Crystallization fouling of CaCO3 – Analysis of experimental thermal resistance and its uncertainty

International Journal of Heat and Mass Transfer 55 (2012) 6927–6937

Contents lists available at SciVerse ScienceDirect

International Journal of Heat and Mass Transfer journal homepage: www.elsevier.com/locate/ijhmt

Crystallization fouling of CaCO3 – Analysis of experimental thermal resistance and its uncertainty T.M. Pääkkönen a,b,⇑, M. Riihimäki a, C.J. Simonson b, E. Muurinen a, R.L. Keiski a a b

University of Oulu, Department of Process and Environmental Engineering, Mass and Heat Transfer Process Laboratory, P.O. Box 4300, FI-90014 Oulu, Finland University of Saskatchewan, Department of Mechanical Engineering, Saskatoon, SK, Canada S7N 5A9

a r t i c l e

i n f o

Article history: Received 1 February 2012 Received in revised form 29 June 2012 Accepted 2 July 2012 Available online 4 August 2012 Keywords: Crystallization fouling Thermal resistance Experimental uncertainty Bulk crystallization

a b s t r a c t Crystallization fouling occurs when dissolved salts precipitate from an aqueous solution. In the case of inversely soluble salts, like calcium carbonate (CaCO3), this may lead to crystal growth on heated walls. Crystallization may also take place in the bulk solution either via homogeneous nucleation or heterogeneous nucleation on suspended material. In this paper, surface crystallization of CaCO3 and crystallization in the bulk fluid and its effect on the fouling rate on a heated wall are studied. The fouling experiments are done in a laboratory scale set-up of a flat plate heat exchanger. Accuracy of the results is analyzed by uncertainty analysis. SEM and XRD are used to determine the morphology and the composition of the deposited material. The uncertainty analysis shows that the bias and precision uncertainties in the measured wall temperature are the largest source of uncertainty in the experiments. The total uncertainty in the fouling resistance in the studied case was found to be ±13.5% at the 95% confidence level, which is considered to be acceptable. Surface crystallization rate is found to be controlled by the wall temperature indicating that the surface integration dominates the fouling process. The flow velocity affects the fouling rate especially at high wall temperature by decreasing the fouling rate with increasing flow velocity. Crystallization to the bulk fluid is found to enhance significantly the fouling rate on the surface when compared to a case in which fouling is due to crystal growth on the surface. Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction Fouling, which is often described as the deposition of unwanted material on a heat transfer surface, diminishes the heat transfer and increases the pressure drop of the process units such as heat exchangers. Due to fouling, operation and maintenance costs are increasing significantly. In addition, heat transfer equipment are often oversized for required duty because of the expected fouling. Therefore, fouling is a major challenge in design and operation of heat exchangers. By decreasing the fouling of heat exchangers, harmful environmental and economic effects can be reduced. One detrimental fouling mechanism in a wide range of industrial applications is crystallization fouling. Crystallization fouling is caused by dissolved salts which precipitate out of the solution due to supersaturation. Calcium carbonate is a common salt causing crystallization fouling especially in cooling water systems, desalination processes, and in drinking water systems. [1,2] ⇑ Corresponding author. at: University of Oulu, Department of Process and Environmental Engineering, Mass and Heat Transfer Process Laboratory, P.O. Box 4300, FI-90014 Oulu, Finland. Tel.: +358 50 3506454; fax: +358 8 5532304. E-mail address: tiina.m.paakkonen@oulu.fi (T.M. Pääkkönen). 0017-9310/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijheatmasstransfer.2012.07.006

Calcium carbonate has inverse solubility which means that supersaturated conditions may be achieved either by heating the solution above the limit temperature in which the supersaturation occurs or by increasing the concentration by evaporating the solution above the solubility limit. Crystallization fouling of calcium carbonate may also be caused by increase in pH, which decreases the solubility of calcium carbonate. In heat exchangers, heated surfaces are therefore easily exposed to the crystallization fouling. Crystallization fouling is a complex phenomenon which is affected by hydrodynamic and thermal conditions of the system, but also related to chemical kinetics, thermodynamics and material properties [3]. Fouling behavior of the same salt may therefore vary in different systems and operation conditions which make it essential to understand the fouling behavior of the process in question in order to reduce fouling in it. Mechanism of crystallization fouling is usually defined to include two main steps: transport of ions from the bulk fluid to the vicinity of the surface, and attachment of the depositing material to the surface [3]. The mass transport step is attributed to the molecular diffusion of ions through the laminar boundary layer at the surface. The driving force for the diffusion is the concentration difference between the bulk fluid (Cb) and the interfacial

6928

T.M. Pääkkönen et al. / International Journal of Heat and Mass Transfer 55 (2012) 6927–6937

Nomenclature

ð1Þ

The attachment step is described by the strongly temperature dependent surface integration of ions into the crystal lattice which has often found to follow Arrhenius type rate equation driven by the nth potential of the concentration difference between the saturation concentration (Csat) and the interfacial concentration (Ci) [2,4–6]:

md ¼ kr ðC i  C sat Þn

ð2Þ

kr ¼ k0 eEa =RT

ð3Þ

In the case of heat exchanger with a constant heat flux, the surface temperature (Ts) at the fouling layer-liquid interphase is depending on the local heat flux and the fluid side heat transfer coefficient. Schematic of the concentration and temperature profiles in the studied system are given in Fig. 1. [7] Crystallization fouling of calcium carbonate in forced convection systems can be controlled by mass transfer, surface integration, or both [8]. Mass transfer of the ions is a prerequisite for crystallization fouling. Mass transfer may control the fouling process particularly when crystallization fouling takes place on a heat exchanger surface and a laminar boundary layer occurs. In the regions further from the surface mass transfer is enhanced due to turbulence and the rate controlling step may be surface attachment. [2] Wall temperature has a clear effect on crystallization fouling of inversely soluble salts. An increase in the wall temperature increases also supersaturation, which enhances the fouling rate both in mass transfer and surface integration controlled fouling. In addition, mass transfer coefficients are linear functions of temperature, which also increases the mass transfer with higher wall temperatures. [3,7] The effect of the wall temperature is though greater with the crystallization fouling process controlled by the surface integration [9]. Flow velocity instead has a complicated role on fouling: On the other hand, an increase in the flow velocity enhances mass transfer resulting from the increased turbulence and further ions transport which promotes deposition. [10] But again, increased shear at the interphase reduces the probability of the adhesion of the depositing material reaching the solution-fouling layer interphase. Therefore, an increase in the flow velocity may

Acronyms SEE Standard Error of Estimate SEM Scanning Electron Microscopy XRD X-ray Diffraction

either increase the fouling rate (mass transfer controls) or decrease it if the interfacial shear has the greater effect (surface integration controls). [2,11–13] Hasson et al. [7] found that CaCO3 deposition in turbulent, constant heat flux conditions at annular test section, with unfiltered, supersaturated tap water is controlled by the diffusion mechanism which is seen as an increased fouling rate with an increasing Reynolds number at the constant surface temperature. Instead, Augustin and Bohnet [14] have defined crystallization fouling of CaCO3 in an annular duct to be controlled by the surface integration. Helalizadeh et al. [3] found out that at low flow velocities of the mixed CaCO3–CaSO4 solution the fouling process in the annular duct is diffusion controlled, but increasing the flow velocity changes the mechanism to the surface integration controlled. Also Fahimia et al. [13] have proposed that the crystallization fouling process is controlled by the mass transfer at low flow velocities, but with increasing velocities the surface integration controls the fouling process. This conclusion was also proposed by Najibi et al. [9] for crystallization fouling of calcium carbonate in convective, singlephase heat transfer conditions. Instead, Mwaba et al. [6,15] proposed crystallization fouling of CaSO4 in a rectangular flow channel to be controlled by the surface integration. This founding was in agreement with the studies of Bansal and Müller-Steinhagen [16] who found out that crystallization fouling of CaSO4 in a plate heat exchanger is surface integration controlled. Therefore, controlling fouling mechanism varies depending on the studied system.

