Applied Thermal Engineering 20 (2000) 12251236

www.elsevier.com/locate/apthermeng

Thermal behaviour of closed wet cooling towers for use with chilled ceilings
Jorge Faca o, Armando C. Oliveira*
Department of Mechanical Engineering and Industrial Management, Faculty of Engineering, University of Porto, Rua dos Bragas, 4050-123 Porto, Portugal Received 22 September 1999; accepted 30 November 1999

Abstract A new closed wet cooling tower, adapted for use with chilled ceilings in buildings, was tested. Experimental correlations were obtained for mass and heat transfer coecients. Existing thermal models for this type of cooling tower were found to predict well thermal performance, if the above correlations are used. 7 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Cooling tower; Chilled ceiling; Experimental testing; Thermal models

1. Introduction There is an increasing demand for cooling in buildings. Better insulated buildings and the increased use of working equipment (computers, etc.) led to higher cooling requirements, particularly in oce and commercial buildings. The need for reducing energy consumption and CO2 emissions, together with the need for using environmental-friendly refrigerants, justify a strong demand for CFC free, ecient and cheap cooling systems. Chilled ceilings are a relatively new approach to cooling, with about 100,000 m2 installed in central European countries in the last ve years. Chilled ceilings oer several advantages over conventional alternatives. The use of water instead of air reduces energy requirements for energy transportation. It also allows a reduction of ventilation rate to a minimum level, for hygiene purposes. Heat transfer from indoor space to chilled ceilings is made by combined
* Corresponding author. Tel.: +351-22-205-2476; fax: +351-22-205-2476. 1359-4311/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved. PII: S 1 3 5 9 - 4 3 1 1 ( 9 9 ) 0 0 0 9 6 - 4

1226

J. Faca o, A.C. Oliveira / Applied Thermal Engineering 20 (2000) 12251236

convection and radiation. Radiative heat transfer allows chilled ceilings to remove considerable heat loads at a relatively small temperature dierence between room air and ceiling. This makes it possible to run the system with ceiling supply temperatures of about 18208C. Heat transfer rates between 25 and 75 W/m2 are possible, with lower air velocity in the rooms and a resulting comfortable indoor environment. Due to the above mentioned moderately high water temperatures used in chilled ceilings, it is possible to deliver cold water with a closed wet cooling tower during most of the cooling period. The cooling tower can be combined with a refrigeration machine or it can be used alone, if some overheating is allowed during short periods, or if energy storage or nightcooling techniques are used. Closed wet cooling towers were conventionally used to remove excess heat from various industrial processes, with a usual range of 32468C hot water temperatures and typical cooling capacities above 40 kW. A recent research work performed in Switzerland, [1], showed that these towers are greatly overpowered in airow and spray water rate, when applied to the range of 22258C hot water temperatures, as needed for chilled ceilings. Tower design for lower cooling loads 10 kW leads to smaller tower dimensions. Existing models to predict thermal performance of closed wet cooling towers were developed for conventional units and operating conditions: high water temperatures and cooling capacities. It is the objective of this work to verify the applicability of existing models to smaller towers, adapted for use with chilled ceilings. For that purpose, a new cooling tower built specically for use with chilled ceilings was tested. Experimental results were used to introduce corrections to existing model correlations.

2. Thermal models for a closed wet cooling tower Thermal models to predict cooling tower performance can be classied as detailed models or correlation based models. Detailed models are based on a CFD-type approach, involving the numerical solution of dierential equations for air/spray water ow, energy and water vapour concentration. After velocity, temperature and humidity elds are calculated, transfer coecients can be calculated as a result. Examples of such models, which need numerical codes and high performance computers, can be found in Ref. [2]. Although a detailed analysis of air and spray water distribution in the tower is possible, they have some limitations, namely regarding boundary conditions in the tubes. Water ow inside the tubes has to be treated separately, which poses some practical problems: for instance, heat ux at tube surface is not uniform. These models and codes also consume considerable computing time and require a certain degree of specialization. Correlation based models may also need the solution of dierential equations, although not necessarily. However these equations result from local energy and mass balances, after some simplications are introduced and assuming mass and heat transfer coecients can be calculated rst. These coecients are calculated through experimental correlations. This approach will be used in this work, since it is more practical and may also lead to improved accuracy. Fig. 1 represents schematically a closed wet cooling tower, with the main variables involved.

