5th European Thermal-Sciences Conference, The Netherlands, 2008

EXPERIMENTAL INVESTIGATION OF FROST

FORMATION ON COLD TUBES

A. Schydlo, P. Stephan
Technische Universitt Darmstadt, Darmstadt, Germany

Abstract
This paper presents results of an experimental investigation of frost formation on cooled tube heat
exchanger surfaces under cross air flow. Frost formation on cold surfaces occurs when humid air is
in contact with cooled surfaces at temperatures below 0C. The frost structure is porous and
consists of ice crystals, solid ice and humid air. An analysis of the frost growth process and the frost
properties for variable air and tube parameters is presented. Frost formation on tube surfaces with a
diameter of 15.0 mm and a length of 85.0 mm was studied for varying air temperatures and air
humidities, tube temperatures, Reynolds numbers of the cross air flow, tube surface roughness and
geometry of the ducts.

Nomenclature
a - reduced geometrical ratio, A, B ,C - function coefficients
d - tube diameter, m
l - reduced geometrical dimension for tube bundle, m
R - roughness factor, Ra - roughness, m
Rd - reduced Number,RH - relative humidity, %
- reduced geometry, -

s - transverse pitch, m
x - mixing ratio, kg/kg
X - exponential coefficients

- frost
K - condensate
L - latent
MA - moist air
S - sensible
T - total

- modified parameter
F

1 Introduction
Frost forms when moist air comes into contact with cooled surfaces with temperatures below the
freezing point of condensed water. Frost is a porous structure composed of ice crystals and humid
air in a saturated condition. The thermal properties of frost, like its thermal conductivity, are
dependent on the local density. When the temperature reaches the freezing point, a melting and
refreezing cycle occurs on a surface covered with frost and moist air. This recurrent cycle changes
the frost structure, increases its density and enhances the thermal conductivity. This process is

5th European Thermal-Sciences Conference, The Netherlands, 2008

known as melt beck phenomenon in the literature and depends on the mass flow of condensate. In
industrial applications, the phenomenon of frost formation is undesirable, as it affects and worsens
the heat transfer and air mass flow rate, particularly in refrigerating machines (e.g. heat
exchangers). The effect of frost formation on cold surfaces has been studied for many years, and
many publications have concentrated on the experimental investigation of frost growing under free
and forced convection and other varying air and cooled surface parameters. Yonko and Sepsy
(1967), stin and Anderson (1991), Lee and Ro (2002), Cheng and Shiu (2002), Beer and Ler
(2002), Fossa and Tanda (2002), Langeheinecke (2005) presented results of experimental work on a
single flat plate or on parallel plates. Similarly, a few papers have presented outcomes of frost
formation on tube geometry; Lee and Ro (2001) published results of frost formation on a horizontal
cylinder during cross air flow. The temperature distribution, the thickness of the frost layer, and the
temperature distribution in the cylinder were measured, and the local heat flux and the effective
thermal conductivity of the frost layer were calculated. This experiment was done for varying
temperatures, air velocities and humidities. Mago and Sherif (2003) presented a semi empirical
model describing heat and mass transfer on a cylindrical surface in humid air cross flow under a
supersaturated frosting condition. Though many papers have been published on frost formation,
there are very few experimental investigations that parametrically study frost formation on
cylindrical geometry. Generally speaking, the experimental investigations afford better
understanding of this process.
This work presents a parametrical study of frost formation on cold tubes. Air parameters, including
temperature ( Tg ) , relative humidity ( RH ) , Reynolds number ( Re ) (or velocity ( v ) ) in the duct
section, the temperature of the cooled tube ( Tw ) and the geometry (height) of the duct, were varied.
Additionally, different roughnesses and structures of tube surfaces were tested. Error analysis is
used to correlate and validate the experimental results. All experimental conditions (with standard
parameters in bold type) are shown in table 1.
Table 1: Experimental conditions, RH relative humidity, Tg air temperature at test section
inlet, Tw temperature of the tube surface, u /Re air velocity/Reynolds number
at test section inlet, C1 plain copper tube surfaces Ra = 0.04m, C2- copper tube
roughness Ra = 1.89 m, C3 - copper tube - roughness Ra =5.17 m, L1 Lotus
hydrophobic surface contact angle (in drop) >160, s - transverse pitch (height
of duct) .
RH
Tg

%
K

Tw
u
(Re )
Ra

K
m/s
-

s (nominal)

mm

30

40

50

70
80
60
295.15 298.15 301.15

266.15
0.28
(700)

263.15 260.15 257.15

0.35
0.39
0.32
(800) (900) (1000)
C2
C3
C1
24

34

0.43
(1100)
L1

44

2 Experimental Setup
Figure 1 shows the experimental setup used in this investigation. The experimental construction
consists of a long square duct with cross section: 114.0 mm 24.0 mm, where a laminar velocity
profile can be established. In a test section of the duct, a measuring unit (cooled tube) is installed.
The tube is made of copper, with an outer diameter of 15.0 mm and a length of 85.0 mm. It is
cooled using 34.0/66.0 Antifrogen N/Water solution. The air is aspirated to the air preparation
section by an aspirator. The section consists of a thermostat and a resistance heater to adjust the air

