Energy consumption and thermal properties of drying banana (musa ssp) under varied relative humidity

The drying of banana slices was investigated under different relative humidity (RH) in a convective hot-air dryer. The experiments were conducted using 10, 20, 30 and 40% RH, 750C and 2.0 m/s air velocity drying conditions. Drying kinetics, energy consumption and thermal properties were investigated. Eight mathematical models describing thin layer drying were employed and results were compared to their goodness of fit in terms of coefficient of correlation (R2), the root mean square error (RMSE) and the reduced chi square (χ2). The Midilli–Kucuk model could satisfactorily describe RH-convective drying of banana slices with R2, RMSE and in the ranges of 0.99947-0.99986, 0.00002-0.00008, and 0.0142-0.01618 respectively. For energy consumption, RH condition increased the energy consumption such that at every 10% increase in RH, a range of 17.9-41.0% increase in energy consumption was observed. In the case of thermal properties, endothermic peak revealed precise composition, fine particle size, and dispersion in a matrix of banana slices thus altering the transition in solid state. In effect, RH drying present new drying concept with prospecting thermodynamics absorbing properties. However, energy consumption issue needs to be addressed in further research.


Energy consumption and thermal properties of drying banana (musa ssp) under varied relative humidity Introduction
Drying technology is a complex thermal process due to unsteady heat and moisture transfer occurring simultaneously; a concept that researchers are still unravelling (Sahin & Dincer, 2005). In order to understand the process, it is critical to model the control parameters to predict, design, and improve the results of the process. In the process of drying thin-layer samples, several mathematical models divided into theoretical, semi-theoretical and empirical models have been proposed. (Doymaz, 2011;Doymaz & Kocayigit, 2012). The theoretical model considers the internal resistance to moisture transfer, while the empirical and semi theoretical models take into account the external resistance of moisture transfer between the air and product (Henderson & Pabis, 1962;Ju et al., 2016). Fick's second law of diffusion continues to be the most widely used drying model whereas other models such as Logarithmic, Modified Page, Henderson-Pabis, Page, Midilli-Kucuk, Two terms and Newton are considered as empirical models (Doymaz, 2011). Mathematical model equations have been used to predict drying kinetics of whole banana (Da-Silva et al., 2014), banana slices (Joardder et al., 2014;Olawoye et al., 2017;Verma et al., 2014), banana cubes (Jiang et al., 2014) and banana paste and flour (Cabrera-Padilha et al., 2014;Rayo et al., 2015) using various drying conditions. However, there remains a gap in literature vis-à-vis the modeling of drying kinetics, thermal energy consumption and thermal properties of dried banana under different relative humidity (RH) that needs to be bridged.
In banana preservation through drying, thermal properties are often altered, transformed and in some cases disperse completely the matrix of dried banana. From literature, the severity of drying impact on thermodynamics is dependent on the drying methods, drying procedures, nature of sample and drying conditions (Nunes et al., 2016).
This study was conducted in order to predict effectively the different RH of drying kinetics using mathematical model, and to assess the thermal energy consumption and thermal properties of banana under varied RH drying conditions.

Preparation of Samples
Ripped Bananas (Cavendish spp.) was obtained from a grocery market, Zhenjiang-China and transported to the School of Food Science and Biological Engineering laboratory in cartoon boxes. Samples were kept at room temperature until further used. After peeling, slices (5 mm thick) were obtained using a cutter (SS-250, SEP Machinery Company Ltd, Guangzhou, China), before drying was performed. The initial moisture content (MC) was determined by the oven method (AOAC, 1990), on wet basis and was noted to be 76.05 ± 0.8% (w.b).
Drying with Relative humidity (RH) convective control hot-air dryer Drying was carried out with an advanced labscale hot-air convective dryer equipped with a humidity controlling device (range 0-99%) as described previously (Sarpong et al., 2018).

