surfaces (Z 6 0.15 and Z P 0.85) are characterized by very small streamwise velocities (and even negative values at X = 0.12 and Y = 0.05 and 0.95) and relatively high spanwise velocities. Elsewhere, distributions of U/Um become almost at along the crosssection with values, sufciently far from the ribbed walls, ranging from 1.3 to 1.6 at lowest X values to 1.251.4 at highest X values. The larger streamwise velocity values occurring at the center of the channel at the lowest X values, as compared with those at the highest X values, can be ascribed to the channel contraction induced by the boundary layer separation which takes place just

downstream of the rib. In the 2RW case, differently from what occurs in the 1RW case, the spanwise velocity distributions show non-zero values within the whole channel, revealing a secondary ow directed along the rib prole in the regions next to the ribbed walls, and along the opposite direction in the middle of the channel. At Y = 0.5 (Fig. 12), U/Um proles are nearly parabolic, with a maximum of 1.41.45 at midheight, irrespective of X coordinate. Even in this Y plane, the V/Um values close to ribbed walls are positive, whereas in the middle of the channel they become negative and insensitive to X coordinate. This result clearly indicates that the secondary ow has the shape of a pair of counter-rotating

1.0 0.9 0.8 0.7 0.6

1.0 0.9 0.8 0.7 0.6

X = 0.12 X = 0.28 X = 0.44

X = 0.12 X = 0.28 X = 0.44 X = 0.60 X = 0.76 X = 0.88

0.5 0.4 0.3 0.2 0.1

X = 0.60 X = 0.76 X = 0.88

0.5 0.4 0.3 0.2 0.1

0.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

0.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

U / Um

V / Um

Fig. 13. Proles of normalized streamwise U/Um and spanwise V/Um velocity components, at selected streamwise X location, in the XZ plane at Y = 0.9. Channel with two ribbed walls (2RW).

F. Satta et al. / Experimental Thermal and Fluid Science 37 (2012) 4656

55

280 260 240 220 200

Y=0.1 Y=0.3 Y=0.5 Y=0.7 Y=0.9

Nu

case, local distributions show progressive spanwise reductions of Nu throughout the entire range of Y (from 0.1 to 0.9). As for the 1RW case, the distribution of the local Nusselt number is consistent with the measured ow eld described by Figs. 1113. Iso-Nu contours reported in Fig. 15 are again characterized by relatively high Nu levels close to the leading side-wall, while Nu is considerably lowered as the opposite region, towards the trailing side-wall, is approached. Comparison of lines at equal Nu with those reported in the literature for similar experimental conditions [6,17] shows reasonable agreement. 4. Conclusions Laser-Doppler velocimetry and liquid crystal thermography have been employed to study the ow and thermal characteristics in a rectangular channel (aspect ratio AR = 5) with repeated, 45 angled, ribs on one or two opposite walls. The rib pitch-to-height ratio P/e was 10, with the Reynolds number Re = 29,000. The region between a pair of consecutive ribs, sufciently far from the test section entrance, has been investigated. The combined ow and thermal analysis made it possible to gain insight into the physical process of the ribbed channel ow and heat transfer; a strong correlation between the ow structure between adjacent ribs and the heat transfer behavior at the wall was found. Moreover, the present measurement data provide a contribution to validation of CFD codes. For the channel with angled ribs on one wall only, a one-cell secondary ow carries the uid from the ribbed wall to the smooth wall close to the trailing side-wall and against the ribbed wall close to the opposite leading side-wall. This ow impingement results in high heat transfer coefcients in the region of the ribbed plate between the rib leading end and the leading side-wall, where the highest streamwise velocities are detected. Marked spanwise reductions in the heat transfer coefcient and streamwise velocity occur as the opposite side-wall is approached. When angled ribs are installed, parallel and in-line, on two opposite walls of the channel, the secondary ow has the form of two counter-rotating cells, transporting uid from the central core region towards the ribbed walls, again in the vicinity of the leading side-wall, as in the 1RW case. Heat transfer characteristics have a shape similar to that encountered for the one-ribbed wall channel, but levels of high transfer coefcient are higher. Acknowledgment This research has been nancially supported by Ansaldo Energia, Genova, Italy.

