Experimental studies on single-phase flow and heat transfer in a narrow rectangular channel
Jian Ma,Longjian. Li,Yanping Huang,Xiaozhong Liu
DOI: https://doi.org/10.1016/j.nucengdes.2011.04.047
IF: 1.7
2011-01-01
Nuclear Engineering and Design
Abstract:The state-of-the-art studies on single-phase flow and hear transfer in narrow rectangular channels shows some difference in terms of the agreement with the conventional theory. To further make clear this issue, the experimental studies on single-phase flow and heat transfer in a narrow rectangular channel with deioned water as test fluid was carried out. The narrow rectangular channel had the following dimensions: depth ( e ) = 2 mm, aspect ratio ( e / b ) = 0.05, length to diameter ratio ( L / D h ) = 300, and mean wall relative roughness ( ɛ / D h ) = 8.3 × 10 −4 . The experiments were performed with the channel oriented uprightly. The parameters that were varied during the experiments included the mass flow rate, inlet temperature and heat flux. Based on the measured temperatures, mass flow rates, pressure drops and heat fluxes, the isothermal and non-isothermal friction factors and the local and mean Nusselt numbers have been calculated. The correlations for the isothermal friction factors and the mean Nusselt numbers have been developed, and have a satisfactory agreement with the conventional theory. Based on the property ratio method, the correlations for non-isothermal friction factors have been proposed, but the new exponent ( m ) for modifying variable-property effect need to be developed. Nomenclature A c cross-sectional area, m 2 A s surface area of the fluid channel, m 2 a thermal diffusivity, m 2 /s b channel width, m C 1 constant C 2 constant C f Fanning friction factor C p specific heat at constant pressure, J/kg K D h hydraulic diameter, D h = 4 A c / P , m e channel depth, mm f Darcy friction factor, f = 4 C f G mass flow flux, kg/m 2 s H specific enthalpy, kJ/kg h heat transfer coefficients, W/(m 2 K) I current, A L total channel length, m L h heated length of channel, m L p total pressure-measuring distance, m m mass flow rate or exponent, kg/s Nu Nusselt number, Nu = hD h / λ f Pr Prandlt number, Pr = Cpμ / λ f P wet perimeter, m p pressure, MPa Q net power gained by the fluid, W q surface heat flux, W/m 2 q internal thermal source per volume, W/m 3 Re Reynolds number, Re = GD h / μ T temperature, °C V voltage or volume, V or m 3 x + dimensionless axial distance x axial distance, m Greek symbols α channel aspect ratio, α = e / b Δ difference δ heating plate thickness, mm ɛ wall absolute roughness, m η energy balance λ thermal conductivity, W/(m K) μ dynamic viscosity, kg/(m s) ν kinetic viscosity, m 2 /s ρ density, kg/m 3 Subscripts 0 room temperature CP constant property cr critical c cross-section F friction f fluid g gravitation in inlet iso isothermal l laminar m mean non non-isothermal out outlet s solid t turbulent wi inner-wall wo outer-wall w wall x local 1 Introduction In recent decade, plate-type fuels assembly which characterize compact configuration, high heat removal capability and many other thermal-hydraulic advantages have been employed in some late-model nuclear reactors in the world. For example, the JRRR of Japan and the CARR of China with such types of fuels assembly are now in commission. In such types of nuclear reactors, the coolant channels have typical narrow rectangular cross-sections with the depth between 1 mm and 3 mm and the large width to depth ratio, and have the hydraulic diameter between 1 mm and 5 mm. According to the classification method for different channel's size proposed by Kandlikar (2002) , Shen et al. (2001) and Guo (2000) , it can be concluded that the continuum theory and the Navier–Stokes equation are still valid, and the conventional theory that satisfactorily predicts single-phase flow and heat transfer in large-sized channels should has been appropriate for such coolant channels. However, a lot of results given in the open literatures exhibited some difference in terms of the agreement with the conventional theory. One of the earliest studies on single-phase flow and heat transfer in a narrow rectangular channel with the depth ( e ) = 2 mm was carried out by Kays and London (1952) . The results of laminar and turbulent heat transfer agreed with the case for conventional channels. You (1989) studied the turbulent heat transfer in a narrow rectangular channel with the depth ( e ) = 3 mm. The results also agreed with the conventional case. Webb and Zhang (1998) studied the single-phase flow and heat transfer in a narrow rectangular channel with hydraulic diameters ranging between 960 μm and 2.13 mm, and found that the conventional correlations could predict heat transfer coefficients and friction factors well. Gao et al. (2002) measured the friction factor and the local Nusselt number along a narrow rectangular channels having a fixed width ( b ) = 25 mm and the depths ( e ) = 100–1000 μm. The critical Reynolds numbers observed ranged between 2500 for e / b equal to 0.04 and 4000 for e / b equal to 0.012. In the hydraulic diameters ranging between 199.2 μm and 1923 μm, the Poiseuille numbers had a good agreement with the conventional values. It