7th International Science, Social Sciences, Engineering and Energy Conference 19-21 November, 2015, Wangchan Riverview Hotel, Phitsanulok, Thailand I-SEEC 2015 http//iseec2015.psru.ac.th Greenhouse solar dryer for drying bael fruits (Aegle Marmelos) in Thailand Jagrapan Piwsaoada,e1, Chayapat Phusampaoa,e2 a Program Physics, Department of Science, Faculty of Science, Loei Rajabhat University, Loei, Thailand, 42000 e1 Jagrapan25@gmail.com , e2joitoji@gmail.com Abstract This paper presents experimental performance and artificial neural network modeling of a greenhouse solar dryer for drying bael fruits. The dryer consists of a polycarbonate sheets on a metal sheet floor. The dryer is 1.5 m in width, 1.0 m in length and 2.0 m in height. Two 15-W DC fans powered by 50-W PV module were used to ventilate the dryer. To investigate its performance, the dryer was used to dry ten batches of bael fruits. For each batch, 5.0 kilograms of bael fruits were dried in the dryer. Results obtained from the experiments showed that drying temperatures varied from 32๐C to 50๐C. In addition, the drying time for drying bael fruits was 3 days, compared to 5 days required for natural sun drying. Bael fruits dried in the dryer were completely protected from rain and high quality bael fruits were obtained. The estimated payback period of the greenhouse solar dryer is about 1 year. A multilayer neural network model was developed to predict the performance of this dryer. The predictive power of the model was found to be high after it was adequately trained. Keyword: solar drying, ANN modeling, bael fruits, greenhouse solar dryer 1. Introduction Bael fruit tree (Aegle marmelos) occurs in dry forests on hills and plains of northern, central, eastern and southern Thailand. Bael fruit (Ma Toom) is Thai herbal fruit growing well in tropical area. Normally, Thai people slice the fruit into rounded pieces and dried well under sunlight. Then, use dried bael fruits for making herbal drink by boiling in hot water and adding sugar to make it sweet and finally straining with clean cloth sheet and keep in the fridge, can drink both hot or cool. Bael biochemical compounds of bael leaves, fruits and seeds have been used in several diseases like diabetes, cardiovascular and antiinflammatory. The bael leaf contain seven monotorpene hydrocarbons (90.7%), three oxygenated monoterpenes (2.9%), four sesquiterpene hydrocarbons (3.1%) and one phenolic compound (0.2%). Limonene (82.4%) was the main constituent of bael [1]. Situated in the tropical area, Thailand receives abundant solar radiation [2-3]. Consequently, the use of solar dryers for bael fruits drying is reasonable. Although several types of solar dryers have been developed in the last 45 years [4-11] but, they could not 2 meet the high demand of bael fruits drying. As a result our research group has developed a solar dryer to dry agricultural products. It was successfully used for drying fruits and vegetable. However, it has not been tested to dry bael fruits. Therefore, the objectives of this research were to investigate the performance of the dryer for drying bael fruits and to develop an artificial neural network (ANN) model to predict the performance of this dryer. Nomenclature a* colour green to red (decimal) b* colour blue to yellow (decimal) Cannual annual cost of the system (USD) Cl labor cost for the construction (USD) Clabour,oplabour cost for operating the dryer (USD) Cm material cost of the dryer (USD) Cmaint,i maintenance cost (USD) Cop operating cost (USD) Cop,i operating cost at the year i (USD) CT total capital cost for the greenhouse solar dryer (USD) C* indicating colour saturation (decimal) h Hue angle (decimal) iin the interest rate (decimal) if inflation rate in percent (decimal) L* lightness colour (decimal) Mdry the dried product obtained from this dryer per year (kg) Mdry annual production of dry product (kg) Mf the amount of fresh product per year (kg) Pd the price of the dry product (USD/kg) Pf the price of the fresh product (USD/kg) Z dying cost (USD/kg) 3 2. Methodology 2.1. Experimental study The greenhouse solar dryer was installed at 29 Moo 4, A. Arawan, Loei (17.48๐N, 101.72๐E), Thailand. The dryer consists of a polycarbonate sheets on a metal sheet floor. The dryer has a width of 1.5 m, length of 1.0 m and height of 2.0 m. Two DC fans operated by 50-W solar cell module were installed in the wall opposite to the air inlet to ventilate the dryer. The pictorial view of the dryer is shown in Fig.1. In this study, bael fruits were dried in the greenhouse solar dryer to investigate its potentials for drying bael fruits. Ten experimental runs were conducted during the period of 25 June 2015-8 August 2015, and 5.0 kg of bael fruits were dried for each run. Solar radiation was measured by a pyranometer (Kipp&Zonen model CMP 