Department of Chemical and Process

Engineering
Faculty of Engineering
University of Peradeniya
Sri Lanka

CP520-MODELLING AND SIMULATION OF

SIMULTANIOUS TRANSPORT PHENOMENA
CONVECTIVE DRYING OF FOOD

K.A. Lakshan
E/16/202
Semester 6
From June to September 2021
1. Introduction
Drying of foods is an essential process to preserve food for a longer time. The term drying can be
defined as the removal of moisture inside the food and there are many methods to carry out this process. Most
of these processes can be classified into two major groups. They are natural convection and forced convection.
For the above-mentioned topic, external forced convection will be considered and a banana slice will be
considered as the food material.
1.1 Dimensions of the food material.
The selected banana slice is considered to be a perfect cylinder with a radius of 15 mm and a thickness
of 4 mm. A sketch of the banana slice is shown in Figure 1.

30 mm
Figure 1- Sketch of the banana slice
1.2 Properties of the material.
The properties of banana are shown in table 1. These properties are entered to a blank material to
represent the banana.

Table 1- Properties of banana

Property Value
Density of banana (dry basis) 980 kg/m3
Initial moisture content 4 kg/kg dry banana
Equilibrium moisture content 0.29 kg/kg dry banana
Density of water 994.59 kg/m3

2. Mathematical model and simulation using Multiphysics platform.

2.1 Translation of model into the software.
The problem was transferred into COMSOL Multiphysics modelling environment using the 2D
axisymmetric model builder. A 2D model was built due to the low availability of computational power.
Selected physics interfaces were heat transfer in solids and the transport of diluted species. Then the food
specimen was simply created using a rectangle with a width of 15 mm and a height of 4 mm. Figure 2 shows
the model which was developed.

Direction of air flow

15 mm
Figure 2- 2D model
2.1.1 Heat transfer in banana.
The governing equation for the heat transfer of solid banana is as following. Consider the control volume in
Figure 3.
 (x+dx,y+dy,z+dy)

T(x,y,z)

(x,y,z)
Figure 3- Considered control area

𝝏𝑻
𝝆𝑪𝒑 + 𝝆𝑪𝒑 𝒖 ∙ 𝛁𝑻 + 𝛁 ∙ (−𝒌𝛁𝑻) = 𝑸 1
𝝏𝒕

ρ – Density of solid
Cp – Specific heat capacity of solid.
T – Temperature
u – Velocity of moisture
k - Thermal conductivity of material
Q – Energy generated within the control volume.

𝜌𝐶 term from equation 1 stands for the energy accumulation. The second term, which is 𝜌𝐶 𝑢 ∙ ∇𝑇
stands for the energy transfer due to convection within the banana slice and the third term ∇ ∙ (−𝑘∇𝑇) stands
for the conductive heat transfer through the control volume.
The terms in the right-hand side of equation 1 is taken as zero since no energy generation takes place
in the control volume. And also, the movement of moisture is not considered in the model. Therefore, Equation
1 can be modified into equation 2 for 2D heat transfer.
𝝏𝑻 𝝏𝟐 (𝒌𝑻) 𝝏𝟐 (𝒌𝑻)
𝝆𝑪𝒑 = + 2
𝝏𝒕 𝝏𝒙𝟐 𝝏𝒚𝟐
Cp and k are originally defined considering the mass of the moisture does not change. Therefore, it creates an
error in the simulation and those terms are manually defined as variables which depend on moisture content.
The software is able to calculate moisture concentration (c) and C p and k are defined from that.
𝒄 ×𝟎.𝟎𝟏𝟖
𝑴𝒘 = 3
𝟗𝟖𝟎 ×𝟎.𝟐

𝒌 = 𝟎. 𝟎𝟎𝟔 × 𝑴𝒘 + 𝟎. 𝟏𝟐𝟎 4

𝑪𝒑 = 𝟎. 𝟖𝟏𝟏 × 𝑴𝒘𝟐 − 𝟐𝟒. 𝟕𝟓 × 𝑴𝒘 + 𝟏𝟕𝟒𝟐 5

Equation 4 and 5 have been taken from previously published studies. (Kumar et al., 2012)
2.1.2 Transport of diluted species
Transport of diluted species physics module was used for the calculation of moisture transfer. The
governing equation for the physics module is,
𝝏𝒄
+ 𝛁 ∙ 𝒋𝒊 + 𝒖𝛁𝒄 = 𝑹𝒊 6
𝝏𝒕

