Professional Documents
Culture Documents
Hermes2010 Used
Hermes2010 Used
Hermes2010 Used
a r t i c l e i n f o a b s t r a c t
Article history: Capillary tube flows have been solved through both numerical and analytical approaches. The former
Received 8 June 2009 requires a reasonable understanding of the governing equations of heat and fluid flow, thermodynamic
Accepted 12 October 2009 relations, numerical methods, and computer programming, and therefore are not the suitable approach
Available online 20 October 2009
for most refrigeration and air-conditioning practitioners. Some simpler procedures based on different
strategies for analytically solving the capillary tube flow have been proposed in the literature, although
Keywords: iterative loops for calculating the mass flow rate are still required. The aim of this work is to advance a
Refrigeration
semi-empirical algebraic model to solve adiabatic capillary tube flows using a relatively simple set of
Capillary tube
Algebraic model
thermodynamic relations and being explicit for the mass flow rate calculation. Comparisons with a com-
Expansion device prehensive experimental data set, taken with the refrigerants HFC-134a and HC-600a, has shown that the
model predicts more than 90% and nearly 100% of all data within ±10% and ±15% error bands,
respectively.
Ó 2009 Elsevier Ltd. All rights reserved.
1359-4311/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.applthermaleng.2009.10.005
450 C.J.L. Hermes et al. / Applied Thermal Engineering 30 (2010) 449–457
Nomenclature
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
boundary conditions and only two equations, but one boundary uftp pi pf pf pe
u 2 þ a þ ab2 ln ap e þb
condition (i.e., the exit pressure) has to be used for the mass flow pD u
t
fsp v f apf þb
w¼ ftp L
ð9Þ
rate calculation. 4
2D
þ lnðvv e Þ
f
The solution algorithm requires the numerical integration of
Eqs. (4) and (5) using a guessed mass flux, G, which is iteratively
where w is given in (kg/s). In the work of Yilmaz and Unal [14], both
corrected since the flow might be choked at the capillary tube exit
single and two-phase friction factors were calculated using the cor-
[13]. In addition, the thermodynamic properties, particularly the
relation proposed by Churchill [15], considering the following aver-
specific volume and its derivatives (@v/@p)h and (ov/oh)p, and the
age viscosity for the two-phase region,
friction factor ought to be calculated at every point of the solution
domain. Because of these, the numerical approach is time consum- " #
8 1 ðpe =pf Þ7=8
ing and requires some programming abilities. As an alternative, gtp ¼ gf ð10Þ
algebraic analytical solutions for capillary tube flows have been
7 1 pe =pf
proposed in the literature, although most of them relied on itera-
tive loops for calculating the refrigerant mass flow rate. where index f refers to the flash-point properties. Such a position is
determined either as shown in Fig. 1a, for subcooled, or Fig. 1b for
two-phase flow inlet conditions.
2. Existing algebraic solutions
The predictions of the formulation introduced by Yilmaz and
Unal [14] have been compared with hundreds of in-house experi-
Yilmaz and Unal [14] proposed an algebraic model for predict-
mental data points taken with adiabatic capillary tubes and with
ing the mass flow rate of pure refrigerant flows through adiabatic
the refrigerants HFC-134a and HC-600a. Table 1 summarized the
capillary tubes. The flow was regarded as isenthalpic, thus allowing
operational and geometric conditions used during the experi-
Eqs. (1)–(3) to be re-written as
ments. As shown in Fig. 2, the Yilmaz–Unal model predicts 86.5%
dz 2D of all data within ±10% error bands.
