ECI Symposium Series, Volume RP2: Proceedings of 6th International Conference on Heat Exchanger Fouling and Cleaning - Challenges

and
Opportunities, Editors Hans Müller-Steinhagen, M. Reza Malayeri, and A. Paul Watkinson, Engineering Conferences International, Kloster Irsee,
Germany, June 5 - 10, 2005

INITIATION OF CaSO4 SCALE FORMATION ON HEAT TRANSFER SURFACES UNDER

POOL BOILING CONDITIONS

M.R. Malayeri and H. Müller-Steinhagen

Institute of Thermodynamics and Thermal Engineering, University of Stuttgart , Pfaffenwaldring 6, D-70550,

Stuttgart, Germany, [email protected]
German Aerospace Research Centre (DLR), Pfaffenwaldring 38-40, D-70569, Germany, [email protected]

ABSTRACT bubbles, which all play a key role in the process of scale
The success of innovative fouling mitigation techniques formation.
such as ion implantation depends upon early stage of scale
formation on heat transfer surface. This is because the first In this investigation, the initial period of fouling during
crystalline nuclei that appear on the surface during the pool boiling heat transfer has been investigated. Published
initial period dictate how fouling would develop in latter experimental results for this region are scarce, and most
stages. In this study, the initial period of deposition of previous investigations have dealt with the full process of
calcium sulphate on heat transfer surfaces has been scale formation without paying detailed attention to the
investigated under pool boiling conditions. The independent initial period. Nevertheless, this topic is of particular
variables were heat flux and calcium sulphate concentration. importance because:
The experimental results show that the time until the heat • The first crystalline nuclei that appear on the surfaces
transfer coefficient reaches its intermediate maximum during the initial period are decisive for analysing and
decreases with increasing concentration and heat flux, and understanding of the fouling mechanism.
is also significantly affected by the surface finish. Neural
network architectures were utilized to correlate the • During the initial period, bubbles are in direct contact
experimental results during the initial deposition period. A with the heat transfer surfaces, while they are
satisfactory agreement between predicted and measured subsequently in contact with the scale layer. Therefore,
heat transfer coefficients has been achieved with an average the success of novel fouling prevention techniques, such
error of 8.7%. as surface coating by ion implantation or sputtering, is
highly depended on the initial surface conditions (such
INTRODUCTION as surface tension, surface energy and roughness) and
the initial period of scale formation.
Boiling heat transfer is one of the most efficient means
of energy transfer per heat transfer unit area in thermal This paper starts with a brief review of previous studies
processes, due to: of this topic. It will then address the influence of different
operating parameters such as heat flux and salt
• high surface temperature above saturation; concentration on scale initiation and formation during
• bubble behaviour resulting in agitation of the thermal boiling. Finally, the experimental results will be correlated
boundary layer. using artificial neural networks. The resulting network
estimates the initial fouling behaviour with good accuracy.
However, the formation of deposit layers on heat
It could ultimately be coupled with the recent model by
transfer surfaces may severely reduce the efficiency of heat
Jamialahmadi and Müller-Steinhagen (2004) that predicts
exchangers, in particular during pool boiling. The
the deposition rate beyond the initial period of scale
deposition of unwanted materials with low thermal
formation during pool boiling heat transfer.
conductivity causes heat transfer coefficients to drop
considerably, leading to additional energy requirements to
maintain the desired process conditions on target. In many BACKGROUND
There have been few serious attempts to understand the
cases, ultimately, operations have to be stopped for
effect of boiling mechanisms on the formation of scale
cleaning.
Such conditions arise in many heat exchangers in which under pool boiling conditions. These studies are generally
divided into two distinct but interrelated categories namely
nucleate boiling is the dominant process, perhaps kettle
single tubes and tube bundles. For single tubes, Partridge
reboilers and steam generators provide the most spectacular
examples. Fouling of heat exchangers with pool boiling is and White (1929); Freeborn and Lewis (1962); Hospeti and
Mesler (1965); Palen and Westwater (1966); Jamialahmadi
more severe and hence more complicated than with forced
et al. (1989, 1993 and 2004) and Bornhorst et al. (1999) are
convective heat transfer. This is mainly due to higher
surface temperatures, bubble behaviour, and rapid among the investigators who have offered explanations for
the mechanisms of fouling. For tube bundles the
supersaturation of salts with inverse solubility beneath