5th European Thermal-Sciences Conference, The Netherlands, 2008

temperature and the humidity, and the adjusted air is directed to the test section via the duct. Water
steam from the humid air condenses on the cooled tube. The water drops solidify on the tube when
the temperature of the cooled surface drops below the freezing point. The coolant from the
refrigerator unit flows into the internal tube with the smaller diameter and flows out through the
annular free space between the internal and external tubes. This technical solution reduces and
minimizes the longitudinal temperature gradient on the tube surface. The test section is made of a
transparent Plexiglas surface. The aluminum duct as well as the test section is thermally isolated
with Armaflex isolation. The temperature of the tube surfaces is measured with a calibrated K
thermocouple (CrNi-Cr). A maximal deviation for the standard thermocouple Pt100 is 0.08K. For
measurements of humidity and temperature of moist air, three transmitters Vaisala HUMICAP
Series HMT330 were adopted. The high accuracy for humidity sensors after the DKD calibration is
smaller than 0.5%RH. A thermocouple (Pt100 PRT DIN IEC 751 class B) integrated into the
humidity sensor has an accuracy of 0.1K. For measurement of the air flow, a thermal flowmeter
Endress+Hauser FL-11427 AT-700F with an adjusted high accuracy 0.63% is used. For the
identification of a frost thickness profile and a frost volume, a non-contact visual optic method is
used. Every 30 seconds during a one hour experiment, the measured data are assembled and
archived on a PC computer.

resistance heating
inlet

refrigerator unit
PC Computer
temperature and
humidity sensors

outlet

IMP

1
aspirator

measuring unit

test section

duct (Aluminium)

(cooled tube)

thermostat
thermal flowmeter

Figure 1: Experimental setup.

3 Post processing of measured value for selected standard

parameters
During the post processing, the following values are computed in Matlab: mass and mass flow of
condensed water steam, a total heat flux, a frost volume, a frost average density and a frost growth
ratio on the tube surface. Furthermore, the optical method allows the calculation and presentation of
the frost thickness rate on the tube surface in time intervals of every 30.0 seconds over an
experimental period of one hour. Each experiment was repeated between 3.0 to 5.0 times. The
results will be presented on the diagrams.

3.1 Measurement of mass and mass flow of condensed steam water

The mass flow of condensed water vapor (M& K ) is calculated from the following mathematical
formula (1):

5th European Thermal-Sciences Conference, The Netherlands, 2008

(x x )
M& K = M& MA 1 3
1 + x3

(1)

To calculate the mass and the mass flow of the condensate, the measured values from the 1st and 3rd
humidity and temperature sensors are used. In the place where the 2nd sensor is installed, the profile
of the humidity air mixing ratio is not homogenous and regular. Measures of the mixing ratio in the
2nd position give incorrect and overestimated values. A result of mass flow rate is shown in figure 2.

Figure 2: Mass flow of condensate vs. time for the standard parameters.
From the mass flow curve, qualitative and quantitative statements can be made about the
condensation process of water vapor. In the first minutes of the experiment, the mass flow rate
reaches the first local maximum value. The solidification process of the water drops and the first
phase of frost formation occur for the maximal mass flow of condensate.
It lasts until the moment when the frost reaches a critical density, which is caused by compaction
phenomena in the porous structure. The thermal conductivity for the frost structure is smaller than
for the liquid or solidified water, reducing the mass flow of the condensate on the frost structure.

3.2 Measurement of total heat flux

The total heat flux consists of a latent and a sensible heat flux. The following well-known formulas
are used for the computation of both heat fluxes (2):
Q&T = Q& L + Q& S

(2)

Q& L = M& K h
Q& S = M& MA (h1 h2 )

(3)
(4)

where:

In figure 3 the sum of both heat fluxes are presented. Generally speaking, the distribution of the
heat flux value on the diagram is proportional on the distribution of mass flow of condensate.

5th European Thermal-Sciences Conference, The Netherlands, 2008

Figure 3: Total heat flux vs. time for the standard parameters.

3.3 Measurement of frost volume

Using a non-contact visual optic method, the frost volume is calculated. A CCD camera installed
parallel to the tube allows observation of the frost structure and profile on the front wall of the duct.
The frost thickness on the tube surface is constant in a longitudinal profile. In order to avoid or
minimize the fringe effect of the frost formation on the cooled surfaces, a Teflon cap was
installed. In figure 4, the basic image of the frost profile on the tube is shown.

Figure 4: Basic image of frost profile

air inlet right.

Figure 5: Black white image of

frost layer.

This image was captured after 50.0 minutes from the beginning of the experiment. The post
processing in Matlab corrects a depth of sharpness and grey scale of this image. Finally, the frost
area on the tube in the cross section is calculated from the image after the post processing.
Therefore, the frost volume (figure 5) is a product of the frost area in the cross section and the
length of the frost structure on the tube. Figure 6 presents the frost volume growth on the tube, and
demonstrates a parabolic progression.

