流场测试与分析

fi d a Be gy, O form tio of ors o as k or by aria rs. � 2003 Elsevier Inc. All rights reserved. enhance the heat transfer process. Thus far most of the publications have obtained heat different rib configurations are distinct from each other. The flow characteristics in rib-roughened ducts are helpful as one likes to understand the complicated heat transfer distributions. In addition, to improve the per- square perpendicular ribs at a Reynolds number of 33,000. Hirota et al. [8] examined the secondary flow valuable correlation between the heat transfer promo- tion and the flow field. However, most available data are limited to perpendicular ribs in square ducts or high aspect ratio rectangular ducts at high Reynolds num- bers. In addition, because LDV usually makes mea- surements at a single point, it has the limitation that it cannot capture the simultaneous spatial structure of a d Scie *Corresponding author. Tel.: +46-46-222-86-05; fax: +46-46-222- transfer results. Experimental techniques such as liquid crystal thermography [1,2], holographic interferometry [3] and thermocouples [4], have been used to demon- strate the detailed heat transfer distributions on the duct walls. Numerical methods using different turbulence models [5] or large eddy simulation [6] are used to pre- dict the flow field and evaluate the cooling configura- tions. The detailed heat transfer distributions caused by patterns caused by perpendicularly arranged square ribs in a square duct. Iacovides et al. [9] reported the tur- bulent flow through a square sectioned U-bend of strong curvature with two walls roughened with square ribs, for both stationary and rotating conditions, through de- tailed LDV measurements. Liou et al. [10] captured the longitudinal vortex pairs induced by 12 differently shaped vortex generators by LDV and provided a very Keywords: Rib-roughened duct; Particle image velocimetry; Heat transfer enhancement; Secondary flow 1. Introduction Artificial roughness elements such as periodic ribs in rectangular ducts for heat transfer enhancement are found in a wide range of applications, e.g., compact heat exchangers, turbine cooling schemes, etc. The presence of periodic ribs changes the flow characteristics in the ducts and induces different secondary flows so as to formance of CFD codes, a validation of the predictions is necessary and detailed measurements of the flow structure in the flow passages are thus required to pro- vide data bases for comparison. Most flow field data available were obtained with laser doppler velocimetry (LDV) or hot wires. Liou et al. [7] measured mean velocities, turbulence intensities and Reynolds stresses by LDV in a 1:2 rectangular duct with PIV measurement of the flow 60� parallel, crosse Xiufang Gao, Division of Heat Transfer, Lund Institute of Technolo Received 12 December 2002; received in revised Abstract The flow characteristics in rectangular ducts with aspect ra locimetry to reveal the function of the inclined ribs as turbulat the wide walls and the rib height-to-hydraulic diameter ratio w showed strong effects of the inclined ribs on the flow behavi streamwise velocity component was detected to have spanwise v corresponding heat transfer results obtained by the same autho Experimental Thermal and Flui 86-12. E-mail address: bengts@emvox2.vok.lth.se (B. Sund�en). 0894-1777/$ - see front matter � 2003 Elsevier Inc. All rights reserved. doi:10.1016/j.expthermflusci.2003.10.005 eld in rectangular ducts with nd V-shaped ribs ngt Sund�en * le Romers Vag 1, Box 118, SE-221 00 Lund, Sweden 23 September 2003; accepted 29 October 2003 1:8 were investigated experimentally using particle image ve- r generators of favorable flows. Circular ribs were attached on ept as 0.06 with an attack angle of 60�. The obtained results the development of secondary flows. The magnitude of the tion. The flow characteristics were correlated with the previous nce 28 (2004) 639–653 www.elsevier.com/locate/etfs flow field. 2. Experimental description 640 X. Gao, B. Sund�en / Experimental Thermal and Fluid Science 28 (2004) 639–653 In the case of inclined ribs, the secondary flow gen- erated along the ribs will act as an additional heat transfer mechanism to elevate the heat transfer process to a higher level (see, e.g., [11]). The uneven distributions of heat transfer were conjectured to be caused by the complicated flow patterns induced by the ribs. However, flow field results regarding inclined ribs in ducts are very limited. Kiml et al. [12,13] performed flow visualization in rectangular ducts (duct aspect ratios were 1/2 and 1/5) and showed strong effects of the ribs on the base flow behavior. Bonhoff et al. [14] investigated experimentally Nomenclature b half width of the wide side wall (m) c half width of the narrow side wall (m) Dh hydraulic diameter (m) e rib height (m) N number of samples p rib pitch (m) Re Reynolds number (dimensionless), Re ¼ UmDh=m Ui instantaneous streamwise velocity compo- nent (m/s) Um mean bulk velocity (m/s) U streamwise mean velocity component along x-direction (m/s) V spanwise mean velocity component along y-direction (m/s) W transverse mean velocity component along z-direction (m/s) and numerically the flow field in a square channel with 45� ribs at a Reynolds number of 50,000. Particle image velocimetry, or PIV, is a nonintrusive optical method. It is possible to obtain instantaneous velocity maps in a flow plane of interest. By post-pro- cessing, the mean velocity maps and instantaneous or average vorticity maps can be obtained (see, e.g., Ref. [15–18]). Son et al. [18] used PIV to study the flow behavior in a two-pass square channel with a smooth wall and a 90� ribbed wall, and then correlated the turbulent flow and the wall heat transfer characteristics. A novel method of generating mean flow-heat transfer correlation and turbulent kinetic energy-heat transfer correlation was adopted to give an insight of the quantitative relationship. The purpose of this study is to implement the PIV measurement technique to study air flow in rectangular ducts with 60� ribs at a relatively low Reynolds number of 5800. The Reynolds number range of lower than 10,000 occurs often in compact heat exchangers such as car radiators, and circular ribs are used by the authors to simulate the round shape in those cases. The results are used to analyze and contribute to the understanding 2.1. Test section A sketch of the experimental setup is shown in Fig. 1. of the heat transfer augmentation mechanism according to the former heat transfer data [1] in ducts and to grasp experimentally the characteristics of the flow field introduced by different rib configurations. u0 rms velocity in streamwise direction, u0 ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 N�1 Pi¼N i¼1 ðUi � UÞ2 q (m/s) v0 rms velocity in spanwise direction, v0 ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 N�1 Pi¼N i¼1 ðVi � V Þ2 q (m/s) w0 rms velocity in transverse direction, w0 ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 N�1 Pi¼N i¼1 ðWi � W Þ2 q (m/s) x coordinate along duct length direction (m) y coordinate along spanwise direction (m) z coordinate along transverse direction (m) a rib angle of attack (�) n coordinate perpendicular to the ribs (m) g coordinate along the rib orientation (m) f coordinate perpendicular to n–g plane (m) Air was used as working medium as in the heat transfer experiments. The plexiglas rectangular duct consists of two parts, each is 700 mm long, with an internal cross- section of 14.5 · 112.5 mm. The second part of duct was roughened by 1.5 mm circular plastic ribs on two opposite wide walls. To facilitate the positioning of the ribs on the wall, 0.10 mm thick plastic film was used as templates and both the ribs and the templates were fixed by double-sided tape. At the measurement location, approximately 50 < x=Dh < 52 from the inlet of the rectangular channel, no template was used instead the ribs were fixed directly on the walls using double-sided tape, so that the camera could take images through the walls. The rib height-to-hydraulic diameter was about 0.06 with a pitch-to-height ratio of 10. The detailed description has been given in [1] and will not be repeated here. 2.2. Seeding particles A TSI model 9306 six-jet atomizer supplied with compressed air was used to generate glycerol particles with a size in the order of 1 lm. The pressure of com- the two walls with both staggered and in-line arrange- ments were included. In the cases of staggered ribs, wall A from x=Dh � 50 to the exit, through which the PIV images were caught, was either roughened by ribs or left Compressed air Lab air RotameterFan Test section Laser sheet Optical mirror CCD camera Camera controller Plenum with grids Plenum with filter Nd:YAG laser Atomizer he ex Flow Ribs on wall A Ribs on wall B 60° Bottom wall Upstream end of ribs Downstream end of ribs Laser sheet x y z Fig. 3. Configuration of ribbed duct. X. Gao, B. Sund�en / Experimental Thermal and Fluid Science 28 (2004) 639–653 641 pressed air was set at about 137895.2 Pa and 1, 2 or 3 jets were open in the current experiments to achieve desired particle density of 5–10 particles in each inter- rogation area for different measurements. The generated glycerol aerosol and laboratory air entered the inlet plenum via a 100 mm tube. The plenum was equipped with two grids to evenly distribute the seeding particles. In PIV measurement, velocities of the particles are measured. Thus the Stokes number of the particles must be sufficiently small. The Stokes number is the ratio between the particle response time and the flow time scale. The particles must, however, be so large that sufficient light is scattered for detection by the CCD camera. There are particles available offering good scattering efficiency and sufficiently small velocity lag. In gas flow PIV measurement aluminum, magnesium and glycerol particles have been employed. In this investi- gation, every effort was taken to satisfy these conditions. 