Produced by The Berkeley Electronic Press, 2016

256 Heat Exchanger Fouling and Cleaning - Challenges and Opportunities [2005], Vol. RP2, Article 37

experimental results are sparse and non-conclusive Fig. 1 illustrates typically how heat transfer coefficient
(Malayeri et al., 2005). In the present study, the focus is, varies as a function of time for a concentration of 1.2 g/L
therefore, only on the initial period for fouling on single and a heat flux of 300 kW/m2 (Malayeri et al., 2005). Open
tubes. circles correspond to data for the initial period up to 20 min,
Jamialahmadi and Müller-Steinhagen (1993), in a which are difficult to recognize in the complete
review paper on fouling during boiling, reported that the experimental run which is shown as solid circles. As it can
primary causes of fouling for solutions with inverse be seen the heat transfer coefficient drops rapidly [region I]
solubility are supersaturation and bubble formation followed by a gradual recovery to an intermediate
behaviour. They also stated that the following steps take maximum [region II] which in turn is followed by a more
place when a deposit layer forms on the surface of a single gradual decline towards an asymptotic value of about 6000
tube during pool boiling: W/m2K [region III]. Firstly, unlike forced convective heat
transfer, one cannot see any induction period of non-fouling
• Bubble nuclei form just above the nucleation sites that
behaviour of zero fouling resistance. This could be due to
are available on the surface. The bubble density is
rapid increase in concentration of salt in the micro-layer
directly proportional to the number of nucleation sites, beneath the growing bubbles. Secondly, most and upright
and hence the surface material and roughness are of
changes occur in regions I and II (called initial period
prime importance.
hereafter) in which significant changes in heat transfer
• Bubble agitation initially produces considerable coefficient happens. Jamialahmadi and Müller-Steinhagen
turbulence that enhances heat transfer. In this phase, (2004) showed that in region I, the number of nucleation
bubble characteristics such as size, shape and rise sites remains almost constant and bubbles are rather
velocity are affected by the physical/chemical properties reluctant to detach due to surface tension. This together with
of the boiling liquid, notably surface tension. rapid supersaturation of solution beneath bubbles would
• Beneath the bubbles, a microlayer exists, in which the give rise to initial reduction of heat transfer coefficients. In
concentration of the dissolved salt is increased due to the region II, however, the number of nucleation sites (known
preferential evaporation of the solvent. If the dissolved as secondary nucleation sites) increases due to formation of
salt has an inverse solubility relationship with first deposit nuclei that would then result in a higher
temperature, the considerable temperature gradient in number of generated bubbles and consequently higher heat
the microlayer leads to further supersaturation of the transfer coefficients. Finally in region III, surface
solution. Under such circumstances, crystal nuclei will temperature and hence bubble generation declines due to the
begin to precipitate on the surface and a deposit layer development of fouling layer which in turn gives rise in
builds up, initially in the form of concentric rings around more moderate decrease in heat transfer coefficient.
the nucleation sites. Both, surface temperature and bulk Nonetheless the initial period, in which also the first
concentration dominate the formation and thickness of crystalline nuclei appear on the surface, is decisive for
the fouling layer in this phase. analysing and understanding of the fouling mechanism.
Initial time, [min]
0 5 10 15 20 EXPERIMENTAL SET-UP AND PROCEDURE
15000 15000
The preliminary experimental results which will be
I II analysed in this study were carried out by Bartlett (1995) as
Initial heat transfer coefficient, [W/m2K]
Heat transfer coefficient, [W/m2K]

13000 12000 part of his MSc thesis (University of Auckland, New

Zealand) and thus more details of the test rig and analysis of
11000 9000 the errors are referred to this thesis. However a brief
description of the experimentation is as follows.
Fig. 2 depicts a schematic diagram of the experimental
9000 6000
equipment. The test rig has been used to measure heat
III
transfer coefficient under pool boiling conditions for
7000 3000 various concentrations and heat fluxes. The fully integrated
q& = 300 kW / m 2
C = 1.2 g / L
experimental rig consists of a boiling vessel with a
5000 0 resistance band heater [1], an enclosed flow loop including
0 200 400 600 800 1000 1200 1400 1600 a co-current condenser [2], a preheater [3], rod heater with
Time, [min] power control units [4] and a data acquisition system [5]
Figure 1. Typical variation of heat transfer coefficient as a connected to a computer [6].
function of time (Malayeri et al, 2005). In this figure, solid The rod heater is mounted horizontally in the centre of
and open symbols correspond to complete run and initial vessel and is manufactured by Ashland Chemical Company
period of 20 min, respectively.

