Applied Thermal Engineering 23 (2003) 22012211

www.elsevier.com/locate/apthermeng

Loss coecient correlation for wet-cooling tower lls

ger Johannes C. Kloppers, Detlev G. Kro
Received 20 February 2003; accepted 10 June 2003
*

Department of Mechanical Engineering, University of Stellenbosch, Stellenbosch 7600, South Africa

Abstract Loss coecient correlations given in the literature for wet-cooling tower lls are relatively simple and generally do not represent the pressure drop accurately over a wide range of operational conditions. A new form of empirical equation is proposed that correlates ll loss coecient data more eectively when compared to other forms of empirical equations commonly found in the literature. 2003 Elsevier Ltd. All rights reserved.
Keywords: Wet-cooling towers; Loss coecient; Splash ll; Trickle ll; Film ll

1. Introduction The loss coecient of a cooling tower ll is determined by measuring the pressure drop across the ll. The results of these tests are correlated by empirical relations which are functions of the air and water mass ow rates. These empirical relations are subsequently employed in the design of cooling towers to determine the draft through the cooling towers. Suitable fans for mechanical draft cooling towers are selected, based to a large extent, on the loss coecient of the ll. The draft in a natural draft cooling tower is also a strong function of the ll loss coecient. It is thus important to represent the ll loss coecient accurately, as inaccurate representation of the loss coecients in the form of empirical relations can have nancial implications if the cooling tower does not meet design specications. Here follows some forms of empirical equations of the ll loss coecient found in the literature. Lowe and Christie [1] used the following form of equation to represent the loss coecients of counterow splash and lm type lls.

Corresponding author: Tel.: +27-21-8084259; fax: +27-21-8084958. ger). E-mail address: dgk@sun.ac.za (D.G. Kro

1359-4311/$ - see front matter 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S1359-4311(03)00201-1

2202

J.C. Kloppers, D.G. Kr oger / Applied Thermal Engineering 23 (2003) 22012211

Nomenclature
A c g G K L l m p r T V v w x D l q a D fd fr i m o v w wb area, m2 constant, or coecient gravitational acceleration, m/s2 mass velocity, kg/m2 s loss coecient length, m characteristic length, m/s mass ow rate, kg/s pressure, Pa correlation coecient temperature, C or K characteristic velocity, m/s velocity, m/s humidity ratio, kg water vapor/kg dry air coordinate dierential dynamic viscosity, kg/m s density, kg/m3 air drag friction and drag eects ll frontal inlet mean outlet vapor water wetbulb

Greek symbols

Subscripts

 Kfi c1

Gw Ga

 c2 1

where c1 and c2 are empirical constants that depend on the ll design. The empirical relations of Lowe and Christie [1] are widely applied and cited by other researchers [24]. Majumdar et al. [3] correlated the data in Kelly [5] by employing Eq. (1). Johnson [6] gives ll loss coecient test results for counterow cellular type lls with variable heights as
c3 c4 2 Kfi c1 Gc w Ga Lfi

J.C. Kloppers, D.G. Kr oger / Applied Thermal Engineering 23 (2003) 22012211

2203

where Lfi is the height of the ll. If the ll height is constant then Eq. (2) becomes
c3 2 Kfi c1 Gc w Ga

Baard [7] conducted extensive testing on expanded metal type lls in various congurations and employed Eq. (3) to correlate his pressure drop data. The correlation coecients obtained by Baard [7] indicates that Eq. (3) does not necessarily correlate the measured data accurately for some ll congurations. He obtained correlation coecients ranging from 0.61 to 0.98. Milosavljevic and Heikkil a [8] tested seven types of counterow lm type lls and correlated their pressure drop data with
c3 2 Dpfi =Lfi c1 1 Gc w Ga