seemed that there was no earlier transition to turbulence compared to conventional channels. It was observed that for the channels with a height larger than 0.4 mm the measured average Nusselt number agreed with the conventional correlations in the laminar regime. For the channels of a height less than 0.4 mm a significant decrease in the Nusselt number was observed. Wang and Peng (1994) analyzed the single-phase forced flow and heat transfer of water or methanol in several parallel narrow rectangular channels. In this experiment, the variable-property effect due to the increase in the temperature along the channels played an important role. The transition region is deduced by observing the trend of the mean Nusselt number. They found that the transition from laminar to turbulent regime occurs for Reynolds numbers ranging between 300 and 800. The mean Nusselt numbers in turbulent regime can be predicted by a simple correlation of Reynolds and Prandlt numbers. The correlation proposed is a modification of the Dittus–Boelter equation where the empirical coefficient 0.023 is replaced with 0.00805; that means that in this case the experimental Nusselt numbers were lower than the predictions of the conventional theory. A further analysis was conducted by Peng and Peterson (1995) using the same test sections analyzed by Wang and Peng (1994) . The effect of the variation of the fluid properties with the temperature for forced convection of water and methanol through narrow rectangular channels was investigated. They observed the laminar region for Reynolds numbers lower than 400. Also in this case, the transition region was deduced by observing the trend of the mean Nusselt number. The authors underlined that the heat transfer behavior in the laminar and transition regions was quite unusual and complicated. The authors evidenced that for a water flow the Reynolds number could double over the length of the channels. They used the Reynolds number at the inlet for their considerations. Toh et al. (2002) numerically investigated the flow and heat transfer in narrow rectangular channels. In their computation, the fluid physical properties were assumed variation with temperature. It was found that the rise of fluid temperature led to the drop of fluid viscosity and friction factor for the case of low Reynolds number. If a high temperature difference existed, the physical properties would fluctuate strongly. Xin and Shi (1994) investigated the forced water flow and heat transfer in narrow rectangular channels with six different sizes. In this case, the channel widths ranged between 0.15 mm and 0.8 mm, and the channel depths ranged between 0.9 mm and 2.7 mm. The measured critical Reynolds numbers of the transition from laminar to turbulent regime ranged between 1400 and 1800 for the case of single-phase isothermal flow, and the mean forced convective heat transfer coefficients were much higher than the results of the conventional channels. Xie et al. (2004) determined experimentally the single-phase flow and heat transfer in narrow rectangular channels that have the same length and the depth between 0.9 mm and 2.7 mm and the width to depth ratios between 3 and 9 for the refrigerant FC-72. It was found that the critical Reynolds numbers for the laminar to turbulent transition were between 750 and 1250, and were much lower than the critical Reynolds number of 2200 that is in conventional-sized rectangular channels. The correlations for friction factors and mean Nusselt numbers in these channels were also found quite deviated from the Blasius correlation and the Dittus–Boelter correlation respectively. Sudo et al. (1985) respectively studied the single-phase heat transfer of water for upward flow and downward flow in a narrow rectangular channel that have a cross-section of 2.25mm × 50 mm. It was found that the Dittus–Boelter correlation could be used to predict the heat transfer coefficients for both upward flow and downward flow for the case of asymmetrically heating and symmetrically heating for Reynolds numbers higher than 4000, and the buoyancy effect could be neglected. In the Reynolds numbers ranging from 2000 to 4000, the fluid flow was still in the laminar to turbulent transitional region. In the Reynolds numbers lower than 700, the heat transfer characteristics for asymmetrically heating, symmetrically heating, upward flow and downward flow were very different from each other, and the buoyancy effect might be significant. Afterwards, Sudo et al. (1990) carried out an experiment for the single-phase flow and heat transfer in a narrow rectangular channel having the depth of 2.5 mm, and found that the heat transfer coefficients were much higher than the conventional values and the forced convective flow and heat transfer were dependent on the developed boundary layer and independent of the buoyancy. The above mentioned studies on the single-phase flow and heat transfer in narrow rectangular channels shows some difference in terms of the agreement with the conventional theory. On the one hand, some results agree with the conventional theory well; on the other hand, some results obviously deviate compared to the conventional theory, and some opposite viewpoints are