2.3. Rib configurations The rib configurations tested are shown in Fig. 2. The Fig. 1. Sketch of t four walls of the duct are designated as wall A, wall B, top wall and bottom wall, as shown in Fig. 3. Parallel ribs, crossed ribs and V-shaped ribs with an angle of 60� to the main flow direction, have been applied on the two wide walls. In the channels with parallel ribs, the ribs on (a) (b) (d) (e) Fig. 2. Rib configurations; (a) Staggered parallel ribs; (b) Staggered-single pa downstream; and (f) V-ribs pointing upstreams. Computer PIV processor perimental setup. 2b 2c 2b=112.5 mm 2c=14.5 mm 700 mm Top wall smooth with only wall B roughened at this location. These will be termed as staggered and staggered-single parallel ribs. The idea was to compare with the heat transfer results in [11], to prove the assumption that the staggered-single parallel ribs at the measurement (c) (f) rallel ribs; (c) In-line parallel ribs; (d) Crossed ribs; (e) V-ribs pointing same applies to the n coordinate for plane C measure- ments. The symmetry plane and the traversing central plane are where the y and z coordinates start, respec- tively. In the cases of parallel and crossed rib-roughened channels, for plane A measurements, the field of view of the PIV images was set approximately as 115 mm · 144 mm to capture the whole field flow behavior between a pair of ribs. The camera was then moved closer to the measurement location to focus on a smaller field of approximately 30 mm · 35 mm in the central area of the channel to capture the detailed flow field between adjacent ribs. In the cases of V-shaped ribs, the view field of plane A was adjusted to be approximately 60 mm · 75 mm, so that the flow in the upper half of the rib-roughened duct could be observed. The view field was about 15 mm · 19 mm for planes B and C for all ribs. The choice of the interrogation area for post- processing is a reasonable compromise between spatial resolution and velocity dynamic range. To maintain a high measurement accuracy the maximum particle dis- rmal and Fluid Science 28 (2004) 639–653 location would not deviate the heat transfer profile far from the case of staggered arrangement on two walls. 2.4. PIV system A Quantel Q-Switched Nd:Yag double cavity pulsed laser provides pulsed light sheets at a wavelength of 532 nm with a maximum output energy of 120 mJ. Each pulse duration is 10 ns. The time interval between two pulses can be adjusted according to the velocity range and for different test planes. The desired laser thickness can be obtained by adjusting the distance between the built-in lenses, and it was set to approximately 0.8 mm during the experiments. The laser head was kept hori- zontal with an optical mirror to turn the laser sheet 90� to enter the window on the top of the channel vertically. The optical mirror was installed on a traversing system with fine scale so that the position of the laser sheet could be precisely controlled by adjusting the traversing system. A Dantec 80C60 HiSense PIV camera with a high performance progressive scan interlines CCD chip hav- ing 1024 · 1280 pixels, together with a Nikon AF Micro 60 f/2.8D lens were used to record the seeding images on double frames. An optical filter of 532 nm was used in front of the camera to allow only the signal illumination wavelength to pass. A PIV 2001 processor and a camera controller were used together with a PC to synchronize, communicate and process. 2.5. Measurement planes For the investigation of the three-dimensional fea- tures of the flow, the test facility was designed in such a way that it allows turning the test section by an angle of 90� without changing the flow conditions. The end of the inclined ribs that is closest to the flow inlet is called the upstream end of the ribs hereafter, and the other end is called the downstream end as illustrated in Fig. 3. The ribs were painted black to decrease the noise caused by light scattered by the ribs, which would then cause invalid velocity vectors. Typically, PIV measure- ments were carried out in three planes in three direc- tions. During the tests of plane A, the wide ribbed walls were placed vertical and the vertical laser sheet illumi- nated the flow field from top to bottom to capture the x– y plane velocity field, as shown in Figs. 3 and 4. The laser sheet position along the z-direction could be ad- justed. The wide ribbed walls were then turned to be horizontal, which gave the possibility to capture the flow field of planes B and C as shown in Fig. 4. The location of plane B could be adjusted along axis y. The ribs themselves blocked some part of the camera view field which made the velocity field right behind the ribs impossible to be obtained in plane B. Therefore, the 642 X. Gao, B. Sund�en / Experimental The laser sheet was adjusted to be at 90� with respect to the rib orientation such as plane C and the camera was put along the rib direction to obtain the flow field between consecutive ribs. The velocity field of plane C was only obtained at the location with the laser sheet crossing the center of the upstream rib. Two coordinate systems are defined in Fig. 4. Because the flow field is fully devel- oped, and also in order to simplify later presentation of the results at different locations, the x coordinate does not originate from a fixed physical location but instead the center of the first rib on wall B is used as x ¼ 0 for plane B at each measured location in all figures. The Plane B Flow x y Flow Plane C Plane A Flow xz wall A wall B ξ η Fig. 4. Laser sheet arrangement for PIV measurements. placement between successive images should be less than accuracy of the final velocity measurements, due to inadequacies of the apparatus and the uncertainties rmal connected with the PIV technique. The resulting errors in the PIV measurements can be estimated from the imperfections of the apparatus, the particular PIV acquisition parameters and the nature of the flow being measured. The dimensions of the plexiglas duct are accurate within ±0.1 mm and the rib height within ±0.05 mm. The position of the ribs on the wall is estimated to be within ±0.5 mm. The positional accuracy of the laser sheet with respect to the duct is a key factor. This effect is expected to be most profound close to the wall, where the velocity magnitude is relatively low while the velocity gradient is high. The laser sheet was adjusted to be vertical by marking the transverse center on both the top and bottom walls, until the laser sheet went through both lines. The V velocity component at the transverse center was then measured in the smooth duct. The magnitude was found to be less than 0.005 m/s over the region of �0:56 y=b6 þ0:5. Thus one can be sure that the laser sheet went through the vertical transverse central plane. The relative position of the vertical laser sheet was then accurately controlled by the traversing system. There are a few inherent factors of PIV that affect the 1/4 of the size of the interrogation area. Based on a pixel displacement level, the dynamic range in a PIV mea- surement may be defined as the displacement divided by the sub-pixel accuracy (see, e.g., Dantec [19]). Using 32 · 32 pixel interrogation areas, the displacement is typically 8 pixels. This corresponds to a displacement where the time interval between the two light pulses is chosen so the displacement is approximately 25% of the interrogation area. This has been adopted for all mea- surement in all planes. The sub-pixel accuracy is a function of many parameters, and as a rule-of-thumb 0.1 pixel accuracy is a realistic value recommended by Dantec to use. This results in a dynamic range of 8/ 0.1¼ 80. The PIV recordings from the CCD camera were processed with the Dantec software FlowManager. A cross-correlation analysis method was used with an interrogation area size of 32 · 32 pixels with 50% overlap between the interrogation areas. The vector field was then evaluated by a predefined velocity magnitude and the invalid ones were replaced by the moving average method. Instantaneous velocity fields were obtained and a series of 100–250 of instantaneous measurements were statistically averaged to get the mean velocity field. 3. Uncertainty analysis There are a number of aspects that can affect the X. Gao, B. Sund�en / Experimental The final accuracy. To name a few, the imperfectness of the camera lens, the slight magnification variation depend- ing on the position on the CCD chip, the deviation of the light sheet from one area to another. Fortunately these errors are small and in practice they are conve- niently ignored. For different measurements in planes A–C, the spatial resolution of the images varied from 9 to 68 pixels/mm. The software interpolation accuracy of 0.1 pixel is expected. The flow field in the duct is three-dimensional, which means the existence of an out-of-plane velocity in the two-dimensional measurement. Due to the parallax ef- fects, the through-plane movement of particles can influence the accuracy of the in-plane measurements. Its magnitude is directly related to the distance from the point in the flow field to the axis of the camera. This error source was controlled by selecting measurement planes which included the high velocity component as an in-plane component and by adjusting the time interval between continuous pulses and the laser sheet thickness (see, e.g., Dantec [19]). The mean velocity field and the turbulence quantities are of great practical interests. However, the accuracy of the obtained flow parameters is affected by the sample size, as analyzed in Grant and Owens [20]. Therefore, the dependence of the mean and rms velocity compo- nents on the number o