Experimental investigation of convection heat transfer of CO2 at supercritical pressures in vertical circular tube

Zhi-Hui Li,Pei-Xue Jiang,Chen-Ru Zhao,YuSong Lin

2010-01-01

Kung Cheng Je Wu Li Hsueh Pao/Journal of Engineering Thermophysics

Abstract:The convection heat transfer characteristics of supercritical CO 2 in a vertical circular tube of 2 mm inner diameter were investigated experimentally for pressures ranging from 78 to 95 bar, inlet temperatures from 25 to 40 °C, and inlet Reynolds numbers from 3800 to 20,000. The effects of the heat flux, thermo-physical properties, buoyancy and thermal acceleration on the convection heat transfer were analyzed. The experimental results show that for high inlet Reynolds numbers (e.g. Re = 9000) and high heat fluxes, a significant local deterioration and recovery of the heat transfer was found for upward flows but not for downward flows. Comparison of the experimental data for inlet Reynolds numbers from 3800 to 20,000 with some well-known empirical correlations showed large differences especially when the heat transfer deteriorates and then recovers when the effect of buoyancy is significant. The experimental data was used to develop modified local turbulent Nusselt number correlations for supercritical CO 2 flowing in vertical small circular tubes. Keywords Supercritical pressures Buoyancy Thermal acceleration Empirical correlations Nomenclature c p specific heat at constant pressure [J/(kg K)] c ¯ p average specific heat [J/(kg K)] d tube inner diameter [m] D tube outer diameter [m] A surface area [m 2 ] G mass flow rate [kg/s] I current [A] l tube length [m] p pressure [N/m 2 ] q v internal heat source strength (4 I 2 R /( π ( D 2 − d 2 ) L )) [W/m 3 ] q w heat flux on the inner tube surface [W/m 2 ] Q loss heat loss from the test section [W] R electrical resistance [Ω] T temperature [°C] or [K] T pc pseudo-critical temperature [°C] or [K] x axial coordinate [m] h f ,0 inlet specific enthalpy [J/kg] h f , b ( x ) local bulk fluid enthalpy [J/kg] h w specific enthalpy at the wall temperature [J/kg] h b specific enthalpy at the bulk temperature [J/kg] h x local heat transfer coefficient [W/(m 2 °C)] Nu Nusselt number Pr Prandtl number Re Reynolds number Gr Grashof number ( Gr = gβd 3 ( T w − T f )/ ν 2 ) Gr * Grashof number ( Gr ∗ = g β d 4 q w / λ ν 2 ) Greek symbols λ thermal conductivity [W/(m K)] ρ density [kg/m 3 ] μ dynamic viscosity [kg/(m s)] Subscripts b bulk f fluid i inner surface in inlet section o outer surface out outlet section w wall x local values 0 tube inlet 1 Introduction Recently, the development of the supercritical pressure water-cooled reactor (SCWR) and other industrial applications such as trans-critical CO 2 air-conditioning and heat pump systems have led to a new interest in the heat transfer to fluids at supercritical pressures. There were many papers published on forced and mixed convection heat transfer in supercritical fluids such as water, carbon dioxide, nitrogen, hydrogen, and helium in large tubes during the late nineteen fifties and throughout the sixties [1–5] . The findings provided a comprehensive understanding of some special features of the heat transfer in large tubes. In contrast, there is little data for narrow geometries (∼2.5 to 3 mm), which are typical of the equivalent diameters of the fuel element sub-channels in an SCWR. The operating conditions for the SCWR are very complex, with little knowledge on the heat transfer mechanisms in narrow geometries, especially the effect of variable physical properties, buoyancy effects, and thermal acceleration on the heat transfer. Song et al. [6] investigated the heat transfer of supercritical pressure CO 2 in vertical tubes having 4.4 mm and 9.0 mm inner diameters. The results showed that a similarity of heat transfer behavior was maintained both in the normal heat transfer mode and deteriorated heat transfer mode with the same length to diameter ratio and the wall heat flux to mass flux ratio. Jiang et al. [7,8] experimentally and numerically investigated the local convection heat transfer of CO 2 at supercritical pressures in vertical mini-tubes