presented by different authors in the same parameter ranges, and the contrary conclusions are even drawn by the same authors in their successive studies. It is noticed that, except the authors in the literatures ( Wang and Peng, 1994; Peng and Peterson, 1995; Toh et al., 2002 ), a great many of authors hardly take the physical property variation with temperature into account in their studies. Furthermore, Mokrani et al. (2009) emphasized that the main causes of some disagreement with the conventional theory can be measurement errors and errors in estimating the parameters governing the convective heat transfer coefficient. They made a further explanation that in most experiments, flows are in the entrance developing zone, and because of the axial variation of the convective heat transfer coefficient the heat flux lines are greatly deformed in the wall's thickness, and some studies have compared Nusselt and Poiseuille numbers estimated for thermally or hydrodynamically undeveloped flows with the Nusselt and Poiseuille numbers of fully developed flows, and the use of one dimensional models to describe heat transfer in the microchannel walls can cause considerable error. With the above reviews being kept in mind, the objectives of this study were to carry out the experiments with deioned water as test fluid in a typical narrow rectangular channel and to clarify whether the single-phase flow and heat transfer agree with the conventional theory or not. 2 Experimental apparatus The schematic of the experimental apparatus is shown in Fig. 1 . The main components of the test apparatus are supply tank, pumps, pressurizer, filter, preheater, flowmeter, heat exchanger, valves and test section. The supply tank and the piston pump are provided for feeding deioned water for the test loop. The pressurizer used keeps the operational pressure in a steady state. The circulating pump forces liquid flow in the test loop duration of experiments. A filter is mounted upstream of the circulating pump for preventing from impurity. The preheater is provided upstream of the test section to preheat water at a required temperature. The liquid flow rate is measured using the flowmeter (Venturitube type). The liquid temperature at the exit of the test section is regulated using the heat exchanger. The test section is upright installed in the test loop, and consists of three layers of components: the core layer is a couple of heating plates with the grooves that make up the narrow rectangular channel (test channel); the middle layer is a couple of mica blocks that insulate the heating plates from the outer layer; the outer layer is a couple of clamps that can be screwed together and keep the test section compacted. Both the heating plates and the clamps are made of stainless steel (0Cr18Ni10Ti). The smooth inlet and outlet transition sections (each 110 mm long) are provided at both the ends of the test section, and are respectively connected with the test loop through the two pipe accessories. Certainly, the connection between the transition section and the accessories is insulated electrically. On the accessories, provisions are made to accommodate inlet and outlet thermocouples. The temperature at the inlet and outlet of the test channel is measured using sheathed thermocouples (N-type, 1 mm). The two temperature measuring locations are 150 mm before the inlet and after the outlet of the test channel. As mentioned prior, the test channel is made up of the two heating plates held together by jointing as shown in Fig. 2 . Both the heating plates are cubic-shaped, and have the same thickness ( δ ) of 3 mm. On the outer side of each heating plate, the two uniform triangles are cut in case of heat concentration at the corners. On the inner side of each heating plates, the two uniform grooves are cut to form the test channel by jointing together. To alleviate the sharp corner effects, the semi-circular transition zones have been machined at the four inner sharp corners in the channel. The test channel has a depth ( e ) of 2 mm and a aspect ratio ( α = e / b ) of 0.05, hence the hydraulic diameter ( D h = 4 A c / P, where A c and P are the cross-sectional area and the wet perimeter of the fluid channel) is about 3.64 mm. The total length to diameter ratio ( L / D h ) is 300, with the effective heated ( L h / D h ) ratio being 285. The mean wall relative roughness in the test channel ( ɛ / D h , where ɛ is the mean wall absolute roughness) is 8.5 × 10 −4 . It is shown in Fig. 3 , that a total of 14 sheathed thermocouples (N-type, 1 mm) are divided into two groups and symmetrically placed on the outer sides of the two heating plates to measure the outer-wall temperatures, and the five vessels with the diameter being 4 mm are installed on the outer side of one heating plate to accommodate measuring pressure drops for four subsections on the test channel. The placement of the thermocouples is at axial distances of 160.0, 360.0, 560.0, 790.0, 860, 930 and 1000 mm from the inlet, respectively. The thermocouples are squeezed on the heating plate's surface and secured in place by thin steel sheets jointing. The installation of the pressure vessels is at axial distances of 170.0, 370.0, 570.0, 870, and 