¼ 2 ½1 þ G2 ð@ v =@pÞh ð6Þ
dp fG v In their work, Yilmaz and Unal [14] assumed the exit pressure
as equal to the evaporating pressure, thus neglecting the occur-
Assuming the liquid specific volume as constant and integrating
rence of choked flow at the capillary outlet. Zhang and Ding [16]
Eq. (6) from the capillary tube entrance until the flash point
improved the Yilmaz and Unal’s [14] model by considering the exit
(Fig. 1a), the liquid region length is then calculated from
Z pressure to be pe = max(pevap, psonic), where the sonic pressure at
f
2D 2D pi pf the capillary exit was obtained setting dz/dp ? 0 in Eq. (6),
Lsp ¼ dp ¼ ð7Þ
i
2
fG v fsp G2 vf qffiffiffiffiffiffiffiffiffiffiffiffi
psonic ¼ G v f pf k ð11Þ
In addition, Yilmaz and Unal [14] demonstrated that the two-
phase specific volume along an isenthalpic line can be calculated Zhang and Ding [16] added several other contributions to Yil-
by v = a + b/p, where a = vf(1 k), b = vfpfk, k = 2.62 105pf0.75, maz and Unal’s [14] formulation. Firstly, they refitted the correla-
pf = pvap(hliq = hi), vf = vliq(pf), and thus (@v/@p)h = b/p2. Therefore, tion for k(pf), proposing k = 1.63 105pf0.72. Also noting that
the two-phase length was calculated integrating Eq. (6) from the Yilmaz and Unal’s [14] formulation was implicit due to the friction
flash point to the capillary tube exit, factor dependence on the mass flow rate, they proposed a two-step
Z !
2D 1 G2 b=p2
e predictor–corrector solution for Eq. (9) based on an approximated
Ltp ¼ 2
dp mass flow rate to calculate the friction factor.
f fG a þ b=p
" ! Fig. 3 compares the predictions of the formulation introduced
#
2D pf pe b ape þ b 2 ve by Zhang and Ding [16] with the entire database, showing that
¼ þ ln G ln ð8Þ the model predicts 82.9% of all data within ±10% error bands.
ftp G2 a a2 apf þ b vf
Later, Yang and Wang [17] derived an empirical p-type dimen-
Noting that the total tube length is given by L = Lsp + Ltp, Eqs. (7) sionless correlation based on the formulation of Zhang and Ding
and (8) can be re-written for the mass flow rate, w, [16], as follows,
C.J.L. Hermes et al. / Applied Thermal Engineering 30 (2010) 449–457 451
inlet
condensing pressure
flash-point
Pressure
outlet
Enthalpy
flash-point
inlet
condensing pressure
Pressure
outlet
Enthalpy
Fig. 1. Schematic representation of flash-point calculation: (a) subcooled liquid and (b) two-phase refrigerant.
Table 1
Range of experimental data.
Fluid # Points Source Range D (mm) L (m) pi (bar) Dtsub (°C) pe (bar) w (kg/h)
R-600a 189 [2,3] Max 0.770 2.926 11.31 15.9 1.08 3.90
Min 2.009 7.13 1.9 0.93 2.01
R-134a 572 Max 1.050 3.020 16.63 20.9 2.12 15.70
Min 0.606 2.009 9.01 1.3 0.92 1.61
!0:7338
w pi v v ;i 0:2220 D 0:4671 3. Proposed explicit solution
pffiffiffiffiffiffiffiffiffiffiffi ¼ 0:2957
2
pD pi =v i pf v l;i L
The formulation introduced in this study was devised consider-
pffiffiffiffiffiffiffiffiffiffiffi!0:1226
D p i =v i DT sub 0:7061 ing that: (i) the flow can be regarded as isenthalpic [14,16]; and (ii)
ð1 xi Þ1:5956 1 þ ð12Þ the pressure drop due to flow acceleration (G2dv) is negligible in
gi Tc
comparison to the friction term (fG2vdz/2D). A brief inspection of
where w is given in (kg/s). Fig. 4 compares Eq. (12) predictions with Eq. (9) shows that the term fL/2D is at least one order of magnitude
all experimental data points, showing that the model predicts 64.7% higher than the term ln (ve/vf). This is also shown in Fig. 5, where
of all data within ±10% error bands. Although simple and explicit in the expression ln (ve/vf)/(fL/2D) is plotted as a function of the Rey-
nature, Eq. (12) does not provide the same level of agreement with nolds number (Re = GD/gf) for all experimental data points, reveal-
experimental data such as the implicit models [14,16]. To put for- ing values of the order of 0.1. It is worth noting that even by
ward an accurate and explicit model for the refrigerant flow eliminating the term Gdv from the momentum conservation equa-
through adiabatic capillary tubes is, therefore, the main focus of this tion, the increase of the specific volume is still accounted for by the
study. frictional term. Eq. (6) can then be simplified to
452 C.J.L. Hermes et al. / Applied Thermal Engineering 30 (2010) 449–457
25
Yilmaz-Unal R-600a
R-134a
20
15
+10%
10
-5
-10
-10%
-15
-20
-25
0 4 8 12 16 20
Measured mass flow rate [kg/h]
Fig. 2. Comparisons of the predictions of Yilmaz–Unal model with experimental data.