3, accuracy ±0.5%) placed on the roof of the dryer. Thermocouples (K type) were used to measure air temperatures in the different positions of the dryer (accuracy ±2%). A hot wire anemometer (Airflow, model TA5, accuracy ±2%) was used to monitor the air speed inside the dryer. The relative humidity of ambient air and drying air was periodically measured by hygrometers (Moisture content meter, model Extech SDL, accuracy±1%). The positions of all measurements are shown in Fig. 2. Solar sell Fan_air out let Polycarbonate cover rh_out let Door 2m rh_inside Solar colector T3 T1 M1 M2 T4 T2 M_ambient 0.5 m 1m T_ambient rh_ambient Air_inlet Figure 1. The pictorial view of a greenhouse solar dryer 1.5 m Figure 2. The structure of the greenhouse solar dryer and the position of the thermocouples ( T), hygrometer ( rh) and product samples ( M). Solar radiation passing through the polycarbonate roof heats the air, the products inside the dryer, as well as the metal sheet floor. Ambient air is drawn in through the air inlets at the bottom of the front side of the dryer and is heated by the floor and products exposed to solar radiation. The heated air, while passing through the products, absorbs moisture from the products. Direct exposure to solar radiation of the products and the heated drying air enhance the drying rate of the products. Moist air is sucked from the dryer by the DC-fan at the top of the rear side of the dryer. 4 The air speed at the inlet and outlet of the dryer was recorded during the drying experiments using the hot wire anemometer. Bael fruits dried in each drying test were 5.0 kg. Bael fruits samples and the pictorial view of bael fruits being dried in the dryer are shown in Fig. 3 and Fig. 4 respectively. Figure 3. Bael fruits samples Figure 4. The pictorial view of bael fruits being dried in the dryer Bael fruits were place on tray inside the dryer. Each day, the experiment was conducted during 8:00 am-6:00 pm. The drying was continued on subsequent days until the desired moisture content was reached. Product samples were placed at various positions in the dryer and were weighed periodically at two-hour intervals using a digital balance (Kern, model 474-42, accuracy ± 0.1g). To compare the performance of the dryer with that of natural sun drying, a control sample of bael fruits were placed near the dryer and dried simultaneously under the same weather conditions. The moisture content during drying was estimated from the weight of the product samples and the estimated dried solid mass of the samples. At the end of the experimental drying, the exact dry solid mass of the product samples was determined by the hot air oven method (103 ๐C for 24 hours, Memmert GMBH, model ULE500, accuracy ± 0.5%). 2.2. Colour measurement of dried bael fruits The colour of dried bael fruits samples was measured by a chromometer (CR–400, Minolta Co., Ltd., Japan) in Commission Internationale d’Eclairage (CIE) chromaticity coordinates. L*, a* and b* represent lightness (0 to100), green to red (−60 to +60) and from blue to yellow (−60 to +60) colours, respectively. Out of five available colour systems, the L*a*b* [12] and L*C*h systems were selected because these are the most-used systems for evaluation of the colour of dried food materials. The instrument was standardized each time with a white ceramic plate. Three readings were taken at each place on the surface of samples and then the mean values of L*, a* and b* were averaged. The different colour parameters were calculated using the following equations [13]. Hue angle (h) indicating colour combination (i.e. browning) is defined as: 1 * * (when a* 0) tan (b / a ) h . (1) 1 * * (when a* 0) 180 tan (b / a ) The chroma (C*) indicating colour saturation is defined as: 2 2 C* = (a* +b* )1/2 . (2) 2.3. Economic analysis The total capital cost for the greenhouse solar dryer (CT ) is given by the following equation: CT = Cm +Cl , (3) 5 Where Cm is the material cost of the dryer and Cl is the labor cost for the construction. The annual cost calculation method proposed by Audsley and Wheeler [17] yields: Cannual CT N i 1 w 1 Cmaint,i Cop,i wi N w w 1 , (4) Where Cannual is the annual cost of the system. Cmaint,i and Cop,i are the maintenance cost and the operating cost at the year i respectively. w is expressed as: w = 100 +iin 100 +i f , (5) Where iin and if are the interest rate and the inflation rate in percent, respectively. The operating cost (Cop) is the labour cost for operating the dryer (Clabour,op). The maintenance cost of the first year was assumed to be 1% of the capital cost. The annual cost per unit of dried product is called the drying cost (Z, USD/kg). It can be written as: C Z = annual , (6) M dry Where Mdry is the dried product obtained from this dryer per year. Payback period CT , M dry Pd M f Pf M dry Z (7) Where Mdry is annual production of dry product (kg), Mf is the amount of fresh product per year (kg), Pd is the price of the dry product (USD/kg) and Pf is the price of the fresh product (USD/kg). 