International Communications in Heat and Mass Transfer: Sciencedirect

International Communications in Heat and Mass Transfer 105 (2019) 84–106
Contents lists available at ScienceDirect
International Communications in Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ichmt
Heat transfer coefficients of laminar, transitional, quasi-turbulent and T

turbulent flow in circular tubes
J.P. Meyer , M. Everts, N. Coetzee, K. Grote, M. Steyn
⁎
Department of Mechanical and Aeronautical Engineering, University of Pretoria, Private Bag X20, Hatfield, 0028, Pretoria, South Africa
ARTICLE INFO ABSTRACT
Keywords: Several well-known correlations to determine the heat transfer coefficients of quasi-turbulent and turbulent flow
Nusselt number in smooth tubes are available in literature. However, when these correlations are compared with each other, the
Friction factor results vary over a considerable range. The purpose of this study was therefore to conduct heat transfer and
Reynolds number pressure drop experiments in the quasi-turbulent and turbulent flow regimes and to develop an accurate heat
Heat transfer
transfer correlation that can be combined with recently developed laminar and transitional flow correlations to
Pressure drop
Circular
obtain a single correlation that is valid for all flow regimes. A total of 1180 experimental data points were
Turbulent collected from careful experiments that were conducted ourselves using two different test section configurations.
Quasi-turbulent The first test section configuration consisted of a tube-in-tube test section on which the wall temperatures were
Transition obtained either indirectly using the Wilson plot method or by direct surface temperature measurements. The
Laminar second test section configuration consisted of single tubes being electrically heated at a constant heat flux.
Smooth tubes Different test sections covering a range of tube diameters from 4 mm to 19 mm and a range of tube lengths from
Uncertainty 1 m to 9.5 m, were used. Experiments were conducted from a Reynolds number of 2445, which corresponded to
the start of the quasi-turbulent flow regime, up to 220,800, which was well into the turbulent flow regime.
Water, as well as different concentrations of multi-walled carbon nanotubes, were used as the test fluid, which
gave a Prandtl number range of 3–10. A new correlation was developed that could estimate 95% of all the
experimental data points within 10% and an average deviation of < 5%. Furthermore, it was able to predict
experimental data in literature with a Prandtl number range of 0.47–276 and Reynolds number range of
3000–401,600 with an average deviation of 14%.
1. Introduction correlations [1]:

p
D
Turbulent heat transfer in circular tubes has been well researched Nu = cRem Pr n
L (1)
and documented over the past 100 years. Fig. 1 compares the number of
publications containing the phrase “turbulent heat transfer” in the title, With the aim of producing a single equation to accurately predict
abstract or keywords per year according to Scopus. This figure indicates the heat transfer coefficients in both heating and cooling applications,
that although the topic of turbulent heat transfer in tubes has been well Dittus and Boelter [2] performed experiments of their own and com-
researched and documented, it is still gaining interest. bined their data with the data of previous studies [3–6]. Because the D/
Although many turbulent heat transfer data have been recorded and L-values in the source data were very small, the D/L term in Eq. (1),
many correlations have been developed in the past 100 years, there is which was associated with entry effects, was neglected. The resulting
still a considerable large discrepancy in the agreement of these studies correlation (Eq. (2) in Table 1) contained an exponent, n, to predict the
and correlations with each other. Table 1 gives a summary of the well- heating and cooling data separately, which enabled good agreement
known and widely used turbulent heat transfer correlations. The ex- with the previous data sets.
perimental data that are readily available in literature and that was Colburn [9] broke new ground by making use of dimensionless
used to develop the correlations in Table 1 are summarised in Table 2. variables such as Reynolds, Prandtl and Nusselt number to reduce the
In 1915, Nusselt suggested the following form for Nusselt number number of variables in the calculation of the heat transfer coefficients.
Abbreviations: BC, Boundary condition; TM, Temperature measurements method; WP, Wilson Plot method
⁎
Corresponding author.
E-mail address: [email protected] (J.P. Meyer).
https://doi.org/10.1016/j.icheatmasstransfer.2019.03.016
Available online 28 April 2019

0735-1933/ © 2019 Elsevier Ltd. All rights reserved.
J.P. Meyer, et al. International Communications in Heat and Mass Transfer 105 (2019) 84–106
Nomenclature Greek symbols
A Area β Thermal expansion coefficient