Laminar boundary layer

Turbulent core

Cb Ci Csat Ts Tb

Concentration

md ¼ bðC b  C i Þ

Subscript/superscript b bulk i interphase sat saturation t at the certain time t w wall/surface 0 initial

Temperature

concentration at the surface (Ci). Mass transport of the ions to the vicinity of the surface can be described as [4,5]:

Greek letters b mass transfer coefficient (m/s)

Fouling layer

v

bias uncertainty concentration (kg/m3) hydraulic diameter (m) activation energy (J/mol) reaction rate constant (m4/kg s) pre-exponential factor (m4/kg s) mass deposition rate (kg/m2 s) order of reaction precision uncertainty heat flux (W/m2) gas constant (J/mol K) fouling resistance (m2 K/W) temperature (K) time (s) total uncertainty in parameter i total heat transfer coefficient between the heated surface and bulk fluid (W/m2 K) flow velocity (m/s)

Heat transfer surface

B C Dh Ea kr k0 md n P q R Rf T t Ûi U

Fig. 1. Schematic of the concentration and temperature profiles at the heat transfer surface [7].

T.M. Pääkkönen et al. / International Journal of Heat and Mass Transfer 55 (2012) 6927–6937

In the case of crystallization fouling, supersaturated conditions may occur also in the bulk fluid. Therefore, crystallization may take place also in the bulk fluid in addition to surface crystallization. [17] Bulk crystallization in the supersaturated solution may take place spontaneously (primary nucleation zone) or it may require seeds or impurities in the solution (metastable zone) depending on the pH value and concentration of the solution [18]. When spontaneous nucleation takes place, minute structures are formed at first from the collision of two molecules which are then collided with a third molecule and so on. Short chains or flat monolayers may be formed and eventually, the lattice structure is built up. This construction process takes place very rapidly, and can only continue if the local supersaturation is very high. Many of these sub-nuclei re-dissolve since they are very unstable. If they manage to grow beyond a certain critical size, they become stable under the average supersaturation conditions occurring in the bulk fluid. After that, they begin to grow into crystals of visible size. In most of the cases, traces of impurities are however the main reason for nucleation in the bulk fluid since the overall free energy change associated to the heterogeneous nucleation is less than in the case of spontaneous nucleation. [5] Solid material, like non-crystallized, inert particles or bulk crystallized particles, in the solution may affect notably not only the formation of nuclei in the bulk fluid but also the fouling rate [19,20]. Effects have been found to vary depending on the particles. Andritsos and Karabelas [1] found out that fine aragonite particles in the solution increased the fouling rate of calcium carbonate whereas relatively large calcite particles and TiO2 particles did not affect notably the fouling rate but they seemed to reduce deposit strength and change the morphology. Also Bansal et al. [21,22] found that bulk crystallized particles augmented the fouling rate of CaSO4 whereas non-crystallizing particles were found to mitigate the fouling [19]. In contrast, Hasson and Karman [23] reported that crystallized CaCO3 particles in the bulk fluid did not affect significantly the surface crystallization rates in pipes in highly supersaturated conditions [24]. Hasson and Zahavi [25] and Bramson et al. [24] suggested that the effect of particles on the crystallization fouling of CaCO3 and CaSO4 may be contradictory. Based on the research reported in the literature, it is clear that many parameters have significant effects on crystallization fouling. For example, the operating conditions (flow velocity, wall and bulk temperatures), solution properties (salt, supersaturation degree) but also the geometry of the system (thickness of the laminar boundary layer, shear forces) affect the formation of the deposition layer on heat transfer surfaces. Therefore, it is not possible to define these effects generally for all the cases but they need to be studied case specifically. Determination of these effects is essential for the prevention of fouling but also for detailed modeling of the fouling process. The aim of this work is therefore to study the effects of different operating conditions on the surface crystallization of calcium carbonate in a heated rectangular flow channel, and further to define the dominating fouling mechanism in the studied process. Since most industrial applications may have impurities in the bulk fluid, the aim is also to study the composite effect of bulk crystallized calcium carbonate particles and surface crystallization of calcium carbonate. Although separate fouling mechanisms have been studied quite extensively in the literature, the uncertainty in the experimental fouling results are rarely documented even though the uncertainty in fouling experiments is especially important because fouling resistances are often small [26]. In this study, the aim is to quantify the experimental uncertainty in order to determine the largest sources of uncertainty and the reliability of the results.

6929

2. Materials and methods 2.1. Experimental set-up Experiments are performed in a laboratory set-up mimicking a plate heat exchanger. Schematic figure of the set-up is given in Fig. 2. A more complete description of the set-up is given in Riihimäki et al. [27]. The set-up contains a vertical test section (flow from above downward) made of two stainless steel (AISI 316L) flat plate test surfaces with heat transfer area of 0.10 m by 0.20 m and thickness of 0.02 m. The nominal flow channel gap between the plates is 15 mm and the hydraulic diameter (Dh) is 26 mm. To ensure fully developed turbulent flow at the test section, a vertical flow channel with the same dimensions as the test section and a length of 0.54 m (20 Dh) precedes the test section. Similarly, a 0.14 m (5 Dh) section of channel after the test section ensure downstream disturbances do not enter the test section. Since turbulent flow becomes fully developed in 10 Dh, the flow entering the test section can be assumed to be hydrodynamically fully developed. However, there is no preheated section in the facility to obtain a fully developed thermal boundary layer. Therefore, the experiments are fully developed hydrodynamically but developing thermally. The surfaces of the test sections are milled, and the average roughness (Ra) value for the surface is 0.7 lm. Only one test section is used in the experiments in order to ensure equal surface conditions in every test. The test surface of the used test section is washed after every run by rinsing it with de-ionized water and brushing it gently with a soft sponge. The test section is heated by ohmic heaters (Watlow Firerod, 230 V/1000 W) embedded in the wall of the test section. The ohmic heaters are connected to a power transmitter (TEKLAB ACP100) to get a constant, but adjustable heat flux. According to the manufacturer, the bias uncertainty in the heat flux is ±2.5%. The test solution is circulated to the test section from a mixing tank (100 dm3) by pumping it through a double pipe heat exchanger which is used to set the inlet temperature of the test section to a desired value. The conductivity of the solution is controlled by adding conductivity control solution (1.5 g/l CaCl2 and 1.5 g/l NaHCO3 salts mixed with deionized water) automatically to the mixing tank during the experiment. The conductivity control is used to keep the concentration level in the test loop approximately constant, and thus to replace the ions that are removed during crystallization of the salts.

Fig. 2. Schematic drawing of the test set-up [27].