Goshayshi and Missenden [9] also tested seven types of counterow lm type lls in various arrangements. Their tests were conducted in a 0.15 m 0.15 m counterow test section where Ga was varied between 0.2 and 1.5 kg/m2 s, and Gw was varied between 0.45 and 2.22 kg/m2 s. These mass velocities are very low and are not typical for industrial applications [10]. Their ll test data is correlated by
c3 2 Dpfi c1 Gc w Ga

where c2 and c3 are constant for all the lls tested. Goshayshi and Missenden [9] reported a maximum error of 3% for Eq. (5). It is stressed that the equations above are only valid for the ranges of Gw and Ga from which they were derived. Eqs. (1)(5) are generally not accurate over a wide range of practical cooling tower operational conditions. A new form of empirical equation is proposed in this study that correlates the measured pressure drop data more accurately.

2. Loss coecient The pressure drop through a ll is coupled to the loss coecient by the following relation Dpfi Kfi qv2 =2 6

The measured static pressure drop across the ll (Dpfi ) is due to viscous drag (frictional drag) and form drag in addition to the acceleration of the air due to heating and mass transfer, while the buoyancy due to the dierence in density of the air in the ll and that in the manometer tube external to the test section will tend to counteract these eects in cases of counterow [2], i.e.
2 Dpfi Dpfd qavo v2 avo qavi vavi qava qavm gLfi

where the subscript fd refers to frictional and drag eects and qava is the density of the ambient air which is essentially equal to the density of the air entering the ll, i.e. qavi . The density of the air leaving the ll is qavo and the harmonic mean density qavm 2=1=qavi 1=qavo . The second term on the right-hand side of Eq. (7) represents the momentum change experienced by the air stream while the third term considers buoyancy eects. This equation assumes that the porosity of the particular ll, which is dened as the ratio of the free ow area at a cross-section to the

2204

J.C. Kloppers, D.G. Kr oger / Applied Thermal Engineering 23 (2003) 22012211

corresponding cross-sectional area of the ll, is unity. In the absence of momentum changes a loss coecient which is determined by frictional and drag eect can be dened, i.e.
2 2 Kfd 2 Dpfd =qv2 2Dpfi qavo v2 avo qavi vavi qavi qavm gLfi =qv

In practice the reference conditions chosen for the denominator in Eq. (8) dier. For example, the loss coecient for a particular ll can be dened in terms of the mean airvapor ow rate and its density through the ll, i.e.
2 2 2 Kfdm 2Dpfi qavo v2 avo qavi vavi qavi qavm gLfi qavm Afr =mavm

where mavm qavm vavm Afr . Per unit height of the ll it follows from Eq. (9) that Kfdm1 Kfdm =Lfi . The following measurements are generally made during ll tests where the transfer coecients and loss coecients are determined: the air inlet drybulb temperature (Tai ), and the air wetbulb temperature (Twb ), the water inlet temperature (Twi ), the water outlet temperature (Two ), the water mass ow rate (mw ) and the air mass ow rate (ma ). The atmospheric pressure (pa ) is also measured to determine the humidity ratio of the inlet air (wi ). The air outlet drybulb temperature (Tao ) is generally not measured since it is relatively dicult to measure accurately because of condensation, drift and supersaturation of the outlet air. The outlet air temperature is not employed in the gener [12] theories to determine the transfer coecient. However, the Merkel [11] or Poppe and Ro outlet temperature can be predicted by these theories. Merkel assumed that the outlet air is saturated which enabled him to determine the approximate outlet air temperature from a simple energy balance. In the case of the Poppe theory the outlet air temperature is accurately determined since Poppe did not make the simplifying assumptions of the Merkel approach. The loss coecient as given by Eq. (9) is dependent on the air outlet temperature. Since the Poppe approach generally predicts higher air outlet temperatures than the Merkel method, the loss coecients will dier. This dierence, however, is generally small.