Cp Specific heat capacity γ Intermittency factor
C Wilson Plot coefficient/ constant used in correlations ε Surface roughness height
c Coefficient used in correlations δ Uncertainty
D Diameter of tube μ Dynamic viscosity of fluid
f Friction factor ν Kinematic viscosity
g Gravitational acceleration ρ Density of fluid
Gr Grashof number
Gz Graetz number Superscripts
h Heat transfer coefficient
j Colburn j-factor m, n, p Exponents used in correlations
K Temperature correction factor (K = (Pr/Prw)0.11) p Wilson Plot variable
k Thermal conductivity
L Length of tube Subscripts
Lt Thermal entrance length
M Measurement b bulk
m Mass flow rate c Cross-section
Nu Nusselt number cor Correlation
∆P Pressure drop exp Experimental
Pr Prandtl number i Inlet/inner/inside
q Heat flux/ constant heat flux boundary condition lam Laminar
R2 Coefficient of determination MCD Mixed convection thermal entrance length
Re Reynolds number o Outlet/outer/annulus
T Temperature ref Reference
Ts Constant surface temperature boundary condition turb Turbulent
V Fluid velocity w Wall
In addition to this, Colburn recognised the relationship between heat theoretical approached combined with experimental pressure drop
transfer and pressure drop, as well as the effect that the wall tem- data, the two correlations are similar. The difference is that the Prandtl
perature has on fluid properties. The proposed correlation (Eq. (3)) was numbers in the correlation of Dittus and Boelter [2] are calculated at
therefore based on the Chilton-Colburn analogy [40] and the fluid the bulk fluid temperature, while Colburn [9] based the Prandtl num-
properties were calculated at the film temperature instead of the bulk bers on the film temperature. Another important observation made by
temperature. It should be noted that although Eq. (2) was obtained by Colburn [9] was that the heat transfer coefficients decreased and de-
correlating experimental data and Eq. (3) was obtained through a viated from the straight line (on a log-log plot) for Reynolds numbers
1200
1000
"Turbulent Heat Transfer" publications
800
600
400
200
0
1960 1970 1980 1990 2000 2010 2020
Year
Fig. 1. Number of publications containing the phrase “turbulent heat transfer” in the title, abstract or keywords per year according to Scopus (Accessed: May 2018).
85
Table 1 between 2300 and 104. A “resume chart”, which could be used to obtain
Summary of well-known and widely used turbulent heat transfer correlations. the heat transfer coefficients in this region, was proposed because a
Dittus and Boelter [2] (2) suitable correlation for this region could not be developed.
Nu = 0.023Re0.8Prn Sieder and Tate [19] incorporated the viscosity ratio (μ/μw), which
n = 0.3 for cooling accounted for the viscosity gradient of the fluid inside the tube. A
n = 0.4 for heating coefficient of 0.14 was found to be suitable for both heating and cooling
3 × 103 < Re < 106, 0.7 < Pr < 120
Source data: [3–6], [7,8]a
conditions. However, other investigations yielded different values for
Colburn [9] (3) the coefficient [41]. For example, Petukhov [26] suggested coefficients
Nu = 0.023Re0.8Pr1/3 of 0.11 and 0.25 for heating and cooling, respectively, while Büyüka-
Re > 104, 0.7 < Pr < 160 laca and Jackson [42] proposed that the coefficient should be a function
Source data: [3–5,10–17], [18]a
of Reynolds number. Similar to Dittus and Boelter [2], Sieder and Tate
Sieder and Tate [19] (4)
0.14 [19] calculated the fluid properties (except for the viscosity at the tube
µ
Nu = 0.027Re0.8 Pr 1/3
µw wall) at the bulk fluid temperature. The decrease in heat transfer
Re > 104, 0.7 < Pr < 17,600 coefficients at lower Reynolds numbers that was observed by Colburn
Source data: [3,16,17,20–24], [18,25] [9], was also observed by Sieder and Tate [19] between Reynolds
Hausen [1] (5) numbers of 2000 and 104 and the authors explained that these data
0.14
Nu = 0.037(Re 0.75 180) Pr 0.42 1+ ()D 2/3
L
µ
µw
formed part of the transitional flow regime. However, according to a
2300 < Re < 106, 0.6 < Pr < 103
recent study conducted by Everts and Meyer [43] these data fell into
Source data: [3,16,17,20–24], [18,25]a the quasi-turbulent flow regime, between the transitional and turbulent
Petukhov [26] (6) flow regimes.
f
RePr To account for the heat transfer coefficients in the transitional flow
8
Nu =
f regime, Hausen [1] proposed the following form for Nusselt number
1.07 + 12.7 (Pr 2/3 1)
8 correlations:
104 < Re < 5 × 106, 0.5 < Pr < 2000
Source data: [27–30] Nu (Re m C) (8)
Gnielinski [31] (7)
f where m and C are constants which can be obtained experimentally.
(Re 1000) Pr
Nu =
8
1+ ()
D 2/3
K This form made it possible to obtain a single correlation for transitional
f L
1 + 12.7
8
(Pr 2/3 1) and turbulent flow (Eq. (5)). However, it was found that Eq. (5) un-
derpredicted the experimental data at lower Reynolds numbers in the
K= ( ) Pr 0.11
Prw transitional flow regime, as well as when short tubes were considered
2 300 < Re < 106, 0.6 < Pr < 105
[31].
Source data: [3,14,16,17,20,23,29,32–34], [18,25,35–38]a
In an effort to predict the turbulent heat transfer behaviour more
a
It was not possible for the authors to confirm these references because of accurately, the next phase of heat transfer research resulted in more
unavailability. work being undertaken in the field where analytical methods were used
to solve for the equations. Petukhov [26] performed a considerable
amount of analyses, using existing data sets as well as producing some
Table 2
Summary of the readily available turbulent experimental data used to develop the well-known turbulent Nusselt number correlations in Table 1. Ts refers to a
constant surface temperature boundary condition and q to a constant heat flux boundary condition.
Test fluid Boundary condition Heating/ cooling Reynolds number Prandtl number Employed by
Webster [39] Water Ts Heating 4256–51,059 3.7–8.8 McAdams and Frost [6]
Dittus and Boelter [2]
Morris and Whitman [3] Water Ts Heating 10,107–38,816 2.9–3.1 Dittus and Boelter [2]
Gas oil Ts Heating 2282–30,011 32.8–47.2 Colburn [9]
Straw oil Ts Heating 754–13,145 57.7–224 Sieder and Tate [19]
Light motor oil Ts Heating 620–4303 203–782 Gnielinski [31]
Gas oil Ts Cooling 3634–44,149 18.2–31.2
Straw oil Ts Cooling 1327–45,265 8.6–99.2
Light motor oil Ts Cooling 947–9363 61.5–229
Clapp and FitzSimons [20] Water Ts Heating 13,307–60,147 1.7–6.4 Sieder and Tate [19]
Velocite B Ts Heating 521–7093 72.1–397 Gnielinski [31]
Water Ts Cooling 19,098–59,650 1.9–5.0
Velocite B Ts Cooling 840–7218 71.1–105
Lawrence and Sherwood [23] Water Ts Heating 4113–140,005 2.2–8.9 Sieder and Tate [19]
Gnielinski [31]
Sherwood et al. [17] Light hydrocarbon oil Ts Heating 280–5570 102–221 Colburn [9]
Sieder and Tate [19]
Gnielinski [31]
Sherwood and Petrie [16] Water Ts Heating 2800–113,000 3.9–367 Colburn [9]
Acetone Ts Heating 944–89,500 3.7–4.4 Sieder and Tate [19]
Benzene Ts Heating 206–86,700 6.3–8.2 Gnielinski [31]
Kerosene Ts Heating 256–31,800 20.6–23.9
n-Butyl alcohol Ts Heating 297–32,500 35.5–897
Stone et al. [32] n-Hexadecane q Heating 12,500–223,800 11.6–18.6 Gnielinski [31]
Di(2-ethylhexyl) adipate q Heating 13,500–124,200 15.1–24.2
Biphenyl q Heating 31,500–401,600 4.1–6.4
Monoisopropylbiphenyl q Heating 18,500–349,800 5.1–15.5
86
data of his own, to obtain a correlation (Eq. (6)) which accounted for Reynolds number. However, at a Reynolds Number of 10,000, the de-
the effect that the friction factor has on heat transfer in the system. viation between the Nusselt numbers obtained using the correlations of
Although Eq. (6) was more complex than the previous correlations (Eqs. Sieder and Tate [19] and Petukhov [26] was > 50%. This deviation
(2)–(5)), it provided an alternative to the Nusselt number correlation gradually decreased to 40% at a Reynolds number of 200,000.
form proposed by Nusselt (Eq. (1)). It should also be noted that in the period of 1922 to 1936, when the
Gnielinski [31] evaluated the existing correlations and experimental majority of the experiments that formed the basis of the work of many
data available for turbulent flow. A new correlation (Eq. (7)) with a scholars in terms of improvements and refinements were conducted, the
form similar to Petukhov [26] was proposed. However, the Reynolds execution of uncertainty analyses was not a requirement in scholarly
number compensation (Eq. (8)) suggested by Hausen [1], a temperature journals. Therefore, the uncertainties of convective heat transfer cor-
correction factor (K = (Pr/Prw)0.11) similar to what was suggested by relations in smooth tubes, which are widely published in heat transfer
Sieder and Tate [19], and a term [1 + (D/L)2/3] were incorporated. textbooks and used for validation and comparison studies today, are in
Similar to Hausen [1], it was found that in relatively short tubes, the general not readily available. Furthermore, the measuring in-
heat transfer coefficients were dependent on the tube length due to the strumentation available today are more accurate than a century ago,
development of the hydraulic and thermal boundary layers, therefore therefore it should be possible to not only conduct more accurate ex-
the term [1 + (D/L)2/3] was included. Eq. (7) was deemed to be con- periments, but also to derive a more accurate correlation with a
siderably accurate because it was able to predict nearly 90% of the quantified uncertainty. The accuracy of existing correlations in litera-
approximately 800 experimental data points in literature within 20%. ture can then also be evaluated using these results.
However, it should be noted that the uncertainties of these experi- The purpose of this study was therefore to conduct heat transfer and
mental data points in literature were not available. Furthermore, at a pressure drop experiments in the quasi-turbulent and turbulent flow
Nusselt number of approximately 700, the experimental data used by regimes and to develop an accurate heat transfer correlation that can be
Gnielinski [31] deviated up to 50%. To account for the transitional flow combined with recently developed laminar and transitional flow cor-
regime, Gnielinski [44] proposed to use linear interpolation for Rey- relations, in order to obtain a single correlation that is valid for all flow
nolds numbers between 2300 and 4000, and Eq. (7) for Reynolds regimes. The main objectives were: (1) To take accurate heat transfer
numbers > 4000. and pressure drop measurements on a smooth tube in the quasi-tur-
Several other turbulent heat transfer correlations were also devel- bulent and turbulent flow regimes and to quantify the uncertainties of
oped in the past decade. These correlations not only accounted for the Reynolds numbers, Nusselt numbers and friction factors. The
variable fluid properties [41,45], but also for heat transfer coefficients pressure drop data were not the focus of this study as these data were
that fell in the quasi-turbulent flow regime [46–49]. However, the only used as complementary data to the heat transfer data. (2) To
majority of these correlations were obtained numerically [46–48,50] or evaluate and compare the existing turbulent correlations in literature
are restricted to certain fluids [45,49,51–54]. with these data. (3) To develop a new heat transfer correlation from this
Table 2 indicates that the experimental data that were used to de- experimental data and compare it to existing correlations and experi-
velop the correlations in Table 1 were obtained using various test fluids, mental data from literature. (4) To link this work in the quasi-turbulent
but the experiments were only conducted up to a Reynolds number of and turbulent flow regimes to recent work, with low uncertainties,
401,600 [32]. Furthermore, in many cases the experiments conducted conducted in the laminar and transitional flow regimes [43,49,57], by
with the highest Prandtl number fluids were not necessarily in the providing a single Nusselt number correlation that is valid for all flow
turbulent flow regime, but in the laminar flow regime (owing to the regimes.
significant increase in pressure drop with increasing Reynolds number
when high viscosity fluids were used). To the authors' best knowledge 2. Experimental set-up and data logging
Morris and Whitman [3] conducted turbulent experiments with the
maximum Prandtl number of 276 using light motor oil. Therefore, al- 2.1. Experimental set-up
though Table 1 indicates that some correlations are valid op to Rey-
nolds numbers of 5 × 106 and Prandtl numbers of 105, these ranges The experimental set-ups were housed in the Clean Energy Research
were obtained by extrapolation and not using experimental data points. Group laboratory at the University of Pretoria and is shown schemati-
In general it is assumed that the correlations of Gnielinski [31] and cally in Fig. 2(a). The experimental set-ups and different test sections
Petukhov [26] are the most accurate as they were most recently de- that were used have been described in detail in references [43, 49,
veloped and are well-known, while the older correlations should be less 57–65] and will only be briefly discussed in this paper. The experi-
accurate and thus be phased out. However, a search (May 2018) on mental set-up consisted of a closed-loop system, which circulated the
Scopus showed that the older correlations are still actively being used test fluid from a storage tank, through the test section and back to the
by many researchers when validating smooth tube experimental data. storage tank, using electronically controlled pumps. The storage tank
Dittus and Boelter [55] had 292 citations, Colburn [56] 49 citations, was maintained at a preselected temperature (20 °C) because it was
Sieder and Tate [19] 1119 citations, Petukhov [26] 865 citations and externally connected to a thermostat-controlled bath that cooled the
Gnielinski [31] had 155 citations. Therefore, based on the number of heated fluid. The fluid in the storage tank was continuously pumped
citations, the correlation of Sieder and Tate [19] seemed to be the most through a filtration cycle to remove any solid particles that might have
utilised correlation. However, a large amount of the initial work of entered the system, as well as to enhance mixing and prevent thermal
Gnielinski and Petukhov was published in the German and Soviet lit- stratification inside the storage tank.
erature and in textbooks, therefore the number of citations to their A bladder accumulator was installed upstream of the flow meters
work were probably underestimated according to cited literature. For and the test section to dampen possible pulsations from the pump. This
example, the work of Gnielinski was published in 1976 in the journal of ensured constant pressures and mass flow rates at the inlet of the test
International Chemical Engineering with permission from the VDI- section. A bypass valve was inserted after the accumulator to allow a
Verlag GmbH; however, this paper is not on the Scopus database. Thus, fraction of the test fluid to flow back to the storage tank. During the
there is no clear evidence from literature (based on number of citations) experiments, the supply valve was partially closed and the bypass valve
which one of the many correlations are in general the most utilised and partially opened, so that the pump was operated close to its maximum
most accurate. speed to ensure a preselected mass flow rate of test fluid to the test
When comparing the different correlations, it was found that for a section. The increased pump speed, as well as the increase in pressure at
Prandtl number of 7, the Nusselt numbers obtained using the correla- the pump inlet (monitored using the pressure gauge), led to decreased
tions of Petukhov [26] and Gnielinski [31] were within 5% at a fixed mass flow rate pulsations [66]. The valve positions were adjusted
87
A Accumulator
Bypass valve
Filter
G P Flow meter
A P
P Storage tank Pt100
G Pressure gage
Thermostat Pressure relief valve
bath P Pump
Test section
Valve
Flow-calming section
(a)
Power supply
Storage tank
(b)
P A
Heater/ Chiller
(c)
Fig. 2. Schematic representation of (a) the experimental set-up used to conduct heat transfer and pressure drop measurements, (b) the constant heat flux test section
and (c) the constant surface temperature test section. Test sections (b) and (c) replace the test section in the dotted blue rectangle in (a).
throughout the experiments to minimise the flow pulsations for all the calming section was used as another bleed valve. The flow-calming
measurements. A pressure relief valve was used to allow the test fluid to section was properly insulated against heat loss using Armaflex in-
flow directly to the storage tank if the pressure exceeded the preselected sulation with a thermal conductivity of 0.034 W/m.K. Peep holes and
threshold. lids were incorporated into the insulation so that any air bubbles could
Because the mass flow rates varied over a wide range, Coriolis mass be detected.
flow meters with different flow rate capacities were installed in parallel.
The mass flow meter that would produce the most accurate mass flow 2.3. Test section
rate measurements was selected during the experiments. The mass flow
rates were controlled by frequency drives, which were connected to the The test sections (Fig. 2(a)) covered a wide range of tube diameters
pump. The required mass flow rate was therefore obtained by in- and tube lengths and the details of the different test sections can be
creasing or decreasing the pump speed. Downstream of the mass flow found in references [43, 49, 57–65]. T-type thermocouples were used to
meters, the fluid flowed through a flow-calming section to the test measure the surface temperatures at selected axial locations on the test
section and mixer, and then back into the storage tank. sections. Depending on the test section configuration and heating
Two different test section configurations were investigated in this method, the thermocouples were either soldered or glued onto the test
study. The first configuration consisted of a single tube that was elec- sections. A 30 mm long capillary tube was silver soldered at each
trically heated to obtain a constant heat flux boundary condition and is pressure tap station. To ensure that the pressure taps did not cause flow
schematically illustrated in Fig. 2(b). A test section in a counter-flow obstructions in the test section [78], a hole of < 10% of the test sec-
tube-in-tube configuration was also used. As schematically illustrated in tion's inner diameter was drilled through the capillary tube and the tube
Fig. 2(c), the counter-flow configuration was serviced with a hot water wall. Care was taken to remove all the burrs from the inside of the test
and a cold water stream. The inner tube of the heat exchanger was used section and the test section was visually inspected using a borescope. A
as the test section and was serviced with cold water and the annulus bush tap with a quick release coupling was fixed to the capillary tube,
was serviced with hot water when the test fluid was being heated. When and nylon tubing was used to connect the pressure taps to the differ-
the test fluid was being cooled, the test section was serviced with hot ential pressure transducers. The test sections were insulated with Ar-
water and the annulus was serviced with cold water. The fluid in the maflex insulation with a thermal conductivity of 0.034 W/m.K.
annulus had a similar closed-loop system than for the test section.
2.4. Mixer
2.2. Flow-calming section
To obtain a uniform outlet temperature, a mixer was inserted after
A flow-calming section, similar to the one used by Ghajar [67–77], the test section. The purpose of the mixer was twofold: to house the
was installed upstream of the test section to straighten the flow. A bleed splitter plates to mix the water exiting the test section, as well as to
valve was installed prior to the inlet section to bleed air that entered the house a Pt100 probe, which was used to measure the outlet tempera-
flow-calming section. The Pt100 probe connection inside the flow- ture. The mixer design was based on work done by Bakker et al. [79],
88
who investigated laminar flow in static mixers with helical splitter The friction factors, f, were calculated from the mass flow rate and
plates. The mixer consisted of four copper splitter plates with a length- pressure drop measurements, ΔP, between two pressure taps, which
to-width ratio of 1.5. The elements were positioned and soldered such were apart from each other a length L:
that the leading edge of an element was perpendicular to the trailing
2 PD P D5 2
edge of the next element. Every splitter plate repeatedly split the f= =
L V2 8m 2 L (13)
thermal boundary layers to ensure a uniform temperature gradient in
the radial direction. The splitter plates were placed inside the acetal In general in this paper, the percentage error of a measurement or
mixer, which directed the fluid to flow over and along the Pt100 probe calculated value was determined as %error = |Mexp – Mcor|/Mref × 100.
after it has been mixed. This ensured that the entire Pt100 probe was When the experimental set-up and data reduction method were vali-
exposed to the mixed fluid and also eliminated any stagnant re- dated, Mref was obtained from existing correlations in literature, Mcor.
circulation zones. The mixer was insulated to prevent any heat loss, and However, when the accuracies of the correlations were determined, Mref
air was bled from the mixer using the Pt100 probe connection to the was obtained from the experimental data, Mexp. The average percentage
mixer housing. error was taken as the average of the absolute errors of the data points.
2.5. Control and data logging 3.1. Surface temperature measurements (TM)
The mass flow rate of the pump was controlled by frequency drives The average surface temperature, Tw, along a tube length, L, mea-
that were connected to a personal computer via a data acquisition sured from the inlet of the test section, was calculated from the local
system. The data acquisition system was used to record the data from surface temperatures, Tw(x), using the trapezoidal rule:
the Pt100 probes (temperatures), thermocouples (temperatures), pres-
1 L
sure transducers (pressure drops) and flow meters (mass flow rates). Tw = Tw (x ) dx
(14)
L 0
The data acquisition system consisted of a personal computer using
National Instruments LabVIEW software as well as SCXI (Signal The heat transfer coefficients, h, were determined from the fol-
Conditioning eXtensions for Instrumentation) hardware, which in- lowing equation, because the heat flux, q, surface temperature, Tw, and
cluded terminal blocks, analogue-to-digital converters and multi- bulk fluid temperature, Tb, were known:
plexers. The measured raw data were saved as .txt files, and Mathworks q
MATLAB scripts were in general used for the data processing. h=
(Tw Tb ) (15)
3. Data reduction The Nusselt numbers, Nu, were determined from the heat transfer
coefficients as follows:
Two different methods were used to obtain the surface tempera- hD
Nu =
tures: (1) direct temperature measurements (TM) using the thermo- k (16)
couples on the test section and (2) the Wilson plot/modified Briggs and
The Grashof numbers, Gr, were determined as
Young method (WP) method [80–82]. The data reduction method used
for the different test section configurations has been described in detail g (Tw Tm ) D3
Gr =
in references [43, 49, 57–65]. Therefore, only the data reduction 2 (17)
method of the main parameters is given in this paper. The bulk fluid 2
where 9.81 m/s was used for the gravitational acceleration, g, and the
temperatures, Tb, were calculated as
kinematic viscosity was obtained from the density and dynamic visc-
Tb =
Ti + To osity (ν = μ/ρ).
2 (9)
The properties of the test fluid (density, ρ, dynamic viscosity, μ, 3.2. Wilson plot method
thermal conductivity, k, specific heat, Cp, Prandtl number, Pr, and
thermal expansion coefficient, β) were determined at the bulk fluid The Reynolds numbers for the inner tube and annulus were calcu-
temperature. lated as follows:
The Reynolds numbers, Re, were calculated as 4mi
Rei =
mD Dii µi (18)
Re =
µAc (10)
4m o
Reo =
where m is the measured mass flow rate inside the tube, D the inner- (Doi Dio ) µo (19)
tube diameter, μ the dynamic viscosity and Ac the cross-sectional area of
For the tube-in-tube test section configurations, the first subscript of
the test section (Ac = π/4D2).
the diameter, D, refers to the tube and the second subscript refers to the
After the Reynolds numbers and Nusselt numbers were calculated
tube surface. For example, Doi indicates the inner surface of the outer
using either surface temperature measurements (Section 3.1) or the
tube. By conducting a wide set of experiments at different mass flow
Wilson plot/modified Briggs and Young method (Section 3.2), the heat
rates [60] for the inner stream, the Nusselt number correlations were
transfer results were also investigated in terms of the Colburn j-factors.
determined as function of Reynolds number, Prandtl number and
This was to account for the variations in the Prandtl numbers of se-
viscosity ratio in the format of the Sieder and Tate equations (Eqs. (20)
quential measurements and to investigate the relationship between heat
and (21)) by using the modified Wilson plot method as prescribed by
transfer and pressure drop:
Briggs and Young [80].
Nu
j= µi
0.14
(11)
1 1
RePr 3 Nui = CiReiPi Pri3
µw (20)
The Graetz numbers, Gz, were determined as
0.14
D 1 µo
Gz = RePr Nuo = CoReoPo Pro3
x (12) µw (21)
89
Table 3
Reynolds number, Nusselt number and friction factor uncertainties of the quasi-turbulent and turbulent experimental data of this study. The uncertainties of the
recent work conducted in the laminar and transitional flow regimes are summarised in Meyer and Everts [57].
Reference Fluid D [m] L [m] BC Heating/ Cooling δRe [%] δNu [%] δf [%]
Grote [59] Water 0.0052 1.0 q Heating 1.1 2.5–3.2 2.0–5.0

