MACHINE LEARNING BASED ESTIMATION OF DRYING CHARACTERISTICS OF APPLE SLICES

Artificial neural networks (ANNs) and multiple linear regression (MLR) for prediction of moisture content for coated pineapple cubes

Nephelium Lappaceum or Rambutan is a seasonal fruits that is very popular in Southeast Asia region. It has short storage life, about 4 to 5 days after harvesting. In this study, an oven was used to dry Rambutan and the effect of chemical pre-treatment on the Rambutan was investigated. Two different sets of pre-treatment were used, namely set B (treatment by using calcium chloride, malic, ascorbic and citric acid, sucrose and trehalose) and C (treatment by using sucrose and citric acid). Untreated sample (Set A) was prepared as a control. Drying temperature of 50oC was used throughout the experiment for 24 hours drying time. Weights of the samples for each set were measured at interval of 10 min, 15 min, 30 min and 1 hour. Drying rate (DR) and moisture ratio (MR) were calculated from the results. Calculated DR and MR were fitted to five different thin layer mathematical models namely, Lewis Model, Page Model, Handerson and Pabis Model, Logarithmic Model and Two-term Model. The value of coefficient of determination (r2), reduced sum square error (SSE) and root mean square error (RMSE) were compared in order to find the best mathematical model for drying kinetics of Rambutan. The results showed that Logarithmic Model gives the highest r2, lowest SSE and RMSE value for set A (r2 = 0.9972, SSE =0.000285 and RMSE = 0.000.016881), set B (r2 = 0.9950, SSE = 0.000535 and RMSE = 0.023138) and set C (r2 = 0.9950, SSE = 0.000624 and RMSE = 0.024989). Thus, Logarithmic model is the best model that suit to represent drying characteristics of Rambutan. The effective moisture diffusivity also estimated by using Fick's diffusion model; set A (3.11 × 10−10), set B (3.32 × 10−10) and set C (2.99 × 10−10).

In this study, the effect of ultrasound pretreatment (UP) and drying methods (convective hot air [CHAD] and microwave oven [MOD]‐assisted foam mat drying [FMD]) on the drying time, behavior and rate, diffusion coefficient, moisture content, color, bulk and tapped densities, flow properties, and reconstitution behavior of taro (Colocasia esculenta) flour were assessed. For this purpose, the taro foams with and without ultrasonic pretreatment (35 kHz, 80°C for 25 min) were dried in a convective oven (70°C, 20% ventilation rate) and microwave oven (460 W). The UP resulted in a higher drying rate, diffusion coefficient, moisture content, Hausner ratio (HR), and wettability time, and a lower drying time compared to the nontreated samples. Combined CHAD and MOD‐assisted FMD improved the disadvantages of the CHAD method for example: long drying time (decreased around 63.33% for CHAD and 59.16% for UP + CHAD), low drying and diffusion rates, high moisture content, low brightness (L*) values, high Carr index (CI), HR, and times of wettability and solubility. Additionally, the CI and HR of the taro flours ranged between 4.29–13.96 and 1.04–1.16 which can also be classified as very good and low levels, respectively.Practical ApplicationsTaro flour is an important alternative flour source in baby food, cake, chips, noodle, and so forth due to its pasting, functional, and nutritional properties. In order to obtain taro flour, the taro corm (root) is washed, peeled, and sliced then generally blanched with distilled water, dried, and ground. The quality of taro flour depends on the preparation method such as the drying method and conditions. Foam mat drying (FMD) is an alternative technique and it has several advantages like low drying time and energy consumption, a high drying rate, increased drying surface area, porous structure, and the desired instant properties of the obtained powders. Ultrasound pretreatment (UP) has also some advantages such as increasing the mass transfer rate, reducing drying time, and energy consumption. In this study, the advantages of these techniques (FMD and UP) were combined and the effects on the drying behavior, time, effective moisture diffusivity, and the physical and powder properties were observed.