3. Improved empirical equation The loss coecient is essentially a drag coecient. Fig. 1 shows the drag coecients of two simple shapes as a function of the Reynolds number. The total drag on a body placed in a stream of uid consists of friction drag and form drag. The sum of the two is called the total drag [13]. It can be seen in Fig. 1 that the drag coecient at low Reynolds numbers falls for increasing Reynolds numbers. This is due to the fact that friction or viscous eects predominate. The curve attens out and remains essentially constant at high Reynolds numbers. Form drag is predominant in this region. The reason for the existence of form drag lies in the fact that the boundary layer displaces the external, potential ow [13]. The Ergun [15] equation for the pressure drop through packed beds is given by dp 150lV 1:75qV 2 dx l2 l 10

where V and l are the characteristic velocity and characteristic length respectively. The rst term accounts for the viscous drag, and the second term accounts for form drag. The characteristic length is constant for a specic packed bed while the characteristic velocity is a function of the air

J.C. Kloppers, D.G. Kr oger / Applied Thermal Engineering 23 (2003) 22012211

10 2

2205

Drag coefficient, CD

10 1

10 0

10-1

Sphere Infinite Cylinder

10-2 10-1

100

101

102

103

104

105

106

Reynolds number, Re

Fig. 1. Drag coecient for bodies of revolution (adapted from Daugherty et al. [14]).

velocity. If a cooling tower ll is approximated by a packed bed, V and l will also be a function of water mass ow rate. Water droplets may be retained in the ll area or be entrained by the air when the drag force acting on the droplets is greater or equal to the weight of the water droplets. This phenomenon is a function of the air velocity and the water droplet size and ultimately on the type and conguration of the ll. Wet-cooling tower lls dier from packed beds as the pebbles (or water droplets in this case) are not static, and of variable shape, quantity and size. The ll, of course, is static. However, Eq. (10) gives a basis of what form a generalized correlation for pressure drop in lls must take. The pressure drop is a sum of two terms where each term is a function of the air and water mass ow rates. Thus, a new general empirical relation is proposed which accounts for the form drag and viscous drag eects as well as the eects that are dependent on the water mass ow rate and the conguration of the ll, i.e.
c3 c5 c6 2 Kfi c1 Gc w Ga c4 Gw Ga

11

4. Experimental setup Trickle, splash and lm type lls were tested in the counterow ll test facility shown in Fig. 2. The splash ll is shown schematically in Fig. 3(a). It consists of 15 layers, spaced at 200 mm, with a total ll height of 3 m. The trickle ll is shown schematically in Fig. 3(b). It consists of 22 layers 90 mm horizontally stacked cylinders with a total ll height of 1.98 m. Each layer is stacked 90 relative to the layer below. The cross-corrugated lm ll is shown schematically in Fig. 3(c). The lm ll consists of four layers of 0.3 m high parallelepipeds with a total ll height of 1.2 m. Each layer is stacked 90 relative to the layer below. The counterow test section has a cross-sectional area of 1.5 1.5 m. The counterow test section can be extended to any practical height by adding 750 mm modules which are bolted together.

2206

J.C. Kloppers, D.G. Kr oger / Applied Thermal Engineering 23 (2003) 22012211

Fig. 2. Experimental ll test facility at the University of Stellenbosch.

Fig. 3. (a) Splash, (b) trickle and (c) lm type lls.

Hot water is pumped from an underground storage tank to the test section. The storage tank has a capacity of 45 m3 . The water is heated by recycling it through a 100 kW diesel-red boiler. During a test, the heated water is pumped from the top of the storage tank to the test section where it is cooled. The cooled water is then fed back to the bottom of the storage tank. This ensures that stratication occurs in the storage tank and that the subsequent supply temperature will remain almost constant for short test runs. The water ow rate is determined from the pressure drop across an orice plate installed in the supply line according to British Standard 1042 [16]. Air is drawn through the tunnel by a 50 kW centrifugal fan with variable speed control. The mass ow rate of the air is determined by measuring the pressure drop across one or more of the ve ASHRAE 51-75 elliptical nozzles mounted in the horizontal section of the windtunnel as shown in Fig. 2. The pressure drop through the nozzles is determined by a calibrated electronic pressure transducer. The temperatures are measured using calibrated copperconstantan thermocouples. The air temperature is measured before the nozzles to accurately predict the density of the air through the