180 160 140 120 100 0.0

0.2

0.4

0.6

0.8

1.0

X
Fig. 14. Nusselt number distributions along the streamwise X direction at various spanwise Y locations. Channel with two ribbed walls (2RW).

trailing side-wall

115 120 125 130 135 145 160 180 200 220 250 280
leading side-wall
Fig. 15. Contours of lines at equal Nusselt number for the channel with two ribbed walls (2RW).

cells, as found in [9] for a similar rib arrangement (rectangular channel with AR = 8 and 60 angled ribs deployed in-line on two opposite walls). Due to mass conservation, the secondary ow velocity close to the ribbed wall is of larger magnitude than the velocity in the middle of the channel. Finally, at Y = 0.9 (Fig. 13), a generalized reduction in the streamwise velocity, at any X and Z locations, is encountered, with each U/Um prole reaching, for any X value, a relative minimum at midheight (U/Um = 0.75), and two peaks (U/Um = 0.80.85) located between the midplane (Z = 0.5) and the ribbed walls (Z = 0 and 1). The large negative spanwise velocity recorded in the middle of the channel, associated with the counter-rotating cells, may explain the local reduction of the streamwise velocity, which is provoked by the momentum transfer operated by the vortices. Local Nu number distributions (Fig. 14) are qualitatively similar to those obtained in the case of one ribbed wall, even if values are typically higher, due to the larger velocities induced by installation of ribs on two opposite sides of the channel; moreover, in the 2RW

References
[1] J.C. Han, L.R. Glickmann, W.M. Rohsenow, An investigation of heat transfer and friction for rib-roughened surfaces, Int. J. Heat Mass Transfer 21 (1978) 1143 1156. [2] J.C. Han, J.S. Park, Developing heat transfer in rectangular channels with rib turbulators, Int. J. Heat Mass Transfer 31 (1988) 183195. [3] J.C. Han, Y.M. Zhang, C.P. Lee, Augmented heat transfer in square channels with parallel, crossed, and V-shaped angled ribs, ASME J. Heat Transfer 113 (1991) 590596. [4] G. Rau, M. akan, D. Moeller, T. Arts, The effect of periodic ribs on the local aerodynamic and heat transfer performance of a straight cooling channel, ASME J. Turbomach. 120 (1998) 368375. [5] C.O. Olsson, B. Sundn, Experimental study of ow and heat transfer in ribroughened rectangular channels, Exp. Thermal Fluid Sci. 16 (1998) 349365. [6] H. Iacovides, G. Kelemenis, M. Raisee, Flow and heat transfer in straight cooling passages with inclined ribs on opposite walls: an experimental and computational study, Exp. Thermal Fluid Sci. 27 (2003) 283294. [7] B. Bonhoff, S. Parneix, J. Leusch, B.V. Johnson, J. Schabacker, A. Blcs, Experimental and numerical study of developed ow and heat transfer in coolant channels with 45 ribs, Int. J. Heat Fluid Flow 20 (1999) 311319.

56

F. Satta et al. / Experimental Thermal and Fluid Science 37 (2012) 4656 [14] W.K. George, Processing of random signals, in: Proc. of the Dynamic Flow Conference, Skovlunde, Denmark, 1978, pp. 757800. [15] F. Satta, D. Simoni, M. Ubaldi, P. Zunino, Experimental difculties in measuring separating boundary layers with the LDV technique, in: Proc. of the XVIII Symposium on Measuring Techniques in Turbomachinery Transonic and Supersonic Flow in Cascades and Turbomachines, Thessaloniki, Greece, 2006, pp. 7582. [16] D. Modarress, H. Tan, A. Nakayama, Evaluation of signal processing techniques in laser anemometry, in: Proc. of IV International Symposium on Application of Laser Anemometry to Fluid Dynamics, Lisbon, Portugal, 1988, pp. 1.20.1 1.20.6. [17] M.E. Taslim, T. Li, D.M. Kercher, Experimental heat transfer and friction in channels roughened with angled, V-shaped, and discrete ribs on two opposite walls, ASME J. Turbomach. 118 (1996) 2028.

[8] D. Chanteloup, Y. Juaneda, A. Blcs, Combined 3-D ow and heat transfer measurements in a 2-pass internal coolant passage of gas turbine airfoils, ASME J. Turbomach. 124 (2002) 710718. [9] X. Gao, B. Sundn, PIV measurement of the ow eld in rectangular ducts with 60 parallel, crossed and V-shaped ribs, Exp. Thermal Fluid Sci. 28 (2004) 639653. [10] G. Tanda, Effect of rib spacing on heat transfer and friction in a rectangular channel with 45 angled rib turbulators on one/two walls, Int. J. Heat Mass Transfer 54 (2011) 10811090. [11] R. Kiml, S. Mochizuki, A. Murata, Effects of rib arrangements on heat transfer and ow behavior in a rectangular rib-roughened passage: application to cooling of gas turbine blade trailing edge, ASME J. Heat Transfer 123 (2001) 675681. [12] G. Tanda, R. Abram, Forced convection heat transfer in channels with rib turbulators inclined at 45, ASME J. Turbomach. 131 (2009) 021012. [13] G. Tanda, Heat transfer in rectangular channels with transverse and V-shaped broken ribs, Int. J. Heat Mass Transfer 47 (2004) 229243.