Abstract-- (in terms of heat pump drying) An optimum drying temperature of 45 °C, air velocity of 1 m/s and slice thickness of 5.04 mm were recommended with following predicted responses close to experimental values: drying time 216.58 min, total energy consumption 2.94 kWh, coefficient of performance heat pump (COPhp) 3.08, coefficient of performance system (COPws) 2.75, specific moisture extraction rate (SMER) 0.22 kg/kWh, drying rate 2.53, L* value 82.53, a* value -1.83 and b* value 25.82. The most suitable models to represent the drying behavior of optimum melon slices was chosen Wang & Sing. Effective moisture diffusivities (Deff) of the melon slices were ranging from 7.075E-10 - 1.843E-07 m2s-1 and increasing drying air temperature, drying air velocity and slice thickness led to an increment of Deff.
Keywords-- Heat pump drying, Melon, Response surface methodology, Drying characteristics, Mathematical modelling
I. INTRODUCTION
Dried fruits function as the concentrated form of their fresh properties in terms of energy, nutrients, dietary fiber and antioxidants (Bennett et al., 2011). Melon (Cucumis melo L.) is a popular fruit grown in European, Asian and African countries (Dias da Silva et al., 2016). In addition to its rich nutritional composition and bioactive substances, melon has excellent flavour properties (Petkova and Antova, 2015). Since melons are highly perishable and rapidly get decomposed, they can easily deteriorate after harvesting (Welti-Chanes et al., 2002). For this reason, drying is a useful process to extend the shelf life of melons (Azuara et al., 2009; Dias da Silva et al., 2016) and to increase their added value by transforming them in to a healthy snack, especially in over production seasons. On the other hand, drying can cause adverse effects on melon by reducing its quality in terms of colour, texture and bioactive compounds (Dias da Silva et al., 2016). It is very important to select a proper drying method in terms of its technology, energy consumption, and frequency of requirements to maintenance (Kırbaş et al., 2019). Due to the prolonged exposure to high temperature and/or heat and oxygen, traditional drying processes adversely affect typical fresh fruit colour, aroma, and overall flavour (Zanoni et al., 1999Abano et al., 2011 Therefore, there is a need to establish a proper drying method to increase the amount of vitamins and flavour in the final dried product in addition to decreasing the rate of pigment deterioration (and The studies in the literature prevailed that Heat Pump Drying (HPD) system played an important method to retain the colour, vitamins, and other heat sensible components of agricultural products (Prasertsan, 2010; Jeyaprakash et al., 2016; Coskun et al., 2017). Moreover, it was proven that HPD consumes the lowest energy among the other drying methods (Tunçkal and Doymaz, 2020). (in terms of
II. METHODS
A. Materials
5 g of melon slices were weighted to measure moisture content in an oven at 105±3 °C. Moisture analysis was ceased once the two successive weights of the sample less than 1% (Koşan et al., 2020).
B. Drying system procedure
Melon slices were dried with a heat pump dryer whose technical properties and detailed figure were given before by Coskun et al. (2017). Before starting the drying process, the dryer was operated for about 30 min. to reach steady state conditions. The process input variables included drying air temperature, velocity and slice thickness ranging by 35 - 45 °C, 0.5 - 1.0 m/s and 5 - 9 mm, respectively. Following weighing of 635 g of melon slices, they were distributed homogeneously on the drying tray, and consequently the drying process was started. During drying process, the mass loss of samples was measured by using the load cell at every four min. and all data were recorded by a computer connected to the system. The temperature and relative humidity of air within the drying system were measured every one minute by a data collector. When the moisture content of the melon slices decreased about 8±0.5 % (wet basis), the drying procedure was finished. Energy consumption of the compressor was measured by a digital wattmeter (Makel, Turkey) every one hour followed by recording the data. Drying system operates with ± 0.8 °C sensitivity by a digital thermostatic control.