having 0.27 mm and 2 mm inner diameters at relatively low Reynolds for upward and downward flows. The experiments investigated the effects of heat flux, mass flow rate, inlet fluid temperature, buoyancy and flow acceleration on the convection heat transfer. However, the length to diameter ratio of the 2 mm inner diameter tube was relatively small ( l / d = 19.5), so the axial conduction effect is significant. Many earlier supercritical heat transfer correlations [9] have been shown to be in good agreement with experimental data when buoyancy is not significant but these correlations differ significantly from the experimental data when buoyancy effects are significant. Protopopov [10] and McEligot and Jackson [11] utilized buoyancy parameters in their correlations, but their correlations were based on data for large tubes so their results need to be verified for small channels. Jong et al. [12] developed a correlation to predict the forced heat transfer and mixed heat transfer over a wide range of operating conditions taking into account the imposed conditions such as mass flow rate and heat flux based on data for circular, triangular, and square cross-sections. Hyungrae et al. [13] developed new correlations to predict the normal heat transfer rate and the reduced heat transfer rate to supercritical pressure CO 2 based on experimental data for tubes having inner diameters of 4.4 mm and 9.0 mm and a concentric annular passage (8 mm × φ 10 mm × L 1800 mm). Further studies of the heat transfer mechanism in supercritical pressure fluids are needed to improve predictions of the heat transfer coefficient in the narrow geometries, especially for the SCWR design. This paper describes experimental investigations of the local convection heat transfer of supercritical pressure CO 2 flowing in a vertical 2 mm inner diameter long mini-tube for upward and downward flows over a wide range of parameters. The local external wall temperatures along the tube were measured directly in the experiments and the corresponding local Nusselt numbers along the tube as well as the dimensionless parameter Bo * for evaluating the buoyancy effect and Kν for evaluating the thermal acceleration effect were presented. The tests investigated the effects of heat flux, flow direction, buoyancy and thermal acceleration on the convection heat transfer. The experimental data for inlet Reynolds numbers from 3800 to 20,000 are then compared with existing correlations for upward and downward flows. Modified local turbulent Nusselt number correlations are presented in this paper. 2 Experimental facility 2.1 Experimental system and test section The experimental system is schematically shown in Fig. 1 which was presented in Jiang et al. [8] . The working fluid, CO 2 with a purity of 99.995%, was fed into the test loop from a compressed CO 2 container. The working fluid was cooled by a cooling bath before being pressurized and driven by a supercritical CO 2 pump (Thar P-50). Then the fluid flowed through a pre-heater in which it was heated to the desired inlet temperature for each test and then into a vertical test section and a cooler and finally returned to the pump. The system pressure was regulated by a decompression valve. The test section was a vertical stainless steel 1Cr18Ni9Ti long tube with inside and outside diameters of 2.018 mm and 3.137 mm. The heated test section was 290 mm long with heating by an AC current. The test section was connected with the test loop by flanges and high-pressure fittings. The test section was insulated thermally and electrically from the test loop by a layer of PTFE placed between the flanges and between the screws and the flanges. CO 2 flowed into the test section from either the bottom for upward flow or from the top for downward flow. The heated length to inner diameter ratio was 145 which is much larger than in the previous work [8] . The tube also included 105 mm long inlet and outlet sections to guarantee fully developed flow. The measurements using a Talysurf 5P showed that the average inner surface roughness of the mini-tube was 0.14 μm to assure the teat section is smooth tube. The