http://dc.engconfintl.org/heatexchanger2005/37
Malayeri and Müller-Steinhagen: 257

to an HTRI design. Four type E thermocouples are along with corresponding surface temperatures for each heat
embedded inside the heater to measure the wall temperature. flux.
The heated zone is 99.1 mm long and the heat flux of the Table 1. Range of operating parameters.
test heater can be varied up to 300 kW/m2. Three of the
embedded thermocouples are used for data acquisition and CaSO4 Concentration, [ g/L ]
the fourth is used as a trip input for the test heater Variac 1.6 1.2 0.8
power supply, should the internal temperature of the heater 55 (104.3oC) +

Heat flux, [kW/m2]

exceed a set limit.
76 (106.6oC) +
105 (107.4oC) + +

(Ts, oC)
Safety valve P Pressure gauge

145 (109.2oC) + + +
200 (110.5oC)
Cooling water, in

+ +
Water vacuum
Band heater
300 (114.1oC) + + +
[6]

[1] RESULTS AND DISCUSSION

Condenser

[2] Data recorder (D) Effect of heat flux

Heater A
A [4]
The variation of heat transfer coefficient during the
[5] initial period for different heat fluxes and a constant salt
concentration of 1.6 g/L is illustrated in Fig. 3. All five data
sets exhibit the same trend as already shown in Fig. 1, i.e. a
Cooling water, out
rapid decrease in heat transfer coefficient, followed by a
Preheater [3]
recovery to a plateau. Apparently in the initial period, heat
Drain transfer coefficients depend strongly on the heat flux.
Inlet/sample line
However Malayeri et al. (2005) showed in the decay period
Figure 2. Schematic representation of test rig. in which heat transfer coefficient reaches its asymptotic
value, this dependency on heat flux declines. For the initial
The average heat transfer coefficient is determined period, it can thus be hypothesised that:
using: 1. The difference between minimum heat transfer
coefficient (end of region I) and clean value (zero time)
q& widens as heat flux increases.
α= (1)
(Ts − Tb ) 2. The time in which the heat transfer coefficient reaches
the minimum value decreases with increasing heat flux
as shown by the solid line in Fig. 3. For instance, for 76
and the fouling resistance due to deposition of a fouling kW/m2, it is about 27 min, while it is only about 1-2 min
layer is then calculated as a function of time from: for 300 kW/m2. This trend is consistent with
observations by Palen and Westwater (1966) and
 T − Tb   Ts − Tb  Jamialahmadi et al. (1989). The effect of heat flux may
R f = s  −   (2) be explained by analysing the mechanisms of bubble
 q&  t  q& 0 nucleation.
The increase of heat flux may lead to firstly the
activation of smaller nucleation sites and secondly break up
In this equation, Ts and Tb are surface and bulk of generated bubbles. Nucleate pool boiling can only occur
temperatures, respectively. The subscripts “t” and “0” when the heated surface contains very small pits and cracks
denote conditions at any time and at the beginning of the known as “cavities”, that are required for bubble nucleation.
experiment when the heat exchanger is considered to be A cavity can ideally be imagined as being conical and
clean, respectively. having a circular opening. For non-fouling pool boiling on
Test solutions of calcium sulphate were prepared by clean metal surfaces, Whalley (1996) proposed the
dissolving calcium sulphate hemi-hydrate (CaSO4.½H2O) in following equation for the prediction of the cavity radius:
a separate container. The saturation concentration of
CaSO4.½H2O at 100°C is 1.6 g/L. The range of operating 2σ Tsat
R= (3)
parameters used in this investigation is given in Table 1 ρ g ∆hv ∆Tsup erheat