MWCNT 1.3–1.7 2.5–3.2 2.0–5.0
Coetzee [60] Water 0.0083 3.75 Ts Cooling 1.0 2.1 0.6–4.5
0.0144 1.0 1.6 0.6–4.5
Steyn [58] Water 0.0084 5.08 Ts Cooling 2.6–2.7 2.1–2.6 11.9–12.2
0.0145 Ts Cooling 1.7–1.8 2.1–4.0 7.0–8.8
0.0145 Ts Heating 1.7–1.9 2.1–4.7 7.0–10.8
Everts [61] Water 0.0115 9.5 q Heating 1.1 3.3–16.7 1.0–2.9
0.004 5.49 1.5 4.9–23.7 9.3–9.4
Abolarin [62] Water 0.019 4.8 q Heating 0.3–0.4 2.1–5.7 0.6–3.5
Bashir [65] Water 0.0051 4.52 q Heating 1.0–1.1 3.4–8.9 2.0–3.8
3.3. Uncertainty analysis and mass flow rate) were captured at a frequency of 10 Hz. The average
value of the 200 measuring points was then used as one data point in
The method proposed by Dunn [83] was used to calculate the un- the calculations. As the mass flow rate was increased with a very small
certainties of the parameters obtained in the data reduction. All un- increment to the next Reynolds number, the time required to reach
certainties were calculated within the 95% confidence interval. The steady-state between Reynolds number increments reduced to ap-
details of the uncertainty analysis method can be found in references proximately 15 min in the quasi-turbulent and turbulent flow regimes.
[43, 49, 57–65] and the results are summarised in Table 3. The Wilson However, the time required to reach steady state depended on the mass
Plot uncertainties were much more challenging than that of the surface flow rate inside the test section [61]. In the laminar flow regime, at very
temperature uncertainties, because the linear regression analysis used low Reynolds numbers, approximately 30 min was required to reach
in the Wilson Plot method had to be incorporated. The details of the steady-state conditions. As the mass flow rate was increased, the time
Wilson Plot uncertainty calculations are given in Coetzee [60]. required for steady state decreased to 20 min. Although the mass flow
Table 3 indicates that the Reynolds number uncertainty was < 3% rates in the transitional flow regime were greater than in the laminar
for all the experimental data. The Nusselt number uncertainties of the flow regime, up to one hour was required to reach steady state owing to
studies that specifically focussed on the turbulent flow regime [58,60] the mass flow rate and temperature fluctuations inside the tube. The
were < 5%. The other studies [59,61,62,65] that focussed more on the Reynolds number was increased by increasing the mass flow rate using
laminar and transitional flow regimes, but also included limited experi- the frequency drives connected to the pumps. The supply and bypass
ments in the turbulent flow regime had higher Nusselt number un- valves were continuously adjusted to ensure that the pumps operated
certainties (with specific reference to Everts [61]). The maximum friction close to their maximum speeds, to reduce mass flow rate pulsations.
factor uncertainties in the different test sections varied between 3% and
12%. The uncertainties of the recent work conducted in the laminar and
transitional flow regimes are summarised in Meyer and Everts [57]. For 5. Experimental test matrix
laminar forced convection, the uncertainties were < 10%, while the
laminar mixed convection uncertainties were < 5%. The Nusselt Table 4 summarises the experiments that were conducted using the
number uncertainties increased to approximately 10% in the transitional different test sections and testing conditions. The test matrix consisted
flow regime. The transitional flow uncertainties were in general higher of 1180 Nusselt number and friction factor data points as a function of
than in the other flow regimes owing to the mass flow rate and Reynolds number and Prandtl number. This covered a wide Reynolds
temperature fluctuations that occurred in this flow regime [43]. number range of 2445–220,818 and a narrower Prandtl number range
of 3–10. Because the maximum Prandtl numbers of the experimental
4. Experimental procedure data of this study were limited to approximately 10, high Prandtl
number experimental data from literature were used to evaluate the
The general experimental procedure is briefly discussed in this performance of the turbulent correlations developed in this study when
paper and more details are given in Everts [61] and Coetzee [60] for the high Prandtl number fluids are used.
constant heat flux (q) and constant surface temperature (Ts) test section The more recent experimental data used from literature between
configurations respectively, as shown schematically in Fig. 2(b) and (c). 2007 and 2016 are summarised in Table 5 and the much older ex-
Wilson-plot experiments were usually operated with the annulus Rey- perimental data used from literature between 1913 and 1964 are
nolds numbers much higher than the Reynolds numbers in the inner summarised in Table 6. The experimental databases obtained from lit-
tube, because this provided accurate results. The regression method erature made available to this study an additional 1147 data points. The
only indirectly produced one average temperature and not a series of total number of Nusselt numbers that were used in Section 9 was thus
temperature measurements in an axial direction, as with the constant 2327. This is a much larger database than was used in previous studies.
heat flux method. Therefore, during these conditions the surface tem- Although the uncertainties of the 1180 data points generated by us for
perature of the inner tube was constant, or assumed to be constant. this study were known, the uncertainties of the 1147 data points from
Steady-state conditions were reached approximately one hour after literature were unfortunately unknown. Recent experimental data in
the start-up of the experimental set-up. Steady-state conditions were the laminar and transitional flow regimes were used complementary to
assumed once there was no increase or decrease in temperatures, the quasi-turbulent and turbulent experimental data of this study, to
pressure drops, and mass flow rates within a period of approximately obtain a single Nusselt number correlation that is valid for all flow
two minutes. Different time periods were considered and a period of regimes. The test matrix is summarised in Meyer and Everts [57] and
approximately two minutes was found to be sufficient. After steady consisted of 1046 mass flow rate measurements, 89,459 temperature
state had been reached, 200 measuring points (temperature, pressure measurements and 2906 pressure drop measurements.
90
Table 4
Experimental test matrix of this study. Ts refers to a constant surface temperature boundary condition and q to a constant heat flux boundary condition.
Author Fluid Data points Reynolds number Prandtl number Boundary condition Heating/ cooling Diameter [mm] Length [m]
Grote [59] Water 140 2656–8356 6.30–7.20 q Heating 5.2 1.0

MWCNT 85 2445–7029 7.57–9.97
Coetzee [60] Water 60 11,905–209,485 3.16–3.88 Ts Cooling 8.3 3.75
63 10,492–218,234 3.32–3.98 14.4
Steyn [58] Water 77 10,679–120,056 3.29–4.42 Ts Cooling 8.4 5.08
135 11,163–220,818 3.08–6.47 Ts Cooling 14.5
62 10,880–91,680 3.91–8.42 Ts Heating
Everts [61] Water 190 2859–10,090 6.09–6.81 q Heating 11.5 9.5
143 2609–8554 4.22–6.11 4.0 5.49
Abolarin [62] Water 156 3225–11,283 5.25–6.58 q Heating 19.0 4.8
Bashir [65] Water 69 2781–7257 4.66–5.90 q Heating 5.1 4.52
Total 1180 2445–220,818 3.08–9.97
Table 5
Experimental data in literature (2007–2016). Ts refers to a constant surface temperature boundary condition and q to a constant heat flux boundary condition.
Author Fluid Data points Reynolds number Prandtl number Testing condition Heating/cooling Diameter [mm] Length [m]
Li et al. [84] #22 lubricating oil 22 7188–19,057 65.0–71.4 Ts Cooling 17.0 2.0
Eiamsa-Ard et al. [85] Air 6 4614–19,451 0.47–0.65 q Heating 4.8 1.5
Buyukalaca et al. [86] Air 15 3099–22,209 0.47–0.72 q Heating 5.6 3.0
Bertsche et al. [87] Water-glycol 18 3319–22,585 7 Ts Cooling 2.6 0.22
7 4041–17,747 10
12 3179–13,497 13
11 3457–12,420 16
Total 91 3099–22,585 0.47–71.4
6. Validation constant heat flux (q) boundary condition [62]. The results correlated
very well with a maximum deviation of 8%, which confirms that the
6.1. Friction factors heat transfer coefficients in the turbulent flow regime are independent
of the boundary condition. The validation of the heat transfer coeffi-
Fig. 3 compares the quasi-turbulent and turbulent friction factors cients are given in Section 8.2 where the prominent turbulent heat
predicted using the correlations of Petukhov [40], Blasius [90], Filo- transfer correlations in literature are compared and evaluated.
nenko [91] and Fang et al. [92], which are the most prominent friction
factor correlations in literature, with the experimental pressure drop 7. Results
data of this study. The performance of the experimental data compared
with the correlations are summarised in Table 7. The results correlated 7.1. Relationship between heat transfer and pressure drop
very well with the Blasius [90] correlation with an average deviation of
only 1.4% and all the data were predicted within 5%. Table 7 also in- Fig. 5 compares the Colburn j-factors and the friction factors as a
dicates that experimental friction factors correlated very well with the function of Reynolds number and the trends of these two parameters
correlations of Filonenko [91] and Fang et al. [92], with > 90% of the were similar (it should be noted that a log-log scale was used). This is as
data within 5% of the correlations and an average deviation of < 2%. It expected because the Chilton-Colburn analogy [40], which was devel-
can therefore be concluded from the experimental results that the ex- oped for laminar and turbulent flow over flat plates, determined that
isting friction factor correlations are adequate and very accurate and the relationship between friction factor (pressure drop) and Colburn j-
that further research on this topic is unnecessary. However, it was ne- factor (heat transfer) was directly proportional. Everts and Meyer [49]
cessary to conduct pressure drop experiments in this study, because the investigated the relationship between pressure drop and heat transfer in
friction factors were required in Section 9. smooth tubes in all flow regimes and found that a direct relationship
between heat transfer and pressure drop existed not only in the laminar
6.2. Heat transfer coefficients and turbulent flow regimes, but also in the transitional and quasi-tur-
bulent flow regimes.
Fig. 4 compares the Nusselt numbers obtained using direct surface To quantitatively investigate the relationship between pressure drop
temperature measurements (TM) [58] with the Nusselt numbers ob- and heat transfer, f/j-factors were obtained by dividing the friction
tained using the Wilson plot method (WP) [60]. Both data sets were factors by the Colburn j-factors (Fig. 6). It is desirable to know this
obtained in test sections with an inner diameter of 14.2 mm and the relationship between heat transfer and pressure drop, because it makes
fluid was cooled in a test section with a relatively constant surface it possible to determine either the heat transfer coefficients or the
temperature (Ts) boundary condition (Fig. 2(c)), thus the Prandtl friction factors when the other variable is known. Everts and Meyer
numbers were approximately the same. [49] concluded that the relationship between heat transfer and pressure
Fig. 4 indicates that the Nusselt numbers of the two data sets cor- drop is a strong function of Grashof number in the laminar flow regime,
related very well with a maximum deviation of only 6%. Therefore, in while it is a strong function of Reynolds number in the other three flow
this paper, both methods generated accurate surface temperatures, and regimes.
thus heat transfer coefficients. Furthermore, these two data sets were A power curve-fit through the f/j-factors between Reynolds numbers
compared with the Nusselt numbers obtained using direct surface of 3000 (quasi-turbulent) and 220,000 (turbulent) in Fig. 6 was per-
temperature measurements in a test section (Di = 19 mm) with a formed to quantify the relationship between heat transfer and pressure
91
Table 6
Experimental data in literature (1913–1964). Ts refers to a constant surface temperature boundary condition and q to a constant heat flux boundary condition.
Author Fluid Data points Reynolds number Prandtl number Testing condition Heating/cooling Diameter [mm] Length [m]
Webster [39] Water 94 4256–51,059 3.71–8.77 Ts Heating 12.7 0.76