whole test section was wrapped by a layer of thick asbestos to reduce heat losses. The orientation of flow includes upward flow and downward flow. 2.2 Measurements The local outer wall temperatures were measured by 33 copper-constantan type thermocouples with diameters of 0.2 mm spot-welded onto the outer surface along the tube. The first two and last two thermocouples were 5 mm apart, with the rest of the thermocouples 9 mm apart. The inlet and outlet bulk temperatures were measured by two accurate RTDs inserted inside the tube through a tee-junction at the inlet and outlet of the heated section. Mixers were installed before the temperature measurement positions to ensure complete mixing before the bulk temperature measurement. The inlet pressure was measured by a pressure transducer (Model EJA430A) with a full range of 14 MPa. The test section pressure drop was measured by a differential pressure transducer (Model EJA110A) with a full range of 10 kPa. The mass flow rate was measured by a Coriolis-type mass flow meter (Model MASS2100-DI1.5) with the full range of 4 kg/h. The electric power inputs to the test section were calculated from the measured current, voltage and electrical resistance. The measured data including the pressure, differential pressure, temperature and mass flow rate were monitored and recorded by a data acquisition system (HP 34970A). The experiments covered a wide range of parameters as listed in Table 1 . 3 Data analysis method and experimental uncertainty The local heat transfer coefficient, h x , and the local Nusselt number, Nu x at each axial location were calculated from the following equation: (1) h x = q w T w , i ( x ) - T f ( x ) , Nu x = h x d λ The heat transport in the tube wall was assumed to be one-dimensional steady thermal conduction with a local uniform volumetric heat source in the tube. The boundary conditions assumed an insulated outer surface with the measured local temperatures for the outer tube surface, T wo ( x ). The local temperatures for the inner tube surface, T wi ( x ), were then calculated with the following equation (Jiang et al. [8] ): (2) T w , i ( x ) = T w , o ( x ) + q v 16 λ ( D 2 - d 2 ) + q v 8 λ D 2 ln d D The heat flux on the inner surface, q w , was calculated as: (3) q w = I 2 R - Q loss A i Q loss represents the overall heat losses as a function of temperature difference between the wall temperature and the surroundings. Q loss was obtained experimentally by changing the heating power without fluid flow in the test system. The local bulk fluid temperature, T f ( x ), was calculated from the local bulk fluid enthalpy, h f ( x ), as: (4) h f ( x ) = h f , 0 + q w π dx G The dimensionless independent parameters Re , Gr and Gr * were calculated from the measured quantities using: (5) Re = ρ ud μ = 4 G π d μ , Gr = g β d 3 ( T w - T f ) ν 2 , Gr ∗ = g β d 4 q w λ ν 2 The accuracies were within ±0.25 °C for the copper-constantan type thermocouples and ±0.1 °C for the RTDs in the temperature range of 0–150 °C. The accuracy of the pressure gage transducer was 0.075% of a full range of 14 MPa. The accuracy of the differential pressure transducer was 0.075% over a range of −5 to 5 kPa. According to the instructions, the accuracy of the Coriolis-type mass flow meter (Model MASS2100-DI1.5) was 0.1% of the actual mass flow rate with 95% confidence (probability) for flows within 5–100% of the sensor’s maximum flow rate. The flow rate during the measurements was about 1.6–3.68 kg/h. Therefore, the relative uncertainty of the mass flow rate was 0.1%. A detailed analysis of the uncertainty showed that the maximum uncertainty of the heating rate into the test section was ±4.0%, the uncertainty of the tube inner surface temperatures was ±0.25 °C and the uncertainty of the local bulk fluid temperature was ±0.24 °C. The root-mean-square uncertainty of the Nusselt number was then estimated to be ±7.9%. The uncertainties in the inlet pressures were estimated to be ±0.13%. 