Microwave-vacuum drying behavior of rosehips: Experimental investigation and mathematical modeling

Investigation of effective moisture diffusivity (Deff) and activation energy (Ea) of cassava were conducted under convective drying at temperature and velocity of 60, 70 and 80 °C, and 1.0, 1.5 and 2.0 m/s, respectively. In the experiment, cassava was sliced into 3 mm-thickness and dried under given conditions until mass was saturated. Deff and Ea were described by Fick’s second law and Arrhenius-type equation, respectively. The experimental results indicated that the increase in Deff was significantly affected by increasing the hot air temperature and velocity. The slope method was used to calculate average Deff, and results were found to range from 3.83 × 10–9 – 9.86 × 10–9 m2/s. The Ea was found to decrease with an increase in hot air velocity, ranging from 21.23– 24.92 kJ/mol. Additionally, Moisture content (Mw) and Drying rate (DR) were also used to describe the drying kinetics. From the experimental results, Mw and DR decreased with an increase in drying time. DR increased with an increase in temperature and velocity causing Mw to rapidly decrease and drying time to reduce. The highest DR was found to be 0.55 gwater/min at temperature of 80 °C and velocity of 2.0 m/s.

In the present study, the freeze drying behavior of apples have been modeled and predicted. Because freeze-drying is a very expensive and complex process, modeling of the freeze-drying process is a challenging task. In this study, a novel data scaling method called multiple output–dependent data scaling (MODDS) has been proposed and combined with an adaptive neuro-fuzzy inference system (ANFIS) to forecast the moisture content (MC), moisture ratio (MR), and drying rate (DR) values, which are outputs of freeze-drying behavior of apples. The input parameters of the freeze drying system are the sample thicknesses, drying time, pressure, relative humidity, chamber temperature, and sample temperature. Using the input parameters, the outputs of the freeze-drying process of apples were predicted using a hybrid system based on MODDS and ANFIS. In the first stage, only input parameters were scaled using MODDS. In the second stage, the outputs of freeze drying of apples were predicted with the scaled input parameters using ANFIS algorithm. Ninety-two samples were included in the data set, including 10-, 7-, and 5-mm samples. In order to evaluate the performance of the proposed model, the mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), coefficient of determination (R 2), index of agreement (IA), and mean absolute percentage error (MAPE) were used. Though MSE values of 2.48, 0.035, and 0.011 and IA values of 0.887, 0.887, and 0.466 were obtained for MC, MR, and DR, respectively, using the ANFIS prediction algorithm the hybrid MODDS-ANFIS model achieved MSE values of 0.003, 0.00005, and 0.00007 and IA values of 0.999, 0.999, and 0.993 for the prediction of MC, MR, and DR, respectively. The results obtained demonstrate that the proposed hybrid system is a robust and efficient method for the modeling and prediction of freeze-drying behavior of apples.

Determination of freeze-drying behaviors of apples by artificial neural network

Heat and Mass Transfer in U-Bend of a Pneumatic Conveying Dryer

Butiá (Butia capitata) is a typical Brazilian Cerrado fruit, rich in several bioactive compounds. This work aimed to study the influence of air temperature on drying kinetics and quality parameters of butiá pulp. The pulps were dried at 50 and 70 °C. Mathematical models were fitted to the moisture ratio data. The effective moisture diffusivity (Deff) and the drying rate (DR) were calculated. The fresh and dried pulps were characterized in terms of moisture content (MC), water activity (aw), total carotenoids content (TCC), yellow flavonoids, total phenolic content (TPC), antioxidant capacity and color. The Page model was the one that best fitted to the moisture ratio data. Drying reduced MC, aw and the bioactive compounds content and altered colorimetric parameters. The higher temperature resulted in lower TCC and higher total color difference. However, it reduced the drying time (from 300 to 180 min), with higher Deff and DR and resulted in samples with higher retention of yellow flavonoids, TPC and total antioxidants and a lower browning index. Therefore, 70 °C was the most suitable temperature for drying butiá pulp.