1070 mm from the inlet, respectively. Here, it needs to be pointed out that the purpose of measuring the four sub-sectional pressure drops is to monitor whether the test channel will distort or not. The four smart type pressure differential transducers (ST3000 Model) are connected to fittings attached to the five pressure vessels. The heat flux is applied using two copper blocks welded at the two ends of the heating plates. Electrical cables are then connected to the copper blocks. The heating plates are energized using a DC power supply (20 kW total power). The temperatures, pressure drops, and power supply voltage and current are recorded using a data acquisition system (AT-96 Model) connected to a computer. Before the experiments, a layer of asbestos is wrapped on the test section to prevent from heat loss. All thermocouples used in the experiments have a calibration accuracy of ±0.3 °C. The accuracy in the measurement of the test channel pressure drops is about ±30 Pa. The flowmeter is calibrated for different inlet liquid temperature and liquid flow rates. The uncertainty in the measured mass flow rate is estimated to be ±2%. The operational pressure is directly read using a manometer installed on the pressurizer. The manometer has an accuracy of about ±1 kPa in the measurement. 3 Experimental procedure and data reduction The experimental procedure adopted for all test cases was as follows: the deioned water was first pumped into the test loop by the piston pump from the supply tank. After the piston pump was switched off and isolated from the test loop, the operational pressure was set to the required value (0.1 MPa). The circulating pump was switched on to force the fluid flow through the needle valve, preheater, flowmeter, test section, heat exchanger, and then the filter. The flow rate of the liquid was measured using the flowmeter and was adjusted as required. The preheater was only used in those few test cases where the liquid temperature at the test channel inlet was found to be below the required value. Once the inlet temperature and the flow rate were at the desired values, the recording of all the temperatures and pressure drops was initiated. Two kinds of experiment were carried out: (1) In the isothermal flow experiments, only the flow rate was adjusted without power input to the test section, and the duration of a single experiment was approximately 1–2 min, and all measurements were taken when the flow rates and the pressure drops were steady for about 0.5–1 min; (2) In the convective heat transfer experiments, the power to the test section was turned on and set to a value dependent on the corresponding flow rate, and the duration of a single experiment was approximately 3–5 min and all measurements were taken when the flow rates, the pressure drops and the wall and outlet temperatures were steady for about 1–2 min. At each new flow rate and each new inlet temperature, the power applied to the test channel was changed accordingly, and the same procedure was repeated. In the isothermal flow experiments, a total of sixty sets of experiments were conducted for varying inlet temperatures ( T f,in = 24–37.5 °C) and mass fluxes ( G = 285–2000 kg/(m 2 s)). The Reynolds number ( Re = GD h / μ f , where μ f is the mean fluid dynamic viscosity and is dependent on the mean fluid temperature) in the channel in these cases was varied from 1090 to 10,200, all of which span the laminar, laminar-turbulent transition regime. The Prandlt number ( Pr = ν f / a f , where ν f and a f are the fluid kinetic viscosity and thermal diffusivity respectively) ranged from 4.6 to 6.2. In the convective heat transfer experiments, a total of twenty-four sets of experiments were conducted for varying inlet temperatures ( T f;in = 27.8–30 °C) and mass fluxes ( G = 306–2320 kg/(m 2 s)). The heat fluxes ( q ″) were varied from 14 to 214 kW/m 2 . The Reynolds number in the channel in these cases was varied from 1460 to 13,000, all of which span the laminar, laminar-turbulent transition regime. The Prandlt number ranged from 3.9 to 5.7. In all the experimental parameters, the temperature differences between channel inlet and outlet are always about 20 °C, and the corresponding density variations are under 8 kg/m 3 . Based on our calculations, the acceleration pressure drops raised by density variations are lower than 0.8 percent of the gravitation pressure drops. Therefore, the acceleration pressure drops in the channel are neglected, the total pressure drops in the channel are calculated as: (1) Δ p = Δ p F + Δ p g where Δ p F and Δ p g are the friction pressure drop and the gravitation pressure drop, respectively. Additionally, the measured total pressure drop has the following expression: (2) Δ p = Δ p 0 + Δ p 0 , g where Δ p 0 represents the reading from the pressure differential transducers, and Δ p 0,g represents the gravitation pressure drop dependent on the fluid density in those pressure vessels mentioned prior. Then, the friction pressure drop and friction factor in the channel are calculated as: (3) Δ p F = Δ p 0 + ( Δ p 0 , g − Δ p g ) (4) f = 2 Δ p F ⋅ D h ⋅ ρ m G 2 ⋅ L p where ρ m is the mean fluid density in the channel and is dependent on the mean fluid