6930

T.M. Pääkkönen et al. / International Journal of Heat and Mass Transfer 55 (2012) 6927–6937

Two different kinds of filters in series (Watman washable nylon 60 lm and Watman polypropene 1 lm) are assembled in-line of the set-up before the test section in order to remove bulk precipitated particles from the test solution when only surface crystallization is studied (later called experiments with filters). For the experiments where the effect of bulk crystallized material is investigated, filters are removed from the set up (later without filters). The control and data acquisition system is built on the National Instruments Inc. components on compact Field point 2100 platform and uses LabView 8.0 software. The controlled variables in the test system are: flow velocity, bulk fluid temperature, inlet temperature to the test section, heat flux, and conductivity. The flow rate is measured using a Bürkert 8045 Electromagnetic Flow Transmitter, and controlled with a control valve (Bürkert 2632). The conductivity is measured (Bürkert 8225) and controlled with a magnetic on-off valve (Bürkert 0124). The temperatures of the inlet and outlet fluids and the heated walls are measured with SKS Automaatio Oy K-type thermocouples. The bias uncertainty in the temperature measurement, based on the manufacturer’s calibration, is ±0.004 T. Therefore, the bias uncertainty varies from 0.1 °C at 25 °C to 0.4 °C at 100 °C. Since the uncertainty increases with temperature, the thermocouples were calibrated by placing them in boiling water and confirming the manufacturer’s calibration. Wall temperature is measured by placing the thermocouples in channels that are drilled through the back of the test section to the surface of the test section. The depth of the channel drilled for the thermocouple is 0.018 m, and thickness of the test section is 0.02 m. There are three temperature measurement points at different axial positions in the test section. Due to relatively low thermal conductivity of the steel test section, a slight variation at the wall temperature is detected. The temperature measurement point in the middle of the test section is estimated to correspond to the average thermal conditions at the wall, and therefore that point is used to calculate the fouling resistance. 2.2. Test solution The test solution is prepared by mixing 50 g of CaCl2  2H2O (M=147.01 g/mol) (BDH Prolabo) and 50 g NaHCO3 (M = 84.0084 g/mol) (BDH Prolabo) into 100 dm3 of pre-heated deionized water. The calcium ion concentration in the solution is therefore 0.0034 mol/l, and the mass concentration is 0.136 g/dm3. Calcium ion concentration corresponds to the samples analyzed by atomic absorption spectrophotometer (AAS) which gave bulk concentration of about 0.132 g/dm3. The saturation concentration of the solution depends on temperature and pH value of the solution. Partial pressure of air, which affects the solubility of CO2 and further pH value of the solution, is considered to be constant during the experiments, and therefore its effect on the saturation concentration is assumed to be minor. The pH value of the solution is approximately 8.1, and the temperature of the solution varies from 30 °C of the bulk fluid up to 95 °C at the wall. According to Plummer and Busenberg [29], the saturation concentration of different polymorphs of CaCO3 depends on the temperature according to Fig. 3. In this study the experiments are done in the conditions where the supersaturated test fluid is on the border of the metastable and primary nucleation zones where induced or even spontaneous nucleation may take place in the bulk fluid [18]. From to polymorphs of CaCO3, calcite is thermodynamically more stable whereas aragonite may change to calcite. The transformation rate increases with increasing temperature, and if the polymorph is in contact with water or solution containing calcium carbonate, the transformation may occur even at room temperature. [30] In process of the time, all the crystals are finally transformed to calcite at any temperature [31]. Scanning Electron

Fig. 3. Solubility of calcium carbonate in water as a function of temperature.

Microscope (SEM, Jeol JSM-6400) images of the deposition layers indicate that the deposits in this study contained mainly two different types of morphologies. The deposits presented in Fig. 4 (a) are orthorhombic, needle-shaped crystals that are up to 300 lm long and about 5–10 lm of diameter. The deposits in Fig. 4 (b) are hexagonal round shaped particles that are up to 5 lm in diameter. Most of the studied samples contain hexagonal and orthorhombic crystal morphologies (examples in Fig. 5), and both calcite and aragonite. According to the X-ray Diffraction (XRD, Siemens D5000) analyses, the amount of calcite is approximately 85– 90%, and aragonite 15–10%. The obtained morphological structures are compared in the future to the results obtained by molecular modeling [32]. 2.3. Experimental procedure In every experiment, the mixing tank is first filled with deionized water. After that, the water is heated up to a desired value by circulating it in the set-up with the test section heaters and tank heaters on. When the desired bulk temperature is obtained, the test section heaters are turned off, and the fluid is circulated only in the mixing tank. The test solution is created by dissolving desired amounts of salts separately to a small amount of heated deionized water. The two concentrated salt solutions are slowly added to the mixing tank in which the heated deionized water is circulated. Then, the fluid is circulated through the whole set-up until the conductivity reaches a steady value in order to ensure full mixing of the salts. Following this, heating of the test section is again turned on, and test begins when the temperature of the test section surface and fluid inlet temperature to the test section are stabilized. The fouling experiments are done with four different fluid velocities (0.20, 0.27, 0.33, 0.40 m/s; average velocity in the test section) indicating Reynolds numbers 5200, 6900, 8900, 10,400, respectively. The uncertainty in the measured velocity and Reynolds number is ±10%. Thus, the flow in the test section is in transition or near to fully turbulent flow regime with a developing thermal boundary layer. Heat fluxes of 49.3, 52.5, 57.6, 59.2, and 61.8 kW/m2 are tested. The bulk fluid temperature entering the test section is set to 30 °C, except in the case where the effect of the increase in bulk temperature to the bulk crystallization is investigated, the bulk temperature of 40 °C is used. The surface temperature changes depending on the flow velocity, heat flux and bulk temperature. The studied conditions correspond to the conditions occurring often in industrial plate heat exchangers. Some results of these experiments can be found also from Pääkkönen et al. [28]. Due to the need to carry out a relatively large number of experiments, only the initial stage of the fouling (up to 350 min) is studied. During this time period, linear growth of the fouling resistance is achieved. Most of the experiments are repeated once or twice to ensure repeatability of the experiments. Since good repeatability was observed in the repeated experiments (increasing the total uncertainty by about 1% unit), the total amount of the experiments was reduced by repeating only some of the test conditions.

T.M. Pääkkönen et al. / International Journal of Heat and Mass Transfer 55 (2012) 6927–6937

6931

Fig. 4. SEM images of the deposited material.

Fig. 5. Examples from the experiments: (a) filters used, (b) filters not used.