Produced by The Berkeley Electronic Press, 2016

258 Heat Exchanger Fouling and Cleaning - Challenges and Opportunities [2005], Vol. RP2, Article 37

In this equation, R is the cavity radius, σ the surface bubble departure diameter increased from 2.1 mm to 3.2
tension, ρg the density of the vapour and ∆hv the latent heat mm. Therefore, the increasing CaSO4 concentration results
of evaporation. The equation indicates that as the wall in lower heat transfer coefficients due to changes in the
superheat increases, more cavities of smaller radius are able mechanisms of bubble formation associated with the
to become active, giving rise to a higher heat transfer rate. changing physical properties of the solution, particularly
The same mechanism applies at the start of the fouling surface tension.
process when both, degree of wall superheat and number of Fig. 4 illustrates the deposition-related changes in heat
nucleation sites, dominate the development of the additional transfer coefficient as a function time for various
thermal resistance. concentrations, at a constant heat flux of 300 kW/m2. It is
The second impact of heat flux is the break-up of big evident that heat transfer coefficients decrease more rapidly
bubbles into much more smaller bubbles. At high heat as the CaSO4 concentration was increased. This is attributed
fluxes (here 300 kW/m2 which, at 1 bar, corresponds to a to the increased supersaturation of the salt solution at the
surface temperature of 114.1°C) bubble break-up occurs, triple interface of vapour, liquid and heat transfer surface.
giving rise to small and almost undistinguishable individual The measurements also confirmed the observations of
rigid bubbles. Formation of these small bubbles and their Najibi et al. (1996), i.e. that an increase in CaSO4
subsequent release into the surrounding liquid enhances concentration results in bigger bubbles. As it was observed
convection currents around the heater and as a result the for flow boiling at low heat fluxes, the gap between
heat transfer coefficient increases. minimum and maximum heat transfer coefficients decreases
as the concentration is reduced.
16000
Symbol Heat flux, kW/m2 18000
55
C=1.6 g/L
Symbol Concentration, g/L
14000 76 1.6
Heat transfer coefficient, W/m2.K

145 1.2
16000
12000 200 0.8
Heat transfer coefficient, W/m2 K

300
10000 14000

8000
12000
6000

10000
4000

2000 8000

0 Heat flux = 300 kW/m2

0 10 20 30 40 50 6000
0 10 20 30 40 50
Initial time, min
Initial time, min
Figure 3. Variation of heat transfer coefficient for a given Figure 4. Heat transfer coefficient as a function of time for
concentration and different heat fluxes in the initial period a given heat flux and different concentrations.
(region I and II in Fig. 1).
Fig. 5 shows the variation of time to reach minimum
Effect of CaSO4 Concentration and maximum heat transfer coefficients as a function of heat
An analysis of CaSO4 concentration effects under clean flux, for a constant CaSO4 concentration of 1.6 g/L. In this
conditions would primarily shed light into heat transfer figure, minimum and maximum points represent the ends of
mechanisms as well as fouling mechanisms. Najibi et al. regions I and II in Fig. 1, respectively. Two distinctive
(1996) investigated the effect of CaSO4 concentration on trends are clearly conspicuous. Firstly, both minimum and
boiling heat transfer from clean surfaces. They observed a maximum values decrease continuously as the heat flux
relatively uniform bubble diameter of around 2 mm and a increases. Secondly, the gap between these two times
low bubble departure frequency during saturated pool expands with decreasing heat flux.
boiling of distilled water for low heat fluxes (38 kW/m2 Experimental data for the time over which the heat
which corresponds to a surface temperature of 103.7°C at 1 transfer coefficient reaches its minimum value are
bar). As the concentration of salts with inverse solubility illustrated in Fig. 6 for different salt concentrations. For
(such as CaSO4) was increased from 0.253 to 1.5 g/L, the both concentrations, similar trends as those in Fig. 5
number of active bubble nucleation sites decreased and the indicate that as concentration decreases, the time to reach

http://dc.engconfintl.org/heatexchanger2005/37
Malayeri and Müller-Steinhagen: 259

the minimum heat transfer coefficient increases for all heat been investigated and correlated. Not surprisingly, the main
fluxes. reason for this was attributed to the complexity of the
fouling behaviour during the initial period. Most recently,
50 Jamialahmadi and Müller-Steinhagen (2004) published a
45
C = 1.6 g/L new phenomenological model for the prediction of fouling
Time to reach min. and max. α, min

resistance of calcium sulphate solutions during boiling heat

40
transfer for region III in Fig. 1. In this model, the heat
35 transfer coefficient is expressed as:
Time to reach maximum α
30
α0
25
α III = (4)
20 1 + φ (ηC − C sat ) 2 (1 − e − kd (t −t0 ) )
b
s