J.C. Kloppers, D.G. Kr oger / Applied Thermal Engineering 23 (2003) 22012211

2207

nozzles. The drybulb and wetbulb temperatures are the average of four thermocouples each, distributed across a vertical plane. The air drybulb and wetbulb temperatures are again measured below the water extraction troughs. These temperatures may dier from the temperatures measured upstream of the nozzles due to inuence of the fan. The average temperatures of the air below the troughs will be used to evaluate the inlet properties to the test section. The drybulb and wetbulb temperatures are the average of four thermocouples each, distributed across a horizontal plane. The pressure drop across the ll and troughs is measured by four static pressure probes. Two are installed below the troughs and two are installed above the ll. The pressure drop across the troughs is subtracted from the total pressure drop to obtain the pressure drop across the ll. Refer to Oosthuizen [17] and Baard [7] for a detailed description of the pressure probes, the water distribution system, the water extraction troughs and psychrometric probes. The data logging system consists of two Schlumberger Isolated Measurement Pods (IMPs). The IMPs are connected to a Pentium Personal Computer (PC) via an S-Net cable and Schlumberger PC card. The data logger has an internal reference point, which eliminates the use of an ice bath needed for temperature measurement purposes. The data logger converts all temperature readings from millivolts to degree celsius before transferring them to the PC. The pressure transducers adapt pressure readings to voltage signals, which are transferred to the data logger. The method of least squares is employed to obtain empirical correlation through the experimentally determined data. To indicate the reliability of the t, the correlation coecient, r, is dened [18]. A perfect t is obtained when r2 1. The sum of the least squares is the objective function that is minimized by the LFOPC [1921], and ETOPC [22,23] optimization algorithms. LFOPC is a gradient method that generates a dynamic trajectory path, from any given starting point to a local optimum. ETOPC is a conjugate gradient method. The algorithms are relatively insensitive to the chosen initial values for this specic application.

5. Experimental results Splash, trickle and lm type ll tests are presented to show the accuracy and generality of Eq. (11) compared to that of Eq. (3) which is commonly found in the literature. 5.1. Splash ll Table 1 presents the results of ts according to Eqs. (3) and (11) for the 200 mm spacing splash ll shown schematically in Fig. 3(a). The equations presented in Table 1 and the measured loss coecients are plotted in Fig. 4.
Table 1 Empirical relations for the loss coecient of the splash ll Equation (3) (11) Empirical relation Kfdm1 Kfdm1
:921207 0:87201 4:30145G0 Ga w 1:083916 1:965418 3:179688Gw Ga

r2 0.9543
:684936 0:642767 0:639088G0 Ga w

0.9932

2208

J.C. Kloppers, D.G. Kr oger / Applied Thermal Engineering 23 (2003) 22012211

25
Gw = 6.76 kg/ms

20

Gw = 4.71 kg/ms Gw = 2.90 kg/ms Equation (3)

Kfdm1, m-1

15

Equation (11)

10

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

G a, kg/ms

Fig. 4. Comparison of experimental data and empirical equations for the splash ll. Table 2 Empirical relations for the loss coecient of the trickle ll Equation (3) (11) Empirical relation Kfdm1 Kfdm1
0:525842 0:107452 10:539809Gw Ga 0:812454 1:143846 7:047319Gw Ga :294827 1:018498 2:677231G0 Ga w