𝑗 – Moisture flux
𝑅 – Moisture generation within control volume
 In Equation 6, stands for the change of moisture concentration in the control area. ∇ ∙ 𝑗 stands for
the diffusive moisture flux while 𝑢∇𝑐 stands for the convective moisture flux within the banana slice which
is neglected. Ri, which is moisture generation within the control volume is also zero. Equation 6 can be
modified as,
𝝏𝒄
= 𝜵(𝑫 ∙ 𝜵𝒄)
𝝏𝒕
𝝏𝒄 𝝏𝟐 𝒄 𝝏𝟐 𝒄
= 𝑫 +𝑫 7
𝝏𝒕 𝝏𝒙𝟐 𝝏𝒚𝟐

The temperature is linked to moisture transport through the temperature dependence of the diffusion
coefficient, D. D is calculated using following equation.
𝑬𝒂
𝑫𝒆𝒇𝒇 = 𝟎. 𝟎𝟏𝟕𝟓𝟏 × 𝒆(𝑹×𝑻) 8

For equation 8
Ea – Activation energy for diffusion of water (51.21 kJ/mol) (Kumar et al., 2012)
R – Universal gas constant

2.2 Boundary conditions.

The heat flux at the considered boundary after neglecting evaporative heat flux was calculated by,
−𝒏·𝒒 = 𝒉 ( 𝑻𝒆𝒙𝒕 − 𝑻) 9

In the equation 9, 𝑛·𝑞 stands for the heat transfer in a normal direction to the boundary while h stands
for the heat transfer coefficient. 𝑇 and T stands for external and boundary temperatures respectively.
Program calculates heat transfer coefficients according to the Reynolds number.

If the Reynolds number ≤ 500000, program uses following equation.

𝒌 𝟎.𝟑𝟑𝟖𝟕 𝑷𝒓𝟏/𝟑 𝑹𝒆𝟏/𝟐
𝒉=𝟐 𝟐 10
𝑳 𝟎.𝟎𝟒𝟔𝟖 𝟑 𝟏/𝟒
(𝟏 )
𝑷𝒓

If the Reynolds number > 500000, equation changes to following.

𝟒
𝒌
𝒉=𝟐 𝑷𝒓𝟏/𝟑 ( 𝟎. 𝟎𝟑𝟕 𝑹𝒆𝒚𝟓 − 𝟖𝟕𝟏 ) 11
𝑳

For the equations 10 and 11,

L – Length of convective surface
Pr – Prandtl number of the air
Re – Reynolds number of the air

The temperature of blowing air was given as 333 K and the speed was given as 0.7 m/s.(Iqbal M et al., 2019)
Moisture flux at the boundary is defined by the equation,

− 𝒏 ∙ (−𝑫 ∙ 𝛁𝒄) = 𝒌𝒄,𝒊 ( 𝒄𝒃,𝒋 − 𝒄𝒊 ) 12

In equation 12,
Kc,I – mass transfer coefficient
Cb,j – Moisture concentration on the surface of the slice
Ci – Moisture concentration of air

The mass transfer coefficient Kc,I is externally defined. (Kumar et al., 2012)
 𝑲𝒄,𝒊 = 𝟎. 𝟎𝟐𝟗𝟔 𝑹𝒆𝟎.𝟓 𝑺𝒄𝟎.𝟑𝟑 𝑫/𝒂 13
Sc – Schmidt number
a – length of convective surface
2.3 Meshing and time step selection
After defining the terms in above mentioned equations and related parameters, the meshing was done.
The extremely fine mesh which is generated by the software was used for the simulation. 1 minute was taken
as the time step.

2.4 Limitations of the model

The model has few limitations which causes the accuracy of the results. Transport of diluted species
Multiphysics module is not the best choice to model the evaporation of moisture. Moisture transport in porous
media would be the ideal choice for the model where moisture transport is more realistically modelled using
parameters such as porosity of banana and permeability of banana.
In the developed model, movement of moisture is only calculated in 2D. But for more accurate results,
a 3D model is suitable. Developing and solving a 3D model requires a lot of computational power. The
velocity of moisture is neglected in the model. This is a major limitation which causes the deviations of results
when compared with experimental data.
In the developed model, flow of hot air is considered to flow from center to edge which is a radial
flow. But when it comes to real dryers, air is not blown radially. To model the air flow more realistically,
boundary has to be defined for airflow too. This is not possible in 2D axisymmetric environment and also,
this requires a lot of computational power to solve.
The heat flux at the boundary is used for the evaporation of moisture too. But this is not included in
the developed model due to the lack of computational power. Modified equation for boundary condition can
be written as,
−𝒏·𝒒 = 𝒉 ( 𝑻𝒆𝒙𝒕 − 𝑻) − 𝒉𝒎 𝝆(𝑴 − 𝑴𝒆 )𝒉𝒇𝒈 14

where M, Me and hfg are standing for moisture concentration, equilibrium moisture
concentration and enthalpy of evaporation respectively.