15
where α0 is the “clean” heat transfer coefficient that could
10 be predicted by Gorenflo’s correlation (2001). Also, in this
5 Time to reach minimum α equation:

0
1
0 50 100 150 200 250 300 350 φ= (5a)
Heat flux, kW/m2 90.2 + 5.7 × 10 −7 q&
Figure 5. Time to reach minimum and maximum heat
transfer coefficients as a function of heat flux for constant η = exp (2.685 × 10 −6 q& ) (5b)
CaSO4 concentration.
−33650 / R g Ts
k d = 1.1 × 10 3 e (5c)
30
Symbol Concentration [g/L]
−124287 / R g Ts
1.6 t 0 = 0.731 × 10 −17 / C b e (5d)
Time to reach minimum α, min

25
1.2
C sb , Csat and t are bulk and saturation concentrations, and
20
operating time, respectively. Eq. (5d) provides an estimate
of the initial time until the beginning of region III.
15 Jamialahmadi and Müller-Steinhagen (2004) reported that
Eq. (4) could predict their experimental results with an error
10 of about 6.5%. However Eq. (4), as stated before, is only
limited to region III, and one should endeavour to correlate
data within the initial period, as well.
5
Recent advances in function approximation methods
have helped researchers to better analyse and predict
0 objective functions for highly complex processes such as
0 100 200 300 400 fouling (Malayeri et al. 2001, 2003). One such technique is
Heat flux, kW/m2 the artificial neural network architecture which may be
Figure 6. Time to reach minimum heat transfer coefficient utilised for analysing processes in which interaction
as a function of heat flux, for different CaSO4 between independent and dependent variables is poorly
concentrations. understood. The present authors have used neural networks
for the prediction of fouling behaviour under subcooled
CORRELATION OF DATA flow boiling conditions and also for CaSO4 precipitation in
Phenomenological correlations for the prediction of cooling towers (Malayeri and Müller-Steinhagen, 2001 &
fouling behaviour under pool boiling heat transfer are rare 2003). In addition, they utilised the same techniques for the
and a review of existing models was given by Jamialahmadi prediction of void fraction data in two-phase flows
and Müller-Steinhagen (1993). They compared their (Malayeri et al., 2003). In this study, the artificial neural
experimental results with previous models but the predictive network architecture is utilised again due to its:
accuracy was clearly not satisfactory. Moreover, the 1) ability to interpolate within the range of introduced
prediction of the fouling behaviour was limited to region III experimental results. Malayeri et al (2003) used the
in Fig. 1, while regions I and II (initial period) have not same technique for extrapolation. However this is