30
Zhang-Ding R-600a
R-134a
25
20
15
+10%
Relative error [%]
10
-5
-10
-10%
-15
-20
-25
0 4 8 12 16 20
Measured mass flow rate [kg/h]
Fig. 3. Comparisons of the predictions of Zhang–Ding model with experimental data.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2D u 5" !#
dz ¼ dp ð13Þ uD pi pf pf pe b ape þ b
fG
2
v w ¼ Ut þ þ 2 ln ð15Þ
L vf a a apf þ b
which may be re-written for the mass flow rate as
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
! where U=p(8f)1/2, a = vf(1 k), b = vfpfk, k = 1.63 105pf0.72,
u 5 Z pe
p u D dp and pe ffi pevap. Fig. 6a plots the experimental values of U as a
w ¼ pffiffiffi t ð14Þ
2 2 fL pi v function of the Reynolds number (Re = 4w/pDgf) for all experi-
mental data points considered in this study. In this case, the
Adopting the approximation introduced by Yilmaz and Unal two-phase specific volume was calculated using the correlation
[14] for the two-phase specific volume, v = a + b/p, Eq. (14) may suggested by Zhang and Ding [16] (i.e., k = 1.63 105pf0.72).
be re-written as, Fig. 6b illustrates a histogram with the frequency distribution
C.J.L. Hermes et al. / Applied Thermal Engineering 30 (2010) 449–457 453
25
Yang-Wang R-600a
R-134a
20
15
+10%
10
-5
-10
-10%
-15
-20
-25
0 4 8 12 16 20
Measured mass flow rate [kg/h]
Fig. 4. Comparisons of the predictions of Yang–Wang correlation with experimental data.
0.0
ln(vf/ve) / fL/2D
-0.1
-0.2
10 3 10 4 10 5 10 6
Re
Fig. 5. Comparison between the acceleration, ln (ve/vf), and the friction terms, fL/2D, in equation (9).
of U, which follows a Gaussian distribution pattern with an and ±15% error bands, respectively. A global RMS error of 6.3%
average value of 6.08 and a standard deviation of 0.49, with was achieved.
most of the U-values (80%) concentrated within the interval To improve the proposed method a step further, the friction fac-
[5.5, 6.5]. It should be noted that assuming U 6.0 and approx- tor in Eq. (15) was expressed as a function of Reynolds number,
imating the exit pressure by its evaporating counterpart, Eq. f = c(4w/pDgf)d, and then inserted into Eq. (15) in order to generate
(15) can be written for mass flow rate, tube length or capillary an explicit equation for calculating the refrigerant mass flow rate,
diameter, based in three thermodynamic relations only, hliq(Tsat),
( " !#)1=ð2dÞ
vliq(p), and p(hliq). p2d 22d3 D5d pc pf pf pe b ape þ b
Fig. 7 compares the predictions of Eq. (15) with all experi- w¼ þ þ 2 ln
c gdf L vf a a apf þ b
mental data points, where a reasonable level of agreement is
shown, with 89.1% and 98.3% of all data being within ±10% ð16Þ
454 C.J.L. Hermes et al. / Applied Thermal Engineering 30 (2010) 449–457
7
R-600a
(a)
R-134a
6.5
5.5
5
103 104 105
Re
50
(b)
45
40
35
Frequency [%]
30
25
20
15
10
0
5.0 - 5.5 5.5 - 6.0 6.0 - 6.5 6.5 - 7.0
Interval of Φ
Fig. 6. Variation of U with Re (a) and frequency distribution of U (b).