J. Faca o, A.C. Oliveira / Applied Thermal Engineering 20 (2000) 12251236

1227

All correlation based models assume that the tube surface is completely wet, through a uniform distribution of spray water over the tube bundle. Therefore, mass and heat transfer occurs between an water lm and air ow. For simultaneous heat and mass transfer, heat ux can be calculated through enthalpy potential [3]: q am hi hair 1

where am is the mass transfer coecient for water vapour and hi is the enthalpy at water lm/ air interface. An enthalpy balance, for an elementary transfer surface dA, can be expressed as mair dhair am hi hair dA 2

which is known as the Merkel equation [4] and which integration for the whole tower gives out am A dhair 3 mair h i hair in The local energy balance, for the elementary surface dA in Fig. 2, can be written as kTw Ti dA am hi hair dA mspray cp, spray dTspray 4

neglecting spray rate variation and with k being the heat transfer coecient between water inside tube and lm/air interface. It can be calculated by adding all thermal resistances: 1 1 dext dext dext 1 ln k aw dint aspray 2ktube dint 5

using external tube surface as reference dext ). Several models can be found in the literature, diering in simplications assumed. Four dierent models will be considered. When calculating tower performance, inlet conditions

Fig. 1. Closed wet cooling tower and main variables.

1228

J. Faca o, A.C. Oliveira / Applied Thermal Engineering 20 (2000) 12251236

(water and air) are known and outlet conditions are a model output. An iterative procedure is usually applied: water outlet temperature is guessed, which allows to calculate outlet air properties and the enthalpy integral in Eq. (3); since the integral value will not be equal to the rst member of Eq. (3) available number of transfer units, NTU a new outlet temperature is used until equality is obtained. Models dier in the method used to calculate the integral in Eq. (3). The rst model to be considered, developed by Mizushina, [5], neglects spray temperature variation in Eq. (4) and considers interface conditions hi , Ti constant throughout the cooling tower. Integration of Eq. (3) then gives am A hi hair, in ln mair hi hair, out The integration of Eq. (4) and substitution in Eq. (6) leads to   Tw, in Ti hi hair, in k mair exp am cpw mw Tw, out Ti hi hair, out 6

which together with the function hi hi Ti forms a non-linear set of equations for calculating Ti , hi ). This is an overall model, since temperature/enthalpy variation inside the tower is not calculated. The second model, from Kals [6] uses a discretisation of the tube transfer surface. The domain is divided in several nodes to allow the calculation of water temperature and air enthalpy. Spray water temperature variation is also neglected. To evaluate local air enthalpy the following equation is used   mw cpw dhair Dhair j 8 exp dTw mair DTw k where j and k are subscripts representing dierent nodes. Interface properties Ti , hi are

Fig. 2. Elementary transfer surface.

J. Faca o, A.C. Oliveira / Applied Thermal Engineering 20 (2000) 12251236

1229

calculated locally in each air node. After calculation of Tw and hair in each node, the enthalpy integral in Eq. (3) is calculated by numerical integration. The third model to be considered was also formulated by Mizushina [5]. Spray water temperature variation is not neglected. It is assumed that spray water lm thermal resistance is negligible, which implies Ti Tspray X Three dierential equations for water temperature, air enthalpy and spray temperature are solved by numerical integration nite dierences [5]. The fourth model was presented by Peterson [7] and uses the concept of cooling eciency. It also uses a global approach. Spray temperature variation is neglected and a linearization is adopted for saturation enthalpy as a function of wet bulb temperature. All these models need experimental correlations for mass and heat transfer coecients. Among others, Mizushina [8] found a relationship between mass transfer coecient, air and spray ow rates:
X9 0X15 2X6 am a 5X028 10 8 Re0 air Respray d ext

where a is the ratio between heat transfer surface and heat exchanger volume. It is valid for Respray between 50 and 240, Reair between 1200 and 14,000 and dext between 12 and 40 mm. In this correlation, mass transfer coecient depends on spray ow rate. Other authors have obtained correlations that just use air ow rate as variable [9,10]. In fact, Niitsu [10] concluded that mass transfer coecient is independent of spray ow rate for Gadext b 0X7, where G is the spray water load, equal to spray mass ow rate divided by tower section. Heat transfer coecient between tube surface and water lm, aspray , can be correlated with spray load and tube diameter. Mizushina [8] presented the following relationship: aspray 2100 Gadext 1a3 valid for Gadext between 0.2 and 5.5 kg/m2/s. 10