5th European Thermal-Sciences Conference, The Netherlands, 2008

Figure 6: Volume of the frost structure vs. time for the standard parameters.

3.4 Measurement of frost density

Using the non - contact optical method to measure the frost thickness and, correspondingly, the
frost volume, only the average frost density can be calculated. It is impossible to estimate the local
frost density. The frost structure contains air bubbles and therefore, has a lower density than the
density of ice. In figure 7, two arbitrary guide lines show the density of liquid water and ice.

Figure 7: Average frost density vs. time for the standard parameters.

4 An example of a correlation equation of the average frost

density
In order to make practical use of the experimental results, correlation equations were prepared. The
results were processed by means of the reduced non-dimensional characteristic number Rd. The
well-known modified Reynolds number and Stefan number were used. For example, the mass of
condensed water vapour is dependent on 8 input values. This 9-D matrix should be reduced to a 2-D
or 3-D matrix. In this case, the mathematical dependence can be described with a line or a surface.

5th European Thermal-Sciences Conference, The Netherlands, 2008

Using the least squares method, the mathematical formula describing the 3-D surface is prepared.
The 3-D diagram reduces to a 2-D graphical dependence, and can be used in practical applications
(figure 8).

Figure 8: Contour diagram of frost density for reduced parameters.

For more precise calculations, the following formula is proposed:

F = A + Rd t + B t 2 tanh(Rd ) + C

(5)

Rd = Re Ste RH R( 0.5 ) E 2

(6)

where:

Re =

u ( g ) l

( g )

s = s X 4 , and d = d X 3
l =

(8)

(9)

s
a =
d

X1

= 1

4a

Ste =

(7)

(10)
X2

(11)

c p ( g ) (Tg Tw )

and:
X 1 = 0.4088, X 2 =2.281, X 3 =0.9414, X 4 =0.89

A = 4.3773E+01, B = -4.9503E+01, C = 9.0502E+01, t 0.11 1 = 60 min.

(12)

5th European Thermal-Sciences Conference, The Netherlands, 2008

The Equation (5) is only available for maximal value of Rd = 8.6. For the verification of the
influence of the tube diameter on the equation (5) for calculating the frost density, additionally the
experiment on 18 mm tube for standard parameters in duct with height s = 44 mm was realised.
Following parameters in the correlation was used: RH 30% 80%, TL 295.15K 301.15K,
TR 257.15K 266.15K, u 0.28 m/s 0.43 m/s, s1 24mm 44mm, da 15mm
18mm, R1 1.9.
The dependences of coefficients R on the roughness Ra are given in table 2:
Table 2: Roughness of the tube and R coefficients.
Ra , m
R,-

0.03 (polished)
1

1.89
1.3

5.17
1.9

The mean standard deviation is smaller than 11.0%.

5 Summary
This publication is one of a few to report on an experimental investigation to describe frost
formation on cold tubes. A full experimental study was performed to better understand the influence
of varying parameters. Moreover, a correlation equation was presented for potential use in industrial
applications.
Maximal air velocity, air humidity, air temperature, surface temperature and roughness are the main
parameters that have influence on the frost porous structure. The experimental results and their
accuracy are satisfactory.

6 References
Beer, H. and Ler, A., 2002, Frost deposition in a parallel plate channel under laminar flow
conditions, Int. J.Therm.Sci., 85-95.
Cheng, C.H. and Shiu, C.S., 2002, Frost formation and frost crystal growth on a cold plate in
atmospheric air flow, Int. Journal of Heat and Mass Transfer, 4289-4303.
Fossa, M. and Tanda, G., 2002, Study of frost formation on a vertical plate, Exp. Thermal and Fluid
Science, 661- 668.
Hosoda, T. and Uzuhashi, H., 1967,Effects of frost on the heat transfer coefficient, Hitachi Review,
16(6).
Langeheinecke, K., 2005, Auswirkungen der Keimbildungsphase auf Reifbildungsvorgnge an
gekhlten und berstrmten ebenen Plate, Ph.D. Thesis, TU-Darmstadt.
Lee, Y.B. and Ro, S.T., 2002, Frost formation on vertical plate in simultaneously developing
flow,Exp. Thermal and Fluid Science, 939-945.
Lee, Y.B. and Ro, S.T., 2001, An experimental study of frost formation an a horizontal cylinder
under cross flow, Int. Journal of Refrigeration, 468-474.
Mago, P.J. and Sherif, S.A., 2003, Heat and mass transfer on a cylinder surface in cross flow under
supersaturated frosting conditions, Int. Journal of Refrigeration, 889-899.
stin, R. and Anderson, S., 1991, Frost growth parameters in a forced air stream, Int. J. Heat Mass
Transfer, 1009- 1017.
Sanders, C. Th., 1974, The influence of frost formation and defrosting on the performance of air
coolers, Ph.D. Thesis, Technische Hogeschool Delft.
Yonko, J.D. and Sepsy, C.F., 1967, An investigation of the thermal conductivity of frost while
forming on a flat horizontal plate ASHRE Trans.