2.4. Fouling thermal resistance The heating system produces a constant heat flux to the test section which means that the surface temperature increases as the resistive fouling layer grows on the surface. The interfacial temperature between the fluid and the deposition remains constant. Fouling resistance is used to describe the formation of the resistive fouling layer on the surface, and can be calculated from the increase in the surface temperature by using the convection heat transfer equation [4,15]:

Rf ¼

1 1  ¼ Ut U0



ðT w  T b Þ q

T t  T tb  T 0w  T 0b ¼ w q



 

t

ðT w  T b Þ q

 0

ð4Þ

The bulk temperature of the fluid entering the test section is automatically controlled to a constant value during an experiment by heating the fluid in the supply tank with a band heater or by cooling the fluid with a double pipe heat exchanger as necessary. The desired heat flux is obtained by using a power transmitter which gives a constant current and voltage. Since the bulk inlet temperature and the heat flux are constant during a test, the bulk temperature of the fluid leaving the test section is also constant during a test. Test data shows that the moving average of the bulk temperatures of the fluid at the inlet and outlet of the test section changes by less than ±0.2 °C during any given test. Therefore, the bulk temperatures at the inlet and outlet can be considered constants during an experiment. The increase in the bulk fluid temperature as the fluid flows through the test section is found to be less than 2 °C during any given test. Therefore, the fouling resistance in Eq. (4) can be calcu-

lated using the arithmetic-mean temperature difference between the bulk fluid temperature and the wall temperature (Tw–Tb) rather than log-mean temperature difference. Since inlet and outlet bulk fluid temperatures remain constant during any experiment, the bulk fluid temperature cancels out from Eq. (4). Some decrease in the heat flux may occur due to an increased heat loss to the environment when a fouling layer forms on the heated plate in the test section [15]. In our case, the experiments are relatively short and the obtained fouling layer is very thin. In addition, the change in the outlet bulk temperature during the experiment is insignificant. Therefore, any decrease in the heat flux is considered to be negligible and the initial heat flux is used in calculation of the fouling resistance. Eq. (4) becomes then

Rf ¼

T tw  T 0w DT ¼ q q

ð5Þ

The obtained fouling curve describes the change in the fouling resistance due to the growth of the fouling layer on the surface as a function of time. Fouling rate is used to express the change in the fouling resistance as a function of time

dRf DT ¼ q Dt dt

ð6Þ

3. Results and discussion 3.1. Uncertainty in experiments An uncertainty analysis [26,34] is performed to study and evaluate the bias and precision errors in the experiments. Bias uncertainties (B) come from sensor calibration and systematic errors

6932

T.M. Pääkkönen et al. / International Journal of Heat and Mass Transfer 55 (2012) 6927–6937

that remain constant throughout the experiment. Precision uncertainties (P) are due to random fluctuations and are observed in repeated measurements which do not agree exactly, and can be determined by statistical methods. Uncertainty analysis is crucial especially when the measured fouling resistances have small values. Detailed results are presented for a single representative case which was found to express typical fouling behavior in the experiments. The fouling resistance Rf (Eq. (5)) is a function of heat flux q and temperature difference DT between initial wall temperature Tw,0 and wall temperature at the specific time Tw,t. The uncertainty in the fouling resistance can be calculated as

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi u" 2  2 # u @Rf @R f BRf ¼ t BDT þ Bq @ DT @q

ð7Þ

In order to find the uncertainty in the fouling resistance, the uncertainty in the temperature difference and heat flux needs to be determined. The operation conditions with their bias uncertainties at the 95% confidence level are presented for the example case in Table 1. Substituting Eq. (5) into Eq. (7), and taking the partial derivatives gives for the example case

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u" 2  2 # u 1 DT  BDT þ  2  Bq BRf ¼ t q q vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi " u  2  2 # u 1 5:4 t ¼  0:48 þ   1440 57600 576002 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi h i ¼

7:1  1011 þ 5:5  1012 ¼ 8:72  106 m2 K=W

ð8Þ

Eq. (8) shows that the most of the uncertainty in the fouling resistance comes from the uncertainty in the temperature measurement. However, uncertainty in the heat flux should not be neglected either. At the 95% confidence level, the fouling resistance is therefore

Rf ¼ 9:38  105  8:72  106

m2 K m2 K ¼ 9:38  105  9:3% W W

The uncertainty in the fouling resistance is found to be less than ± 30% for most of the experiments. However, in the cases where the fouling rate is very low, the temperature difference remains also low which leads to high uncertainty in the fouling resistance. Therefore, a high enough (DT P 2 °C) temperature difference leads to acceptable uncertainty. In addition, accuracy of the experiments could be increased by using temperature sensors with lower uncertainty. In addition to the bias uncertainties, precision uncertainties exist in the experiments. Precision uncertainties are due to random fluctuations during the experiments. These fluctuations may be for example due to fluctuations in the flow velocity which further changes the wall temperature. Repeatability of the experiments is also considered in the calculation of the precision uncertainty by including all the repeated experiments in the calculation of the precision uncertainty.

The precision uncertainties are quantified by using linear regression and standard error of estimate (SEE). The SEE is multiplied by two to get 95% confidence level for more than 30 measurements [34]. Therefore, the precision uncertainty is P = 2 ⁄ SEE. Replicates done in the same conditions are included to the calculation of the SEE to get a common precision uncertainty for the experiments done in the same conditions. Standard error of estimate (SEE) can be used to study fouling curves in which growth in fouling resistance is linear because the method gives the standard deviation of the fouling resistance to the linear fit. Induction time, during which the fouling resistance does not increase but may even decrease, at the beginning of the experiment or asymptotic behavior of the fouling curve at the end of the experiment differ from the linear growth rate. Therefore, the linear fit in these areas cannot be used to predict fouling resistance. Instead, these areas need to be excluded from the calculation of the standard error of estimate since they increase misleadingly the value of the standard error of estimate because of the large difference to the linear fit in these areas. For that reason, the period of linear growth can reasonably be used to determine the uncertainty in the fouling resistance. For the example case, the 35 min induction period is excluded from the beginning of the experiment, whereas the remaining 267 min long linear growth period is used in linear regression. A replicate for the example case is also included to the calculation of the precision uncertainty. A common regression line for both the experiments and the measured values with their bounds of precision uncertainty are presented in Fig. 6. The obtained SEE for the 2 individual tests are 4.0  106 m2 K/W and 3.3  106 m2 K/W, while SEE for the common linear fit is 4.6  106 m2 K/W. This shows that the replicate corresponds very well with the studied case because SEE increases very little (about 15 %) when both fouling curves are included in the analysis. The correlation coefficient for the linear fit line is 0.96, which indicates that for 426 measurements, the correlation is highly significant [34]. The total uncertainty in the fouling resistance is a square root of the sum of the squares of the precision and bias uncertainties. In the precision uncertainty, the standard error of estimate is multiplied by two to get the total uncertainty at the 95% confidence level because more than 30 measurement points are used in the regression [34].

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ½ðBmeasurement Þ2 þ ð2  SEEÞ2  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ ½7:61  1011 þ 8:34  1011  ¼ 1:26  1005

b tot ¼ U Rf

ð9Þ

Therefore, for the 95% confidence level, the fouling resistance is

Rf ¼ 9:38  105  1:26  105

m2 K m2 K ¼ 9:38  105  13:5% W W

It should be noted that the total uncertainty in the fouling resistance would decrease from 13.5% to 12.7% if the replicate test was

Table 1 Conditions in the example case. Parameter

Numeral value

Bias uncertainty

q [kW/m2] DT [°C] v [m/s] Re t[s] linear

57.6 5.4 0.267 6721 17880

±1.44 ±0.48 ±0.027 ±672 ±18

Fig. 6. Excluded induction periods, linear fit and 95 % precision uncertainty bounds for the fouling resistance for the example case and its replicate.