2.4. Neural Network Modeling The neurocomputing methods are shaped after biological neural functions and structure. As a result, they are generally known as artificial neural network (ANN). Similar to biological neural network, the function of ANN are developed not by programming them, but by exposing them to given sets of input and output data on which they can learn how to perform a required task. In such modeling approach, a formulation of analytical description of a process is not required. Instead, a black-box process model is created by interacting the network with representative samples of measurable quantities characterizing the process. In this work, a multilayer ANN model of the greenhouse solar dryer for drying the bael fruits was developed; the model has four and five-layered network. This network consists of a large number of processing elements, called neuron (Fig. 5 and Fig. 6). Hidden Layer-1 Hidden Layer-1 Input Layer Air temperature (T) Hidden Layer-2 Input Layer Air temperature (T) Output Layer Relative humidity (rh) Relative humidity (rh) Airflow rate ( m ) Final moisture content (Mf) Solar radiation (I t ) Initial moisture content (Mi) Figure 5. The structure of the artificial neural network (5:10:5:1)of the greenhouse solar dryer for drying bael fruits Airflow rate ( m ) Solar radiation Hidden Layer-2 Hidden Layer-3 Output Layer Final moisture content (Mf) (I t ) Initial moisture content (Mi) Figure 5. The structure of the artificial neural network (5:10:5:3:1)of the greenhouse solar dryer for drying bael fruits 6 The neural networks with various structures were investigated, including 4 and 5 layers with different number of neurons in each hidden layer, different values of learning rate and momentum. The best ANN structure was selected on the basis of the lowest error on the training and verification of ANN. The best ANN model and optimum values of network parameters were obtained by trial and error. The performances of the various ANN configurations were compared using the coefficient of determination (R2) and the root mean square error, RMSE. The input layer of the model comprises five neurons which correspond to solar radiation (It), airflow rate ( m ), air relative humidity (rh), air temperature (T) and initial moisture content (Mi). The output layer has one neuron which represents the final moisture content (Mf). A selection of the number of neurons for hidden layers is optional. A large number of neurons can represent the system more precisely but it is more complicated to obtain proper training of the network. In this work, (model 1) the selected number of neuron in hidden layer 1, 2 and 3 of the model are 5, 10 and 3 respectively, (model 2) the selected number of neuron in hidden layer 1, 2, 3 and 4 of the model are 5, 10, 5 and 3 respectively . ANN is able to modify its behavior in response to its environment. Unlike analytical model, the structure of ANN cannot represent the system behavior, unless it is properly trained. The aim of training the network is to adjust the weights of the interconnecting neurons of the network so that an application of a set of inputs produces a desired set of output. Initially, random values are given as weights. One input-output set can be referred to as a vector. Training assumes that each input vector is paired with target vector representing the desired output and these are called a training pair. In general, a network is trained over number of training pairs. In this work, the ANN model of the greenhouse solar dryer was trained by the back propagation algorithm [14-16]. The procedures of the training are as follows: (1) An input vector was applied. (2) The output of the network was calculated and compared to the corresponding target vector. (3) The difference (error) between the calculated and the target outputs was fed back though the network. (5) Weights were changed. This procedure was repeated over the entire training set until the error was within an acceptable value or until the outputs did not significantly change any more. The ANN model was programmed in C++. 3. Results and Discussions 3.1. Drying characteristic of bael fruits Drying experiments of bael fruits in the greenhouse solar dryer were carried out in 25 June 2015-8 August 2015. Ten batches of experimental run were carried out. The typical results are shown in Fig. 6Fig. 9. Solar radiation (W/m2)aa 1000 6/08/2015 800 7/08/2015 600 400 200 0 Time (hr) Figure 6. Variation of solar radiation with time of the day during drying of bael fruits 8/08/2015 7 Temperature (๐C) Solar radiation from Fig. 6, it was observed that solar radiation was strongly fluctuated on the second day of the experiment due to cloud. On the first, third days, solar radiation was low because of rain (Fig. 6.). The comparison of air temperature at three different locations inside the dryer and the ambient air temperature for the experimental runs of solar drying of bael fruits are shown in Fig. 7. 