As illustrated in Fig. 8, the friction factor was fitted using all error around 7%. In addition, Eq. (16), with c = 0.14 and d = 0.15,
experimental data points, providing c = 0.14 and d = 0.l5. Fig. 9 presents a similar accuracy than that obtained through the distrib-
compares the predictions of Eq. (16) with all experimental data uted model proposed by Hermes et al. [13]. It should be mentioned
points, where a good level of agreement is shown, with 91.3% that all thermodynamic properties used in this work were com-
and 99.1% of all data being within ±10% and ±15% error bands, puted using the REFPROP software [18] linked to EES platform [19].
respectively. A global RMS error of 5.5% was observed. It is worth
noting that Eq. (16) requires an extra thermodynamic relation for 4. Concluding remarks
calculating gf = gliq(pf), which is a low price to pay for its better
accuracy. The conclusions of this study can be summarized as follows:
The accuracy of all formulations discussed in this paper was
also compared in Fig. 10 and in Table 2. It can be seen that Eq. An explicit semi-empirical algebraic equation was proposed to
(15), with U = 6.0, and the implicit formulation [14] show a similar calculate the refrigerant mass flow rate, the tube length or the
level of agreement with the experimental data, both with an RMS capillary diameter for adiabatic flows.
C.J.L. Hermes et al. / Applied Thermal Engineering 30 (2010) 449–457 455
25
Φ -formulation R-600a
R-134a
20
15
+10%
10
-5
-10
-10%
-15
-20
-25
0 4 8 12 16 20
Measured mass flow rate [kg/h]
Fig. 7. Comparisons of the predictions of the U-formulation with experimental data.
10-1
R-600a
R-134a
2x10-2
10-2
10 3 10 4 10 5
Re
The model predictions were compared to a comprehensive It should be emphasized that the internal diameters of all cap-
experimental dataset, predicting 91% of all data within ±10% illary tubes used in this work were measured with a maximum
error bands. experimental uncertainty of ±0.03 mm, which translates in varia-
Eq. (15) and the implicit formulation [14] showed a similar tions up to ±5% in the mean diameters. Considering that w D2.5
accuracy, in the same way Eq. (16) presented the same accu- in Eq. (15), then Dw/w 2.5 DD/D, i.e., a 5% variation in the capil-
racy as that obtained using a sophisticated distributed model lary tube diameter leads to errors up to 12.5% in the mass flow rate
[13]. predictions, a value of similar order of magnitude as those obtained
The proposed model requires few thermodynamic relationships, during the model validation exercise. Therefore, model predictions
i.e., only relations for hliq(Tsat), vliq(psat), p(hliq), and gliq(psat) are with deviations of ±10% in relation to the experimental data can be
actually needed. understood as having a satisfactory level of performance.
456 C.J.L. Hermes et al. / Applied Thermal Engineering 30 (2010) 449–457
25
friction factor formulation R-600a
R-134a
20
15
+10%
10
-5
-10
-10%
-15
-20
-25
0 4 8 12 16 20
Measured mass flow rate [kg/h]
Fig. 9. Comparisons of the predictions of the f-formulation with experimental data.
100
<10%
90 <15%
80
Error frequency distribution [%]
70
60
50
40
30
20
10
0
Yilmaz & Unal [14] Zhang & Ding [16] Yang & Wang [17] Proposed, Φ Proposed, f
Table 2
References
Error frequency distribution obtained through all methodologies.
Error Yilmaz and Zhang and Yang and Hermes Proposed, Proposed, [1] P.K. Bansal, A.S. Rupasinghe, An empirical model for sizing capillary tubes,
band (%) Unal [14] Ding [16] Wang [17] et al. [13] U f International Journal of Refrigeration 19 (1996) 497–505.
[2] C. Melo, R.T.S. Ferreira, C. Boabaid Neto, Empirical correlations for the
<10 86.5 82.9 64.7 91.5 89.1 91.3 modelling of HFC-134a flow through adiabatic capillary tubes, ASHRAE
<15 94.0 92.2 87.4 97.5 98.3 99.1 Transactions 105 (1999) 51–59.