4 Experimental results and discussion 4.1 Parameters for buoyancy effect on the convection heat transfer The buoyancy parameter, K , introduced by Protopopov [10] can be used to estimate the influences of buoyancy on the heat transfer. The buoyancy number, K , was defined as: (6) K = ( 1 - ρ w / ρ b ) · Gr / Re 2 According to Protopopov [10] , for upward flow, when K < 0.01, the buoyancy does not significantly affect the heat transfer, when 0.01 < K < 0.4, the buoyancy will significantly reduce the heat transfer; however, when K > 0.4, the buoyancy will enhance the heat transfer. For downward flow, the buoyancy will always enhance the heat transfer. Another buoyancy parameter, Bo * , introduced by Jackson et al. [4] can also be used to estimate the influences of buoyancy on the heat transfer. The buoyancy number, Bo * , was defined as: (7) Bo ∗ = Gr ∗ / ( Re 3.425 Pr 0.8 ) According to McEligot and Jackson [11] , the buoyancy is negligible for Bo * < 6 × 10 −7 for upward and downward flows. For upward flow with 6 × 10 −7 < Bo * < 1.2 × 10 −6 , the buoyancy will reduce the heat transfer while for 1.2 × 10 −6 < Bo * < 8.0 × 10 −6 , the heat transfer reduction will be gradually weakened as the Bo * increases but the buoyancy will still negatively affect the heat transfer. For Bo * > 8.0 × 10 −6 the buoyancy will enhance the heat transfer. For downward flow, the buoyancy will always enhance the heat transfer for Bo * > 6 × 10 −7 4.2 Parameter for thermal acceleration effect on the convection heat transfer The thermal acceleration parameter, Kν , introduced by McEligot et al. [14] to estimate the influence of the thermal acceleration on the heat transfer: (8) K ν = 4 q w d β Re 2 μ c p McEligot et al. [14] suggested that for turbulent flows, when Kν < 3 × 10 −6 the fluid flow remains turbulent, but when Kν ⩾ 3 × 10 −6 the turbulence may be reduced which results in the deteriorated heat transfer. 4.3 Convection heat transfer at relatively high Reynolds numbers ( Re in = 9000) 4.3.1 Convection heat transfer for low heat fluxes Fig. 2 a and b shows the local wall temperatures (solid symbols) and local bulk fluid temperatures (hollow symbols) along the tube for upward and downward flows for low heat fluxes ( q w = 6498 W/m 2 , 7926 W/m 2 , 13,626 W/m 2 in Fig. 2 a and q w = 6494 W/m 2 , 7930 W/m 2 , 13,399 W/m 2 in Fig. 2 b) at an inlet pressure of 8.80 MPa, inlet temperature of 25.0 °C, and inlet Reynolds number of 9000. The inlet fluid temperature is less than the pseudo-critical temperature ( T pc = 39.1 °C for p = 8.80 MPa). As can be seen from these figures, the local wall temperature increases continually along the tube for both upward and downward flows without abnormal temperature variations with the local wall temperatures and fluid temperatures both lower than the pseudo-critical temperature. Fig. 3 a and b shows the local Bo * distributions along the tube for upward and downward flows with the same conditions as in Fig. 2 a and b. The local Bo * are less than 6 × 10 −7 for the lower heat fluxes ( q w = 6498 W/m 2 and 7926 W/m 2 in Fig. 2 a and q w = 6494 W/m 2 and 7930 W/m 2 in Fig. 2 b) , Bo * approaches but is still less than 6 × 10 −7 for the higher heat fluxes ( q w = 13,626 W/m 2 in Fig. 2 a and q w = 13,399 W/m 2 in Fig. 2 b). According to McEligot and Jackson [11] the influence of buoyancy on the heat transfer is still negligible. However, this is not true as will be analyzed later when discussing Fig. 5 . Fig. 4 a and b shows the local Kν distributions along the tube for upward and downward flows for the same conditions as in Fig. 2 a and b. The local Kν are much less than 3 × 10 −6 for all the heat fluxes, which means that the influence of thermal acceleration on the heat transfer is negligible. Fig. 5 a and b shows the local Nusselt number distributions for upward and downward flows for the same conditions as in Fig. 2 a and b. The local Nusselt number first decreases along the flow direction in the entrance region as the boundary layer develops along the tube and then increases further down the tube. Since the wall and fluid temperatures along the tube are lower than the pseudo-critical temperature, Fig. 2 , as the fluid bulk temperature increases along the tube, the viscosity and thermal conductivity decrease while the specific heat ( c p ) significantly increases, as shown in Fig. 6 . The decreased thermal conductivity reduces the heat transfer, but the decreased viscosity makes the viscous sub-layer thinner which enhances the heat transfer. The sharp increase of the specific heat significantly enhances the heat transfer due to the increased convection transport capacity. Therefore, the local Nusselt number decreases along the tube only in a short entrance region, and then increases further down the tube. As can be seen from Fig. 5 , the local Nusselt number increases with increasing heat flux due to the influence of the thermo-physical properties with the buoyancy having little effect with q w = 6498 W/m 2 and q w = 7926 W/m 2 for upward flow and q w = 6494 W/m 2 and q w = 7930 W/m 2 for downward flow. However, at the higher heat flux of q w = 13,626 W/m 2 the heat transfer was impaired relative to that with the lower heat flux for upward flow, when the heat transfer was increased at the heat flux of q w = 13,399 W/m 2 , relative to that with the lower heat flux for downward flow. These results illustrate that the influence of buoyancy on the heat transfer occurs earlier than Bo * = 6 × 10 −7 given by McEligot and Jackson [11] . 4.3.2 Convection heat transfer for high heat fluxes Fig. 7 a and b shows the local wall temperatures (solid symbols) and local bulk fluid temperatures (hollow symbols) along the tube for four higher heat fluxes ( q w = 26,784 W/m 2 , 31,882 W/m 2 , 39,389 W/m 2 , 51,955 W/m 2 in Fig. 7 a and q w = 26,703 W/m 2 , 31,786 W/m 2 , 38,706 W/m 2 , 50,378 W/m 2 in Fig. 7 b) at an inlet pressure of 8.80 MPa, inlet temperature of 25.0 °C, and inlet Reynolds number of 9000 for upward and downward flows. For these conditions, the local wall temperature is higher than T pc and the local bulk fluid temperature approaches the pseudo-critical temperature along the tube. For upward flows, the local wall temperature first increases, then decreases, then increases again along the tube, and finally decreases in the outlet region due to the longitudinal conduction. For downward flows, the local wall temperature increases continuously along the tube without abnormal temperature variations, except for the temperature decreasing slightly at the outlet due to the end effect. Fig. 8 a and b shows the local Bo * distributions along the tube for upward and downward flows for the same conditions as in Fig. 7 a and b. These local Bo * are greater than 6 × 10 −7 and even larger than 1.2 × 10 −6 , which means that the buoyancy has a significant influence on the heat transfer. Fig. 9 a and b shows the local Kν distributions along the tube for upward and downward flows for the same conditions as in Fig. 7 a and b. The local Kν are less than 3 × 10 −6 for both upward and downward flows which means that the thermal acceleration has little influence on the heat transfer. Fig. 10 a and b shows the local Nusselt number distributions for upward and downward flows for the same conditions as in Fig. 7 a and b. For the upward flows, the local Nusselt number first decreases to a local minimum along the tube, then increases to a maximum, then decreases again, and then finally increases downstream in the tube illustrating the significant local deterioration and recovery of the heat transfer. The turbulent diffusivity is reduced for upward flow when the low-density wall layer which results in strong buoyancy becomes thick enough to reduce the shear stress in the region where energy is typically fed into the turbulence; thus, significantly reducing the heat transfer. Therefore, The Nusselt number decreases to a local minimum as the wall temperature increases to a maximum as shown in Figs. 10 and 7 a. Further along the tube, the increasing wall temperature further increases the density difference between the near wall region and the bulk. This density difference becomes more significant which causes upward acceleration of the fluid in the boundary layer and enhances the turbulence production and, hence, the heat transfer. As a result, The Nusselt number increases to a local maximum as the wall temperature decreases along the tube to a