temperature, and L p is the total pressure-measuring distance at the axial direction( x direction as shown in Fig. 3 .). The measurement uncertainty in the friction pressure drops is estimated to about ±4%. The mean fluid temperature in the channel is calculated as: (5) T m = ( T in + T out ) / 2 where T in , and T out are the channel inlet and outlet temperature, respectively. In the convective heat transfer experiments, the energy balance ( η ) on the test channel is checked using the following expression: (6) η = Q V ⋅ I × 100 % where V , I are the power voltage and current supplied to the heating plates respectively, and Q is the net power gained by the fluid and is calculated as: (7) Q = m ⋅ ( H out − H in ) where m is the mass flow rate, and H in , and H out are the fluid enthalpy respectively at the inlet and outlet of the channel. Before the experiments, Eqs. (6) and (7) were employed to verify the reliability of each measured parameters. In the case of the inlet mass flow rate being 643 kg/h and the input power being 41.7 kW, the value of η was about 96.5%, hence each measured parameter in the experiments was reliable. Because the actual heat loss varied with the different test cases, the actual η obtained in all the test cases ranged from 80% to 99%. During the experiments, the heating plates were directly supplied with electrical power, hence the thermal conduction in the heating plates could be viewed a progress with a internal thermal source. Additionally, the length and width of the heating plates are much higher than the thickness, the thermal conduction in the heating plates mostly occurs at the z direction as shown in Fig. 3 . Therefore, the governing equation for the thermal conduction in the heating plates is written as: (8) 1 d z λ s ( T ) d T d z + q ‴ = 0 where the thermal conductivity ( λ s ) is calibrated as a linear function for the independent variable T , i.e. (9) λ s ( T ) = C 1 + C 2 T and has a calibration uncertainty of ±0.5W/(m °C), where C 1 , C 2 are two constants. The internal thermal source per volume ( q ′′′) is calculated as: (10) q ‴ = Q V s where V s is the volume of the heating plates. The outer-wall of the heating plates is approximated to satisfy adiabatic boundary condition, so the inner-wall temperature ( T wi ) of the heating plates is obtained as: (11) T wi = f ( C 1 , C 2 , q ‴ , T wo , δ ) where T wo is the measured outer-wall temperature of the heating plates, and δ is the thickness of the heating plates. The wall heat flux ( q ″) is assumed to be uniform and is calculated as: (12) q ″ = Q A s where A s is the surface area of the fluid channel. The calculation on the mean fluid temperature employs Eq. (5) . The local fluid temperature at any axial location ( x ) is calculated from the energy balance as: (13) T x = q ″ C p,m ⋅ m ⋅ P ⋅ x + T in where C p,m is the mean fliud specific heat at constant pressure and is dependent on the mean fluid temperature. The local heat transfer coefficient ( h x ) is calculated as: (14) h x = q ( T wi , x − T x ) The measurement uncertainty in the heat transfer coefficients is estimated to about ±6%. The corresponding local Nusselt number ( Nu x ) is calculated as: (15) N u x = h x ⋅ D h λ f where λ f is the mean fluid thermal conductivity. Each local Nusselt number is processed as the following dimensionless expression: (16) N u x + = φ ( x + ) = φ x / D h RePr The mean Nusselt number ( Nu m ) in the channel is obtained from the following integral equation: (17) N u m = ∫ φ ( x + ) ⋅ d x + L h + where L h + = L h / D h RePr is the dimensionless effective heated length of the channel. 4 Results and discussion 4.1 Isothermal flow It is well known according to hydrodynamics that, if single-phase isothermal flow has been fully developed in a channel, each measured sub-sectional pressure drop should be proportional to the corresponding sub-sectional length. Actually, each measured sub-sectional pressure drop in the test channel has a approximate ratios of 1.0:1.0:1.5:1.0, which is directly proportional to the corresponding sub-sectional length (200, 200, 300, and 200 mm, respectively), and the result is well demonstrated in Fig. 4 . It is noticed that the first pressure vessel in the test channel has a dimensionless length ( x / D h ) of 46.7, so the result shown in Fig. 4 implies that both the laminar and turbulent regime in the channel have been hydrodynamically developed before the dimensionless length ( x / D h ) of 46.7. Therefore, it can be concluded that the test channel is well in shape in the experiments. Also shown in Fig. 4 indicates that the laminar-to-turbulent transitional regime occurs for Reynolds numbers ranging between 2500 and 4000, approximately. The variation of friction factor for isothermal flow as a function of the Reynolds number is shown in Fig. 5 . It can be seen from Fig. 5 that the fluid flow falls into three kinds of typical regime, viz. laminar regime, laminar-to-turbulent transitional regime and turbulent regime, and the critical Reynolds number for laminar-to-turbulent transition is about 2500 and the turbulent regime occurs at the region for the Reynolds numbers being higher than 4000. This result has a good agreement with that shown in Fig. 3 . It