3. Results and discussion.

3.1 Results comparison with experimental data.
Two simulations were done to generate results. One simulation had air flow over top surface, bottom
surface and cylindrical wall while other simulation was run considering stagnant air at the cylindrical wall.
Aim of the study was to determine the effect of the air flow at the cylindrical wall. Figure 13 shows the
comparison of temperature of two simulations with experimental data while figure 14 shows the comparison
of moisture content of two simulations with experimental data.
 Temperature variation with time
335
330
325
Temperature (K)

320
315 Stagnant air at cylinderical wall
310 With flux from cylinderical wall
305
Experimental temperature
300
295
0 100 200 300 400 500 600
Time (min)

Figure 13- Comparison of temperature variation.

Variation of moisture content with time

4.5
4
Dry basis moisture content

3.5
3
2.5
Stagnant air at cylinderical wall
2
With flux from cylinderical wall
1.5
1 Experimental results
0.5
0
0 100 200 300 400 500 600
Time (min)

Figure 14- Comparison of moisture content variation

According to figure 13 and 14, there is a remarkable difference between the results and the
experimental data. This is expected because of the limitations which were mentioned in 2.4 and if those
limitations were addressed with a modified model, more accurate result would appear. In figure 13, simulation
quickly approaches the 333K mark while experimental data shows a slower approach. Major reason for this
is that the software takes all the incoming heat to increase the temperature of banana while in real case, a large
fraction of heat is occupied by the moisture evaporation.
In figure 14, the difference between results and experimental data are caused by the lack of
modelling for the moisture. Model does not calculate 3D movements and velocities of moisture and that causes
the moisture content to deviate from experimental data

3.2 Empirical model

Page model was used to validate the results from 2 simulations (Onwude et al., 2016). The page
model can be written as,
𝒏
𝑴𝑹 = 𝒆( 𝒌 𝒕 ) 15
where,
MR - Moisture ratio
 MR is calculated using,
𝑴 𝑴
𝑴𝑹 = 𝒕 𝒆 16
𝑴𝒐 𝑴𝒆
Mt – Moisture content when t = t
Me – Equilibrium moisture content
Mo – Initial moisture content
t- Drying time.
Equation 14 is linearized to obtain a more accessible equation which is Equation 16.

𝒍𝒏 −𝒍𝒏(𝑴𝑹) = 𝒏𝒍𝒏(𝒕) + 𝒍𝒏(𝒌) 17

3.3 Validation of results with an empirical model.

Aim of the validation is to obtain values for k and n in Equation 17 for both simulations and to
compare those. The variation of ln −ln(MR) was plotted against ln(t) for both simulations and Figure 15
was generated. Then n and k constant values were calculated using Figure 15 and those values were taken for
the comparison (Table 5).

Validation of empirical model

2

0
0 2 4 6 8
ln(-ln(MR))

-1 With flux at cylinderical

wall
-2
Stagnant air at
-3 cylinderical wall

-4

-5
ln(t)

Figure 15- Linearized page model.

Table 5- Results of the linearization.

Parameter With flux from cylindrical wall Without flux from cylindrical wall Experimental

k 0.011 0.011 0.008

n 0.89 0.91 1.0849

R2 0.9926 0.9936 0.994

According to figures 13, 14 and 15 alongside with table 5, it is clear that the impact from the
behavior of air at the cylindrical wall is neglectable. In the model validation, constant k is same for both
scenarios while constant n shows a slight difference. Therefore, it is fair to say that the behavior of air at the
cylindrical wall can be neglected.
4. Error analysis, sensitivity analysis and discretization.
4.1 Sensitivity analysis.
Sensitivity analysis for the model was done by changing 3 parameters of the model during 6
experimental runs. Air speed, air temperature and slice thickness were changed to obtain the standard
deviation of drying time. Table 2 shows the reasonable changes of drying time due to the variation of
parameters.