minimum as shown in Figs. 10 and 7 a. As the wall temperature further increases along the latter part of the tube, the fluid temperature near the wall region is greater than the pseudo-critical temperature. The variation of the thermo-physical properties, especially the specific heat, then strongly reduces the heat transfer. As a result, The Nusselt number decreases again as the wall temperature increases along the tube. In the outlet region, The Nusselt number increases as the wall temperatures decrease somewhat due to the effect of the axial thermal conduction. For downward flows, the local Nusselt number decreases along the flow direction in the entrance region as the boundary layer develops and then increases to a maximum due to the increase of the specific heat and the decrease of the viscosity as the bulk temperature approaches T pc and the turbulence is enhanced by the buoyancy. Further downstream, the near-wall layer fluid temperature increases and passes through T pc as the wall temperature increases, the specific heat decreases and the viscosity increases so that the local Nusselt number decreases as a result. At higher heat fluxes, the maximum local Nusselt number is closer to the inlet as shown in Fig. 10 b. Deterioration of the heat transfer was not observed for downward flows. The Nusselt number increases along the tube for downward flow mainly due to the variations of the thermo-physical properties and the strong buoyancy which increase the turbulence energy and, hence, the heat transfer. The decrease of the local Nusselt number with increasing heat flux was due to the decreased specific heat, density, and thermal conductivity when the wall temperature was much larger than T pc as shown in Figs. 6 and 7 . 5 Comparison with existing empirical correlations 5.1 Experimental results Fig. 11 shows the ratio of the local Nusselt number for mixed convection to that for forced convection for pressures ranging from 78 to 95 bar, inlet temperatures from to 25 to 40 °C, and inlet Reynolds numbers from 3800 to 2 × 10 4 plotted against local values of the buoyancy parameter. In determining Nu / Nu f and Bo * , the fluid properties were evaluated at the bulk temperature with Nu f equal to Nu b in Eq. (12) multiplied by ε l from Eq. (9) . The results show the contrasting behavior for upward and downward flows. For downward flow, the heat transfer is enhanced by the buoyancy influence as has been found in larger circular tubes. For upward flow, the heat transfer decreases with increasing buoyancy for 2 × 10 −7 < Bo * < 6 × 10 −7 followed by recovery of the heat transfer for 6 × 10 −7 < Bo * < 2 × 10 −5 and improved heat transfer for Bo * > 2 × 10 −5 which is similar to that found in larger circular tubes, but with different ranges of Bo * . 5.2 Heat transfer without buoyancy effects Krasnoshchekov and Protopopov [2] and Grigoriev et al. [15] developed empirical correlations for the local Nusselt numbers for turbulent and transition convection heat transfer in vertical tubes: (9) Nu ( x / d ) = Nu 0 ε l ( x / d ) ε ϕ ( x / d ) where ε l = 1 + 2.35 Pr f - 0.4 Re f - 0.15 ( x / d ) - 0.6 exp ( - 0.39 Re f - 0.1 · x / d ) ε φ = ( ρ w / ρ b ) m · ( c ¯ p / c pb ) n upward flow : m = 0.4 ; downward flow : m = 0.3 c ¯ p = ∫ T b T w c p dT / ( T w - T b ) = ( h w - h b ) / ( T w - T b ) when c ¯ p / c pb > 1 , n = 0.7 when c ¯ p / c pb < 1 ′ , n = 0.4 for T w < T pc or T b ⩾ 1.2 T pc n 1 = 0.22 + 0.18 ( T w / T pc ) for T b < T pc and 1 < T w T pc < 2.5 n 1 - ( 5 n 1 - 2 ) [ T b / T pc - 1 ] for 1 < T b T pc < 1.2 where Nu 0 is calculated using the formulas proposed by Petukhov et al. [16] and Gnielinski [17] : (10) Nu 0 = ( ζ / 8 ) ( Re - 1000 ) Pr 1 + 12.7 ζ / 8 ( Pr 2 / 3 - 1 ) ( for 2300 ⩽ Re ⩽ 0 4 ; 0.5 ⩽ Pr ⩽ 200 ) (11) Nu 0 = ( ζ / 8 ) Re · Pr 1 + 900 / Re + 12.7 ζ / 8 ( Pr 2 / 3 - 1 ) ( for 10 4 ⩽ Re ⩽ 5 × 10 6 , 0.5 ⩽ Pr ⩽ 5 × 10 5 ) ζ = ( 1.82 log Re - 1.64 ) - 2 The Jackson correlation [18] for flow without buoyancy is given by: (12) Nu b = 0.0183 Re b 0.82 Pr b 0.5 ( c ¯ p / c pb ) n ( ρ w / ρ b ) 0.3 where n = 0.4 T b < T w < T pc , 1.2 T pc < T b < T w 0.4 + 0.2 ( T w / T pc - 1 ) T b < T pc < T w 0.4 + 0.2 ( T w / T pc - 1 ) [ 1 - 5 ( T b / T pc - 1 ) ] T pc < T b < 1.2 T pc , T b < T w Fig. 12 a and b compares Eqs. (9)–(11) with the current experimental data for K less than 0.01 for upward and downward flows. Here, the ordinate, Nu p , is the value predicted by the correlations and the abscissa, Nu e , is the experimental value. The Krasnoshchekov and Protopopov correlation, Eq. (9) , captures 95% of the data with an accuracy of ±15%. Fig. 13 a and b compares Eq. (12) with the current experimental data for B o * less than 6 × 10 −7 for upward and downward flows. The Jackson correlation captures 95% of the data with an accuracy of ±20%. The differences in Fig. 13 are greater than with Eqs. (9)–(11) in Fig. 12 , one reason is that the effect of the thermal entry region is not considered in Eq. (12) . Eqs. (9) and (12) can accurately predict the heat transfer coefficients without buoyancy as shown by the good agreement with the present experimental data. 5.3 Heat transfer with buoyancy The correlation proposed by Protopopov [10] for flow with buoyancy is given by: (13) Nu b ( 2 ) = Nu b 1 ϕ ( 2 ) ( K ) ( 0.01 < K < 0.4 ) (14) Nu b ( 3 ) = Nu b 1 ϕ ( 3 ) ( K ) , ( K > 0.4 ) where Nu b 1 is obtained from Eqs. (9)–(11) . φ (2) ( K ) and φ (3) ( K ) were given by Protopopov [10] . The semi-empirical model proposed by McEligot and Jackson [11] is given by: (15) Nu Nu f = 1 ± 8 × 10 4 Gr ∗ Re 3.425 Pr 0.8 Nu Nu f - 2 0.46 where “+” is for downward flow, “−” is for upward flow and Nu f is obtained from Eq. (12) . Fig. 14 a and b compares Eqs. (13), (14) with the experimental data for K greater than 0.01 for upward and downward flows. The Protopopov correlation captures 95% of the data with an accuracy of ±35%. Large differences occur between the predicted and experimental values, especially where the heat transfer deteriorates and then recovers for upward flow where the bulk fluid temperature and wall temperature are near the pseudo-critical temperature. Fig. 15 a and b compares Eq. (15) with the experimental data for Bo * greater than 6 × 10 −7 for upward and downward flows. The Jackson correlation captures 95% of the data with an accuracy of ±30% for upward flow and ±25% for downward flow. 5.4 Modified local Nusselt correlations for turbulent heat transfer The current experimental data was analyzed to develop new local Nusselt number correlations for turbulent heat transfer with buoyancy effects for carbon dioxide flow at supercritical pressures flowing in vertical small circular tubes: For downward flow: (16) Nu Nu f = 1 + ( Bo ∗ ) 0.1 c ¯ p c pb - 0.3 ρ w ρ b 0.5 Nu Nu f - 2 0.46 For upward flow: (17) Nu Nu f = 1 - ( Bo ∗ ) 0.1 c ¯ p c pb - 0.009 ρ w ρ b 0.35 Nu Nu f - 2 0.46 where Nu f is equal to Nu b from Eq. (12) multiplied by ɛ l from Eq. (9) . Fig. 16 a and b compares Eqs. (16) and (17) with the experimental data for upward and downward flows. The correlations capture 95% of the data with an accuracy of ±20%, with relatively accurate predictions of the deterioration and recovery of the convection heat transfer for upward flow and the heat transfer enhancement for downward flow. 6 Conclusions (1) For a higher inlet Reynolds number, Re in = 9000, and lower heat fluxes, the local wall temperatures increase along the tube for upward and downward flows. However, for high heat fluxes, the local wall temperatures vary in a complex, nonlinear fashion with deterioration and recovery of the heat transfer observed in upward flows but not in downward flows. (2) For upward flow, the heat transfer decreases with increasing buoyancy for 2 × 10 −7 < Bo * < 6 × 10 −7 followed by recovery of the heat transfer for 6 × 10 −7 < Bo * < 2 × 10 −5 and improved heat transfer for Bo * > 2 × 10 −5 . (3) The Krasnoshchekov correlation and the Jackson correlation are applicable for turbulent heat transfer without buoyancy effects and are in good agreement with the current experimental data. (4) The current experimental data was used to develop local Nusselt number correlations to predict the heat transfer with buoyancy effects for upward and downward turbulent flows of carbon dioxide