C. HPD system analyses
The performance of HPD system is expressed in terms of specific
moisture extracted ratio (SMER) and the coefficient of performance (COP). SMER
is defined as the water extracted from the product per consumed energy; and COP
is the ratio between the heat delivered to the drying air through the condenser
) and
energy consumed at the compressor (
(Jia et al.,
1990). COP for HPD system was calculated by Eq. 1.
(1)
where 
represents
coefficient of performance of heat pump.
The
COP of the overall system (
) was based on Eq. 2:
(2)
where 
and 
represents
power consumption (kJ/s) of the internal and external condenser fan,
respectively.
Heat transfer rate (kJ/s) of the internal condenser was used to increase the enthalpy of the drying air as given in Eq. 3.
(3)
where 
and 
represents
mass flow rate of air (kg/s), specific enthalpy of air at the exit and at the
inlet (kJ/kg) of the condenser, respectively.
SMER (specific moisture added ratio) represents the total amount of energy consumed per the moisture absorbed from the product and calculated using the following equation (Jia et al., 1990).
(4)
where 
represents
amount of water evaporated from the melon during drying process (kg/s).
D. Mathematical modelling of drying kinetics
Seven drying models were applied to determine the best model for drying characteristics of melon slices. These models were given as following equations (Özkan Karabacak et al., 2020):
Page: 
(5)
Modified Page: 
(6)
Logarithmic: 
(7)
Lewis: 
(8)
Henderson & Pabis: 
(9)
Two Term Exponential: 
(10)
Wang & Singh: 
(11)
where 
, 
, and 
represent
model constant, 
is the coefficient in the drying model (1/s) and 
is the drying
period (s).
The moisture ratio (
) and drying
rate of the samples (
) (g/s) were
estimated from the following equations:
(12)
In Eq. 12, , 
, 
, 
are the moisture ratio, the moisture content
at a specific time (g water/g dm.), moisture content (g water/g dm.) of the
sample prior to drying, moisture content (g water/g dm.) at equilibrium state,
respectively (Tüfekçi and Özkal, 2017). Equilibrium moisture content was
accepted as zero since is much lower
than and values.
(13)
where 
and 
represent the weight (g) of the melon before drying and after
drying, respectively.
refers to the
duration of the drying process (s).
The highest correlation coefficient (R2) and lowest root mean square error (RMSE), Chi square (χ2) were used as the statistical parameters to identify the integrity of the fit models describing the drying kinetics of melon slices (Gamlı et al., 2018). Subsequent equations were used for the calculation of these values:
(14)
(15)
Here, 
and 
are the experimental and
dimensionless moisture ratios, respectively for the test 
. 
is the number
of observations and is the
constant number for the models.
E. Effective moisture diffusivity (Deff)
Fick’s second diffusion law is broadly utilized for the estimation of effective moisture diffusivities of agricultural products (Süfer and Palazoğlu, 2019). The mathematical explanation of Fick’s law is given in Eq. 16, based on the assumptions including uniform initial moisture diffusion, negligible shrinkage, constant diffusion coefficient throughout the process and symmetric mass transfer in an infinite slab,
(16)
In Eq. (16) , 
, 
and represents
dimensionless moisture ratio, effective moisture diffusivity (m2/s),
half thickness of the slab in the melon slices (m) and positive integer,
respectively. For sufficiently extended drying periods, Eq. (17) can be
abbreviated to the first term in the series by using a logarithmic form as
follows:
(17)
was computed through the slope of a straight line which was plotted 
versus time
in drying experiments. The equation was expressed as follows:
Table 1. Box–Behnken matrix of independent variables used in the RSM design
|
Values |
Parameter |
Levels |
||
|
-1 |
0 |
1 |
||
|
|
Drying air temperature (ᵒC) |
35 |
40 |
45 |
|
|
Slice thickness (mm) |
5 |
7 |
9 |
|
|
Drying air velocity (m/s) |
0.5 |
0.75 |
1 |
Table 2. Experimental Design
(18)
F. Colour analysis
Colour measurements were carried out with a chroma meter (CR-5, Konica Minolta, Osaka-Japan). L* value indicates the change from vertical luminous intensity to darkness, (+a*) redness, (−a*) greenness, (+b*) yellowness, (−b*) blueness. Measurements were performed in triplicate, and the average was used for further calculations.