6933

not included in the uncertainty analysis. This small change in the total uncertainty shows that the tests are very repeatable and the uncertainty due to repeatability of the experiment increases the total uncertainty by less than 1% unit. As a result, more focus is placed on conducting tests at different conditions rather than repeating tests at the same conditions. Nevertheless, many of the tests are repeated to ensure that the tests are repeatable and the reported uncertainty values include the repeatability. The total uncertainty in the fouling resistance is found to be acceptable for the most of the experiments during the linear period of the growth of the fouling layer. However, if the entire experiment is considered, the cases which include long induction periods at the beginning of the experiment have higher total uncertainties due to large deviations from a linear fit during the induction period. This shows the importance of only using the portrait of the experiment where the growth in the fouling resistance is linear in quantifying the uncertainties. Since the length of each experiment varies and the fouling resistance increases with the time, comparison between the experiments cannot be done based on fouling resistances without taking into account the length of the experiment. Therefore it is better to compare fouling rates (Eq. (6)). The fouling rate is obtained by dividing the fouling resistance with the length of the experiment which is for the studied case 17880 s ± 18 s or ±0.1% based on dozens of experiments. Uncertainty in fouling rate is obtained by taking the partial derivatives of Eq. (10) with respect to fouling resistance and time

b dR =dt U f

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u" 2  2 # u @ðdRf =dtÞ @ðdRf =dtÞ t tot b ¼  U Rf þ  Bt @Rf @t qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ ½4:988  1019 þ 2:734  1023  ¼ 7:06  1010 m2 K=Ws

ð10Þ

Therefore, at the 95% confidence level, the fouling rate is

dRf m2 K m2 K ¼ 5:24  109  13:5% ¼ 5:24  109  7:06  1010 Ws Ws dt Examples of the fouling rates with their uncertainty bounds in different flow velocities when the heat flux is constant (q = 57.6 kW/ m2) are presented in Fig. 7. The highest uncertainty value in the fouling rate when the whole experimental set is considered is ± 3. 4  109 m2K/Ws which is 16.8% of the calculated fouling rate. It can be seen from Fig. 7 that the error bounds are smaller with higher velocities, but the fouling rate has also a lower value. In addition, with the higher flow velocities the fouling rate is so low that the experimental facility can barely detect the formation of a fouling layer on the surface which also increases the uncertainty. An important part of the uncertainty analysis for a heat transfer experiment is an energy balance. The experimental data should satisfy the energy balance within experimental uncertainty and within a reasonable fraction. For the tests presented in this paper, the difference between the total energy flow into and out of the test section is less than ±3% when normalized by the energy flow into the test section, which is less than the uncertainty in the energy imbalance and therefore acceptable. 3.2. Surface crystallization Surface crystallization fouling experiments are carried out with the filters in-line of the set-up to avoid the effects of bulk fluid particles. The experiments are done with different flow velocities, heat fluxes, and surface temperatures to determine their effects on crystallization fouling. Fig. 8 presents examples of the fouling resis-

Fouling rate x109 [m2K/Ws]

T.M. Pääkkönen et al. / International Journal of Heat and Mass Transfer 55 (2012) 6927–6937

18 16 14 12 10 8 6 4 2 0 0.15 -2

0.2

0.25

0.3

0.35

0.4

0.45

Velocity [m/s]

Fig. 7. Surface crystallization fouling rates with their 95% uncertainty bounds for the experiments done with q = 57.6 kW/m2.

tances as a function of time. The fouling resistance is calculated from Eq. (5) by using the initial wall temperature and the measured temperature at every minute detected at the middle of the wall of the test section. In Fig. 8 (a) the flow velocity is varied with a constant heat flux (59.2 kW/m2) and in Fig. 8 (b) the heat flux is varied with a constant flow velocity (0.27 m/s). At the initial phase of the experiments, an induction time is detected. During this delay period, the first crystals start to grow on the heat transfer surface. During the induction time, the fouling resistance does not increase but may even decrease when first crystals nucleate on the surface increasing the wall roughness. The increased roughness enhances turbulence on the zone near to the heat transfer surface which then leads to enhanced convective heat transfer and further lowered surface temperature and fouling resistance. [3,6,9,16] The duration of the induction time seem to vary with the operating conditions: a decrease in the flow velocity or an increase in the heat flux shortens the induction time. When the flow velocity is lower, convective heat transfer is lower and further the wall temperature is higher. Higher wall temperature caused by either a low flow velocity or a high heat flux increases the crystallization fouling rate due to the temperature effect on the reaction rate constant (Eq. (3)), and increases the supersaturation at the wall. Therefore, thickness of the fouling layer reaches quickly the limit in which the advantage of the high heat transfer coefficient due to enhanced roughness is overcome by the added heat transfer resistance of the fouling layer. This is seen as a short induction time with low flow velocities and high heat fluxes. The wall temperature and the thickness of the laminar boundary layer decrease both when the flow velocity is increased. Temperature has an exponential effect on the surface integration which makes the fouling process much slower with the higher flow velocity and lower wall temperature. Therefore, it takes longer time for the crystals to compensate the effect of enhanced turbulence even the laminar boundary layer is thinner with the higher flow velocities. This can be observed as a longer induction time with the higher flow velocities. [16] When more crystals grow on the surface, the heat transfer resistance caused by the deposition layer overruns the effect of enhanced heat transfer due to turbulence, leading to linear elevation of the fouling resistance and ending of the induction time. During the linear increase in the fouling resistance, the deposition layer grows steadily on the surface. Results show that increasing the heat flux slightly increases the linear growth rate of the fouling resistance. This is caused by the increase in the initial wall temperature which is about 9 °C between the heat fluxes 52.5 kW/m2 and 61.8 kW/m2. The increased wall temperature decreases solubility of the inversely soluble calcium carbonate increasing the supersaturation at the wall which further leads to higher fouling rate. In addition, the rate constant for the crystallization fouling process

6934

T.M. Pääkkönen et al. / International Journal of Heat and Mass Transfer 55 (2012) 6927–6937

(b)

100

v=0.20m/s v=0.27m/s

80

v=0.33m/s v=0.40m/s

60 40 20 0 -20

0

50

100

150

200

250

300

350

400

Time [min]

Fouling resistance x106 [m2K/W]

Fouling resistance x106 [m2K/W]

(a)

100

q=62kW/m2 q=59kW/m2

80

q=53kW/m2 60 40 20 0 0 -20

100

200

300

400

Time [min]

Fig. 8. Fouling resistance (a) with different velocities when the heat flux is 59.2 kW/m2 and (b) with different heat fluxes when the velocity is 0.27 m/s (Re = 6900) [28].