at supercritical pressures in vertical circular tubes. The correlations more accurately represent the data and accurately predict the complex Nusselt number variations resulting from complex thermo-physical property variation and buoyancy effects. Acknowledgements This project was supported the Key Project Fund from the National Natural Science Foundation of China (No. 50736003) and the National High Technology Research and Development Program of China (863 Program) (No. 2006AA05Z416). We thank Professor J.D. Jackson in the School of Mechanical, Aerospace and Civil Engineering, the University of Manchester, UK for many suggestions for this paper. We also thank Prof. David Christopher for editing the English. References [1] M.E. Shitsman Impairment of the heat transmission at supercritical pressures High Temperature 1 1963 237 244 [2] E.A. Krasnoshchekov V.S. Protopopov Experimental study of heat exchange in carbon dioxide in the supercritical range at high temperature drops Teplofizika Vysokikh Temperature 4 3 1966 389 398 (in Russian) [3] B.S. Petukhov Heat Transfer and Friction in Turbulent Pipe Flow with Variable Physical Properties, Advances in Heat Transfer vol. 6 1970 Academic Press New York [4] J.D. Jackson M.A. Cotton B.P. Axcell Studies of mixed convection in vertical tubes International Journal of Heat and Fluid Flow 10 1 1989 2 15 [5] V.A. Kurganov A.G. Kaptilnyi Flow structure and turbulent transport of a supercritical pressure fluid in a vertical heated tube under the conditions of mixed convection International Journal of Heat and Mass Transfer 36 13 1993 3383 3392 [6] J.H. Song H.Y. Kim H. Kim Y.Y. Bae Heat transfer characteristics of a supercritical fluid flow in a vertical pipe The Journal of Supercritical Fluids 44 2008 164 171 [7] P.X. Jiang Y. Zhang C.R. Zhao R.F. Shi Convection heat transfer of CO 2 at supercritical pressures in a vertical mini tube ar relatively low Reynolds numbers Experimental Thermal and Fluid Science 32 2008 1628 1637 [8] P.X. Jiang Y. Zhang Y.J. Xu R.F. Shi Experimental and numerical investigation of convection heat transfer of CO 2 at supercritical pressures in a vertical tube at low Reynolds numbers International Journal of Thermal Sciences 47 2008 998 1011 [9] I.L. Pioro H.F. Khartabil R.B. Duffey Heat transfer to supercritical fluids flowing in channels – empirical correlations (survey) Nuclear Engineering and Design 230 1–3 2004 69 91 [10] V.S. Protopopov Generalized correlations for local heat transfer coefficient for turbulent flow of water and carbon dioxide at supercritical pressure in uniformly heated tubes Teplofizika Vysokikh Temperature 15 4 1977 815 821 (in Russian) [11] D.M. McEligot J.D. Jackson “Deterioration” criteria for convective heat transfer in gas flow through non-circular ducts Nuclear Engineering and Design 232 2004 327 333 [12] K.K. Jong K.J. Hong S.L. Joon Wall temperature measurement and heat transfer correlation of turbulent supercritical carbon dioxide flow in vertical circular/non-circular tubes Nuclear Engineering and Design 237 2007 1795 1802 [13] K. Hyungrae Y.K. Hwan H.S. Jin Y.B. Yoon Heat transfer to supercritical pressure carbon dioxide flowing upward through tubes and a narrow annulus passage Progress in Nuclear Engineering 50 2008 518 525 [14] D.M. McEligot C.W. Coon H.C. Perkins Relaminarization in tubes International Journal of Heat and Mass Transfer 13 1970 431 433 [15] V.S. Grigoriev A.F. Polyakov S.V. Rosnovsky Heat transfer of fluids at supercritical pressures with variable heat flux along length in tubes Teplofizika Vysokikh Temperature 15 6 1977 1241 1247 (in Russian) [16] B.S. Petukhov L.G. Genin S.A. Kovalev H eat Transfer in Nuclear Power Equipment 1986 Energoatomizdat Press Moscow (in Russian) [17] V. Gnielinski New equations for heat and mass transfer in turbulent pipe and channel flows International Chemical Engineering 16 1976 359 368 [18] J.D. Jackson, Consideration of the heat transfer properties of supercritical pressure water in connection with the cooling of advanced nuclear reactors, in: Proceedings of the 13th Pacific Basin Nuclear Conference, Shenzhen, China, October 21–25, 2002.