Cooling Tower Fill Efficiency

Applied Thermal Engineering 23 (2003) 2201–2211 www.elsevier.com/locate/apthermeng

Loss coefficient correlation for wet-cooling tower fills Johannes C. Kloppers, Detlev G. Kr€ oger

*

Department of Mechanical Engineering, University of Stellenbosch, Stellenbosch 7600, South Africa Received 20 February 2003; accepted 10 June 2003

Abstract Loss coefficient correlations given in the literature for wet-cooling tower fills 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 fill loss coefficient data more effectively when compared to other forms of empirical equations commonly found in the literature. 2003 Elsevier Ltd. All rights reserved. Keywords: Wet-cooling towers; Loss coefficient; Splash fill; Trickle fill; Film fill

1. Introduction The loss coefficient of a cooling tower fill is determined by measuring the pressure drop across the fill. The results of these tests are correlated by empirical relations which are functions of the air and water mass flow 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 coefficient of the fill. The draft in a natural draft cooling tower is also a strong function of the fill loss coefficient. It is thus important to represent the fill loss coefficient accurately, as inaccurate representation of the loss coefficients in the form of empirical relations can have financial implications if the cooling tower does not meet design specifications. Here follows some forms of empirical equations of the fill loss coefficient found in the literature. Lowe and Christie [1] used the following form of equation to represent the loss coefficients of counterflow splash and film type fills.

*

Corresponding author: Tel.: +27-21-8084259; fax: +27-21-8084958. E-mail address: [email protected] (D.G. Kr€ oger).

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) 2201–2211

Nomenclature A c g G K L l m p r T V v w x

area, m2 constant, or coefficient gravitational acceleration, m/s2 mass velocity, kg/m2 s loss coefficient length, m characteristic length, m/s mass flow rate, kg/s pressure, Pa correlation coefficient temperature, C or K characteristic velocity, m/s velocity, m/s humidity ratio, kg water vapor/kg dry air coordinate

Greek symbols D l q

differential dynamic viscosity, kg/m s density, kg/m3

Subscripts a D fd fi fr i m o v w wb

air drag friction and drag effects fill frontal inlet mean outlet vapor water wetbulb

Kfi ¼ c1

Gw Ga

þ c2

ð1Þ

where c1 and c2 are empirical constants that depend on the fill design. The empirical relations of Lowe and Christie [1] are widely applied and cited by other researchers [2–4]. Majumdar et al. [3] correlated the data in Kelly [5] by employing Eq. (1). Johnson [6] gives fill loss coefficient test results for counterflow cellular type fills with variable heights as Kfi ¼ c1 Gcw2 Gca3 Lcfi4

ð2Þ


2203

where Lfi is the height of the fill. If the fill height is constant then Eq. (2) becomes Kfi ¼ c1 Gcw2 Gca3

ð3Þ

Baard [7] conducted extensive testing on expanded metal type fills in various configurations and employed Eq. (3) to correlate his pressure drop data. The correlation coefficients obtained by Baard [7] indicates that Eq. (3) does not necessarily correlate the measured data accurately for some fill configurations. He obtained correlation coefficients ranging from 0.61 to 0.98. Milosavljevic and Heikkil€ a [8] tested seven types of counterflow film type fills and correlated their pressure drop data with Dpfi =Lfi ¼ c1 ð1 þ Gcw2 ÞGca3

ð4Þ

Goshayshi and Missenden [9] also tested seven types of counterflow film type fills in various arrangements. Their tests were conducted in a 0.15 m · 0.15 m counterflow 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 fill test data is correlated by Dpfi ¼ c1 Gcw2 Gca3

ð5Þ

where c2 and c3 are constant for all the fills 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 coefficient The pressure drop through a fill is coupled to the loss coefficient by the following relation Dpfi ¼ Kfi qv2 =2

ð6Þ

The measured static pressure drop across the fill (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 difference in density of the air in the fill and that in the manometer tube external to the test section will tend to counteract these effects in cases of counterflow [2], i.e. Dpfi ¼ Dpfd þ ðqavo v2avo qavi v2avi Þ ðqava qavm ÞgLfi

ð7Þ

where the subscript fd refers to frictional and drag effects and qava is the density of the ambient air which is essentially equal to the density of the air entering the fill, i.e. qavi . The density of the air leaving the fill is qavo and the harmonic mean density qavm ¼ 2=ð1=qavi þ 1=qavo Þ. 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 effects. This equation assumes that the porosity of the particular fill, which is defined as the ratio of the free flow area at a cross-section to the