70 60 50 40 30 20 10 0 6/08/2015 T1 T2 7/08/2015 8/08/2015 T3 T_ambient T_outlet Time (hr) Figure 7. Variation of ambient temperature and the temperature at different positions inside the greenhouse solar dryer during of bael fruits. Relative humidity (%)aa The pattern of temperature change in different positions was comparable for all locations. Temperatures in different positions at these five locations varied within a narrow band. In addition, temperatures at each of the locations differed significantly from the ambient air temperature. The relative humidity two different locations inside the dryer and ambient air relative humidity during solar drying of bael fruits are shown in Fig. 8. 100 80 7/08/2015 6/08/2015 8/08/2015 60 40 20 Inlet Ambeint Outlet 0 Time (hr) Figure 8. Variation of relative humidity with time of the day during drying of bael fruits Relative humidity decreased over time at different locations inside the dryer during the first half of the day while the opposite is true for the other half of the day. No significant difference was found between relative humidity of different positions inside the dryer. However, there was a significant difference in relative humidity for all locations inside the dryer compared to the ambient air. The relative humidity of the air inside the dryer was lower than that of the ambient air. Hence, the air leaving the dryer had lower relative humidity than that of the ambient air and this indicated that the exhaust air from the dryer still had drying potential for recirculation to dry the product. 8 Moisture content (%, wb.)aa The comparison of moisture content at three different locations inside the dryer and the open sun drying for the experimental runs of solar drying of bael fruits are shown in Fig. 9. 70 60 50 40 30 20 10 0 6/08/2015 7/08/2015 Greenhouse dryer 8/08/2015 Open sun drying Time (hr) Figure. 9 Comparison of the moisture contents of bael fruits at different positions inside the greenhouse dryer with those obtained by the open air sun drying method The moisture content of bael fruits in the greenhouse solar dryer was reduced from an initial value of 59-60% (w.b.) to a final value of 2.0-4.0% (w.b.) within 3 days whereas the moisture content of the natural sun-dried samples was reduced to 10.8% (w.b.) in the same period. 3.2. Colour measurement of dried bael fruits The colour of bael fruits dried using the greenhouse solar dryer was compared with that of bael fruits obtained by the open air sun drying. The bael fruits dried using the greenhouse dryer has the colour changed from dense/white colour to bright gel /yellow after having dried (L* = 44.57). The colour of bael fruits dried using open air insolation is dark – yellow (L* = 30.47). This result indicated that drying bael fruits using the greenhouse dryer has better than the open air insolation. 3.3. Economic evaluation As there are now several units of this type of dryer are being used for production of dried bael fruits, information used for economic evaluation is based on the field level data and recent prices of the materials used for construction of the dryers. Data on costs involved and economic parameters are shown in Table 1. 9 Table 1. Data on costs and economic parameters Items Costs and Economic Parameters Polycarbonate plates 160 USD Solar modules and fans 266 USD Materials of constructions 100 USD Labour costs for constructions 88.9 USD Repair and maintenance cost 1% of capital cost per year Operating cost 171 USD per year Price of fresh bael fruits 0.889 USD kg-1 Price of dried bael fruits 5.334 USD kg-1 Expected life of the dryer 15 years Interest rate 1.5% (Bank of Thailand) Inflation rate 1.0% (Bank of Thailand) In term of economic evaluation, the capital cost for construction and installation of the solar greenhouse dryer is estimated to be USD 1,000 (1USD=33.74 baht). It is estimated that 1,800 kg of dry bael fruits are produced annually. Based on these production scales, capital and operating costs, the payback period of the greenhouse solar dryer for drying bael fruits is estimated to be about 1 year. Moisture content (%, wb.)aa 3.4. Performance prediction by ANN model The ANN model of the solar dryer developed for bael fruits drying were trained with the experimental data from nine experiments. The data from the tenth experiment were reserved for testing the model. After 100,000 times of iteration step of training, the square sum of difference (error) between the observed and the predicted output reached a significant low level. The comparison between the modelpredicted and measured moisture contents of the dryer is shown in Fig. 10. 