G. Experimental design
Response Surface Methodology (RSM) was applied to
evaluate the effect of independent variables including air temperature (
), slice
thickness (
) and drying
air velocity (
) on depended
variables which include drying time (
), total
energy consumption (
), 
, 
, 
, drying rate (
), and colour
parameters (L*, a*, b*) (, 
, 
). A three
level of Box-Behnken design was used for RSM study. Three factors and their
levels (-1, 0, 1) were determined based on the preliminary experimental trials.
1, 0, and -1 represent high, optimal and low value. This experimental design
involved 17 experimental trials carried out in a randomized order. Factors,
their levels and experimental design including coded and uncoded values were
illustrated in Table 1 and Table 2.
All experiments were performed in triplicate, and each sample was analyzed in duplicate. Average analysis results of each trial were used to draw three dimensional contour plots and consequently were evaluated to establish an optimization model. The contour surface as a bi-dimensional projection of RSM was a proper tool to determine a relationship between drying parameters of the process and the studied variables.
A second order polynomial equation (Eq. 19) was fitted to model each responses, i.e. depended variables as a function of independent variables.
(19)
In this equation, coefficients of the model were 
(constant), 
,
, 
(linear), 
,
, 
(interaction),
,
, 
(quadratic).
Second-order coefficients of model equations were generated by regression analysis with Minitab statistical software (version 17, Minitab Inc., Pennsylvania, USA). The goodness of fit of the model was evaluated by the coefficient determination (R2). The significance of the estimated regression coefficient for each response variable was assessed by using a 5% test of significance at a probability (p) of 0.05. Moreover, optimum levels were determined by optimizing the responses.
III. RESULTS AND DISCUSSION
A. Analysis the Response Surface Models
The analytical results obtained from all the experiments designed based on Box–Behnken model for drying of melon slices were given in Table 3. Those experimental data were used to determine the coefficients of the quadratic polynomial equations given Eq. 19. The resulting model parameters were demonstrated in Table 4 providing regression coefficients for each dependent variable. The determination coefficient (R2) for most of the responses except colour ranged from 0.9772 and 0.9928, indicating a high correlation between predicted and observed values. The alterations in moisture content of melon slices over time were illustrated in Fig. 3. Drying time of the melon samples changed between 244 - 792 min.
Drying time:
Drying time () of the melon
samples changed between 244 and 792 min. The quadratic model for drying time
can be expressed by the following equation.
(20)
From the Eq. (20), it can be
observed that while linear terms of drying air temperature () and
drying air velocity () had negative
impact on drying time, slice thickness () had positive
effect on it. Quadratic term of slice thickness (
) had the highest positive effect. However, drying time significantly
negative influenced by the quadratic terms of drying air velocity (
) (p<0.05). Maximum drying time was determined for in Run 11.
However, minimum drying time was found in Run 3. It was prevailed that drying
time decreased by reducing the slice thickness and increasing the drying air
velocity and temperature (Fig. 1a). Those results were in line with the
previous studies reported in the literature in which increasing air temperature
(Phahom et al., 2021) and drying air velocity (Ndukwu, 2009) and reduced
slice thickness (Sadin et al., 2014) led to a reduction on drying time.