Produced by The Berkeley Electronic Press, 2016

260 Heat Exchanger Fouling and Cleaning - Challenges and Opportunities [2005], Vol. RP2, Article 37

not possible in this study due to limited number of to these constraints, further process information is available
experimental results that are just enough for the to improve the outcome of the resulting network, such as
training phase. that shown in Figs. 5 and 6:
2) capability, as will be discussed later, to adopt process
prior knowledge as a tool to make sure that the
∂t
resultant network complies with process constraints. <0 (8)
This could be considered as an advantage for ∂q& α min
possible intelligent heat exchanger monitoring which
cannot be possible by conventional regression ∂t
models (Thompson and Kramer, 1994). <0 (9)
The construction of a neural network in its most basic ∂q& α max
form involves three entirely different layers. The input
layer is made of input nodes. The layer between the input All these constraints will be introduced into the network as
and output layers is known as hidden layer, which may “Prior Knowledge (PK)”, and the network output should
consist of only one or of several sub-layers. In some neural primarily comply with these constraints in order to obtain
networks such as Radial Basis Functions (RBF), only one the correct qualitative results. The available experimental
hidden layer exists, where in contrast to the multi-layer data are usually divided into two sets, for the training and
perceptron, the transformation from input space to the generalisation phases respectively. In the present case,
hidden layer space is non-linear, while the transformation however, all experimental results were used for training
from the hidden layer space to the output space is linear. because only few data sets have been available the initial
The output layer supplies the response of the network to the deposition period. All input data have been normalised
activation patterns applied to the input layer. A detailed before introducing them to the network. The resulting
introduction into the foundations of the radial basis function network was validated by comparison of the network output
(RBF) is given by Haykin (1999). with those experimental results which have been used in the
In the present study, for correlating the data, the radial training phase. The principle of the modelling can be
basis function neural network architecture is investigated. described in more detail with the help of the flow chart in
Three input dimensionless groups ( q& / q& arb , Cb/Csat, t/t0) Fig. 7.
are defined as vector x and one output (heat transfer
coefficient) as vector y. q& arb is an arbitrary value of heat
flux in order to make the input which is related to heat flux Inputs (xi)
dimensionless and hence here is taken as the maximum
investigated heat flux, which was 300kW/m2.
The network consists of three layers, which are input
layer, hidden layer and output layer. The number of nodes Normalization
of the input layer is equal to the dimension of the input
vector x, and here is equal to three. The hidden layer is
composed of non-linear units that are connected directly to
all of the nodes in the input layer. There is one hidden unit
Neural
for each data point xi=1 to N, where N is the number of Network
experimental data. The output layer consists of a single
linear unit and its output is the heat transfer coefficient. In
order to protect the network from under- or over-fitting the PK
following constraints are applied:

@ t=0 α = α0 (6) Satisfaction index

@ t ≥ t0 α = αIII (7)

Heat transfer Coefficient (α)

Here, α0 and αIII are determined by the Gorenflo
correlation (2001) in conjunction with the model proposed Figure 7. Explanation of the neural network modelling in
by Jamialahmadi and Müller-Steinhagen (2004). In addition stepwise layout.

http://dc.engconfintl.org/heatexchanger2005/37
Malayeri and Müller-Steinhagen: 261

Fig. 8 illustrates a comparison of the network output against such as surface modification (Müller-Steinhagen and
the experimental data shown in Fig. 3. The absolute mean Zhao, 1997). It is the time in which the first crystalline
average error is 8.7 % for regions I and II and the maximum nuclei become active on the heat transfer surface,
error of 23.4%. Even though the accuracy of the established whereas in the decay period secondary nuclei are in
network is good, one can see slightly under- and over- contact with the deposit layer.
prediction after the minimum heat transfer coefficient. The
initial under-estimation is partly related to errors in the • The time until the heat transfer coefficient reaches its
intermediate minimum and maximum values decreases
prediction of the clean heat transfer coefficient by the
with increasing heat flux and salt concentration, but
Gorenflo correlation (2001). It is also noticeable that the
accuracy of the resulting network increases for higher heat more dominantly for higher heat fluxes. For any given
concentration, the time span between reaching the
fluxes, notably for 200 kW/m2 and above.
minimum and maximum heat transfer coefficients
widens as heat flux decreases.
16000
Symbol Heat flux, kW/m2
55
C=1.6 g/L • The use of artificial neural networks has helped to
14000 145 correlate the experimental results with good accuracy,
Heat transfer coefficient, W/m2.K

200 even though more reliable data would further improve

12000 300
the capability of the network.
10000
• As a point of precaution, the authors would like to
8000
emphasize that the present results are only indicative and
that further investigations are required, particularly with
6000 respect to the influence of surface finish.

4000
NOMENCLATURE
2000 C concentration, g L-1
kd deactivation rate constant, s-1
0 q& heat flux, W m-2
0 10 20 30 40 50
R cavity radius, m
Initial time, min
Rg universal gas constant, Jmol-1 K-1
Rf fouling resistance, m2 K W-1
Figure 8. Comparison of network output with experimental
T temperature, K
data.
t time, s
t0 transition time until the beginning of region
It should also be pointed out that noisy data have not III, s
been removed from the data sets and that the effect of other
Greek Letters
parameters that may improve the performance of the
α heat transfer coefficient, Wm-2 K-1
network, such as surface finish in terms of surface
roughness, were not considered. Surface roughness could ∆hv latent heat of evaporation, J/kg
not be considered as an input because it was assumed to be ∆Tsuperheat degree of wall superheat = Ts-Tsat, K
constant for all experimental runs. However, for φ defined by Eq. (5a)
investigations with different surface finishes, this parameter η defined by Eq. (5b)
will definitely have to be included as an input vector. ρg vapour density inside bubble nucleus, kg m-3
Obviously, the quality of the network will improve if more σ surface tension, N m-1
reliable information is available. Subscripts and Superscripts
0 zero time or clean conditions
b bulk
CONCLUSIONS s surface
The principle conclusions that can be made by sat saturation
considering the results of this investigation are:
• The rather short initial period in which significant ACKNOWLEDGMENTS
changes in heat transfer coefficient occurs is decisive for The authors are grateful to Mr T.H. Bartlett whose
investigation of novel fouling mitigation techniques experimental data were used in this study.