Morris & Whitman [3] Water 12 10,107–38,816 2.83–3.15 Ts Heating 15.7 0.27
Gas oil 11 3708–30,011 32.8–41.4
Straw oil 13 3820–13,145 57.7–133
Light motor oil 2 3212–4303 203–276
Gas oil 15 3634–44,149 18.1–31.2 Cooling
Straw oil 23 3038–45,265 8.58–99.3
Light motor oil 6 4265–9363 87.0–164
Clapp & FitzSimons [20] Water 7 13,307–60,147 1.74–6.36 Ts Heating 12.5 1.42
Velocite B 10 3100–7094 72.1–76.4
Water 5 19,098–59,651 1.91–4.98 Cooling 1.33
Velocite B 13 3113–7218 71.1–97.5
Lawrence & Sherwood [23] Water 32 4746–132,530 2.16–7.44 Ts Heating 15.1 3.38
30 5375–140,005 2.16–8.18 2.75
31 6868–98,034 2.32–8.34 1.84
34 4113–97,366 2.83–8.87 0.89
Sherwood & Petrie [16] Water 131 3050–113,000 1.94–3.91 Ts Heating 12.5 1.22
Acetone 54 3730–89,500 3.69–4.45
Benzene 65 3500–86,700 6.37–8.17
Kerosene 52 3430–31,800 10.4–24.2
n-Butyl Alcohol 24 4280–32,500 19.4–35.1
Sherwood et al. [17] Light hydrocarbon oil 26 3130–5570 102–130 Ts Heating 15.1 3.53
24 3100–5260 110–115 2.74
14 3050–4930 132–141 1.83
5 3000–3480 172–174 0.91
Barnes & Jackson [88] Air 84 4200–130,000 0.72 q Heating 6.3 0.30
Helium 69 4000–65,000 0.67
Carbon dioxide 28 8000–40,000 0.75
Stone et al. [32] n-Hexadecane 24 12,500–223,800 11.6–18.6 q Heating 7.9 0.61
Di(2-ethylhexyl) adipate 14 13,500–124,200 15.1–24.2 7.9
Biphenyl 13 56,200–401,600 4.1–6.4 7.9
6 57,400–163,800 4.1–4.2 7.8
8 31,500–164,900 4.2–6.3 7.8
Monoisopropylbiphenyl 27 18,500–349,800 5.1–15.5 7.8
McEligot [89] Air 11 3406–192,271 0.68–0.71 q Heating 3.1 0.51
Helium 3 9628–56,347 0.66
Allen & Eckert [29] Water 36 13,000–111,000 7.94 q Heating 19.1 0.57
Total 1056 3000–401,600 0.66–276
0.045
Petukhov [40]
0.04 Blasius [90]
Filonenko [91]
0.035 Fang et al . [92]
Steyn [58]
Grote [59]
0.03 Coetzee [60]
Everts [61]
Abolarin [62]
Bashir [65]
f
0.025
0.02
0.015
2000 5000 10000 20000 50000 100000 2e+05

Re
Fig. 3. Comparison of the quasi-turbulent and turbulent experimental pressure drop data of this study with the friction factors predicted using the correlations of
Petukhov [40], Blasius [90], Filonenko [91] and Fang et al. [92].
92
Table 7
Performance of the experimental data of this study compared with the most prominent friction factor correlations in literature.
Eq. Range Error
± 5% [%] ± 10% [%] Ave [%]
3 5
Blasius [90] (22) 4 × 10 < Re < 10 100 100 1.4
f = 0.3125Re−0.25
Petukhov [40] (23) 3 × 103 < Re < 5 × 106 78 99 3.1
f = (0.79 ln Re − 1.64)−2
Filonenko [91] (24) 3 × 103 < Re < 106 92 100 2.0
f = (1.8 log Re − 1.5)−2
3 8
Fang et al. [92] (25) 3 × 10 < Re < 10 96 100 1.7
2
f = 0.25 ( log
150.39
Re0.98865 ) 152.66
Re
drop in the quasi-turbulent and turbulent flow regimes as data points within 20% errors. The average deviation was 6.4%. Fur-
thermore, Eq. (28) was able to predict 34% of the experimental data
f
= 24.475Re 0.117
from literature [3,16,17,20,23,29,39,84–89] within errors of 10% and
j (26)
63% of the data points within 20% errors.
By substituting the Colburn j-factor with Eq. (11), the following It should also be noted from the very strong relationship between
correlation was obtained to calculate the Nusselt numbers as a function friction factors and Colburn j-factors (as was shown by Everts and
of friction factor: Meyer [49] and Meyer and Abolarin [62]), that it can be postulated that
Eq. (27) can be considered as a general relationship that will not only
Nu = 0.041Re1.117 Pr 1/3f (27) be valid for the smooth tubes of this study, but also for rough tubes.
Thus, if the friction factor in Eq. (27) of a rough tube can be determined
Therefore, Eq. (27) makes it possible to calculate the Nusselt num-
from the Moody chart [40], correlations, or from experiments, then it
bers when the Reynolds number, Prandtl number and friction factor are
should be possible to directly estimate the Nusselt number from Eq.
known. In the case when friction factor data are not available, the
(27) without additional experiments or from developing additional
friction factor in Eq. (27) can be substituted by the correlation of
correlations, tables or graphs.
Blasius (f = 0.3125Re−0.25) for a smooth tube, because it is shown in
Table 7 that this correlation was able to predict the friction factors of
this study with an average deviation of 1.4%. The resulting correlation 7.2. Evaluation of existing correlations
(without explicitly the friction factor) is then:
The existing correlations in literature were evaluated using the ex-
Nu = 0.013Re0.867 Pr 1/3 (28)
perimental data of this study (Fig. 7) as well as experimental data from
Eq. (28) performed very well and was able to predict 79% of the literature [3,16,17,20,23,29,39,84–89] (Fig. 8). Table 1 indicates that
experimental data of this study within errors of 10% and 96% of the the correlations are functions of fluid properties, such as Prandtl
3
10
2
Nu
10
Temperature measurements, constant surface temperature [58]

Wilson plot method, constant surface temperature [60]
Temperature measurements, constant heat flux [62]
1
10
3 4 5 6
10 10 10 10
Re
Fig. 4. Comparison of the Nusselt numbers obtained using direct surface temperature measurements and the Wilson plot method for both constant surface tem-
perature and constant heat flux boundary conditions.
93
-1 8.3 mm, Ts [60]

10
14.4 mm, Ts [60]
. 2
11.5 mm, q = 1 kW/m [61]
.
11.5 mm, q = 2 kW/m
2
[61]
.
11.5 mm, q = 3 kW/m
2
[61]
f
-2
10
j
3 4 5
10 10 10
Re
Fig. 5. Comparison of the pressure drop and heat transfer results in terms of the friction factors and Colburn j-factors as a function of Reynolds number.
number and viscosity, and test section dimensions. The consequence is correlations. Table 8 summarises the ranges of the correlations, as well
that a single correlation produced several ‘lines’ for different Prandtl as the performance of the correlations compared with the experimental
numbers when plotted in terms of Nusselt number as a function of data of this study and literature.
Reynolds number. It was therefore not feasible to compare the different The black dotted ovals A, C, and D in Fig. 7 indicate that the cor-
correlations in terms of the Nusselt numbers as a function of Reynolds relations of Dittus and Boelter [2], Sieder and Tate [19], and Hausen
number in a single graph, but rather to compare the experimental [1] could not accurately predict the lower Nusselt numbers which fell in
Nusselt numbers to the Nusselt numbers obtained using the different the quasi-turbulent flow regime. Although Table 8 indicates that the
16
8.3 mm, Ts [60]

14.4 mm, Ts [60]
14 . 2
11.5 mm, q = 1 kW/m [61]
. 2
11.5 mm, q = 2 kW/m [61]
.
11.5 mm, q
2
= 3 kW/m [61]
12
10
f/j
Transition Quasi-turbulent
4
Laminar Turbulent
3 4 5
10 10 10
Re
Fig. 6. Comparison of the friction factors divided by the Colburn j-factors as a function of Reynolds number.
94
Fig. 7. Comparison of the experimental heat transfer data of this study with the Nusselt numbers predicted using the correlations of (a) Dittus and Boelter [2], (b)
Colburn [9], (c) Sieder and Tate [19], (d) Hausen [1], (e) Petukhov [26] and (f) Gnielinski [31].
95
±10% ±10%
±20% ±20%
±10% ±10%
±20% ±20%
±10%
±20%
±10%
±20%
Fig. 8. Comparison of the experimental heat transfer data in literature [3,16,17,20,23,29,32,39,84–89] with the Nusselt numbers predicted using the correlations of
(a) Dittus and Boelter [2], (b) Colburn [9], (c) Sieder and Tate [19], (d) Hausen [1], (e) Petukhov [26] and (f) Gnielinski [44].
96
Table 8
Performance of existing Nusselt number correlations compared with the experimental data of this study and literature [3,16,17,20,23,29,32,39,84–89].
Eq. Error
This study Literature [3,16,17,20,23,29,32,39,84–89]
± 10% ± 20% Ave% ± 10% ± 20% Ave%
Dittus and Boelter [2] (2) 38 76 14 40 70 17