increases exponentially with the increasing temperature. On the other hand, increasing the flow velocity decreases the fouling rate because the wall temperature decreases with the increasing flow velocity due to enhanced convective heat transfer. The increase in the flow velocity from 0.20 m/s to 0.40 m/s changes the initial wall temperature about 15 °C which explains the stronger changes in the fouling rate when the flow velocity is varied. At some point of the fouling process, fouling resistance may start to reach a constant value leading to an asymptotic fouling curve. Decreased fouling rate may be caused by the decrease in the temperature at the interphase between the deposition and the fluid, increased flow rate due to narrowed flow channel or removal of the deposits from the surface. [6,8] Because the heat flux in the studied system is constant, the decrease in the temperature at the interphase is minor [35]. Since the obtained fouling layer is very thin, it’s effect on the cross-sectional area of the flow channel and further on the flow velocity is also minor. Therefore, the asymptotic fouling behavior would be caused by the removal of the deposits from the surface which becomes equal to the deposition rate to the surface. Asymptotic fouling behavior cannot be detected in the fouling curves presented in Fig. 8. Experiments presented here are rather short (up to 350 min) which denotes early events of the fouling process. At the initial stage of the fouling crystals are small and they do not affect the cross sectional area of the flow channel, and therefore the removal rate at the beginning of the fouling process is minor [8,16]. In addition, pure calcium carbonate deposition forms a rather tenacious fouling layer which is hard to remove from the surface [1,33]. Therefore, it is likely that hardly any removal occurs at the early stage of fouling, and the growth of the fouling layer is seen as a linear fouling resistance curve. When the fouling layer grows further, crystals become longer and more fragile, and they eventually affect the local shear forces [8]. These effects could be seen as asymptotic fouling curve if longer experiments are done [36]. In order to define the controlling mechanism for the studied fouling process, the effect of the flow velocity needs to be studied at constant initial wall and bulk temperatures when the flow velocity is varied [7]. In these conditions, the rate for the surface integration remains constant, and the effect of the flow velocity becomes visible. The initial wall temperature of the heated plate is a result of the heat flux and the flow velocity when the bulk temperature is kept constant. Therefore, to keep the initial wall temperature constant, the flow velocity and the heat flux is changed at the same time. When the fouling takes place, the measured wall temperatures change, and these changes are used in calculation of the fouling rates according to Eq. (6) in which the temperature difference between the initial and final temperatures and the length of the experiment are used. The dependency of the fouling rate on the

constant initial wall temperature as a function of (a) flow velocity and (b) heat flux is presented in Fig. 9 with the 95% uncertainty bounds. In the studied case, the fouling rate decreases with the increasing flow velocity (Fig. 9(a)) when the wall temperature remains constant. If the mass transfer controls the fouling process, the fouling rate should instead increase with the increasing flow velocity at constant surface temperature. The opposite effect shows that in our case the surface integration controls over the mass transfer. This finding is in line with many studies on crystallization fouling found in the literature [6,8,14,16]. In our case, the fouling rate is even decreasing with the flow velocity. This can be explained by increased shear forces at the fouling layer-fluid interphase which prevent the depositing material to adhere to the crystal lattice at the surface at higher velocities. This phenomenon can be associated with the formation of the laminar boundary layer at the surface. The flow velocity at the laminar boundary layer increases with the increasing bulk flow velocity and residence time of the fluid at the wall shortens. Therefore, probability of the depositing material to adhere to the surface decreases as the residence time of the fluid at the wall decreases. Eventually at higher flow velocities the fouling rate reaches a rather steady value. The initial wall temperature has a significant effect on the fouling rate when the flow velocity is low, but when the flow velocity is increased, the effect of the initial wall temperature becomes minor. Reason for that could be that the stronger is the surface integration (meaning the higher is the wall temperature at the surface integration controlled fouling process) the higher flow velocity is needed to prevent the foulants to attach to the surface. Therefore, attachment of the crystals to the surface is more probable with higher wall temperature at the same flow velocity. As the flow velocity increases, the shear forces overcome the effect of wall temperature. The same trends are clearly seen when results for different heat fluxes are investigated (Fig. 9(b)): The fouling rate decreases at constant initial wall temperature even though the heat flux increases. This is due to an increasing flow velocity to keep the wall temperature constant. The increased flow velocity then prevents the depositing material to attach to the surface. The decrease in the fouling rate is the highest with the highest wall temperature (Tw = 85 °C) but minor with lower wall temperatures (Tw = 75 °C). Therefore these results show clearly the effect of increased shear near the wall at higher flow velocities. These phenomena are also detected by other researchers for example for the initial chemical reaction fouling [11,12] and for the initial stages of the calcium sulphate crystallization fouling [13]. They have proposed that at low flow velocities mass transfer is controlling the fouling process, but at some flow velocity, a maximum fouling rate is achieved. The flow velocity, in which the

6935

T.M. Pääkkönen et al. / International Journal of Heat and Mass Transfer 55 (2012) 6927–6937

(b)

6

Tw,init=85

5

Tw,init=80 Tw,init=75

4 3 2 1 0 0.15 -1

0.2

0.25

0.3

0.35

0.4

0.45

Velocity [m/s]

Fouling rate x109 [m2K/Ws]

Fouling rate x109 [m2K/Ws]

(a)

6

Tw,init=85

5

Tw,init=80

0.20

Tw,init=75

4 3 2 1 0 45000 -1

0.27 0.20 0.27 0.27

50000

0.33

55000

0.33 0.40

60000

0.33 0.40

65000

Heat flux [W/m2 ]

Fig. 9. Fouling rates on constant initial wall temperature with different (a) flow velocities and (b) heat fluxes (labels denote the flow velocity of the experiment) with their 95% uncertainty bounds.

maximum fouling rate occurs, is the higher the higher is the wall temperature, and the maximum fouling rate is the higher the higher is the wall temperature, which corresponds to our results. After the maximum, the fouling rate begins to decrease with an increasing flow velocity denoting the surface integration controlled fouling. The fouling rate decreases even the surface temperature remains constant because increasing shear forces at increasing flow velocities prevent the depositing material to attach to the surface. [12] Eventually, fouling rates reach nearly constant values. [12,16] Therefore, the experiments in this study are performed in the surface integration controlled area which occurs after the maximum fouling rate. Decreasing the flow velocity would most probably show the maximum fouling rates for each wall temperatures, and further mass transfer controlled area in which the fouling rate decreases with decreasing flow velocity. The obtained results show that the flow velocity has a clear effect on fouling rate especially at high wall temperatures. The experimental data shows that fouling rate decreases as the velocity increases, even though the transport of ions to the surface increases with the velocity. This means that the effect of flow is not due to enhanced mass transfer at high velocities, but likely due to the shear force effects of the flow. Therefore, the experiments seem to be performed in the area controlled by the surface integration reaction but the fouling process is affected by the shear forces which prevent the foulants to attach to the surface in the vicinity of the surface. 3.3. Bulk crystallization Bulk crystallization may take place when the test solution is supersaturated also in the bulk fluid, which is the case in our experiments. The pH value of the test fluid is about 8.1 and the calcium-ion concentration about 3 mmol/l, which indicates that the system is on the border of metastable and primary nucleation zones where induced or even spontaneous nucleation may take place in the bulk fluid [18]. To study the effect of the bulk crystallization on the fouling rate, the bulk temperature is set to 30 °C, heat flux is varied between 52.5 kW/m2 and 59.2 kW/m2, and velocity between 0.27 m/s and 0.40 m/s. The surface temperatures for the studied conditions are between 69 °C and 85 °C. Fig. 10(a) presents the effect of the heat flux when the flow velocity is set to 0.27 m/s and (b) the effect of the flow velocity when the heat flux is set to 52.5 kW/m2 for the experiments performed with (thin lines) and without (thick lines) in-line filters. Fig. 10 shows that the effect of the bulk crystallization is very significant in every studied case. The deposition of the bulk crystallized material increases the fouling rate by 76% and 81%, when the