is noticed that the critical Reynolds number in the test channel is slightly higher than the number of 2300 for the case of the conventional channel. A comparison of the measured laminar friction factor with that predicted by the prior investigators is shown in Fig. 6 . An analytical solution given by Kays and Clark (1953) predicted that the laminar friction factor as a function of Reynolds number has an expression of approximate f = 90/ Re for the channel aspect ratio ( α ) of 0.05. The measured laminar friction factors in the experiments are in good agreement with the prediction. Also shown in Fig. 6 is the best-fit curve for the experimental data and the correlation developed for f in the laminar region is given as: (18) f iso , l = 89.3 Re Eq. (18) is valid for 1090 ≤ Re ≤ 2500. The measured turbulent friction factor is compared with the following two correlations given by Blasius (1913) and Kakaç et al. (1987) , respectively. Blasius (1913) : a widely used correlation for turbulent flow in conventional channel. (19) f = 0.3164 R e 0.25 Eq. (19) is applicable for 4000 ≤ Re ≤ 10 5 . Kakaç et al. (1987) : a correlation for turbulent flow in rectangular channel. (20) f = 4 ( 1.0875 − 0.1125 α ) f ′ where ( α = e / b ) is the channel aspect ratio, and f ′ is the friction factor for circular duct given by Techo et al. (1965) and has the following correlation: 1 f ′ = 1.7372 ln Re 1.964 ln Re − 3.8215 Eq. (20) is applicable for 5000 ≤ Re ≤ 10 7 . Fig. 7 shows a typical comparison of the experimental turbulent friction factor with the two correlations listed above. It can be seen that, except the limited cases for Reynolds number slightly lower than 5400, the other measured turbulent friction factors are located in the prediction band. Also shown in Fig. 7 is the best-fit curve for the experimental data and the correlation developed for f in the turbulent region is given as: (21) f iso , t = 0.0426 − 2.48 × 10 − 6 R e 0.9 Eq. (21) is applicable for 4000 < Re ≤ 10,200. 4.2 Non-isothermal flow convective heat transfer The variations of friction factor and mean Nusselt number for non-isothermal flow convective heat transfer as a function of the Reynolds number are shown in Fig. 8 . It can be seen from Fig. 8 that the fluid flow regimes denoted by the trend of friction factor and by that of mean Nusselt number respectively are in good agreement with each other, and synchronously falls into three kinds of typical regime same as shown in Fig. 5 . Also, the critical Reynolds number of 2500 for laminar-to-turbulent transition and the Reynolds number of 4000 for turbulent occurrence are same as the case of isothermal flow. Based on the results, the characteristic of non-isothermal flow regimes is same as that of isothermal flow regimes. Additionally, it needs to be emphasized that the used Reynolds number in present studies employs the mean value of the temperatures at the channel inlet and outlet to determine the physical property. 4.2.1 Non-isothermal laminar and turbulent flow It is well known that the non-isothermal flow differs from the isothermal flow mainly in the following two aspects: (1) the temperature difference exists between the fluid bulk and the channel wall; and (2) the fluid temperature varies along the channel. Therefore, the correlations based on the constant property assumption need to be modified and/or corrected appropriately for the case of temperature-dependent fluid property. For most liquids, the viscosity variation is the most important factor in variable-property effects, so the non-isothermal friction factor for the water in present studies can be modified as the following correlations using the property ratio method ( Kakaç et al., 1987; Kays et al., 2003 ): (22) C f C f , cp = μ w μ f m where μ w is the dynamic viscosity evaluated at the channel wall temperature, C f is Fanning friction factor, and the subscript cp refers to the constant-property solution. It is shown in Fig. 9 that the measured non-isothermal laminar friction factor is obviously lower than the predicted value given by Kays and Clark (1953) . Variable-property effect being taken into account and the exponent ( m ) in Eq. (22) being taken as 0.58 according to Deissler (1951) , the friction factors modified become higher and closer to the prediction line. Also shown in Fig. 9 is the best-fit curve for the modified data and the correlation developed for f is given as: (23) f non , l = 85.5 Re μ w μ f 0.58 Eq. (23) is applicable for 1460 ≤ Re ≤ 2500 and 0.72 ≤ ( μ w / μ f ) ≤ 0.82. Compared to the corresponding results for the isothermal flow, the present Poiseuille number ( fRe = 85.5) is lower than that ( fRe = 89.3) in Eq. (18) . For water, the viscosity decreases markedly with increasing temperature, thus leads to lower friction resistance. Accordingly, a satisfactory modification for variable-property effect should have counteracted such influence. However, the exponent ( m ) being 0.58 proposed by Deissler (1951) refers to the case for circular ducts originally, so the modification for the variable-property effect in the present test channel is not enough. It is shown in Fig. 10 that the measured non-isothermal turbulent friction factor is even lower than the lower