Colburn [9] (3) 74 99 7.0 25 57 19
Sieder and Tate [19] (4) 27 68 17 39 61 19
Hausen [1] (5) 38 89 12 44 76 14
Petukhov [26] (6) 72 89 8.5 54 79 13
Gnielinski [31] (7) 70 92 8.0 47 76 15
Range 2445 < Re < 220,818 3000 < Re < 401,600
3 < Pr < 10 0.47 < Pr < 273
correlation of Colburn [9] was able to predict almost all the data of this are not available, because the art of uncertainty analyses was not es-
study within 20% and with an average deviation of only 7%, the black tablished as a requirement for the publication of experimental data in
dotted circle, B, in Fig. 7(b) indicates that the results began to deviate at scholarly journals. Table 8 indicates that the correlations of Hausen [1],
lower Nusselt numbers, thus lower Reynolds numbers in the quasi- Petukhov [26], and Gnielinski [31] gave the best results and were able
turbulent flow regime. Fig. 7(e) and (f) indicate that the correlations of to predict > 70% of the data within 20%. Although the Reynolds
Petukhov [26] and Gnielinski [31] were able to accurately predict the number range of the correlation of Gnielinski [31] was wider, the
majority of the Nusselt numbers in the quasi-turbulent flow regime and overall performance of the correlation of Petukhov [26] was slightly
the average deviation of both correlations was approximately 8%. The better.
correlation of Gnielinski [31] performed slightly better than the cor-
relation of Petukhov [26], because it was able to predict 92% of the
7.3. New Nusselt number correlations
experimental data of this study within 20%. It should be noted that
although the average deviation between the experimental data of this
A new Nusselt number correlation for quasi-turbulent and turbulent
study and the correlation of Gnielinski [31] was 8%, only 70% of the
flow is presented in Section 8.3.1, while a Nusselt number correlation
data were predicted within 10%.
for transitional flow is presented in Section 8.3.2. In Section 8.3.3, a
Comparing the experimental data of this study in Fig. 7 with the
single correlation that can be used for laminar, transitional, quasi-tur-
experimental data from literature [3,16,17,20,23,29,39,84–89] in
bulent and turbulent flow is presented. Section 8.3.4 gives a summary
Fig. 8, significant scatter can be observed in Fig. 8. This is as expected,
of the correlations presented in this paper.
because the experiments were conducted across a period of > 100 years
using a wide range of tube diameters, tube lengths, test fluids, heating
or cooling methods, as well as measuring instrumentation. Un- 7.3.1. Quasi-turbulent and turbulent flow
fortunately, the uncertainties of these experimental heat transfer results Fig. 9 compares the experimental heat transfer data of this study in
terms of the Nusselt number, as a function of Reynolds number. This
Fig. 9. Comparison of the experimental heat transfer data of this study (obtained using different heat fluxes, heating methods, test fluids and test sections) in terms of
the Nusselt number as a function of Reynolds number.
97
figure indicates that the Nusselt numbers increased with increasing points, which produced a coefficient of determination of R2 = 0.999, as
Reynolds number, as can be expected; however, the Nusselt numbers shown in Fig. 11(a):
did not form a single diagonal line and scatter existed. This is because
Nu
the Prandtl numbers of the different experimental data points varied = 0.0575(Re 500)1.071
and, as summarised in Table 4, the experiments were conducted in test Pr 0.42 ( )
Pr 0.11
Prw
f
(29)
sections with a wide range of diameters and lengths, at different heat
fluxes, using different test fluids, heating methods and boundary con- Which can be simplified to,
ditions. 0.11
Pr
Friend and Metzner [93] found that for turbulent flow with fluids Nu = 0.058(Re 500)1.07 Pr 0.42 f
with Prandtl numbers between 0.46 and 346, the effect of Prandtl
Prw (30)
number can be accounted for by using an exponent of 0.42 for the By substituting the friction factor with the Blasius correlation (Eq.
Prandtl number instead of an exponent value of 1/3. Furthermore, (22)), the following Nusselt number for quasi-turbulent and turbulent
Gnielinski [31] introduced the Prandtl number ratio, (Pr/Prw)0.11, to flow was obtained:
account for the variable fluid properties across the cross-section.
0.11
Therefore, to account for different Prandtl numbers (due to the different Nu = 0.018Re 0.25 (Re 500)1.07 Pr 0.42
Pr
fluid temperatures and test fluids), as well as variable fluid properties Prw (31)
across the test section, Fig. 10 compares the heat transfer results in
Fig. 11(b) indicates that Eq. (31) performed very well and was able
terms of Nu/[Pr0.42(Pr/Prw)0.11] as a function of Reynolds number. This
to predict 95% of our own experimental data within 10% and the
figure indicates that the Prandtl number correction, Pr0.42(Pr/Prw)0.11,
average deviation was only 4.4% (which is in the same range as our
was sufficient because the scatter due to the different test sections and
Nusselt number uncertainties of approximately 5%).
operating conditions in Fig. 9, reduced significantly in Fig. 10. The
As discussed previously, Eqs. (27) and (30) will most probably not
dotted circle in Fig. 10 indicates that the results began to deviate at
only be valid for smooth tubes, but also for rough tubes if the friction
Reynolds numbers > 130,000. The decreasing temperature differences
factors of the rough tube can be determined. It should be noted that
with increasing mass flow rates, led to higher uncertainties at very high
with rough tubes, we refer to tubes with significant relative roughness
Reynolds numbers.
values (ε/D) and not to tubes with enhanced surfaces. The measured
The relationship between heat transfer and pressure drop in Fig. 6
relative roughness of the tubes used in our own experiments varied
makes it possible to calculate the Nusselt numbers as a function of
between 1.89 × 10−5 and 3.45 × 10−5. These tubes could con-
friction factor (Eq. (27)). To incorporate this relationship, the heat
servatively be considered as smooth tubes as the Moody chart [40]
transfer results in Fig. 10 were divided by Blasius friction factor (Eq.
indicates that the relative roughness should increase by two orders of
(22)). The correlation of Blasius [90] was used, because friction factor
magnitude to 0.002, before the friction factor will increase by 5% at a
data might not necessarily be available and it was concluded from
Reynolds number of 10,000.
Table 7 that this correlation is both simple and accurate. Furthermore,
The Prandtl numbers of the experimental data of this study were
Gnielinski [31] found that by plotting the heat transfer results as a
limited to 3–10. Therefore, experimental data from literature
function of (Re – 1000), better correlation was obtained at lower Rey-
[3,16,17,20,23,29,39,84–89], with a Reynolds number range of
nolds numbers close to the transitional flow regime. For this study, it
3000–401,600 and a wider Prandtl number range of 0.47–276, were
was found that best results were obtained when the heat transfer results
used to evaluate the performance of Eq. (31) when high Prandtl number
were plotted in terms of Nu/[Pr0.42(Pr/Prw)0.11f] as a function of (Re –
fluids, or fluids with a Prandtl number of < 1, are used. Fig. 12(a)
500). The following turbulent Nusselt number correlation was obtained
compares the heat transfer data of this study and literature
by performing a power curve fit regression through the 1180 data
[3,16,17,20,23,29,39,84–89] in terms of Nu/[Pr0.42(Pr/Prw)0.11f] as a
High uncertainties
Fig. 10. Comparison of the experimental heat transfer data of this study in terms of Nu/[Pr0.42(Pr/Prw)0.11] as a function of Reynolds number.
98
R2 = 0.999
See Fig. 13
±10%
±20%
Fig. 11. Comparison of (a) the experimental heat transfer data of this study in terms of Nu/[Pr0.42(Pr/Prw)0.11f] as a function of Re – 500 and (b) deviation between
Eq. (31) and the experimental data of this study.
function of (Re – 500), while Fig. 12(b) compares the deviation between literature (Table 1) using the experimental data of this study and lit-
Eq. (31) and the experimental data in literature. erature [3,16,17,20,23,29,39,84–89] and the results are summarised in
Eq. (31) was able to predict 50% of the data within errors of 10% Table 9. This table indicates that when the experimental data of this
and 77% of the data within 20% errors and the average deviation was study (relatively low Prandtl numbers) were used to compare the cor-
15%. It should be noted that Eq. (31) was able to predict the 216 ex- relations, the correlations of Colburn [9], Petukhov [26], and Gnielinski
perimental data points obtained using gases [85,86,88,89] with an [31] produced similar Nusselt numbers than Eq. (31). However, when
average deviation of only 11%. Furthermore, Eq. (31) predicted the 87 the experimental data in literature (Prandtl numbers less than unity, as
high Prandtl number experimental data points of Morris and Whitman well as very high Prandtl numbers) were used, the correlations of Dittus
[3] and Li et al. [84] with an average deviation of 9%. It can therefore and Boelter [2], Hausen [1], and Petukhov [26] produced similar
be concluded that Eq. (31) is a suitable correlation to use for gases Nusselt numbers than Eq. (31).
(Prandtl numbers < 1), water (low Prandtl numbers), as well as oils It can therefore be concluded that the agreement of the different
(high Prandtl numbers between 20 and 276). Because Eq. (31) also correlations with each other is, as expected, influenced by the Prandtl
predicted the experimental data of Stone et al. [32] with a Reynolds number and Reynolds number. To investigate this further, Eq. (31) was
number range of 12,000–401,600 with an average deviation of 13%, it compared with the existing correlations in literature (Table 1) between
was concluded that this correlation is suitable for very high Reynolds Reynolds numbers of 2000 and 106 for Prandtl numbers of 0.7, 7 and
numbers. 70. In general, it was found that between Reynolds numbers of 10,000
The Nusselt numbers predicted using Eq. (31) were also compared and 47,000, the correlations were within 20% of each other for all
with the Nusselt numbers predicted with the existing correlations in Prandtl numbers. However, as the Reynolds number was decreased
99
±10%
±20%
Fig. 12. Comparison of (a) the experimental heat transfer data of this study and literature [3,16,17,20,23,29,32,39,84–89] in terms of Nu/[Pr0.42(Pr/Prw)0.11f] as a
function of Re – 500 and (b) deviation between Eq. (31) and the experimental data in literature [3,16,17,20,23,29,39,84–89].
below 10,000 (to the quasi-turbulent flow regime), or increased above than for fully developed flow, Gnielinski [31] used the term [1 + (D/
47,000, the deviation increased up to 60%. In the quasi-turbulent flow L)2/3] to account for very short tubes. From the experimental data of
regime, the correlation of Gnielinski [31] was in good agreement with this study, it was found that the same term [1 + (D/L)2/3] can be added
Eq. (31), while the correlation of Dittus and Boelter [2] correlated to Eq. (31). Therefore, the following general correlation was obtained
better with Eq. (31) at very high Reynolds numbers. to calculate the Nusselt numbers of quasi-turbulent and turbulent flow
Because the heat transfer coefficients of developing flow are higher in short tubes:
Table 9
Comparison of existing Nusselt number correlations compared with Eq. (31) using the experimental data of this study and literature [3,16,17,20,23,29,32,39,84–89].
Eq. Error
This study Literature [3,16,17,20,23,29,32,39,84–89]
± 10% [%] ± 20% [%] Ave [%] ± 10% [%] ± 20% [%] Ave [%]
Dittus and Boelter [2] (2) 33 80 14 75 96 6.8