The moisture content and shrinkage of a sample are often ignored when determining its effective moisture diffusivity (EMD), which will result in a larger error in simulating the dynamic moisture diffusion process during drying. This work aimed to determine the EMDs of cortex and core in axial and radial directions considering the combined influence of moisture content and shrinkage of material. The drying and shrinkage characteristics of cortex and core were investigated under different hot air drying conditions, and the EMDs of the cortex and core in different drying stages were determined by the slope method considering the moisture content and shrinkage. The results showed that the EMDs were dependent on the carrot component rather than its directions. The EMDs of the core and cortex were different. It decreased slowly at the early drying stage, and decreased sharply at the later drying stage, and increased with the hot air temperature increasing. A third order polynomial relationship was established to correlate the EMDs with the moisture content and hot air temperature (R2 > 0.9460), and was verified by experiments. The maximum weighted absolute percentage error of the moisture content from simulation and experiment was only 6.11%, more accurate than that based on a constant effective moisture diffusivity (CEMD) for characterizing the intrinsic moisture diffusion in carrot during drying.Practical applicationsThe EMD is an important parameter that characterizes intrinsic moisture diffusion in material and often used for modeling and calculation in food drying. The investigation methods developed in this study can be applied for determining the EMD of samples with shrinkage and variable moisture content. This study would be helpful for understanding the moisture transfer mechanism and optimizing operation conditions in carrot drying industry.

In the broader healthcare domain, the prediction bears more value than an explanation considering the cost of delays in its services. There are various risk prediction models for cardiovascular diseases (CVDs) in the literature for early risk assessment. However, the substantial increase in CVDs-related mortality is challenging global health systems, especially in developing countries. This situation allows researchers to improve CVDs prediction models using new features and risk computing methods. This study aims to assess nonclinical features that can be easily available in any healthcare systems, in predicting CVDs using advanced and flexible machine learning (ML) algorithms. A gender-matched case-control study was conducted in the largest public sector cardiac hospital of Pakistan, and the data of 460 subjects were collected. The dataset comprised of eight nonclinical features. Four supervised ML algorithms were used to train and test the models to predict the CVDs status by considering traditional logistic regression (LR) as the baseline model. The models were validated through the train-test split (70:30) and tenfold cross-validation approaches. Random forest (RF), a nonlinear ML algorithm, performed better than other ML algorithms and LR. The area under the curve (AUC) of RF was 0.851 and 0.853 in the train-test split and tenfold cross-validation approach, respectively. The nonclinical features yielded an admissible accuracy (minimum 71%) through the LR and ML models, exhibiting its predictive capability in risk estimation. The satisfactory performance of nonclinical features reveals that these features and flexible computational methodologies can reinforce the existing risk prediction models for better healthcare services.

Response surface methodology was employed to investigate the effects of drying temperature (15.0–45.0C), air velocity (1.00–2.00 m/s), moisture content at conversion point (40.0–60.0%) and microwave (MW) power (1.0–3.0 W/g) on the average drying rate (DR), specific moisture evaporation rate (SMER), total color difference (ΔE), rehydration ratio (RR) and shrinkage ratio (SR) of yacon dried by combined heat pump (HP) and MW methods. A central composite rotatable design was used to develop models for responses. The coefficients of determination R2 of DR, SMER and ΔE were higher than 0.7. However, R2 of RR and SR were 0.540 and 0.343, respectively. Based on response surface and desirability functions, the optimum conditions for combined HP and MW drying of yacon were drying temperature of 42.7C, air velocity of 1.69 m/s, moisture content at conversion point of 50% and MW power of 2.0 W/g. At this optimum point, DR, SMER and ΔE were 0.262 kgH2O/kg(d.b.)·h, 0.222 kg/kW·h and 23.59, respectively. Practical Applications Yacon are rich in essential amino acids, minerals and fructooligosaccharides (FOS). However, FOS hydrolyzation, browning and decay are the common phenomena during postharvest storage and transportation. Hybrid drying techniques are being developed to minimize weakness and maximize merits of different drying techniques to produce better quality of foodstuffs, reduce drying time, decrease energy consumption and increase drying efficiency. Therefore, response surface methodology was employed to investigate the effects of drying temperature, air velocity, moisture content at conversion point and MW power on the average DR, SMER, ΔE, RR and SR of yacon dried by combined HP and MW methods, and the optimum process conditions were determined to obtain the criteria of maximum DR and SMER, and minimum ΔE. This research will provide valuable information for industrialization of yacon dried by combined HP and MW methods.