3. New closed wet cooling tower and experimental results A new closed wet cooling tower was designed in order to be used with chilled ceilings. Design conditions were a cooling capacity of 10 kW, for an inlet water temperature of 218C, a water ow rate of 0.8 kg/s and an air wet bulb temperature of 168C. The tower has a section of 0.6 1.2 m and a height of 1.55 m. The tube bundle has 228 staggered tubes of 10 mm outside diameter, with a total transfer area of 8.6 m2. This corresponds to a much smaller size than usual towers. The load/volume ratio of this tower is equal to 29 kW/m3 (per heat exchanger volume), with a ratio between heat transfer surface and heat exchanger volume of 25 m2/m3. The tower was manufactured by Sulzer Escher Wyss (Lindau, Germany). It is much smaller than cooling tower models usually built by this manufacturer. A forced draft conguration was chosen with a crossow fan located at air entrance. This arrangement has a lower noise level, and also leads to a lower pressure drop. It was also chosen to facilitate air ow measurements. A test facility was assembled at Porto to test this cooling tower. It is shown schematically in Fig. 3, including main instrumentation used. The thermal load was modeled with an electrical

1230

J. Faca o, A.C. Oliveira / Applied Thermal Engineering 20 (2000) 12251236

heater located in a water tank. Tower inlet water temperature was controlled by varying heating power PID controller. Fan speed was also controlled by means of a frequency controller, which allowed changing air ow rate. Spray water ow rate could be changed manually (valve with seven xed positions). Cooling water ow rate could also be changed, by changing pump speed and by using regulating valves. Air ow rate was measured with one vane anemometer, located in dierent points of tower outlet section. In order to measure cooling water temperature evolution, several thermocouples

Fig. 3. Experimental facility for cooling tower testing.

J. Faca o, A.C. Oliveira / Applied Thermal Engineering 20 (2000) 12251236

1231

were connected to the tubes. The data acquisition system used a data logger HP 34970A and HP VEE as software. Tower thermal performance was expressed by means of tower thermal eciency. This is dened as e Tw, in Tw, out Tw, in Twb 11

where Tw, in is inlet water temperature, Tw, out outlet water temperature and Twb the inlet wet bulb air temperature. Experiments were carried out to analyse the inuence of inlet water temperature. It was found that this parameter has a very little inuence in tower eciency. The eect of spray water ow rate can be seen in Fig. 4. An increase in spray rate increases eciency up to a certain level. Above a rate of about 1 kg/s, an increase in spray rate does not improve signicantly tower performance, because tube surface is almost completely wet. This means that an optimum spray rate can be found for this tower, regarding cooling capacity and water/ energy (pumping) consumption. This analysis can only be done through experimental measures, since all thermal models assume tube surface is completely wet. Then, if spray rate is kept close to the maximum or optimum value, tower eciency is a function of three parameters: air ow rate, cooling water ow rate and wet bulb temperature. Fig. 5 shows experimental eciency values for dierent air and water rates, for the same wet bulb temperature. As expected, eciency increases with air ow rate and decreases with an increase in water ow rate (due to a decrease in water temperature dierence). Air wet bulb temperature has an inuence in tower eciency which is shown in Fig. 6.

Fig. 4. Tower eciency as a function of spray water ow rate, for dierent air and water rates: mmax spray 1X39 kg/s, wet bulb temperature = 12.68C.

1232

J. Faca o, A.C. Oliveira / Applied Thermal Engineering 20 (2000) 12251236

Fig. 5. Tower eciency as a function of air and water mass ow ratios, for a wet bulb temperature of 15.88C; mmax water 0X8 kg/s, mmax air 1X7 kg/s.

Eciency increases slightly with wet bulb temperature: about 8% (absolute value) for temperatures between 10 and 208C. The increase is linear and the inuence is similar for dierent air and water ow rates. A global coecient of performance (COP) can be dened for the cooling tower. It is equal to cooling capacity divided by energy input (pumping and fan consumption). By measuring electricity input in the fan motor and spray ow rates, with Tw, in 218C and Twb 168C. Mass and heat transfer coecients am and aspray were calculated for dierent operating conditions. These coecients were correlated to, respectively, air ow rate and spray ow rate, following the same approach from other authors [8,9,10]. The following correlations were obtained:

Fig. 6. Inuence of air wet bulb temperature in tower eciency for dierent air and water mass ow rates.

J. Faca o, A.C. Oliveira / Applied Thermal Engineering 20 (2000) 12251236

1233

12 13

valid for 0X4 ` mw ` 0X8 kg/s, 10 ` Twb ` 208C, 15 ` Tw, in ` 288C, with mmax air 1X7 kg/s and mmax spray 1X39 kg/s. 4. Model results The dierent thermal models described previously were applied to the new cooling tower geometry. Model 2 was implemented with a xed number of nodes, equal to 13 number of tubes in one column + 1. Model 3, a nite dierences model, can be implemented with any number of nodes. However, above 25 nodes no signicant dierence was noted in the calculated tower eciency. Fig. 7 shows experimental and calculated eciencies using Mizushina's correlations Eqs. (9) and (10). Results for models 1, 3 and 4 are quite close, but all calculated eciencies are above experimental values. The lower experimental eciencies must be due to the particular geometrical conditions of the cooling tower: smaller tower dimensions and non-uniformity in air velocity in the tower section, which is caused by fan position. This non-uniformity of the air ow over the tube bundle was noted during air velocity measurements. Fig. 8 shows experimental and calculated eciencies when using experimental correlations for mass and heat transfer coecients Eqs. (12) and (13). Calculated results are closer to experimental ones, and the dierence between the 4 models is smaller. Model 1 and 2 seem to give the best results in this case.