Colburn [9] (3) 86 100 6.0 37 75 14
Sieder and Tate [19] (4) 29 66 16 91 76 13
Hausen [1] (5) 42 93 12 44 100 4.3
Petukhov [26] (6) 60 98 7.3 69 93 8.1
Gnielinski [31] (7) 67 100 6.9 27 88 13
100
0.25 Pr
0.11
D 2/3 experimental data of this study [61,62,65] were average values across
Nu = 0.018Re (Re 500)1.07 Pr 0.42 1+
Prw L (32) the tube length, while the data of Strickland [94] was obtained at a
single measuring station in the fully developed region of the test sec-
It was found that for the experimental data of this study, Eq. (32) tion. Because fully developed flow conditions existed in the experi-
performed slightly better than Eq. (31) and was able to predict 95% of mental data of this study [61,62,65], it was considered appropriate to
our own experimental data within 10% and the average deviation was incorporate the fully developed experimental data of Strickland [94] as
only 4.4%. Furthermore, Eq. (32) was able to predict 50% of the ex- well.
perimental data from literature [3,16,17,20,23,29,39,84–89] within The Nusselt numbers were divided by Pr0.33 and Gr-0.08, to account
errors of 10% and 76% of the data within 20% errors. The average for different Prandtl number fluids as well as free convection effects,
deviation was 15%. and plotted as a function of Reynolds number in Fig. 14(a). Although
the exponent of the Prandtl number in the turbulent correlations was
7.3.2. Transitional flow 0.42, it was found that better results were obtained in the transitional
Although the results in Fig. 11(a) collapsed onto a single line and flow regime when an exponent of 0.33 was used. Furthermore, the
yielded a correlation with a coefficient of determination, R2, of 0.999, Grashof number exponent of −0.08 might seem negligible; however,
this was only true for the quasi-turbulent and turbulent flow regimes. the Grashof numbers ranged between 1.19 × 103 and 1.75 × 105,
Once the heat transfer results in the transitional flow regime were therefore Gr-0.08 ranged between 0.38 and 0.60, which is significant. It
added, as shown in Fig. 13, it became clear that other parameters need was not possible to exclude the Grashof number from the transitional
to be included to account for the heat transfer characteristics in the flow correlation, because free convection effects significantly affect the
transitional flow regime (Re < 3000). heat transfer characteristics in this flow regime [43]. A linear curve fit
Everts and Meyer [43,49] investigated the heat transfer character- was performed through the 119 data points in Fig. 14(a) to obtain the
istics of transitional flow and found that it was dependent on devel- following correlation that can be used to calculate the heat transfer
oping flow (tube length or axial position), free convection effects coefficients in the transitional flow regime with a square-edged inlet:
(Grashof number), and the type of fluid (Prandtl number). Previous Nu = (0.017Re 30.3) Pr 0.33Gr 0.08
(33)
work by Ghajar and Tam [69] also found that the heat transfer coeffi-
cients in the transitional flow regime were significantly affected by the The results in Fig. 14(b) indicate that Eq. (33) was able to predict
inlet geometry. Taler [47] gives a comprehensive overview of the 43% of the data within 10% errors and 70% of the data within 20%
available correlations in the transitional flow regime. Although several errors and the average deviation was 15%. It should be noted that these
correlations exist in literature, these correlations were either limited to errors are significantly larger than for Eq. (31), which is valid for quasi-
specific fluids, or did not account for free convection effects. The lack of turbulent and turbulent flow regime. This is as expected, because the
experimental data in the transitional flow regime, is probably one of the uncertainties of the heat transfer coefficients in the transitional flow
main reasons for the limited understanding and correlations that are regime are much larger (usually between 10% and 20%) than in the
available [44]. To develop a correlation for transitional flow, the ex- other flow regimes owing to the temperature and mass flow rate fluc-
perimental data obtained using a square-edged inlet [61,62,65] were tuations that occur in this flow regime [43]. Furthermore, due to the
used for this study. As these experiments were conducted using water, limited transitional flow experimental data that are readily available in
the Prandtl number range was 4–6.6. Therefore, to extend the Prandtl literature, significantly less data points were used to develop Eq. (33)
number range, the square-edged inlet transitional flow data of Strick- than Eq. (31). The transitional flow experimental data that are available
land [94], obtained using ethylene glycol-water mixtures in literature were mainly for mixed convection conditions. Limited
(21.9 < Pr < 49), were also used. It should be noted that the experiments were conducted for forced convection conditions;
Transitional flow regime
Gr
Fig. 13. Comparison of the experimental heat transfer data of this study in the transitional, quasi-turbulent and turbulent flow regimes in terms of Nu/[Pr0.42(Pr/
Prw)0.11f] as a function of Re – 500.
101
±10%
±20%
Fig. 14. Comparison of (a) the transitional experimental heat transfer data of this study and of Strickland [94], obtained using a square-edged inlet, in terms of Nu/
(Pr0.33Gr-0.08) as a function of Reynolds number and (b) deviation between Eq. (31) and the experimental data of this study and of Strickland [94].
however, the uncertainties are relatively high [43,57,63]. Furthermore, Nu = 4.36 + Nu1 + Nu2
the available data were for a constant heat flux boundary condition
1
only and not for a constant surface temperature boundary condition. 0.2Lt 0.46
Nu1 = ( 0.84 Pr MCD + 0.72(ReD )0.54Pr 0.34LtMCD )
L
Although the boundary condition had a negligible influence in the
turbulent flow regime, the laminar Nusselt numbers were affected 1
Nu2 = (0.207Gr 0.305 1.19) Pr 0.42 (ReD) 0.08 (L LtMCD )
(Nu = 4.36 for a constant heat flux boundary condition and Nu = 3.66 L
for a constant surface temperature boundary condition). Furthermore, 2.4RePr 0.6D
LtMCD = for L > LtMCD.
the influence of free convection effects were stronger with a constant Gr 0.57
heat flux boundary condition than with a constant surface temperature LtMCD = L for L < LtMCD (34)
boundary condition [70]. It can therefore be expected that the
By making use of the method of Churchill and Usagi [95], a single
boundary condition will also affect the heat transfer characteristics in
correlation that is valid for all flow regimes was obtained:
the transitional flow regime, especially when the mixed convection
conditions exist. 10
Nu = [NuEq 8 8
.(34) + (NuEq .(33) + NuEq .(31))
10/8]0.1
(35)
Eq. (34) (laminar flow), Eq. (33) (transitional flow) and Eq. (31)
7.3.3. Single correlation for laminar, transitional, quasi-turbulent and (quasi-turbulent and turbulent flow) should be used to calculate
turbulent flow NuEq. (34), NuEq. (33) and NuEq. (31), respectively. The exponents of −8
Meyer and Everts [57] developed the following correlation to cal- and 10 that were used in Eq. (35) may appear to be high (as exponents
culate the average laminar Nusselt numbers for developing and fully of 3 and 4 are usually used). However, these high exponents were
developed flow in both forced and mixed convection conditions: chosen to ensure that the error at the intersection of Eqs. (34) and (31)
102
(at the start of the transitional flow regime) and Eqs. (31) and (33) (at transition starts is not necessarily 2300, but will depend on the inlet
the end of the transitional flow regime) was always < 10%, irrespective geometry, tube diameter, developing flow, Prandtl number, as well as
of the Reynolds number, Prandtl number and Grashof number. It should free convection effects [43]. (3) The Reynolds number at which tran-
be noted that although Eqs. (31) and (34) are valid for low and high sition ends is also affected by free convection effects, developing flow
Prandtl number fluids, different inlet geometries and test sections with and fluid properties [43]. (4) The gradient of the Nusselt number in the
a wide range of diameters and lengths, Eq. (33) was developed using transitional flow regime is significantly affected by free convection ef-
data with a Prandtl number range of 4–49 and a square-edged inlet fects and developing flow [43,49,57]. Therefore, Eq. (36) might be a
only. The performance of Eq. (35) was evaluated using 837 data points simple correlation to use to get an indication of the Nusselt numbers in
with a Reynolds number range of 597–46,001, Prandtl number range of the transitional flow regime, but Eq. (35) will give a better re-
3–140 and Grashof number range of 334–4 × 105. It was found that Eq. presentation of the actual transitional Nusselt numbers.
(35) was able to predict 60% of the data within errors of 10% and 79% Furthermore, it should be noted that the following correlation can
of the data within 20% errors. The average deviation was 18%. be extracted from Eq. (35) to obtain a single correlation that is valid for
Gnielinski [44] proposed a different, but simpler, method to obtain transitional to turbulent flow:
a correlation that links the laminar and turbulent Nusselt number cor-
relations: Nu = (NuEq8.(33) + NuEq8.(31) ) 1/8
(37)
Nu = (1 ) Nulam,2300 + Nuturb,4000 where Eq. (33) for transitional flow, and Eq. (31) for quasi-turbulent
and turbulent flow can be used to calculate NuEq. (33) and NuEq. (31),
=
Re 2 300 respectively. Because Eq. (33) was developed using data for a constant
4 000 2 300 (36) heat flux boundary condition, a square-edged inlet and Prandtl numbers
less than 49, Eq. (37) may be accurate for fluids with Prandtl numbers
Separate correlations for laminar and turbulent flow were presented
between 4 and 49 but will become less accurate as the Prandtl number
and the Nusselt number at a Reynolds number of 2300 was calculated
is increased. For fluids with Prandtl numbers between 4 and 49, it was
using the laminar correlation, while the turbulent correlation was used
found that Eq. (37) was able to predict 64% of the data within errors of
to calculate the Nusselt number at a Reynolds number of 4000. Eq. (36)
10% and 90% of the data within 20% errors. The average deviation was
was then used to calculate the Nusselt numbers in the transitional flow
9%.
regime by means of interpolation.
Although this method is simple and easy to implement, the fol-
lowing should be kept in mind when using this method: (1) The laminar 7.3.4. Summary of correlations
correlation that was provided are suited for forced convection condi- Table 10 gives a summary of the correlations and their ranges that
tions and did not account for free convection effects. The Nusselt were presented in this paper. The laminar correlation (Eq. (34)) was
number at the start of transition will thus be underpredicted if mixed developed by Meyer and Everts [57]. This correlation is valid for de-
convection conditions exist [57]. (2) The Reynolds number at which veloping and fully developed laminar flow in both forced and mixed
Table 10
Summary of the correlations presented in this study.
Eq. Range Data ± 10% [%] ± 20% [%] Ave [%]
points
Laminar [57] (34) 48 ≤ Re ≤ 3217, 495a 98 100 3.6

Nu = 4.36 + Nu1 + Nu2 2.9 ≤ Pr ≤ 282, 2.6 ≤ Gz ≤ 1.14 × 105,
Nu1 =
1
( 0.84 Pr 0.2LtMCD + 0.72(ReD)0.54Pr 0.34LtMCD 0.46
) 5.5 ≤ Gr ≤ 4.51 × 105
L
1 0.305 0.42 0.08
Nu2 = (0.207Gr 1.19) Pr (ReD) (L LtMCD )
L
2.4RePr 0.6D
LtMCD = for L > LtMCD
Gr 0.57
LtMCD = L for L < LtMCD
Transitional (33) 2115 ≤ Re ≤ 3586, 119 43 70 15
Nu = (0.017 Re − 30.3)Pr0.33Gr−0.08 4 ≤ Pr ≤ 49, 101b 46 73 14
1.19 × 103 ≤ Gr ≤ 1.75 × 105
Quasi-turbulent and turbulent (Simple) (27) 2445 < Re < 401,600 2351 57 80 12
Nu = 0.041Re1.117Pr1/3fc (28) 0.5 < Pr < 276 1180b 79 96 6.4
Nu = 0.013Re0.867Pr1/3 0.85 < (Pr/Prw)0.11 < 1.17
Quasi-turbulent and turbulent (Lengthy, but more accurate) (30) 2445 < Re < 401,600 2351 72 88 9.5
(31) 0.5 < Pr < 276 1180b 95 100 4.4
Nu = 0.058(Re 500)1.07 Pr 0.42 ( )
Pr 0.11 c
Prw
f
0.85 < (Pr/Prw)0.11 < 1.17
Nu = 0.018Re 0.25 (Re 500)1.07 Pr 0.42 ( )