heat flux is 52.5 and 59.2 kW/m2, respectively. The increase in the fouling rate due to bulk crystallization for the velocities of 0.27 and 0.4 m/s is 90% and 96%, respectively. When the solution is not filtered, some impurities may occur in the set-up, and act as seeds for bulk crystallization. In addition, in the studied conditions spontaneous nucleation to the bulk fluid may take place. [18] Bulk crystallized particles may affect the fouling rate in two different ways: Significant quantities of particles may attach to the surface by particulate fouling especially if the particulate concentration is large; or only a few particles attach to the surface and increase the rate of surface crystallization by providing additional nucleation sites on the surface. [8] When the flow velocity is decreased (in Fig. 10(b)) or the heat flux is increased (Fig. 10(a)) the surface temperature increases which further increases the rate of surface integration but also the supersaturation at the wall. Therefore the bulk crystallization near to the wall is enhanced when the flow velocity is decreased, and more bulk crystallized particles are likely to attach to the surface and provide more nucleation sites for the surface crystallization. Combined particulate and crystallization fouling is found to weaken the structure of the deposition layer and make the removal from the surface more significant [1,33]. The removal rate from the surface increases with the flow velocity, and may therefore become more significant with the higher flow velocity when bulk the crystallization occurs. If the filters are not used in the experiments, the amount of the bulk crystallized material increases with the bulk temperature because of the increasing degree of supersaturation. This was observed visually from turbidity of the test fluid in the mixing tank. During the experiments with lower fluid temperature without the filters, the bulk fluid was quite clear having lower turbidity, while in the experiments with higher fluid temperatures the bulk fluid was very turbid. During all the experiments with the filters, the bulk fluid was visually observed clear. The bulk temperature is varied between 30 °C and 40 °C in order to determine the effect of bulk crystallization on the surface deposition (Fig. 11) in the experiments done without filters. In these experiments, other parameters are kept constant. For the studied bulk temperatures (Fig. 11), the bulk temperature does not have a strong effect on the fouling rate even the surface temperature is higher with the higher bulk temperature. However, the shape of the fouling curve with the higher bulk temperature is more asymptotic whereas with the lower bulk temperature, the shape is more linear. An increase in the bulk temperature increases the supersaturation in the bulk fluid, and more crystals are likely to form to the bulk fluid. Therefore, at higher bulk temperatures, more particulate material is available in the solution, and it is likely that also the

6936

T.M. Pääkkönen et al. / International Journal of Heat and Mass Transfer 55 (2012) 6927–6937

(b)

140 120 100 q=59kW/m2_no filtered q=53kW/m2_no filtered q=59kW/m2_filtered q=53kW/m2_filtered

80 60 40 20 0 -20

0

50

100

150

200

250

300

350

400

Time [min]

Fouling resistance x106 [m2K/W]

Fouling resistance x106 [m2K/W]

(a)

140

v=0.27m/s, no filtered

120

v=0.4m/s, no filtered v=0.27m/s, filtered

100

v=0.4m/s, filtered

80 60 40 20 0 -20

0

50

100

150

200

250

300

350

400

Time [min]

Fig. 10. Effect of the bulk crystallization on the fouling resistance when (a) the heat flux and (b) the flow velocity is varied [28].

Fouling resistance x106 [m2K/W]

deposition layer includes more particles which make the deposition layer weaker. A weakened deposition layer facilitates the removal process, which is seen as an asymptotic fouling curve at higher bulk temperature [1,33]. Enhanced bulk crystallization may also decrease the ion concentration of the solution which further decreases the surface crystallization due to lowered supersaturation. Ion concentration may be lowered due to the fact that even the conductivity of the solution is controlled, the amount of calcium ions in the solution decreases due to crystallization when sodium and chloride ions accumulate to the solution with the time because the test fluid is not renewed continuously. In the case of surface crystallization, the decrease in the calcium ion concentration of the solution with the time is found to be minor and to have a negligible effect on the fouling rate. However, when strong bulk crystallization takes place, the amount of calcium ions in the solution may decrease substantially, which leads to lowered supersaturation of the solution and further decreased bulk and surface crystallization. Therefore, in order to study bulk crystallization at higher bulk temperatures, the solution should be renewed during the experiment to ensure the supersaturation of the solution and to avoid accumulation of sodium and chloride ions. Therefore, it seems that with lower bulk temperature the bulk crystallized particles provide more nucleation sites to the surface and therefore enhance the surface crystallization. The amount of particles in the deposition layer remains rather low and removal from the surface is not significant. When the bulk temperature is higher, the amount of particles at the bulk solution increase significantly which also increases the amount of particulate material in the deposition layer, may weaken the structure of the deposition and therefore enhance the removal from the surface. Therefore, controlling fouling mechanism seems changing from the surface

140 Tw=77ºC_Tb=30ºC

120

Tw=85ºC_Tb=40ºC

100 80 60

crystallization to particulate fouling when bulk temperature is increased. 4. Conclusions The experimental uncertainty in the crystallization fouling of CaCO3 on a flat heat transfer surface was studied by quantifying the bias and precision uncertainties in measurements. The total uncertainty in the fouling resistance in the example case was found to be ±13.5% at the 95% confidence level, which is considered to be acceptable. The highest uncertainty in the whole experimental set was defined to be ±3.4  109 m2 K/Ws (±16.8% from the calculated fouling rate). The uncertainty analysis shows that the bias and precision uncertainties at the measured wall temperature are the largest source of uncertainty in the experiments. Reducing those uncertainties would have the greatest impact on the total uncertainty in the fouling rate. This could be done by prolonging the duration of the experiments in order to obtain larger difference between the initial and final wall temperatures. Effects of different operating parameters on crystallization fouling were studied to find out their effects on controlling fouling mechanism. Surface crystallization was found to be controlled by the strongly temperature dependent surface integration step. Increasing the flow velocity at the constant initial wall temperature decreased the fouling rate which indicates that mass transfer is not restricting the surface crystallization fouling processes in these conditions but the shear forces at the wall may prevent the crystals to attach to the surface when the flow velocity is high. Bulk crystallization was found to increase the fouling rate significantly by increasing the surface crystallization via an increased amount of nucleation sites at the surface, and by increasing particulate fouling of bulk crystal to the surface. Increase in the bulk temperature seems to promote particulate fouling and removal of the deposition from the surface but it may also decrease the supersaturation of the fluid which decreases crystallization fouling. More experiments are needed to quantify the portions of the surface crystallization and particulate fouling of the bulk crystallized material in the composite fouling. Acknowledgements

40 20 0 -20

0

100

200

300

Time [min]

Fig. 11. Effect of the bulk temperature on the fouling resistance when the filters are not used [28].

Financial support from the Graduate School for Energy Science and Technology (EST), the MATERA ERA-Net network, Finnish Funding Agency for Technology and Innovations (Tekes), Jenny and Antti Wihuri Foundation, Tauno Tönning Foundation, NSERC (Canada), and industrial partners are greatly acknowledged. The authors also express their thanks to all the research partners. The corresponding author greatly appreciates the Department of

T.M. Pääkkönen et al. / International Journal of Heat and Mass Transfer 55 (2012) 6927–6937

Mechanical Engineering at the University of Saskatchewan for the fruitful co-operation during her researcher phase in Canada.