This study aimed to evaluate the influence of drying air temperature on drying kinetics and the physicochemical properties of dried bananas. Banana slices were dried at 40, 60, and 80°C. Drying was terminated when the samples had a moisture of 20%. Mathematical models were fitted to the moisture ratio. A generalized model of moisture was developed, to predict the moisture of samples as a function of drying time and temperature. The effective moisture diffusivity (Deff), activation energy and the drying rate (DR) were calculated. Electrical energy consumption was measured. The moisture, water activity, reducing and total sugar content, acidity, hardness, and color of the dried bananas were evaluated. The moisture decreased during drying, obtaining a mean value of 20.132%. All mathematical models fitted well to the MR data, with a determination coefficient greater than 0.95. The Midilli model was that which best fitted. The higher temperature resulted in higher Deff (3.538 × 10−9 m2 s−1) and DR, less drying time (120 min) and electrical energy consumption (4.319 kWh), higher reducing sugar content (47.51%) and hardness (28.187 N), lower acidity (1.038%), more yellow tonality (78.04°), and higher chromaticity (18.49) of the dried bananas. The optimum temperature for drying bananas was 80°C.Practical ApplicationsBanana is a fruit widely consumed in the world. However, it is very perishable, causing great waste and financial loss. Convective drying is a simple and low‐cost method, widely used in fruit processing to produce new products and extend the shelf life of food. The dried banana is a product of good sensory acceptance and therefore drying is a good alternative for processing the banana pulp. The temperature of the drying air is one of the principal parameters of this process since it influences the drying kinetics and the physicochemical properties of the dry product. Therefore, it is important to study different drying air temperatures to determine the best drying condition for dried banana production, to optimize the dried banana process and properties.

In order to improve the drying characteristics and to optimization of drying conditions, machine learning (ML) and response surface methodology (RSM) were applied in air-convective drying of orange slices (Washington Navel and Valencia cultivars). Interactions of temperature (T, 50–60°C), sample thickness (ST, 5–9 mm), and drying time (DT, 8–10 h) like independent variables with specific moisture extraction rate, effective moisture diffusivity, energy efficiency, and energy consumption like dependent variables were determined. In addition, five machine learning algorithms (random forest-RF; artificial neural network-ANN; gaussian processes-GP support vector regression-SVR, and k-nearest neighbors-kNN) were used to predict moisture ratio and drying rate. In Washington Navel and Valencia cultivars, the greatest correlation coefficients (R) for prediction of moisture ratio were obtained k-NN algorithm with values of 0.9944 and 0.9898, respectively. Also, drying rate prediction results showed that k-NN achieved higher R with values of 1.0000 and 0.9954, respectively. Experimental findings were adapted by a second-degree polynomial model through variance analysis to identify model fitness and optimal drying conditions. Combined desirability value was calculated as 0.8812 for Valencia and 0.8564 for Washington. Increasing energy consumption was encountered with increasing drying time and sample thickness. Besides, energy consumption had a decreasing trend at higher temperatures. Practical applications Machine learning models are novelty and rapid methods that have been successfully utilized to solve such challenges agricultural commodities. Drying is common process to preserve the food quality. This study provides optimum conditions for drying orange slices in single unit air-convective dryer and improves the effect of drying system on some drying characteristics energy aspects. In addition, this study can be able to present a technical knowledge for orange slice drying and related equipment design.