r2 0.7779 0.9684

Splash lls with ll spacings of 100, 300, and 400 mm were also tested. Similar trends were observed as shown in Fig. 4 for the loss coecient. The same order of ts for the data by Eqs. (3) and (11) were obtained as the correlation coecients were approximately the same as those presented in Table 1 for each equation type. 5.2. Trickle ll Table 2 presents the results of ts according to Eqs. (3) and (11) for the 1.98 m high trickle ll shown schematically in Fig. 3(b). The equations presented in Table 2 and the measured loss coecients are plotted in Fig. 5. Exactly the same trends were observed for 1.08 and 1.53 m high trickle lls. 5.3. Film ll Table 3 presents the results of ts according to Eqs. (3) and (11) for the 1.2 m high lm ll shown schematically in Fig. 3(c). The equations presented in Table 3 and the measured loss coecients are plotted in Fig. 6. Exactly the same trends were observed for 0.6 and 0.9 m high trickle lls. 5.4. Data of Kelly correlated by Majumdar et al. [3] Majumdar et al. [3] correlated the data in Kelly [5] for employment in their VERA2D program for the heat and mass transfer analysis of wet-cooling towers. As already mentioned, they em-

J.C. Kloppers, D.G. Kr oger / Applied Thermal Engineering 23 (2003) 22012211

40
Gw = 6.71 kg/ms

2209

35 30

Gw = 4.50 kg/ms Gw = 2.82 kg/ms Equation (3) Equation (11)

Kfdm1, m-1

25 20 15 10 1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

G a, kg/ms

Fig. 5. Comparison of experimental data and empirical relations for the trickle ll.

Table 3 Empirical relations for the loss coecient of the lm ll Equation (3) (11) Empirical relation
:281255 0:175177 Kfdm1M 19:658921G0 Ga w :777271 2:114727 0:215975 0:079696 Ga 15:327472Gw Ga Kfdm1 3:897830G0 w

r2 0.8561 0.9562

35.0
Gw = 5.96 kg/ms

32.5 30.0 27.5 25.0 22.5 20.0 1.0

Gw = 4.34 kg/ms Gw = 2.80 kg/ms Equation (3) Equation (11)

Kfdm1, m-1

1.5

2.0

2.5

3.0

3.5

4.0

4.5

G a, kg/ms

Fig. 6. Comparison of experimental data and empirical relations for the lm ll.

ployed Eq. (1). Fig. 7 shows Kellys [5] data for a type F ll correlated by Majumdar et al. [3] by employing Eq. (1). The air ow range employed in the experiments of Kelly is relatively narrow compared to the experiments conducted in this investigation. Correlations of Kellys data by

2210

J.C. Kloppers, D.G. Kr oger / Applied Thermal Engineering 23 (2003) 22012211

7 6 5
Gw = 1.36 Gw = 6.78 Gw = 12.03 Kelly Gw = 4.07 Gw = 9.49 Gw = 16.27 Equations (7) and (10)

Kfi1, m-1

4 3 2 1 0 2.4 2.5 2.6 2.7 2.8 2.9 3 3.1 3.2 3.3

G a. kg/ms

Fig. 7. Data by Kelly [5] correlated by Majumdar et al. [3] (Eq. (1)) and Eqs. (3) and (11).

employing Eqs. (3) and (11) are also shown in Fig. 7. In this instance, Eqs. (3) and (11) give virtually identical results with correlation coecients for both equal to 0.9991.

6. Conclusion Eq. (11) will generally correlate measured pressure loss coecients accurately for all types of lls under all types of practical operational conditions as it make provision for a spectrum of forces due to shear and drag. Other types of equations may correlate observed trends accurately, but they generally lack generality and are only applicable for limited ranges of water and air ow rates. Eq. (11) is also more accurate than Eq. (3) if it is extrapolated. It is recommended that as much information as possible be supplied with the empirical relations of the loss coecients, such as the ranges of applicability of Ga and Gw where Ga and Gw are greater than zero. The goodness of t must also be supplied in the form of a correlation coecient. This will enable the designer of wet-cooling systems to take the necessary precautions to compensate for any uncertainties. If possible, the same water spray system must be employed in the ll test and the subsequent cooling tower application of the ll. This will eliminate the eects of droplet size and distribution on the loss coecient.

Acknowledgements The authors gratefully acknowledge Sasol Ltd. for their nancial support, Prof. J.A. Snyman from the University of Pretoria for the optimization algorithms and Industrial Water Cooling Pty. Ltd. for supplying the ll materials and the nancial support for the tests.