Experimental design for drying
The sample (60 g) was loaded into a sieve loading tray into the dryer with aluminum foil cover after tarring to zero. Drying was carried out using a combination of four RH (10, 20, 30 and 40%) at 70°C with m/s air velocity and continued until a constant weight was obtained. To ensure stable drying conditions, the drying system was run for 1 h before loading the sample. For the control (drying with no RH), the RH device was switched off to allow only the convective drying system (HAD) to dry banana at 70°C with m/s air velocity and continued until a constant weight was obtained which ranged from 280 -410 min due to different drying conditions used.

Modeling of drying kinetics
According to Newton's heat transfer cooling law (Lewis, 1921), the moisture ratio (MR) of the sample is defined and expressed in Eq (1).
Where M e M, and M o and are the equilibrium moisture content of the initial moisture content and dry basis respectively at time t; k is the drying constant. For the analysis of MR, the equilibrium moisture content was assumed to be zero (Man et al., 2014). At time (t), the drying rate (DR) of banana slices was calculated as follows: Where t 1 and t 2 are the drying times (min) at different times during drying; M t1 and M t2 are the moisture content of samples (g water per g dry matter).
The 8 thin-layer drying model was fitted for the (MR, t) drying kinetics of banana slices, which are commonly used in scientific literature, as described in Table 1. Using MATLAB R2012 (MathWorks, USA) tool, regression analysis was performed. Three primary parameters such as coefficient determination (R 2 ), root mean square error (RMSE) and reduced chi-square (χ2) were used to evaluate the adjustment/ goodness of fit to the models using Eq. (3-5). The best model was chosen based on highest and lowest χ2 and RMSE.
Where MR expt,i and MR pred,i are the experimental and predicted dimensionless MR respectively, N is the number of observations, and is the number of constants. The best model to describe the drying kinetics of banana slices under RH conditions was chosen as the one with the highest R 2 and least RMSE and X 2 (Doymaz, 2011;Wu et al., 2014).

Effective moisture diffusivity calculation
Based on Fick's second law (law of diffusion), the theoretical model in thin-layer for most agricultural products is expressed in Eq 6.
Where Deff is effective moisture diffusivity (m 2 /s), is the rate of change, is concentration (mol/m 3 ), M is the length (m). For slab geometry calculations, the diffusion coefficient, (m 2 /s) can be expressed based on the following assumptions (constant diffusion coefficient, single-dimensional moisture movement, volume change, consistent temperature, and negligible external resistance) (Crank, 1979) F. Sarpong et al. (2021) Ghana Jnl. Agric. Sci. 56 (2), 22 -33 Where M, Mo and Me are the average, initial and equilibrium moisture contents (g/g d.b) of the sample respectively at time t, is the constant effective diffusivity (m 2 /s). L and t represent half the thickness of the banana slice and the drying time (s) respectively. Only the first term of the Eq.(7) can be used for long drying times (Xiao et al., 2012) The slope (k 0 ) is calculated by plotting ln (MR) against time

Energy consumption
The total energy consumption for drying banana under different RH conditions was measured using an electric energy meter with 0.01 kWh accuracy while the specific energy consumption (the energy required to dry 1 kg of banana) was calculated by Eq. (3.10) (Jiang et al., 2017)

……………[10]
Where is the specific energy consumption, is the total energy consumption, and is the initial weight of banana.

Differential scanning calorimetry (DSC)
The thermal behavior of the samples was studied by using a μDSC 7 (Setaram, Caluire, France) furnished with a cooling bath (JULABO, Labortechnik GmbH, 77960 Seelbach, Germany) and adjusted with Naphthalene. Nitrogen was used to prevent water condensation at low temperatures, during the experiment. To obtain the spectra, samples (2 mg) were placed on μDSC cell, sealed with a crimper and placed in the scanning calorimetry device. The spectra (data) were obtained using Data Acquisition (version 4.1D-TA Instruments) and Calisto Processing (version 1.065-Setaram, Caluire, France) software. The DSC was taken at the temperature range 20-120°C and scanned at 1.2°C min -1 and isotherms time were set at 15 and 10 min for initial and the final temperatures respectively.