References [1] N. Andritsos, A.J. Karbelas, Calcium carbonate scaling in a plate heat exchanger in the presence of particles, Int. J. Heat Mass Transfer 46 (2003) 4613–4627. [2] T.R. Bott, Aspects of fouling, Exp. Therm. Fluid Sci. 14 (1997) 356–360. [3] A. Helalizadeh, H. Müller-Steinhagen, M. Jamialahmadi, Mixed salt crystallization fouling, Chem. Eng. Process 39 (2000) 29–43. [4] S. Krause, Fouling of heat transfer surfaces by crystallization and sedimentation, Int. Chem. Eng. 33 (3) (1993) 335–401. [5] J.W. Mullin, Crystallization, Butterworths, London, 1961. 268 p. [6] M.G. Mwaba, M.R. Golriz, J. Gu, A semi-empirical correlation for fouling of heat exchange surfaces, Appl. Therm. Eng. 26 (2006) 440–447. [7] D. Hasson, M. Avriel, W. Resnick, T. Rozenman, S. Windreich, Mechanism of calcium carbonate scale deposition on heat-transfer surfaces, Ind. Eng. Chem. Fundament. 7 (1) (1968) 59–65. [8] B. Bansal, X.D. Chen, H. Müller-Steinhagen, Analysis of ‘classical’ deposition rate law for crystallization fouling, Chem. Eng. Process. 47 (2008) 1201–1210. [9] S.H. Najibi, H. Müller-Steinhagen, M. Jamialahmadi, Calcium carbonate scale formation during subcooled flow boiling, J. Heat Transfer 110 (1997) 767–775. [10] B.O. Hasan, G.J. Nathan, P.J. Ashman, R.A. Craig, R.M. Kelso, The effect of temperature and hydrodynamics on the crystallization fouling under cross flow conditions, Appl. Therm. Eng. 36 (2012) 210–218. [11] N. Epstein, A model of the initial chemical reaction fouling rate for flow within a heated tube, and its verification, in: G. F. Hewitt (Ed.), Proceedings of 10th International Heat Transfer Conference, Taylor & Francis, Brighton, UK (1994) pp. 225–229. [12] I.C. Rose, P. Watkinson, N. Epstein, Testing a mathematical model for initial chemical reaction fouling using a dilute protein solution, Canad. J. Chem. Eng. 78 (2000) 5–11. [13] F. Fahiminia, A.P. Watkinson, N. Epstein, Early events in the precipitation fouling of calcium sulphate dehydrate under sensible heating conditions, Canad. J. Chem. Eng. 85 (2007) 679–691. [14] W. Augustin, M. Bohnet, Influence of the ratio of free hydrogen ions on crystallization fouling, Chem. Eng. Process. 34 (1995) 79–85. [15] M.G. Mwaba, C.C.M. Rindt, A.A. Steenhoven, M.A.G. Vorstman, Experimental investigation of CaSO4 crystallization on a flat plate, Heat Transfer Eng. 27 (3) (2006) 42–54. [16] B. Bansal, H. Müller-Steinhagen, Crystallization fouling in plate heat exchangers, Trans. ASME 115 (1993) 584–591. [17] P.J.R. Schreier, P.J. Fryer, Heat exchanger fouling: a model study of the scale up of laboratory data, Chem. Eng. Sci. 50 (8) (1995) 1311–1321. [18] H. Elfil, H. Roques, Prediction of the limit of the metastable zone in the ‘‘CaCO3–CO2–H2O’’ system, AIChE J. 50 (8) (2004) 1908–1916. [19] B. Bansal, X. Chen, H. Müller-Steinhagen, Use of non-crystallizing particles to mitigate crystallization fouling, Int. Commun. Heat Mass Transfer 5 (30) (2003) 695–706.

6937

[20] R. Sheikholeslami, Composite fouling of heat transfer equipment in aqueous media – a review, Heat Transfer Eng. 21 (2000) 34–42. [21] B. Bansal, H. Müller-Steinhagen, X. Chen, Comparison of crystallization fouling in plate and double-pipe heat exchangers, Heat Transfer Eng. 22 (5) (2001) 13– 17. [22] B. Bansal, H. Müller-Steinhagen, X. Chen, Effect of suspended particles on crystallization fouling in plate heat exchangers, J. Heat Transfer 8 (119) (1997) 568–574. [23] D. Hasson, M. Karman, Novel process for lining water mains by controlled calcite deposition, Corros. Prevent. Control 31 (1984) 9–17. [24] D. Bramson, D. Hasson, R. Semiat, The roles of gas bubbling, wall crystallization and particulate deposition in CaSO4 scale formation, Desalination 100 (1995) 105–113. [25] D. Hasson, J. Zahavi, Mechanism of calcium sulphate scale deposition on heat transfer surface, Ind. Eng. Chem. Fundament. 9 (1) (1970). [26] M. Izadi, D.K. Aidun, P. Marzocca, H. Lee, Experimental investigation of fouling behavior of 90/10 Cu/Ni tube by heat transfer resistance monitoring method, J. Heat Transfer 133 (2011). [27] M. Riihimäki, U. Ojaniemi, T. Pättikangas, T.M. Pääkkönen , M. Manninen, E. Puhakka, E. Muurinen, C.J. Simonson, R.L. Keiski, Fouling in high solid content suspension – effect of nucleating air and background electrolyte, in: H. MüllerSteinhagen, M.R. Malayeri, A.P. Watkinson (Eds.), Proceedings of 8th International Conference on Heat Exchanger Fouling and Cleaning – 2009, EUROTHERM Seminar No. 86, Schladming, Austria, 2010 pp. 192–199. [28] T.M. Pääkkönen, M. Riihimäki, C.J. Simonson, E. Muurinen, R.L. Keiski, Crystallization fouling of CaCO3– Effect of bulk precipitation on mass deposition on the heat transfer surface, in: H. Müller-Steinhagen, M.R. Malayeri, and A.P. Watkinson (Eds.), Proceedings of 8th International Conference on Heat Exchanger Fouling and Cleaning – 2009, EUROTHERM Seminar No. 86, Schladming, Austria, 2010, pp. 209–216. [29] L.N. Plummer, E. Busenberg, The solubilities of calcite, aragonite and vaterite in CO2 H2O solutions between 0 °C and 90 °C, and evaluation of the aqueous model for the system CaCO3 – CO2 – H2O, Geochim. Cosmochim. Acta 1 (46) (1982) 1011–1040. [30] R. Sheikholeslami, M. Ng, Calcium sulphate precipitation in the presence of nondominant calcium carbonate: thermodynamics and kinetics, Ind. Eng. Chem. Res. 40 (2001) 3570–3578. [31] K. Savada, The mechanism of crystallization and transformation of calcium carbonates, Pure Appl. Chem. 69 (1997) 921–928. [32] E. Puhakka, M. Riihimäki, T.M. Pääkkönen, R.L. Keiski, Density functional theory studies on the formation of CaCO3 depositions on cristobalite, diamond, and titanium carbide surfaces, Heat Transfer Eng. 32 (2011) 282–290. [33] G.J. Lee, L.D. Tijing, B.C. Pak, B.J. Baek, Y.I. Cho, Use of catalytic materials for the mitigation of mineral fouling, Int. Commun. Heat Mass 33 (2006) 14–23. [34] H.W. Coleman, W.G. Steele, Experimentation and Uncertainty Analysis for Engineers, Wiley Interscience, New York, 1989. 275 p. [35] D. Hasson, J. Zahavi, Mechanism of calcium carbonate scale deposition on heat-transfer surface, Ind. Eng. Chem. Fundament. 7 (1) (1968) 59–65. [36] W. Li, R.L. Webb, Foulig in enhanced tubes using cooling tower water, Part II Combined particulate and precipitation fouling, Int. J Heat Mass Transfer 43 (2000) 3579–3588.