HighlightsThis research investigated the feasibility of using a microwave (MW) set at 915 MHz frequency to dry high moisture content (MC) parboiled rough rice at 44.3% MC dry basis (d.b.).The research evaluated the impacts of specific power delivered during the drying of parboiled rough rice using a MW on post-drying milling characteristics.The volumetric heating phenomenon provided by MW offered a method to quickly remove 23.1% points of moisture from parboiled rough rice in one-pass to a MC of 21.2% d.b., with minimal impacts on the kernel quality.The findings suggest that increased MW specific powers have a positive effect on rice MC reduction but negatively affects the rice milling characteristics, especially the head rice yield.The study recommended that MW specific powers exceeding 2.92 kW.[kg-DM]-1 should not be exceeded during drying of parboiled rough rice to preserve the rice milling yields.ABSTRACT. The objectives of this research were to study the impacts of specific power of MW generated at 915 MHz frequency to dry high MC parboiled rough rice on moisture removal and milling characteristics of the parboiled rough rice. Long-grain rough rice of the cultivar (cv.) Mermentau at harvest MC of 31.6% dry basis (d.b.) was parboiled by soaking at 73°C for 3 h and then steamed at 67 kPa for 10 minutes. Following the parboiling process the sample was subjected to the MW drying. The drying was accomplished at MW specific powers that ranged from 1.10 to 8.77 kW. [kg-DM] -1 and 0.37 to 2.92 kW. [kg-DM] -1 (power per unit dry matter mass of the grain). These treatment levels of MW specific power were varied by heating parboiled rough rice for 2 and 6 minutes (min) at MW powers that ranged from 1 to 24 kW. The process of parboiling increased the rough rice MC to 44.3% dry basis (d.b.). During the MW drying, as the specific power increased, the general tendency was for rough rice final moisture content (FMC), milled rice yield (MRY) and head rice yield (HRY) to decrease while the drying rate increased. Parboiled rough rice samples treated with a specific power of 8.77 kW.[kg-DM] -1 while maintaining specific energy input at 0.29 kWh.[kg-DM] -1 had least-square means FMC, drying rate, MRY and HRY of 19.7% d.b. (S.D ± 1.1%), 12.3% d.b. [min-1] (S.D ± 0.8%) (2 min drying duration), 68.18% (S.D ± 1.70%) and 67.51% (S.D ± 0.73%) respectively. However, treatment at a lower specific power of 2.92 kW.[kg-DM] -1 while maintaining the same specific energy input of 0.29 kWh.[kg-DM]-1) resulted in least-square means FMC, drying rate, MRY and HRY of 21.2% d.b. (S.D ± 0.5%), 3.9% d.b. [min-1] (S.D ± 0.1%) (2 min drying duration), 73.22% (S.D ± 0.84%) and 73.21% (S.D ± 0.21%) respectively. The increased drying rates for treatments with higher specific power was associated with higher treatment powers and shorter treatment durations. Higher specific powers negatively impacted the observed MRY and HRY. The findings suggest that increased MW specific powers have a positive effect on rice MC reduction but above a certain threshold of specific power (2.92 kW.[kg-DM]-1) may negatively affect the milling characteristics of the parboiled rice. When used to dry high MC parboiled rough rice, rice processors should know that there exists an optimum drying rate that if exceeded the milled rice quality is negatively affected thus generating an economic loss to the parboiled rice industry. The volumetric heating phenomenon provided by microwave (MW) offers a means to quickly dry high MC parboiled rough rice. This can translate to considerable economic savings for the rice processor who often experiences low drying rates because of limited drying capacity, especially at peak rice harvest times. When drying rates are optimized, rice processors can expect minimal impacts on the kernel quality which can also be translated to considerable economic savings for the rice processor. Keywords: 915 MHz microwave, Microwave drying, Milling Quality, Parboiled rice, Specific power.