Produced by The Berkeley Electronic Press, 2016

262 Heat Exchanger Fouling and Cleaning - Challenges and Opportunities [2005], Vol. RP2, Article 37

REFERENCES Malayeri, M.R., H. Müller-Steinhagen, and J.M. Smith,

Bartlett, T.H., 1995, Fouling of boiler tubes, MSc 2003, Neural network analysis of void fraction in air/water
thesis, University of Auckland, New Zealand two-phase flows at elevated temperatures, Chem. Eng. &
Bornhorst, A., Zhao, Q. and Müller-Steinhagen, H., Processing, Vol. 42, No. 8/9, pp. 587-597.
1999, Reduction of Scale Formation by Ion Implantation and Malayeri, M.R., Müller-Steinhagen, H. and Bartlett,
Magnetron Sputtering on Heat Transfer Surfaces. Heat T.H., 2005, Fouling of tube bundles under pool boiling
Transfer Eng., 20 (2), 6-14. conditions, Chem. Eng. Sci., Vol. 60, No.6, pp. 1503-1513.
Freeborn, J. and Lewis, D., 1962, Initiation of boiler Müller-Steinhagen, H. and Zhao, Q., 1997,
scale formation. J. Mech. Sci., 4, 46-52. Investigation of low fouling surface alloys made by ion
Gorenflo, D., 2001, State of art in pool boiling heat implantation technology. Chem. Eng. Sci., 52, 3321-3332.
transfer of new refrigerants, Int. J. Refrig., 24, 6-14. Najibi, S.H., Müller-Steinhagen, H. and Jamialahmadi,
Haykin, S.S., 1999, Neural networks: a comprehensive M., 1996, Boiling and non-boiling heat transfer to
foundation, 2nd edition, Prentice Hall, London, UK. electrolyte solutions. Heat Transfer Eng., 17 (4), 46-63.
Hospeti, N.B. and Mesler, R.B., 1965, Deposits formed Palen, J.W. and Westwater, J.W., 1966, Heat transfer
beneath bubbles during nucleate boiling of radioactive and fouling rates during pool boiling of calcium sulphate
calcium sulphate solutions. AIChE J, 11, 662-665. solutions. Chem. Eng. Prog. Symp. Series, 62(64), 77-86.
Jamialahmadi, M., Blöchl, R., and Müller-Steinhagen, Partridge, E.P. and White, A.H., 1929, Mechanism of
H., 1989, Bubble dynamics and scale formation during formation of calcium sulphate boiler scale. Ind. Eng. Chem.,
nucleate boiling of aqueous calcium sulphate solutions. 21, 834-838.
Chem. Eng. Process., 26, 15-26. Thompson M.I. and Kramer, M.A., 1994, Modelling
Jamialahmadi, M. and Müller-Steinhagen, H., 1993, chemical processes using prior knowledge and neural
Scale formation during nucleate boiling - A review. network, AIChE J., Vol. 40, pp. 1328-1340.
Corrosion Reviews, 11 (1&2), 25-54. Whalley, P.B., 1996, Two-phase flow and heat transfer.
Jamialahmadi, M. and Müller-Steinhagen, H., 2004, A Oxford University Press, NY, USA.
new model for the effect of calcium sulfate scale formation
on pool boiling heat transfer. ASME J. Heat Transfer, 126,
507-517.
Malayeri, M.R., and Müller-Steinhagen, H., 2001,
Neural network analysis of heat transfer fouling data, The
4th United Engineering Foundation Conference on Heat
Exchanger Fouling: Fundamental Approaches & Technical
Solutions, Davos, Switzerland, pp. 145-150.

http://dc.engconfintl.org/heatexchanger2005/37