Statistical analysis
Data were processed and presented as Means ± SD using OriginPro 9.2 (Origin Lab Corporation, Northampton, MA, USA). Further comparisons were done with one-way ANOVA, Pearson's coefficient of correlation and turkey test using XLSTAT (Addinsoft Inc USA) respectively.

Effect of Relative Humidity on Drying Curves
The moisture content of sliced banana samples was reduced until constant moisture content was obtained from the initial moisture content of 76.05%. Moisture ratio (MR) decreased exponentially with drying time for all RH samples. The results of the kinetic curves of drying banana slices at 75 o C under different relative humidity (10, 20, 30 and 40%) are shown in Fig. 1a. The longest drying time of 410 min was obtained at 40% RH to achieve constant moisture content under 70 0 C drying condition whilst HAD observed the shortest drying time (280 min) as a result of dried atmospheric condition in the drying chamber for convective drying system. However, 10%, 20% and 30% RH achieved this constant moisture content in 290, 320 and 390 min which are equivalents to 29.3, 21.0 and 4.8% drying time reduction respectively when compared with 40% RH. These results were similar to the observation made by Ju et al. (2016) who concluded that lower relative humidity (RH) predicts a lower moisture content than higher RH. This phenomenon is caused by a rapid increase in the internal temperature of the product (Fig.  1c) of the sample under 10% RH condition which reached a maximum temperature of 73.5 ± 1.5 0 C compared with 70.4 ± 1.4, 68.5 ± 1.8 and 70.1 ± 2.80 0 C for 20, 30 and 40% RH respectively. This high internal temperature of banana slice at 10% RH provided an additional driving force for water diffusion and promoted the internal moisture movement toward achieving highest moisture reduction and drying rate (Fig 1c).
Drying rates (DR), after 30 min was in the range of 0.0056-0.0069 kg/kg for RH samples whilst HAD observed 0.00733 kg/kg as depicted in Fig. 1b. DR peaked at 120 min for all the drying conditions with higher value of 0.0172 kg/kg for 10% RH whilst 20, 30 and 40% RH recorded 0.015, 0.0125 and 0.0092 kg/kg respectively. Meanwhile, the highest DR (0.01873 kg/kg) was observed in HAD conditions affirming the highest moisture loss in the drying of banana sample. There was no constant period, and most of the drying occurred during the rate of decline. This effect was related to the availability of a large amount of free moisture which was removed during the initial stages of drying (Doymaz, 2014;Mghazli et al., 2017). The falling period was governed by water diffusion in banana sample; a typical behavior observed for other fruits (Lee et al., 2010;Mghazli et al., 2017;Sarker et al., 2013). The slow pace of the falling period is a result of the evaporation of moisture from the surface of banana slices at the initial stages due to a higher heat transfer and mass transfer coefficient. Moreover, Sarker et al. (2013) suggested that the gel that is formed as a result of protein denaturing impeded water movement inside the banana matrix to the surface for evaporation, thus slowing the dehydration process.

Fitting of the Drying Curves
For the various RH drying conditions of each run, the experimental data fitted to eight models ( Table 1) and R 2 , and RMSE were used to evaluate the performance of each model. The obtained model constant parameters for the eight models are summarized in Table 2. All the models gave consistently R 2 values in the range of 0.97007-0.99983; indicating that all the models could be used to describe the drying behavior of banana slices under 10-40% RH conditions. However, Midilli-Kucuk displayed the highest average value of R 2 , lowest and RMSE ( Table 2). The changes in the R 2 , and RMSE range between 0.99947-0.99986, 0.00002-0.00008, and 0.0142-0.01618 respectively. Consequently, Midilli-Kucuk model is the best among the tested models that accurately expressed the thin-layer drying behavior of banana slices under the studied drying conditions.
Good conformity is observed between the experimental and predicted moisture ratio shown in Fig. 1d under 10