Pr 0.11
Prw
Quasi-turbulent and turbulent (Short tubes) (32) 2445 < Re < 401,600 2351 73 88 9.8
0.5 < Pr < 276 1180b 95 100 4.4
Nu = 0.018Re 0.25 (Re 500)1.07 Pr 0.42 ( )
Pr 0.11
Prw
1+ ()
D 2/3
L 1.0081 < [1 + (D/L)2/3] < 1.15
0.85 < (Pr/Prw)0.11 < 1.17
Transitional, quasi-turbulent and turbulent (37) 2282 ≤ Re ≤ 46,001, 584 64 90 9.3
Nu = (NuEq. (33)−8 + NuEq. (31)−8)−1/8 4 ≤ Pr ≤ 49, 542b 63 89 9.5
334 ≤ Gr ≤ 1.75 × 105
1.0 < (Pr/Prw)0.11 < 1.02
All flow regimes (35) 597 ≤ Re ≤ 46,001, 837 60 79 18
Nu = [NuEq. (34)10 + (NuEq. (33)
−8
+ NuEq. (31)
−8 −10/8 0.1
) ] 3 ≤ Pr ≤ 139, 768b 60 79 18
334 ≤ Gr ≤ 4.04 × 105
a
Meyer and Everts [57].
b
Performance evaluated using the experimental data of this study.
c
Suitable for rough tubes when appropriate friction factor is used.
103
convection conditions and was therefore selected to be used in the la- accurate, correlation, (2) a more accurate correlation and (3) a corre-
minar flow part of Eq. (35). For continuity from laminar to turbulent lation specifically for short tubes. All three correlations performed very
flow, Eq. (33) was developed to be used in the transitional flow part of well. The more accurate correlation was able to predict 95% of the data
Eq. (35). within 10% and the average deviation was only 4.4%. Furthermore, it
Because this study focussed on quasi-turbulent and turbulent flow, was able to predict a wide range of experimental data in literature (with
three sets of correlations are presented. Eqs. (27) and (28) are easy to unknown uncertainties) with a Prandtl number range of 0.47–276 and
use and quite accurate, while Eqs. (30) and (31) are more accurate. Eq. Reynolds number range of 3000–401,600 with an average deviation
(32) is specifically for short tubes, but can of course be used for short was 15%. It can therefore be concluded that this correlation is able to
and long tubes. Furthermore, Eqs. (27) and (30) are a function of accurately predict the Nusselt numbers in the quasi-turbulent and tur-
friction factor, while the Blasius friction factor for smooth tubes has bulent flow regimes. For continuity from laminar to turbulent flow, a
been incorporated in Eqs. (28) and (31). Therefore, Eqs. (28) and (31) correlation to predict the heat transfer coefficients in the transitional
are valid for smooth tubes only, but it can be postulated that Eqs. (27) flow regime was also developed using the experimental data of this
and (30) may be suitable for rough tubes if the appropriate friction study. This correlation was able to predict 41% of the data within 10%,
factors are used. As Eq. (31) is the most accurate correlation for quasi- 80% of the data within 20%, and the average deviation was 13%.
turbulent and turbulent flow, it is suggested that this correlation is used Finally, a single Nusselt number correlation that is valid for laminar,
in the turbulent flow part of Eq. (35). If no experimental data in the transitional, quasi-turbulent and turbulent flow was also presented.
laminar flow regime are available and only data in the transitional,
quasi-turbulent and turbulent flow regimes are available, Eq. (37) can Acknowledgements
be used to calculate the Nusselt numbers of transitional to turbulent
flow. The funding obtained from both the DST and NRF from South Africa
are duly appreciated. We would also like to acknowledge the assistance
8. Conclusions and recommendations of Professor Volker Gnielinski (Karlsruher
Institut für Technologie) during a visit to our laboratory in 2010, as
In the past century, several correlations to determine the heat well as his feedback as external examiner of Coetzee [60] and his
transfer coefficients in smooth tubes in the turbulent flow regime were subsequent email correspondence. Coetzee, Grote and Steyn were
developed. Unfortunately, when these equations were developed, no master's degree students and Everts a PhD student/postdoctoral fellow.
uncertainty analyses were conducted. The purpose of this study was to All studied under the supervision of the first author.
conduct heat transfer and pressure drop experiments in the quasi-tur-
bulent and turbulent flow regimes and to develop an accurate heat References
transfer correlation, that can be combined with recently developed la-
minar and transitional flow correlations, to obtain a single correlation [1] H. Hausen, Heat Transfer in Counterflow, Parallel-Flow, and Cross-Flow, McGraw-
that is valid for all flow regimes. The main objectives were: (1) To take Hill, New York, 1983.
[2] F. Dittus, L.M.K. Boelter, Heat transfer in automobile radiators of the tubular type,
accurate heat transfer and pressure drop measurements on a smooth Univ. Calif. Publ. Eng. 2 (12) (1930) 443–461.
horizontal tube in the quasi-turbulent and turbulent flow regimes and [3] F.H. Morris, W.G. Whitman, Heat transfer for oils and water in pipes, Ind. Eng.
to quantify the uncertainties of the Reynolds numbers, Nusselt numbers Chem. 20 (3) (1928) 234–240.
[4] W. Nusselt, Der wärmeübergang on rohrleitungen, Z. Ver. Dtsch. Ing. 191-192
and friction factors. (2) To evaluate and compare the existing turbulent (1910) 1750–1755.
correlations in literature with this data. (3) To develop a new correla- [5] E. Josse, Versuche über oberklächenkondensation, insbesondere für dampfturbinen,
tion from this data and to compare it to the existing correlations and Z. Ver. Dtsch. Ing. 53 (1909) 322–330.
[6] W.H. McAdams, T.H. Frost, Heat transfer for water flowing inside pipes, Refrig.
experimental data in literature. (4) To link this work in the quasi-tur- Eng. 10 (9) (1924) 323–334.
bulent and turbulent flow regimes to recent work conducted in the la- [7] B. Hilliger, Untersuchungen über die wirkung von einlagekörpen in den
minar and transitional flow regimes by providing a single Nusselt rauchröhren von lokomiikesselu, Z. Ver. Dtsch. Ing. 60 (1916) 889.
[8] E. Schulze, Versuche Zur bestimmung der WärmeüberGangszahl von luft und
number correlation that is valid for all flow regimes. Heat transfer and
Rauchgas in Technischen rohren, Wärmestelle des Vereines deutscher
pressure drop measurements were taken using two different test section Eisenhüttenleute, vol. 117, (1928).
configurations. The first configuration consisted of a tube-in-tube heat [9] A.P. Colburn, A method of correlating forced convection heat transfer data and a
exchanger to obtain a constant surface temperature boundary condi- comparison with fluid friction, Trans. Am. Inst. Chem. Eng. 29 (1933) 174–210.
[10] R. Poensgen, Ueber die wämeübertragung von strömendem überhitztem wasser-
tion, for both heating and cooling conditions. The second test section dampf an rohrwandungen und von heizgasen as wasserdampf, Z. Ver. Dtsch. Ing. 60
configuration consisted of single tubes being electrically heated at a (1916) 27–32.
constant heat flux. Different test sections covering a range of tube [11] W.H. McAdams, T.H. Frost, Heat transfer by conduction and convection II- liquids
flowing through pipes, J. Ind. Eng. Chem. 14 (12) (1922) 1101–1104.
diameters from 4 mm to 19 mm and a range of tube lengths from 1 m to [12] C.S. Keevil, W.H. McAdams, How heat transmission affects fluid friction in pipes,
9.5 m, were used. A total of 1180 experimental data points were col- Chemi. Mettall. Eng. 36 (8) (1929) 464–467.
lected from careful experiments that were conducted between Reynolds [13] A.P. Colburn, O.A. Hougen, III-flow of fluids at low velocities, Ind. Eng. Chem. 22
(5) (1930) 534–539.
numbers of 2445 and 220,800. [14] A. Eagle, R.M. Ferguson, On the coefficient of heat transfer from the internal surface
The surface temperatures were obtained by either direct tempera- of tube walls, proceedings of the Royal Society, Series A, Containing Papers of a
ture measurements or by making use of the Wilson plot method. It was Mathematical and Physical Character, 1930, pp. 540–566.
[15] R. Hermann, T.H. Burbach, Strömungswiderstand und Wärmeübergang in Rohren,
found that both methods yielded accurate heat transfer coefficients and
(1930).
confirmed that the turbulent heat transfer coefficients were in- [16] T.K. Sherwood, J.M. Petrie, Heat transmission to liquids flowing in pipes, Ind. Eng.
dependent of the boundary condition. An uncertainty analysis was Chem. 24 (7) (1932) 736–745.
[17] T.K. Sherwood, D.D. Kiley, G.E. Mangsen, Heat transmission to oil flowing in pipes
conducted to quantify the uncertainties of the experimental data of this
effect of tube length, Ind. Eng. Chem. 24 (3) (1932) 273–277.
study. The Reynolds number, Nusselt number and friction factor un- [18] H. Kraussold, Gebiete ingenieurw. 2, Forschungsheft, vol. 351, 1931.
certainties were approximately 2%, 5% and 10% respectively. [19] E.N. Sieder, G.E. Tate, Heat transfer and pressure drop of liquids in tubes, Ind. Eng.
By making use of the relationship between heat transfer and pres- Chem. 28 (1936) 1429–1435.
[20] M.H. Clapp, O. FitzSimons, The Effect of Heat Transfer on Friction Factors in
sure drop and accounting for different Prandtl number fluids, variable Fanning's Equation, Masters Dissertation, Massachusetts Intitute of Technology,
fluid properties with temperature, as well as different test section di- Massachusetts, 1929.
mensions and configurations, heating methods and boundary condi- [21] C.G. Kirkbride, W.L. McCabe, Heat transfer to liquids in viscous flow, Ind. Eng.
Chem. 23 (6) (1931) 625–631.
tions, three versatile Nusselt number correlations for flow in the quasi- [22] T.B. Drew, J.J. Hogan, W.H. McAdams, Heat transfer in stream-line flow, Ind. Eng.
turbulent and turbulent flow regimes, were obtained: (1) a simple, yet
104
Chem. 23 (8) (1931) 936–945. [56] A.P. Colburn, A method of correlating forced convection heat-transfer data and a
[23] A.E. Lawrence, T.K. Sherwood, Heat transmission to water flowing in pipes, Ind. comparison with fluid friction, Int. J. Heat Mass Transf. 7 (12) (1964) 1359–1384.
Eng. Chem. 23 (3) (1931) 301–309. [57] J.P. Meyer, M. Everts, Single-phase mixed convection of developing and fully de-
[24] T.B. Drew, Heat transfer in stream-line flow II, Exp. Glycer Ind. Eng. Chem. 24 (2) veloped flow in smooth horizontal circular tubes in the laminar and transitional
(1932) 152–157. flow regimes, Int. J. Heat Mass Transf. 117 (2018) 1251–1273.
[25] C.S. Keevil, PhD Thesis, Massaschusettes Institute of Technology, Cambridge, 1930. [58] M. Steyn, J.P. Meyer, Heat Transfer Coefficients for Tubes in the Turbulent Single
[26] B.S. Petukhov, Heat transfer and friction in turbulent pipe flow with variable Phase Flow Regime with a Focus on Uncertainty, in: 15th International Heat
physical properties, in: J.P.I.T.F. Hartnett (Ed.), Advances in Heat Transfer, Transfer Conference (IHTC-15), Kyoto, Japan, 2014.
Academic Press, New York, 1970, pp. 503–564. [59] J.P. Meyer, T.J. McKrell, K. Grote, The influence of multi-walled carbon nanotubes
[27] V.V. Yakovlev, Local and mean heat-transfer for a turbulent flow o nonboiling water on single-phase heat transfer and pressure drop characteristics in the transitional
in a tube with high heat loads, Sov. J. At. Energy 8 (3) (1960) 221–224. flow regime of smooth tubes, Int. J. Heat Mass Transf. 58 (1–2) (2013) 597–609.
[28] B.S. Petukhov, L.I. Roizen, Teplofiz. Vysok. Temp. 1 (3) (1963) 416–424. [60] N. Coetzee, Heat Transfer Coefficients of Smooth Tubes in the Turbulent Flow
[29] R. Allen, E. Eckert, Friction and heat-transfer measurements to turbulent pipe flow Regime, Masters Dissertation, University of Pretoria, Pretoria, 2015.
of water (Pr = 7 and 8) at uniform wall heat flux, J. Heat Transf. 86 (3) (1964) [61] M. Everts, Single-phase Mixed Convection of Developing and Fully Developed Flow
301–310. in Smooth Horizontal Circular Tubes in the Laminar, Transitional, Quasi-turbulent
[30] L.S. Sterman, V.V. Petukjov, Investigation of the heat transfer to organic liquids, in: and Turbulent flow Regimes, PhD thesis University of Pretoria, Pretoria, 2017.
A.V. Lykov, B.M. Smol'skii (Eds.), Heat and Mass Transfer, 1965. [62] J.P. Meyer, S.M. Abolarin, Heat transfer and pressure drop in the transitional flow
[31] V. Gnielinski, New equations for heat and mass-transfer in turbulent pipe and regime for a smooth circular tube with twisted tape inserts and a square-edged inlet,
channel flow, Int. Chem. Eng. 16 (2) (1976) 359–368. Int. J. Heat Mass Transf. 117 (2018) 11–29.
[32] J.P. Stone, C.T. Ewing, R.R. Miller, Heat-transfer studies on some stable organic [63] M. Everts, J.P. Meyer, Flow regime maps for smooth horizontal tubes at a constant
fluids in a forced convection loop, J. Chem. Eng. Data 7 (4) (1962) 519–525. heat flux, Int. J. Heat Mass Transf. 117 (2018) 1274–1290.
[33] W. Hufschmidt, E. Burck, W. Riebold, Die bestimmung örtlicher und mittlerer [64] J.P. Meyer, M. Everts, A.T.C. Hall, F.A. Mulock Houwer, M. Joubert, L.M.J. Pallent,
wärmeübergangszahlen in rohren bei hohen wärmestromdichten, Int. J. Heat Mass E.S. Vause, Inlet tube spacing and protrusion effects on multiple circular tubes in
Transf. 9 (6) (1966) 539–565. the laminar, transitional and turbulent flow regimes, Int. J. Heat Mass Transf. 118
[34] W. Hufschmidt, E. Burck, Der einfluss temperaturabhängiger stoffwerte auf den (2018) 257–274.
wärmeübergang bei turbulenter strömung von flüssigkeiten in rohren bei hohen [65] A.I. Bashir, J.P. Meyer, Experimental Investigation of Convective Heat Transfer in
wärmestromdichten und prandtlzahlen, Int. J. Heat Mass Transf. 11 (6) (1968) the Transitional Flow Regime of an Inclined Smooth Tube, in: 16th International
1041–1048. Heat Transfer Conference (IHTC16), Beijing, China, 2018.
[35] M.N. Ivanovskii, Rapid Method of Measuring the Average Heat Transfer Coefficient [66] M. Everts, Heat Transfer and Pressure Drop of Developing Flow in Smooth Tubes in
in a Tube, Department of Commerce, Washington, 1962. the Transitional Flow Regime, Master's Dissertation, University of Pretoria,
[36] I.T.A.E. Alad'jev, N.A. Vel'tishchev, N.S. Kondrat'ev, Heat transfer by Convection at Pretoria, 2014.
high pressure, Foreign Technol. Div., Wright-Patterson AFB, Ohio. [67] A.J. Ghajar, L.M. Tam, Laminar-Transition-Turbulent Forced and Mixed Convective
[37] M.A.M. Mikheyev, Mean heat transfer of fluids flowing in tubes, Foreign Technol. Heat Transfer Correlations for Pipe Flows with Different Inlet Configurations, in:
Div.,Wright-Patterson AFB, Ohio. Winter Annual Meeting of the American Society of Mechanical Engineers, Publ by
[38] H. Reinecke, Uber den Warmeubergang von Kurzen Durchstromten Rohren Und ASME, New York, United States, 1991, pp. 15–23.
Querumstromten Zylindren Verschiedener Anordnung an Zahe Flussigkeiten [68] A.J. Ghajar, K.F. Madon, Pressure drop measurements in the transition region for a
Verschiedener Prandtl-Zahl Bei Kleinen Temperaturdifferenzen (Heat Transfer for circular tube with three different inlet configurations, Exp. Thermal Fluid Sci. 5 (1)
Vsicous Liquids of Different Prandtl Number Flow through Tubes and across (1992) 129–135.
Cylinders of Different Orientation for Small Temperature Differences), Dissertation [69] A.J. Ghajar, L.M. Tam, Heat transfer measurements and correlations in the transi-
D17, Techn, Hocshule Darmstadt, 1969. tion region for a circular tube with three different inlet configurations, Exp.
[39] G.C. Webster, Some experiments on the condensation of steam, Trans. Inst. Eng. Thermal Fluid Sci. 8 (1) (1994) 79–90.
Shipbuild. Scotl. 57 (1913) 58–105. [70] A.J. Ghajar, L.M. Tam, Flow regime map for a horizontal pipe with uniform wall
[40] Y.A. Cengel, A.J. Ghajar, Heat and Mass Transfer: fundamentals and Applications, heat flux and three inlet configurations, Exp. Thermal Fluid Sci. 10 (3) (1995)
5th ed., McGraw-Hill, 2015. 287–297.
[41] H. Zhao, X. Li, X. Wu, New friction factor and Nusselt number equations for tur- [71] L.M. Tam, A.J. Ghajar, Effect of inlet geometry and heating on the fully developed
bulent convection of liquids with variable properties in circular tubes, Int. J. Heat friction factor in the transition region of a horizontal tube, Exp. Thermal Fluid Sci.
Mass Transf. 124 (2018) 454–462. 15 (1) (1997) 52–64.
[42] O. Büyükalaca, J.D. Jackson, The correction to take account of variable property [72] L.M. Tam, A.J. Ghajar, The unusual behavior of local heat transfer coefficient in a
effects on turbulent forced convection to water in a pipe, Int. J. Heat Mass Transf. circular tube with a bell-mouth inlet, Exp. Thermal Fluid Sci. 16 (3) (1998)
41 (4–5) (1998) 665–669. 187–194.
[43] M. Everts, J.P. Meyer, Heat transfer of developing and fully developed flow in [73] L.M. Tam, A.J. Ghajar, H.K. Tam, S.C. Tam, Development of a Flow Regime Map for
smooth horizontal tubes in the transitional flow regime, Int. J. Heat Mass Transf. a Horizontal Pipe with the Multi-Classification Support Vector Machines, in: 2008
117 (2018) 1331–1351. ASME Summer Heat Transfer Conference (HT2008), Jacksonville, USA, 2009, pp.
[44] V. Gnielinski, On heat transfer in tubes, Int. J. Heat Mass Transf. 63 (2013) 537–547.
134–140. [74] A.J. Ghajar, C.C. Tang, W.L. Cook, Experimental investigation of friction factor in
[45] P.K. Namburu, D.K. Das, K.M. Tanguturi, R.S. Vajjha, Numerical study of turbulent the transition region for water flow in minitubes and microtubes, Heat Transf. Eng.
flow and heat transfer characteristics of nanofluids considering variable properties, 31 (8) (2010) 646–657.
Int. J. Therm. Sci. 48 (2) (2009) 290–302. [75] H.K. Tam, L.M. Tam, A.J. Ghajar, C.W. Cheong, Development of a unified flow
[46] J.P. Abraham, E.M. Sparrow, J.C.K. Tong, Heat transfer in all pipe flow regimes: regime map for a horizontal pipe with the support vector machines, 2nd
laminar, transitional/intermittent, and turbulent, Int. J. Heat Mass Transf. 52 (3–4) International Symposium on Computational Mechanics (ISCM II), 2010, pp.
(2009) 557–563. 608–613 Hong Kong.
[47] D. Taler, A new heat transfer correlation for transition and turbulent fluid flow in [76] H.K. Tam, L.M. Tam, A.J. Ghajar, S.C. Tam, T. Zhang, Experimental investigation of
tubes, Int. J. Therm. Sci. 108 (2016) 108–122. heat transfer, friction factor, and optimal fin geometries for the internally microfin
[48] D.M. McEligot, X. Chu, R.S. Skifton, E. Laurien, Internal convective heat transfer to tubes in the transition and turbulent regions, J. Enhanc. Heat Transf. 19 (5) (2012)
gases in the low-Reynolds-number “turbulent” range, Int. J. Heat Mass Transf. 121 457–476.
(2018) 1118–1124. [77] H.K. Tam, L.M. Tam, A.J. Ghajar, Effect of inlet geometries and heating on the
[49] M. Everts, J.P. Meyer, Relationship between pressure drop and heat transfer of entrance and fully-developed friction factors in the laminar and transition regions
developing and fully developed flow in smooth horizontal circular tubes in the of a horizontal tube, Exp. Thermal Fluid Sci. 44 (2013) 680–696.
laminar, transitional, quasi-turbulent and turbulent flow regimes, Int. J. Heat Mass [78] R. Rayle, influence of orifice geometry on static pressure measurements, Am. Soc.
Transf. 117 (2018) 1231–1250. Mech. Eng. (1959) ASME Paper 59-A-234; New York.
[50] D. Taler, Simple power-type heat transfer correlations for turbulent pipe flow in [79] A. Bakker, R.D. LaRoche, E.M. Marshall, Laminar flow in static mixers with helical
tubes, J. Therm. Sci. 26 (4) (2017) 339–348. elements, The Online CFM Book, 2000.
[51] W. Yu-ting, L. Bin, M. Chong-fang, G. Hang, Convective heat transfer in the laminar- [80] D.E. Briggs, E.H. Young, Modified Wilson plot techniques for obtaining heat transfer
turbulent transition region with molten salt in a circular tube, Exp. Thermal Fluid correlations for shell and tube heat exchangers, Chem. Eng. Symp. Ser. 92 (65)
Sci. 33 (7) (2009) 1128–1132. (1969) 35–45.
[52] A.R. Sajadi, M.H. Kazemi, Investigation of turbulent convective heat transfer and [81] R.K. Shah, Assessment of the modified Wilson Plot techniques for obtaining heat
pressure drop of TiO 2/water nanofluid in circular tube, Int. Commun. Heat Mass exchanger design data, 9th International Heat Transfer Conference, 1990, pp.
Transf. 38 (10) (2011) 1474–1478. 51–56 Jerusalem.
[53] M.M. Heyhat, F. Kowsary, A.M. Rashidi, S. Alem Varzane Esfehani, A. Amrollahi, [82] E.E. Wilson, A basis for rational design of heat transfer apparatus, ASME J. Heat
Experimental investigation of turbulent flow and convective heat transfer char- Transf. 37 (1915) 47–82.
acteristics of alumina water nanofluids in fully developed flow regime, Int. [83] P.F. Dunn, Measurement and Data Analysis for Engineering and Science, 2nd ed.,
Commun. Heat Mass Transf. 39 (8) (2012) 1272–1278. CRC Press, United States of America, 2010.
[54] Y. Li, K. Fukuda, Q. Liu, M. Shibahara, Turbulent heat transfer with FC-72 in small [84] X.-W. Li, J.-A. Meng, Z.-X. Li, Experimental study of single-phase pressure drop and
diameter tubes, Int. J. Heat Mass Transf. 103 (2016) 428–434. heat transfer in a micro-fin tube, Exp. Thermal Fluid Sci. 32 (2007) 641–648.
[55] F.W. Dittus, L.M.K. Boelter, Heat transfer in automobile radiators of the tubular [85] S. Eiamsa-Ard, V. Kongkaitpaiboon, P. Promvonge, Thermal performance assess-
type, Int. Commun. Heat Mass Transf. 12 (1985) 3–22. ment of turbulent tube flow through wire coil turbulators, Heat Transf. Eng. 32
105
(11−12) (2011) 957–967. [91] G.K. Filonenko, Hydraulischer Widerstand von Rohrleitungen (Orig. Russ.),
[86] O. Buyukalaca, V. Ozceyhan, S. Gunes, Experimental investigation of thermal per- Teploenergetika 1 (4) (1954) 40–44.
formance in a tube with detached circular ring turbulators, Heat Transf. Eng. 33 (8) [92] X. Fang, Y. Xu, Z. Zhou, New correlations of single-phase friction factor for tur-
(2012) 682–692. bulent pipe flow and evaluation of existing single-phase friction factor correlations,
[87] D. Bertsche, P. Knipper, T. Wetzel, Experimental investigation on heat transfer in Nucl. Eng. Des. 241 (2011) 897–902.
laminar, transitional and turbulent circular pipe flow, Int. J. Heat Mass Transf. 95 [93] W.L. Friend, A.B. Metzner, Turbulent heat transfer inside tubes and the analogy
(2016) 1008–1018. among heat, mass, and momentum transfer, AICHE J. 4 (4) (1958) 393–402.
[88] J.F. Barnes, J.D. Jackson, Heat transfer to air, carbon dioxide and helium flowing [94] D.T. Strickland, Heat Transfer Measurements in the Transition Region for a
through smooth circular tubes under conditions of large surface/gas temperature Horizontal Circular Tube with a Square-Edged Entrance, Master's Dissertation,
ratio, Proc. Inst. Mech. Eng. C J. Mech. Eng. Sci. 3 (4) (1961) 303–314. Oklahoma State University, Stillwater, 1990.
[89] D.M. McEligot, Effect of Large Temperature Gradients on Turbulent Flow of Gases [95] S.W. Churchill, R. Usagi, A general expression for the correlation of rates of transfer
in the Downstream Region of Tubes, PhD thesis Stanford University, Stanford, 1963. and other phenomena, AICHE J. 18 (6) (1972) 1121–1128.
[90] F.M. White, Fluid Mechanics, 6th ed., McGraw-Hill, Singapore, 2009.
106