>15 6.0 7.8 12.6 2.5 1.7 0.9 [3] C. Melo, R.T.S. Ferreira, C. Boabaid Neto, J.M. Goncalves, M.M. Mezavilla, An
RMS (%) 6.9 7.7 9.9 6.0 6.3 5.5 experimental analysis of adiabatic capillary tubes, Applied Thermal
Engineering 19 (1999) 669–684.
[4] C. Melo, L.A.T. Vieira, R.H. Pereira, Experimental study on adiabatic flow of R-22
alternatives in capillary tubes, in: International Refrigeration Conference at
Purdue, Paper R-075, West Lafayette-IN, USA, 2004.
Acknowledgements [5] J. Choi, Y. Kim, J.T. Chung, An empirical correlation and rating charts for the
performance of adiabatic capillary tubes with alternative refrigerants, Applied
Financial support from the Brazilian funding agencies CNPq and Thermal Engineering 24 (2004) 29–41.
[6] C. Zhang, Generalized correlation of refrigerant mass flow rate through
FINEP is duly acknowledged. Thanks are also due to Embraco S.A., adiabatic capillary tubes using artificial neural network, International Journal
for sponsoring this research project. of Refrigeration 28 (2005) 506–514.
C.J.L. Hermes et al. / Applied Thermal Engineering 30 (2010) 449–457 457
[7] C. Melo, R.T.S. Ferreira, R.H. Pereira, Modelling adiabatic capillary tubes: a [13] C.J.L. Hermes, C. Melo, J.M. Gonçalves, Modeling of non-adiabatic capillary tube
critical analysis, in: International Refrigeration Conference: Energy Efficiency flows: a simplified approach and comprehensive experimental validation,
and New Refrigerants, West Lafayette-IN, USA, 1992, pp. 113–122. International Journal of Refrigeration 31 (2008) 1358–1367.
[8] A.L. Seixlack, A.T. Prata, C. Melo, Modeling the HFC-134a flow through capillary [14] T. Yilmaz, S. Unal, General equations for the design of capillary tubes, ASME
tubes using a two-fluid model, in: International Refrigeration Conference at Journal of Fluids Engineering 118 (1996) 150–154.
Purdue, West Lafayette, IN, USA, 1996, pp. 89–94. [15] S.W. Churchill, Friction factor equations spans all fluid flow regimes, in:
[9] M. Chung, A numerical procedure for simulation of Fanno flows of refrigerants International Congress of Refrigeration, Madrid, 1977.
or refrigerants mixture in capillary tubes, ASHRAE Transactions 104 (1998) [16] C. Zhang, G. Ding, Approximate analytic solutions of adiabatic capillary tube,
1031–1043. International Journal of Refrigeration 27 (2004) 17–24.
[10] O. Garcia-Valladares, C.D. Perez-Segarra, A. Olivia, Numerical simulation of [17] L. Yang, W. Wang, A generalized correlation for the characteristics of
capillary tube expansion devices behaviour with pure and mixed refrigerants adiabatic capillary tubes, International Journal of Refrigeration 31 (2008)
considering metastable region. Part I: mathematical formulation and 197–203.
numerical model, Applied Thermal Engineering 22 (2002) 173–182. [18] E.W. Lemmon, M.O. McLinden, M.L. Huber, NIST Reference Fluids
[11] O. Garcia-Valladares, C.D. Perez-Segarra, A. Olivia, Numerical simulation of Thermodynamic and Transport Properties REFPROP 7.0, Standard
capillary tube expansion devices behaviour with pure and mixed refrigerants Reference Database 23, National Institute of Standards and Technology,
considering metastable region. Part II: experimental validation and parametric Gaithersburg-MD, USA, 2002.
studies, Applied Thermal Engineering 22 (2002) 379–391. [19] S.A. Klein, F.L. Alvarado, EES – Engineering Equation Solver User’s annual, F-
[12] P.K. Bansal, G. Wang, Numerical analysis of choked refrigerant flow in Chart Software, Middleton, WI, USA, 2004.
adiabatic capillary tubes, Applied Thermal Engineering 24 (2004) 851–863.