Determination of Effective Moisture Diffusivity (D eff )
Isothermal temperature at 75 0 C was maintained during all the RH drying processes. values were within the range of 1.805 x 10 -8 -1.956 x 10 -8 m 2 /s (Table 2) indicating that moisture movement in the banana slice was in the liquid form (Saravacos & Kostaropoulos, 2002). decreased with increase in RH except for 20% RH which recorded the highest value. Sample internal temperature (Table 2) revealed that 20% RH recorded highest temperature even though Isothermal temperature (75 0 C) was used for all RH. So RH with highest sample internal temperature recorded highest value. This was confirmed by other authors who used different drying temperatures in determining value (Doymaz, 2011;Hacıhafızoğlu et al., 2008;Mghazli et al., 2017). Fig. 2 shows the energy consumption and specific energy consumption in RH convection drying of banana under varied RH conditions. The energy consumption needed to obtain a constant dry weight of banana ranged from 10.21-17.08 kWh. Similarly, 8-12 kWh energy consumption was observed in convection drying of pomegranate arils at 70 o C at varied air velocity (Motevali et al., 2011). In their research, air velocity, temperature and drying time impacted energy consumption significantly (p<0.05). Zhao et al. (2018) also observed that different thicknesses of sample could play a role in energy consumption of the drying system. Observably, RH and drying time played a significant (p<0.05) role in the energy consumption of drying banana with a direct relation. In effect, the RH condition increased the energy consumption such that at every 10% increase in RH, a range of 17.9-41.0% increase in energy consumption was observed. In similar manner, specific energy consumption increased with increasing RH which corresponded with an increase in drying time. In the case of 40% RH, maximal value of 284.72 kWh/kg was observed whilst 201.39 kWh/kg was observed in 10% RH. However, HAD drying condition recorded the minimal specific energy consumption of 170.13 kWh/ kg with a significant difference (p<0.05) when compared with all four RH drying conditions.

Effect of RH on characteristics of DSC thermogram
Thermal properties of the banana slices were examined through a DSC thermogram as shown in Fig. 3. In this concept, DSC measured heat absorbed or released during the transition state of the sample when subjected to measurable heat range (in this case, 20-120 o C). From Fig. 3, the first endothermic peak for the four RH and HAD conditions ranged from 30-35 o C and the variation was caused by precise composition, particle size, and dispersion in a matrix of banana slices (Izidoro et al., 2011) as a result of varied RH drying condition. When anhydrous, the banana slices showed endothermic reaction and again samples exhibited varied peaks (Tp).
For HAD, 10% and 20% RH, Tp were 85 o C while 79 o C and 90 o C were observed for 30% and 40% RH respectively. The intensity of the endothermic phenomenon also revealed the transition in solid state of the sample and from Fig. 3, was very pronounced in lower RH. This is an indication of high degree of crystallinity, thus, making the sample more brittle. Also, other texture profile properties such as stiffness, hardness, tensile and melting point were highly related to the crystallinity of the sample. The 10% RH and HAD did not exhibit another endothermic peak. However, a third peak was observed in 20%, 30% and 40% RH at 105 o C, 110 o C and 120 o C which is ascribed to the caramelization (Izidoro et al., 2011).

Conclusion
The drying kinetics, energy consumption and thermal properties of banana slices were investigated under four RH and HAD (used as control) conditions at a temperature of 75°C. A significant reduction in MC was observed in lesser RH as a result of increased rise in sample internal temperature. Among the eight thin-layer drying models studied, the Midilli-Kucuk model was judged the best model from its R2, χ2 and RMSE values. The model constant and drying time for various model regressions for MR calculations was determined and reported. RH increased the energy and specific energy consumption of drying banana with a direct relation. The DSC thermogram revealed that the RH affected the precise composition, particle size, and dispersion in a matrix of dried banana such that the transition in solid state was also altered. Again, lower RH increased the crystallinity of banana, thus, making the sample more brittle when compared with higher RH.