one of the predicted value given by Blasius (1913) and Kakaç et al. (1987) . Variable-property effect being taken into account and the exponent ( m ) in Eq. (22) being taken as 0.25 according to Allen and Eckert (1964) , the friction factors modified become higher, but only those of Reynolds number higher than 9700 are located in the prediction band. Also shown in Fig. 10 is the best-fit curve for the modified data and the correlation developed for f is given as: (24) f non , t = ( 0.039 − 4.36 × 10 − 6 R e 0.81 ) μ w μ f 0.25 Eq. (24) is applicable for 4000 ≤ Re ≤ 13,000 and ( μ w / μ f ) ≈ 0.83. Compared to the corresponding results for the isothermal flow, the non-isothermal turbulent friction factors are lower. Similarly, because the exponent ( m ) being 0.25 proposed by Allen and Eckert (1964) refers to the case for circular ducts originally, the modification for the variable-property effect in the present test channel is not enough. 4.2.2 Laminar and turbulent convective heat transfer The variation of local outer-wall temperature and Nusselt number for Reynolds number of 1500 (Corresponding to a typical laminar flow convective heat transfer) as a function of the dimensionless axial distance ( x + = ( xD h )/( RePr ) is shown in Fig. 11 . It is clear that, at the first dimensionless axial location ( x + = 0.00716), the local outer-wall temperature is lower than the predicted value raised from the linear trend line dependent on the other outer-wall temperatures, and the local Nusselt number is far over the distribution band dependent on the other Nusselt numbers with close values. The results indicate that the laminar regime before the first dimensionless axial location is still thermally developing. A comparison of the measured mean laminar Nusselt number with that predicted by prior investigators is shown in Fig. 12 . An analytical solution given by Shah and London (1978) predicted that the mean laminar Nusselt number is a constant of approximate Nu m = 7.46 for the channel aspect ratio ( α ) of 0.05. The measured mean laminar Nusselt number is very close to the prediction line, viz. (25) N u m , l ≈ 7.46 Eq. (25) is applicable for 1460 ≤ Re ≤ 2500 and 4.1 ≤ Pr ≤ 5.5. The variation of local outer-wall temperature and Nusselt number for Reynolds number of 4700 (Corresponding to a typical turbulent flow convective heat transfer) as a function of the dimensionless axial distance ( x + = ( xD h )/( RePr ) is shown in Fig. 13 . It is clear that, all the local outer-wall temperature is distributed at the same linear trend line, and all the local Nusselt number is located in the same distribution band between 50 and 54. The results indicate that the turbulent regime in the test channel is thermally developed. The measured mean turbulent Nusselt number is compared with the following two correlations given by Dittus and Boelter (1985) and Gnielinski (1976) , respectively. Dittus and Boelter (1985) : a widely used correlation for turbulent convective heat transfer in conventional channel. (26) N u = 0.023 R e 0.8 P r 0.4 Eq. (26) is applicable for 10 4 ≤ Re ≤ 1.2 × 10 5 , 0.7 ≤ Pr ≤ 120, L / De ≥ 60. Gnielinski (1976) : a correlation for both transitional and turbulent convective heat transfer in conventional channel. (27) N u = ( f / 8 ) ( Re − 1000 ) P r f 1 + 12.7 f / 8 ( P r f 2 / 3 − 1 ) where f is the friction factor for circular duct given by Filonenko (1954) and has the following correlation: 1 f = 1.58 ln Re − 3.28 Eq. (27) is applicable for 2300 ≤ Re ≤ 5 × 10 6 , 0.5 ≤ Pr ≤ 2000. Fig. 14 shows a typical comparison of the experimental mean turbulent Nusselt number with the two correlations listed above. It can be seen that there is a good agreement between the predicted values by Dittus and Boelter (1985) and the experimentally measured values for Reynolds number higher than 10,000, and the predicted values are about 4–40% higher than the measured values for the case of Reynolds number ranging between 4000 and 10,000. It can be also seen that the predicted values by Gnielinski (1976) are about 1–15% higher than the measured values for the whole experimental turbulent regime (4000 ≤ Re ≤ 13,000). Therefore in general, Gnielinski correlation has a higher accuracy for predicting the present experimental turbulent Nusselt number in compare with Dittus–Boelter correlation. Also shown in Fig. 14 is the best-fit curve for the experimental data and the correlation developed for Nu m in the turbulent region is given as: (28) N u m , t = 0.00354 ReP r 0.4 Eq. (28) is applicable for 4000 ≤ Re ≤ 13,000, 3.9 ≤ Pr ≤ 4.1. In the present narrow rectangular channel, some limited cases for the isothermal and non-isothermal friction factors immediately after turbulent occurrence have the lower experimental values than the predicted values by conventional Blasius correlation, as shown in Figs. 7 and 10 , respectively. This phenomenon may be ascribed to the absence of sharp corners effects in the test channel. In fact, the sharp corners effects in a normal rectangular channel will induce secondary flows and hence increase pressure loss. In addition, the modification for variable-property effect in the present studies appears to be not enough, and the new exponent ( m ) for the present test channel needs to be developed. Despite these mentioned cases, the present experimental results have a satisfactory agreement with the conventional theory. 