70 60 50 40 30 20 10 0 6/08/2015 7/08/2015 ANN model 1 ANN model 2 Minside1 M_ANN2 M_ANN1 M_ambeint 8/08/2015 Open sun drying Time (hr) Figure 10. Predicted and measured moisture contents of bael fruits From Fig. 10, it is found that the agreement between the predicted (model 1, model 2) and measured moisture contents is good and the root mean square difference (RMSD) is 0.996% and 0.998% with respect to the mean measured value. Thus, if the model is adequately trained, it can appropriately predict the performance of the solar dryer for drying bael fruits. 10 4. Conclusion In order to investigated the performance of a greenhouse solar dryer, ten batches of bael fruits were dried in the greenhouse solar dryer at Loei Province, Thailand. Solar drying in the greenhouse solar dryer resulted in considerable reductions in drying time as compared with the natural sun drying and the colour of products dried in the greenhouse solar dryer are better than natural sun dried samples. The estimated payback period of the greenhouse solar dryer is about 1 year. The ANN model was able to predict variations of MC quite well with determination coefficients (R2) of 0.996 and 0.998 for training, validation and testing, respectively. The prediction mean square error was obtained as 0.003 and 0.0028 for training, validation and testing, respectively. Results show good agreement between the experimental data on the one hand and mathematical models as well as the ANN model on the other. Acknowledgements This research was supported by Faculty of Science and Technology, Loei Rajabhat University, Loei Province, Thailand. References [1] Maity, P., Hansda, D., Bandyopadhyay, U., & Mishra, D.K. (2009). Biological Activity of crude extracts and chemical constituents of bael. Aegle marmelos (L.) Correa. Ind J Experimental Biology, 47, 849-861. [2] Janjai, S., & Bala, B.K. (2012). Solar drying technology. Food Engineering Reviews, 4,16-54. [3] Duffie, J.A.,& Beckman, W.A. (1991). Solar Engineering of thermal Processes. John Wiley and sons, New York. [4] Janjai, S., Laksanaboonsong, J., Nunez, M., & Thongsathitya, A. (2005). Development of a method generating operational solar radiation maps from satellite data for a tropical environment. Solar Energy, 78, 739-751. [5] Murthy, MV. (2005). A review of new technologies, models and experimental investigations of solar dryers. Renewable Energy Reviews, 13, 835-844. [6] Janjai, S., & Kaewparsert, T. (2006). Design and performance of a solar tunnel dryer with a polycarbonate cover. International Energy Journal, 7, 187-194. [7] Janjai, S., Khamvongsa, V., & Bala, B.K. (2007). Development, Design and Performance of a PV-Ventilated Greenhouse Dryer. International Energy Journal, 8, 249-258. [8] Janjai, S., Lamlert, N., Intawee, P., Mahayothee, B., Boonrod, Y., Haewsungcharern, M., Bala, B.K., Nagle, M., & Muller, J. (2009). Solar drying of peeled longan using a side loading type solar tunnel dryer: Experimental and simulated performance. Drying Technology, 27, 595–605. [9] Janjai,S., Lamlert, N., Intawee, P., Mahayothee, B., Bala, B.K., Nagle, M., & Muller, J. (2009). Experimental and simulated performance of a PV-ventilated solar greenhouse dryer for drying of peeled longan and banana. Solar Energy, 83, 1550-1565. [10] Fudholi, A., Sopian, K., Ruslas, M.H., Alghoul, M.A., & Sulaiman, M.Y. (2010). Review of solar dryers for agricultural and marine products. Renewable and Sustainable Energy Reviews, 14, 1-30. [11] Janjai, S., (2012). A greenhouse type solar dryer for small-scale dried food industries: Development and dissemination. International Journal of Energy and Environment, 3, 383-398. [12] Krokida, M.K., Tsami, E., & Maroulis, Z.B. (1998). Kinetics on color changes during drying of some fruits and vegetables. Drying Technology, 16 (3–5), 667–685, [13] Lopez Camelo, A.F., Gomez, P.A. (2004). Comparison of colour indexes for tomato ripening. Horticultural Brassica, 22(2), 534-537, [14] Patterson, D.W. (1996). Artificial Neural Networks; Theory and applications. Prentice Hall, New York. [15] Janjai, S., Piwsaoad, J., Nilnont, W., & Pankaew, P. (2015). Experimental performance and Neuron Network modeling of a Large-scale Greenhouse solar dryer for drying Natural rubber sheets. Journal of Control Science and Engineering, 1(1), 48-53. [16] Piwsaoad, J., & Phusampao, C. (2015). Experimental Performance and Neural Network Modeling of Solar Dryer for Drying Bamboo shoot strips in Thailand. International Journal of Scientific & Engineering Research, 6 (7), 1306-1310. [17] Audsley, E., & Wheeler, J. (1978). The annual cost of machinery calculated actual cash flows. Journal of Agricultural Engineering Research, 23, 189-201.