Table 2- Variation of drying time with the selected parameters.

min nominal max

Air speed (m/s) 0.4 0.7 1
Air temperature ( ͦC) 55 60 65
Slice thickness (mm) 3 4 5
Drying time (min) 512 1016 1625

Table 3- Uncertainty analysis

run 1 run2 run 3 run 4 run 5 run 6
Air speed 1 0.4 0.7 0.7 0.7 0.7
Air temperature 60 60 65 55 60 60

Slice thickness 4 4 4 4 5 3
time to reach 0.04 1012 1017 865 1220 1412 318
moisture / min
uncertainty range 5 355 1094
parameter x range 0.6 10 2
Sensitivity coefficient 8.33 35.5 547
(SC)
Standard deviation (SD) 0.18 3.03 0.606
SC^2*SD^2 2.29 11572.54 109902.3
Time variance 121477.1
SD of time 348.5357
Parameter value 974 ± 1.645x 348.5357
uncertainty
minimum 400.6588
Maximum 1547.341
 Variation of drying time with air speed
1018

Drying time (min)

1017
1016
1015
1014
1013
1012
1011
0 0.2 0.4 0.6 0.8 1 1.2
Air speed (m/s)

Figure 4- Variation of drying time with air speed

Variation of drying time with air temperature

1400
1200
Drying time (min)

1000
800
600
400
200
0
54 56 58 60 62 64 66
Air temperature ( C
ͦ )

Figure 5- Variation of drying time with air temperature

variation of drying time with slice thickness

1500
Drying time (min)

1000

500

0
0 1 2 3 4 5 6
Slice thickness (mm)

Figure 6- Variation of drying time with slice thickness

 impact (%)
80
70
60
50
40
30
20
10
0
Air speed Air temperature Slice thickness

Figure 7- Impact from the changed parameters to drying time.

According to Table 3, which contains the uncertainty analysis, air speed has a lower impact towards
the drying time. But air temperature has a much higher impact towards the drying time when compared with
air speed. Even a change of 15 0C has a high impact. But when it comes to slice thickness, even a slight change
of slice thickness can change the drying time dramatically. In simulation runs, a change of 1mm of thickness
has changed the drying time by more than 400 minutes. This variation can be inspected using figures 4,5,6
and 7. Therefore, the model is sensitive to the slice thickness and air temperature more than to the air speed.

4.2 Error analysis.

Error analysis was done by changing the mesh preset to obtain the drying time for the moisture content
to reach 4%. Then the drying times of each mesh were compared with the previous mesh preset to select the
mesh with least difference. Table 3 shows the data gathered during error analysis

Table 3- Error analysis data

Time (min) Percentage

Coarse 1054
Normal 1068 1.33

Fine 1079 1.03

Finer 1086 0.65

Extra fine 1095 0.83

Extremely fine 1091 0.37

4.3 Discretization.
Discretization was done by changing the behavior of moisture concentration and temperature between
linear model and quadratic model. Drying time for the moisture content to become 4% was calculated by
changing the linear-quadratic combination to obtain best combination. Table 4 shows the data gathered during
the discretization and since there is no considerable difference, linear-linear combination was selected.

Table 4- Data of discretization.

Concentration Temperature Time (min)

Linear linear 1016
linear quadratic 1021
quadratic linear 1020
quadratic quadratic 1018

For further analysis, the variation of moisture content at 4% surface moisture content was plotted along
a cutline. The cutline can be observed in Figure 8 while the moisture variations can be observed from Figures
9,10,11 and 12.

Figure 8- Cutline
 Figure 9- Both linear Figure 10- Temperature
quadratic and moisture linear

Figure 12- Both quadratic

Figure 11- Temperature linear
and moisture quadratic

According to figures 9,10,11 and 12 it is fair to mention that the behavior of 2 above mentioned
parameters does not have a considerable impact on the results. Therefore linear-linear combination can be
used for the simulation.

5. Conclusion.
The objective was to inspect the characteristics of food drying and the drying process of a banana slice
was simulated using COMSOL Multiphysics platform. Drying was simulated using heat transfer in solids
physics module and transport of diluted species physics module. Results were generated considering 2
different scenarios where the heat and mass transfer through the cylindrical wall was considered in one
simulation and wasn’t considered in the other. The results were generated for the respective scenarios and
they were compared with each other and experimental data too. Major reasons for the variations of the results
with experimental data were identified and methods to carry out a more realistic simulation were identified
6. References.
i. Iqbal M, J., Akbar M, W., Aftab, R., Younas, I., Jamil, U., 2019. Heat and mass transfer modeling for
fruit drying: a review. MOJ Food Process. Technol. 7, 69–73.
https://doi.org/10.15406/mojfpt.2019.07.00222
ii. Kumar, C., Karim, A., Saha, S.C., Joardder, M.U.H., Brown, R.J., Biswas, D., 2012. Multiphysics
Modelling of convective drying of food materials. Proc. Glob. Eng. Sci. Technol. Conf. 1–13.
iii. Onwude, D.I., Hashim, N., Janius, R.B., Nawi, N.M., Abdan, K., 2016. Modeling the Thin-Layer
Drying of Fruits and Vegetables: A Review. Compr. Rev. Food Sci. Food Saf. 15, 599–618.
https://doi.org/10.1111/1541-4337.12196