J.C. Kloppers, D.G. Kr oger / Applied Thermal Engineering 23 (2003) 22012211

2211

References
[1] H.J. Lowe, D.G. Christie, Heat transfer and pressure drop data on cooling tower packings and model studies of the resistance of natural draft towers to airow, in: Proceedings of the International Heat Transfer Conference, Colorado, Part V, 1961, pp. 933950. ger, Cooling Systems for Power, Petrochemical and Process Plants, Pennwell, Tulsa, 2003. [2] D.G. Kro [3] A.K. Majumdar, A.K. Singhal, D.B. Spalding, VERA2D: program for 2-D analysis of ow, heat, and mass transfer in evaporative cooling towers, EPRI Report CS 2923, vols. 1 and 2, March 1983. [4] A.K.M. Mohiuddin, K. Kant, Knowledge base for the systematic design of wet cooling towers. Part II: Fill and other design parameters, International Journal of Refrigeration 19 (1) (1996) 5260. [5] N.W. Kelly, Kellys Handbook of Crossow Cooling Tower Performance, Neil W. Kelly and Associates, Kansas City, MO, 1976. [6] B.M. Johnson (Ed.), Cooling tower performance prediction and improvement, vol. 1: Applications Guide, EPRI Report GS-6370, vol. 2: Knowledge Base, EPRI Report GS-6370, EPRI, Palo Alto, 1989. [7] T.W. Baard, Performance characteristics of expanded metal cooling tower ll, M.Eng. Thesis, University of Stellenbosch, Stellenbosch, South Africa, 1998. [8] N. Milosavljevic, P. Heikkil a, A comprehensive approach to cooling tower design, Applied Thermal Engineering 21 (2001) 899915. [9] H.R. Goshayshi, J.F. Missenden, The investigation of cooling tower packing in various arrangements, Applied Thermal Engineering 20 (2000) 6980. [10] K.W. Li, A.P. Priddy, Power Plant System Design, John Wiley & Sons, 1985. hlung, VDI-Zeitchrift 70 (70) (1925) 123128. [11] F. Merkel, Verdunstungsku gener, Berechnung von Ru ckku hlwerken, VDI-W [12] M. Poppe, H. Ro armeatlas (1991) Mi1Mi15. [13] H. Schlichting, Boundary Layer Theory, fourth ed., McGraw-Hill, New York, 1960. [14] R.L. Daugherty, J.B. Franzini, E.J. Finnemore, Fluid Mechanics with Engineering Applications, SI Metric ed., McGraw-Hill, Singapore, 1989. [15] S. Ergun, Fluid ow through packed columns, Chemical Engineering Progress 48 (1952) 8994. [16] British Standard 1042, Measurement of Fluid Flow in Closed Conduits, Part 1, Section 1.1, 1981. [17] P.C. Oosthuizen, Performance characteristics of hybrid cooling towers, M.Eng. Thesis, University of Stellenbosch, Stellenbosch, South Africa, 1995. [18] T.G. Beckwith, R.D. Marangoni, J.H. Lienhard, Mechanical Measurements, fth ed., Addison-Wesley Publishing Company, 1993. [19] J.A. Snyman, A new and dynamic method for unconstrained minimization, Applied Mathematical Modelling 6 (1982) 449462. [20] J.A. Snyman, An improved version of the original leap-frog dynamic method for unconstrained minimization LFOP1 (b), Applied Mathematical Modelling 7 (1983) 216218. [21] J.A. Snyman, Unconstrained minimization by combining the dynamic and conjugate gradient methods, Quaestiones Mathematicae 8 (1985) 3342. [22] J.A. Snyma, ETOPC: a fortran program for solving general constrained minimization problems by the conjugate gradient method without explicit line searches, Research Report, Department of Mechanical Engineering, University of Pretoria, 1998. [23] J.A. Snyman, The LFOPC leap-frog method for constrained optimization, Computers and Mathematics with Applications 40 (8/9) (2000) 10851096.