International Communications in Heat and Mass Transfer: Sciencedirect

Uploaded by

Copyright:

Available Formats

International Communications in Heat and Mass Transfer: Sciencedirect

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

International Communications in Heat and Mass Transfer: Sciencedirect

Uploaded by

Copyright:

Available Formats

International Communications in Heat and Mass Transfer 105 (2019) 84–106

Contents lists available at ScienceDirect

International Communications in Heat and Mass Transfer

Heat transfer coefficients of laminar, transitional, quasi-turbulent and T

ARTICLE INFO ABSTRACT

1. Introduction correlations [1]:

Available online 28 April 2019

Nomenclature Greek symbols

A Area β Thermal expansion coefficient

Grote [59] Water 0.0052 1.0 q Heating 1.1 2.5–3.2 2.0–5.0

Grote [59] Water 140 2656–8356 6.30–7.20 q Heating 5.2 1.0

Webster [39] Water 94 4256–51,059 3.71–8.77 Ts Heating 12.7 0.76

0.035 Fang et al . [92]

2000 5000 10000 20000 50000 100000 2e+05

± 5% [%] ± 10% [%] Ave [%]

Temperature measurements, constant surface temperature [58]

-1 8.3 mm, Ts [60]

8.3 mm, Ts [60]

This study Literature [3,16,17,20,23,29,32,39,84–89]

± 10% ± 20% Ave% ± 10% ± 20% Ave%

Dittus and Boelter [2] (2) 38 76 14 40 70 17

This study Literature [3,16,17,20,23,29,32,39,84–89]

Dittus and Boelter [2] (2) 33 80 14 75 96 6.8

Transitional flow regime

Laminar [57] (34) 48 ≤ Re ≤ 3217, 495a 98 100 3.6

Nu = 0.018Re 0.25 (Re 500)1.07 Pr 0.42 ( )

You might also like