2204


corresponding cross-sectional area of the fill, is unity. In the absence of momentum changes a loss coefficient which is determined by frictional and drag effect can be defined, i.e. Kfd ¼ 2 Dpfd =ðqv2 Þ ¼ 2½Dpfi ðqavo v2avo qavi v2avi Þ þ ðqavi qavm ÞgLfi =ðqv2 Þ

ð8Þ

In practice the reference conditions chosen for the denominator in Eq. (8) differ. For example, the loss coefficient for a particular fill can be defined in terms of the mean air–vapor flow rate and its density through the fill, i.e. Kfdm ¼ 2½Dpfi ðqavo v2avo qavi v2avi Þ þ ðqavi qavm ÞgLfi qavm A2fr =m2avm

ð9Þ

where mavm ¼ qavm vavm Afr . Per unit height of the fill it follows from Eq. (9) that Kfdm1 ¼ Kfdm =Lfi . The following measurements are generally made during fill tests where the transfer coeffcients and loss coefficients 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 flow rate (mw ) and the air mass flow 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 difficult to measure accurately because of condensation, drift and supersaturation of the outlet air. The outlet air temperature is not employed in the Merkel [11] or Poppe and R€ ogener [12] theories to determine the transfer coefficient. However, the 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 coefficient 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 coefficients will differ. This difference, however, is generally small.

3. Improved empirical equation The loss coefficient is essentially a drag coefficient. Fig. 1 shows the drag coefficients of two simple shapes as a function of the Reynolds number. The total drag on a body placed in a stream of fluid 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 coefficient at low Reynolds numbers falls for increasing Reynolds numbers. This is due to the fact that friction or viscous effects predominate. The curve flattens 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 flow [13]. The Ergun [15] equation for the pressure drop through packed beds is given by dp 150lV 1:75qV 2 þ ¼ dx l2 l

ð10Þ

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


2205

Drag coefficient, CD

10 2

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 coefficient for bodies of revolution (adapted from Daugherty et al. [14]).

velocity. If a cooling tower fill is approximated by a packed bed, V and l will also be a function of water mass flow rate. Water droplets may be retained in the fill 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 configuration of the fill. Wet-cooling tower fills differ from packed beds as the pebbles (or water droplets in this case) are not static, and of variable shape, quantity and size. The fill, of course, is static. However, Eq. (10) gives a basis of what form a generalized correlation for pressure drop in fills must take. The pressure drop is a sum of two terms where each term is a function of the air and water mass flow rates. Thus, a new general empirical relation is proposed which accounts for the form drag and viscous drag effects as well as the effects that are dependent on the water mass flow rate and the configuration of the fill, i.e. Kfi ¼ c1 Gcw2 Gca3 þ c4 Gcw5 Gca6

ð11Þ

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

2206


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

Fig. 3. (a) Splash, (b) trickle and (c) film type fills.

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-fired 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 stratification occurs in the storage tank and that the subsequent supply temperature will remain almost constant for short test runs. The water flow rate is determined from the pressure drop across an orifice 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 flow rate of the air is determined by measuring the pressure drop across one or more of the five 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 copper–constantan thermocouples. The air temperature is measured before the nozzles to accurately predict the density of the air through the


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 differ from the temperatures measured upstream of the nozzles due to influence 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 fill and troughs is measured by four static pressure probes. Two are installed below the troughs and two are installed above the fill. The pressure drop across the troughs is subtracted from the total pressure drop to obtain the pressure drop across the fill. 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 fit, the correlation coefficient, r, is defined [18]. A perfect fit is obtained when r2 ¼ 1. The sum of the least squares is the objective function that is minimized by the LFOPC [19–21], 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 specific application.

5. Experimental results Splash, trickle and film type fill 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 fill Table 1 presents the results of fits according to Eqs. (3) and (11) for the 200 mm spacing splash fill shown schematically in Fig. 3(a). The equations presented in Table 1 and the measured loss coefficients are plotted in Fig. 4. Table 1 Empirical relations for the loss coefficient of the splash fill Equation (3) (11)

r2

Empirical relation Kfdm1 ¼ Kfdm1 ¼

4:30145G0:921207 G0:87201 w a 3:179688G1:083916 G1:965418 w a

0.9543 þ

0:639088G0:684936 G0:642767 w a

0.9932

2208

J.C. Kloppers, D.G. Kr€oger / Applied Thermal Engineering 23 (2003) 2201–2211 25 Gw = 6.76 kg/m²s

20

Gw = 4.71 kg/m²s

Kfdm1, m-1

Gw = 2.90 kg/m²s Equation (3)

15

Equation (11)