Total energy consumption
Total energy consumption () in kWh
changed between 3.008 - 9.258 kWh. The quadratic model for total energy
consumption can be expressed by the following Eq. (21):
(21)
From Eq. (21), it can be monitored that while linear terms of drying
air temperature () and
drying air velocity () had negative
impact on total energy consumption, slice thickness () had positive
effect on it. Similar to drying time, quadratic term of slice thickness ( had the
highest positive effect on total energy consumption. However, total energy
consumption was significantly influenced by the quadratic terms of drying air
velocity () in a
negative manner (p<0.05). Maximum total energy consumption was determined
for in Run 3. In contrast, the lowest total energy consumption was found in Run
11. The graphical representation of the influence of slice thickness () and drying
air temperature () on
total energy consumption (Fig. 1b) revealed that when drying air temperature
increased to 40 – 45 °C and minimum slice thickness was applied, total energy
consumption was the lowest. Also it can be observed that lower drying air
velocity (< 0.8 m/s) and lower drying air temperature (35 °C) had negative
influence on total energy consumption. The effect of slice thickness () and drying
air velocity () at center
point of drying air temperature (40°C) demonstrated that increase of slice
thickness resulted in higher total energy consumption.
HPD system analyses
(
) changed
between 1.430 and 3.074. The quadratic model for can be
expressed by the following Eq. (22):
(22)
From the Eq. (22), it can be monitored that while linear
terms of drying air temperature () and
drying air velocity () had positive
impact on , slice
thickness () had negative
effect on it. Quadratic term of drying air temperature and slice thickness (
and had negative
effect on . However, positive
influenced by the quadratic terms of drying air velocity () (p<0.05).
Maximum determined
for in Run 8. However, minimum was found in
Run 5. Factorial plots for demonstrated
that when drying air temperature increased, was also
increased (Fig. 1c). It can be also observed that lower drying air velocity
(< 0.8 m/s) had negative influence on . The effect
of slice thickness () and drying
air velocity () at center
point of drying air temperature (40 °C) was represented in Figure 1c,
demonstrated that increase of drying air velocity (> 0.75 m/s) resulted in
higher .
(
) changed
between 1.359 and 2.749. The quadratic model for COPhp can be expressed
by the following Eq. (23):
(23)
From the Eq. (23), it can be determined that both linear and
quadratic terms of slice thickness ( and ) had negative effect on (p < 0.05). Although maximum determined
for in Run 8, minimum was found in
Run 9.
Those results were consistent with previous studies
performed with 
in which 2.72 
value was
obtained for banana at 48 – 52 ᵒC drying air
temperature and 0.4 m/s drying air velocity conditions (Kuan et al.,
2019) and values
between 2.11-2.96 were obtained for
grated carrots at 56 °C drying air temperature (Aktaş
et al., 2019). On the other hand values of the
melon slices were found lower than mint leaves were dried by heat pump dryer at
40 °C drying air temperature and 1.8 m/s
drying air velocity conditions. Factorial plots for showed that
when drying air temperature () was
increased, was also increased (Fig. 1d). It can be seen that drying air
velocity () (> 0.8
m/s) had positive influence on . The effect of slice thickness () and drying
air velocity () at center
point of drying air temperature (40°C) was showed in Fig. 1d,demonstrated that
increase of drying air velocity (> 0.75 m/s) resulted in higher .
(
) changed
between 0.125 and 0.215 kg/kWh. The quadratic model for can be
expressed by the following Eq. (24):
(24)
According to Eq. (24), both drying air temperature () and drying
air velocity () had positive
correlation, but slice thickness () had negative
correlation on SMER value. Moreover, it was significantly negative influenced
by the quadratic term of drying air temperature (). Quadratic
terms of slice thickness () and drying
air velocity () had highest
positive effect on it (p < 0.05). Similar to drying time and total energy
consumption, maximum was
determined for in Run 11. However, minimum was found in
Run 3. In previous studies values were
obtained as 2.05 kg/kWh for mango by Wang
et al. (2019), 0.6 kg/kWh for banana
by Kuan et al. (2019), 1.64 kg/kWh
for red jujube by Yuan et al. (2019) and 0.06
kg/kWh for ginger by Chapchaimoh et al. (2016).
The graphical representation of
the influence of slice thickness () and drying
air temperature () on value (Fig. 1e)
revealed that when both drying air temperature decreased and slice thickness
increased (> 6 mm), value was
reduced. From Fig. 1e it can be seen that lower drying air velocity (< 0.8
m/s) and lower