Fig. 7. Comparison of experiment and models using Mizushina's correlations for mass transfer and heat transfer between tube surface and water lm. Water rate = 0.8 kg/s, spray rate = 1.37 kg/s, wet bulb temperature = 168C.

1234

J. Faca o, A.C. Oliveira / Applied Thermal Engineering 20 (2000) 12251236

Fig. 8. Comparison of experiment and models using experimental correlations for mass and heat transfer. Water rate = 0.8 kg/s, spray rate = 1.37 kg/s, wet bulb temperature = 168C.

Fig. 9 shows experimental and calculated eciencies while varying air wet bulb temperature, using experimental correlations for mass and heat transfer coecients. A good accuracy is also obtained, particularly with models 1 and 3. Water temperature prole along the tubes (water ow direction) was calculated with model 3. As Fig. 10 shows, the calculated temperature prole is quite close to experimental one.

Fig. 9. Inuence of air wet bulb temperature in tower eciency; air rate = 0.6 kg/s, water rate = 0.8 kg/s, spray rate 1.37 kg/s. Models with experimental correlations.

J. Faca o, A.C. Oliveira / Applied Thermal Engineering 20 (2000) 12251236

1235

Fig. 10. Variation of water temperature along tube banks. Air rate = 0.7 kg/s, water rate = 0.8 kg/s, spray rate = 1.39 kg/s, wet bulb temperature = 15.68C. Model with experimental correlations.

The use of the thermal models also conrmed one experimental conclusion: the inuence of water inlet temperature in tower eciency is negligible. 5. Conclusions Thermal models are very useful both for designing and predicting cooling tower performance. This is not always possible with experiment, since it is dicult to reproduce all possible operating conditions. Dierent thermal models to predict the performance of closed wet cooling towers were presented. A new closed wet cooling tower, adapted for use with chilled ceilings, was tested and experimental correlations for mass and heat transfer coecients were obtained. Dierent models were used and their results compared to experimental ones. It was found that, for such a small tower, the models give better results when using the new correlations. It was also found that simpler models, with a global approach, can give as good, or even better, results as models based on nite dierence techniques. Acknowledgements The authors wish to thank the Commission of the European Union (DGXII) for partially funding the work done, under Joule IV research programme. All partners of the ECOCOOL research project are greatly acknowledged: Sulzer Infra Lab (CH), Sulzer Escher Wyss Lindau (DE), University of Nottingham (UK) Institute of Building Technology, Helsinki University of Technology (FIN) HVAC Laboratory.

1236

J. Faca o, A.C. Oliveira / Applied Thermal Engineering 20 (2000) 12251236

References
[1] P. Sprecher, B. Gasser, M. Koschenz, P. Kofoed, NENOP Energieoptimaler Betrieb Von Kuehldecken. Technical Report 575. Swiss National Energy Research Fund, Switzerland, 1996. [2] G. Gan, S.B. Riat, Numerical simulation of closed wet cooling towers for chilled ceiling systems, Applied Thermal Engineering 19 (12) (1999). [3] ASHRAE, Handbook of Fundamentals. American Society Heating Refriger. Air Condit. Eng., USA, 1998. [4] F. Merkel, Verdunstungskuehlung, VDI Forschungsarbeiten No. 275, Berlin, 1925. [5] R.I.T Mizushina, H. Miyashita, Characteristics and methods of thermal design of evaporative coolers, Int. Chemical Engineering 8 (3) (1968). [6] W.A. Kals, Wet-surfaces air coolers, Chemical Engineering 78 (1971) 9094. [7] J.L. Peterson, An efectiveness model for indirect evaporative coolers, ASHRAE Transactions 99 (part II), 1993. [8] R.I.T Mizushina, H. Miyashita, Experimental study of an evaporative cooler, Int. Chemical Engineering 7 (4) (1967). [9] R.O. Parker, R.E. Treybal, The heat and mass transfer characteristics of evaporative coolers, Chemical Engineering Progress Symposium 57 (32) (1961). [10] Y. Niitsu, K. Naito, T. Anzai, Studies on characteristics and design procedure of evaporative coolers, Journal of SHASE 43 (7) (1969).