10

5

0

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

G a, kg/m²s

Fig. 4. Comparison of experimental data and empirical equations for the splash fill. Table 2 Empirical relations for the loss coefficient of the trickle fill Equation (3) (11)

r2

Empirical relation Kfdm1 ¼ Kfdm1 ¼

10:539809Gw0:525842 G0:107452 a 7:047319Gw0:812454 G1:143846 þ a

0.7779 2:677231G0:294827 G1:018498 w a

0.9684

Splash fills with fill spacings of 100, 300, and 400 mm were also tested. Similar trends were observed as shown in Fig. 4 for the loss coefficient. The same order of fits for the data by Eqs. (3) and (11) were obtained as the correlation coefficients were approximately the same as those presented in Table 1 for each equation type. 5.2. Trickle fill Table 2 presents the results of fits according to Eqs. (3) and (11) for the 1.98 m high trickle fill shown schematically in Fig. 3(b). The equations presented in Table 2 and the measured loss coefficients are plotted in Fig. 5. Exactly the same trends were observed for 1.08 and 1.53 m high trickle fills. 5.3. Film fill Table 3 presents the results of fits according to Eqs. (3) and (11) for the 1.2 m high film fill shown schematically in Fig. 3(c). The equations presented in Table 3 and the measured loss coefficients are plotted in Fig. 6. Exactly the same trends were observed for 0.6 and 0.9 m high trickle fills. 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-


2209

40 Gw = 6.71 kg/m²s Gw = 4.50 kg/m²s

35


Kfdm1, m-1

30

Equation (11)

25 20 15 10 1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

G a, kg/m²s

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

Table 3 Empirical relations for the loss coefficient of the film fill Equation

Empirical relation

r2

(3)

Kfdm1M ¼ 19:658921G0:281255 G0:175177 w a

0.8561

(11)

G2:114727 þ 15:327472Gw0:215975 Ga0:079696 Kfdm1 ¼ 3:897830G0:777271 w a

0.9562

35.0 Gw = 5.96 kg/m²s

32.5

Gw = 4.34 kg/m²s

Kfdm1, m-1


30.0

Equation (11)

27.5 25.0 22.5 20.0 1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

G a, kg/m²s

Fig. 6. Comparison of experimental data and empirical relations for the film fill.

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

2210

J.C. Kloppers, D.G. Kr€oger / Applied Thermal Engineering 23 (2003) 2201–2211 7 Gw = 1.36 Gw = 6.78 Gw = 12.03 Kelly

6

Gw = 4.07 Gw = 9.49 Gw = 16.27 Equations (7) and (10)

Kfi1, m-1

5 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/m²s

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 coefficients for both equal to 0.9991.

6. Conclusion Eq. (11) will generally correlate measured pressure loss coefficients accurately for all types of fills 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 flow 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 coefficients, such as the ranges of applicability of Ga and Gw where Ga and Gw are greater than zero. The goodness of fit must also be supplied in the form of a correlation coefficient. 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 fill test and the subsequent cooling tower application of the fill. This will eliminate the effects of droplet size and distribution on the loss coefficient.

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


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 airflow, in: Proceedings of the International Heat Transfer Conference, Colorado, Part V, 1961, pp. 933–950. [2] D.G. Kr€ oger, Cooling Systems for Power, Petrochemical and Process Plants, Pennwell, Tulsa, 2003. [3] A.K. Majumdar, A.K. Singhal, D.B. Spalding, VERA2D: program for 2-D analysis of flow, 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) 52–60. [5] N.W. Kelly, KellyÕs Handbook of Crossflow 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 fill, 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) 899–915. [9] H.R. Goshayshi, J.F. Missenden, The investigation of cooling tower packing in various arrangements, Applied Thermal Engineering 20 (2000) 69–80. [10] K.W. Li, A.P. Priddy, Power Plant System Design, John Wiley & Sons, 1985. [11] F. Merkel, Verdunstungsk€ uhlung, VDI-Zeitchrift 70 (70) (1925) 123–128. [12] M. Poppe, H. R€ ogener, Berechnung von R€ uckk€ uhlwerken, VDI-W€ armeatlas (1991) Mi1–Mi15. [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 flow through packed columns, Chemical Engineering Progress 48 (1952) 89–94. [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, fifth ed., Addison-Wesley Publishing Company, 1993. [19] J.A. Snyman, A new and dynamic method for unconstrained minimization, Applied Mathematical Modelling 6 (1982) 449–462. [20] J.A. Snyman, An improved version of the original leap-frog dynamic method for unconstrained minimization LFOP1 (b), Applied Mathematical Modelling 7 (1983) 216–218. [21] J.A. Snyman, Unconstrained minimization by combining the dynamic and conjugate gradient methods, Quaestiones Mathematicae 8 (1985) 33–42. [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) 1085–1096.

Cooling Tower Fill Efficiency

Recommend Documents