5 Summary Experimental single-phase flow and heat transfer data in the narrow rectangular channel have been obtained. From the measured wall temperatures, heat fluxes, inlet and outlet fluid temperatures, and pressure drops, the isothermal and non-isothermal friction factors and convective heat transfer Nusselt numbers have been calculated. The correlations for the isothermal friction factors and the mean Nusselt numbers have been developed, and have a satisfactory agreement with the conventional theory. Based on the property ratio method, the correlations for non-isothermal friction factors have been proposed, but the new exponent ( m ) for modifying variable-property effect need to be developed. Due to the limited data, more extensive experiments will be required to further validate the present results. References Kandlikar, 2002 S.G. Kandlikar Fundamental issues related to flow boiling in minichannels and microchannels Experimental Thermal and Fluid Science 26 2002 389 407 Shen et al., 2001 X.Z. Shen Study on boiling heat transfer in a narrow annular channel oriented uprightly Nuclear Science and Engineering 21 3 2001 244 251 (In Chinese) Guo, 2000 Z.Y. Guo Frontier of micro heat transfer Advances in Mechanics 30 1 2000 1 6 (In Chinese) Kays and London, 1952 W.M. Kays A.L. London Convective heat transfer and flow friction behavior of small cylindrical tubes-circular and rectangular gross section Journal of Heat Transfer-Transactions of the ASME 74 1952 1179 1189 You, 1989 H.I. You Experimental determination of heat transfer coefficient between a microstructure and fluid International Communications in Heat and Mass Transfer 16 1989 537 Webb and Zhang, 1998 R.L. Webb M. Zhang Heat transfer and friction in small diameters channels Microscale Thermophysical Engineering 2 1998 189 202 Gao et al., 2002 P.Z. Gao S. Le Person M. Favre-Marinet Scale effects on hydrodynamics and heat transfer in two-dimensional mini and microchannels The International Journal of Thermal Sciences 41 2002 1017 1027 Wang and Peng, 1994 B.X. Wang X.F. Peng Experimental investigation on liquid forced convection heat transfer through microchannels International Journal of Heat and Mass Transfer 7 Suppl. 1 1994 73 82 Peng and Peterson, 1995 X.F. Peng G.P. Peterson The effect of thermofluid and geometrical parameters on convection of liquids through rectangular microchannels International Journal of Heat and Mass Transfer 38 1995 755 758 Toh et al., 2002 K.C. Toh X.Y. Chen J.C. Chai Numerical computation of fluid flow and heat transfer in microchannels International Journal of Heat and Mass Transfer 45 2002 5133 5141 Xin and Shi, 1994 M.D. Xin J.S. Shi Experiments on forced convective heat transfer in micro rectangular channel Journal of Chongqing University 17 3 1994 117 122 (In Chinese) Xie et al., 2004 Y.Q. Xie J.Z. Yu Experimental study on single-phase flow and heat transfer for FC-72 Journal of Beijing University of Aeronautics and Astronautics 30 8 2004 740 743 (In Chinese) Sudo et al., 1985 Y. Sudo K. Miyata H. Ikawa Experimental study of differences in single-phase forced-convection heat transfer characteristics between upflow and downflow for narrow rectangular channel Journal of Nuclear and Technology 22 3 1985 202 212 Sudo et al., 1990 Y. Sudo T. Usui Heat transfer characteristics in narrow vertical rectangular channels heated from both sides JSME International Journal Series II 22 4 1990 Mokrani et al., 2009 O. Mokrani B. Bourouga C. Castelain H. Peerhossaini Fluid flow and convective heat transfer in microchannels International Journal of Heat and Mass Transfer 52 2009 1337 1352 Kays and Clark, 1953 Kays, W.M., Clark, S.H., 1953. TR no. 17, Department Engineering, Stanford University, Stanford, California, August 15. Blasius, 1913 Blasius, H., 1913. Das Ähnlichkeitsgesetz bei Reibungvorgägen in Flüssigkeiten, Forchg. Arb. Ing. -Wes., No. 131, Berlin. Kakaç et al., 1987 S. Kakaç R.K. Shah W. Aung Handbook of Single-Phase Convective Heat Transfer 1987 Wiley-Interscience Techo et al., 1965 R. Techo R.R. Tickner R.E. James An accurate equation for the computation of the friction factor for smooth pipes from the Reynolds number Journal of Applied Mechanics 32 1965 443 Kays et al., 2003 W. Kays C. Michael B. Weigand Convective Heat Transfer and Mass Transfer Fourth ed. 2003 Mc Graw Hill Education Deissler, 1951 Deissler, R.C., 1951. NASA TN2410, Washington. Allen and Eckert, 1964 R.W. Allen E.R.C. Eckert Journal of Heat Transfer 85 1964 301 310 Shah and London, 1978 R.K. Shah A.L. London Laminar Flow Forced Convection in Ducts, Advances in Heat Transfer 1978 Academic Press New York Dittus and Boelter, 1985 P.W. Dittus L.M.K. Boelter Heat transfer in automobile radiators of tubular pipe International Communications in Heat and Mass Transfer 12 1985 3 22 Gnielinski, 1976 V. Gnielinski New equations for heat and mass transfer in turbulent pipe and channel flow International Chemical Engineering 16 1976 359 368 Filonenko, 1954 G.K. Filonenko Hydraulic resistance in pipes Teploenergetika 1 4 1954 40 44 (in Russian)