Journal of Food Quality

Volume 2018, Article ID 3278595, 8 pages

https://doi.org/10.1155/2018/3278595

## Artificial Neural Network Modeling of Drying Kinetics and Color Changes of Ginkgo Biloba Seeds during Microwave Drying Process

Correspondence should be addressed to Cun-Shan Zhou; moc.361@uohznahsnuc

Received 21 September 2017; Revised 16 January 2018; Accepted 28 January 2018; Published 21 February 2018

Academic Editor: Egidio De Benedetto

Copyright © 2018 Jun-Wen Bai et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

Ginkgo biloba seeds were dried in microwave drier under different microwave powers (200, 280, 460, and 640 W) to determinate the drying kinetics and color changes during drying process. Drying curves of all samples showed a long constant rate period and falling rate period along with a short heating period. The effective moisture diffusivities were found to be 3.318 × 10^{−9} to 1.073 × 10^{−8 }m^{2}/s within the range of microwave output levels and activation energy was 4.111 W/g. The and values of seeds decreased with drying time. However, value decreased firstly and then increased with the increase of drying time. Artificial neural network (ANN) modeling was employed to predict the moisture ratio and color parameters (, , and ). The ANN model was trained for finite iteration calculation with Levenberg-Marquardt algorithm as the training function and tansig-purelin as the network transfer function. Results showed that the ANN methodology could precisely predict experimental data with high correlation coefficient (0.9056–0.9834) and low mean square error (0.0014–2.2044). In addition, the established ANN models can be used for online prediction of moisture content and color changes of ginkgo biloba seeds during microwave drying process.

#### 1. Introduction

Ginkgo biloba (GB) is the oldest relict plant of the extant gymnosperms and is referred to as an archaic living fossil. GB is native to China, and the cultivation area in China accounts for 90% of the worldwide cultivated land of GB trees [1]. In some oriental countries, including China, Japan, and Korea, GB seeds are regarded as delicious food and tonic medicine, which is a rich source of health-promoting compounds such as flavonoids, ginkgo acid, bilobol, and ginkgolides as well as carbohydrates, protein, fats, vitamins, and mineral elements [2]. As a seasonal nut, in China, GB is generally harvested in late September to October, which is a typical seasonal crop. Although it is wrapped by a hard shell, GB seed cannot be stored for a long time due to its relative high moisture content. Drying is one of the most frequently used methods to prolong the shelf life of GB seeds. The dried products can be used as materials for further processing for other products such as GB seeds powers.

The traditional drying methods for GB seeds are natural sun drying and hot air drying. Although both methods are relatively simple and inexpensive, they have several disadvantages such as long drying time, low energy efficiency, and substantial deterioration of food quality, such as degradation of color and loss of nutrients. Microwave drying is a technique that can be used as an alternative to shorten the drying time, improve the quality of the dried products, and reduce energy consumption [3]. Microwaves can penetrate into the material with the effect known as volumetric hearting [4], which can increase the drying rate in the falling drying rate period [5]. Microwave drying technology has been applied to several fruits and vegetables, such as carrot [6], nut seed [7], blueberry [8], apple slices [9], thyme leaves [10], and jujube [11].

Drying is a complex, dynamic, highly nonlinear, strongly interactive, and multivariable thermal process [12]. Therefore, the prediction of moisture content and quality parameters are very useful and necessary to improve the overall performance of drying process. Researchers usually develop mathematical models which can be classified as theoretical, semitheoretical, and empirical models to describe the drying kinetics and quality changes. Although these models can give good regression to experimental data in very specific conditions, there is no way to obtain general equations to describe the drying process of every product [13].

Artificial neural networks (ANN) offer several advantages over conventional modeling techniques because of the learning ability and being suitable to the nonlinear process. ANN models have been developed to model the moisture content and quality parameters in drying process. Jafari et al. [14] observed that ANN model was more productive and precise than mathematical modeling method for predicting changes in the moisture ratio of green bell pepper during hot air fluidized bed drying. Sarimeseli et al. [10] used ANN to describe microwave drying kinetics of thyme leaves. Behroozi Khazaei et al. [15] applied machine vision and ANN for modeling and controlling of the grape drying process in hot air dryer. Nadian et al. [16] developed an ANN model to predict the color changes of apple slices during hot air drying. Guiné et al. [17] employed ANN to characterize the antioxidant activity and phenolic compounds degradation kinetics of bananas under different drying conditions.

The objectives of current work are (i) to explore the drying characteristic and color changes kinetic of GB seeds at different microwave powers, (ii) to calculate effective moisture diffusivity and the activation energy to highlight the effect of microwave power, and (iii) to model the experimental drying kinetics and color changes of GB seeds during its microwave drying process using ANN methodology.

#### 2. Materials and Methods

##### 2.1. Materials

Fresh GB seeds were purchased from a local market in Taixing, China. The cultivar of GB is Dafozi. To ensure uniformity of physical characteristics of the experimental materials, the samples were carefully selected with the same size (average major axis, middle axle, and minor axis were 22.07 mm, 13.84 mm, and 12.03 mm, resp.). The initial moisture content of samples was determined by vacuum drying at 70°C for 24 h following the standard method (AOAC, 1990). The initial moisture content of the samples was reported as 53.02% in wet basis (w.b.) or 1.13 kg/kg in dry basis (d.b.). Prior to experiments, the kernel (nut meat) of the GB seed was obtained by shelling and removing the bronzing pellicle. All the GB seeds were stored in a refrigerator at °C and 90% relative humidity before the experiments were carried out.

##### 2.2. Drying Experiments

Drying experiments were carried out in a domestic digital microwave oven with maximum power output capacity of 700 W at 2450 MHz (P70d2otl, Galanz, China). The microwave oven has a capability to operate at four different microwave output powers (200, 280, 460, and 640 W), with measurement accuracy of ±10 W. Processing time and microwave output power were adjusted with the digital control on the microwave oven. GB seeds of 80 g were placed in a single layer on a rotating glass plate in the oven. The weight loss was periodically recorded by taking out the rotating glass and weighing it on an electronic balance within the accuracy of ±0.01 g during drying. Drying was stopped when the moisture content of the samples reached the final moisture content of 0.15 kg/kg (d.b.). All the drying experiments were conducted in triplicate.

##### 2.3. Calculation of Moisture Ratio and Drying Rate

The moisture ratio () of the samples was calculated according to [23]where , , and are moisture content at any time of drying (kg water/kg dry matter), initial moisture content (kg water/kg dry matter), and equilibrium moisture content (kg water/kg dry matter), respectively. The equilibrium moisture content was assumed to be zero for microwave drying as stated by Maskan [24].

The drying rate (DR) of samples during drying experiments was computed using [18]where and are the moisture content at and moisture content at (kg water/kg dry matter), respectively, and is drying time (min).

##### 2.4. Calculation of Effective Moisture Diffusivity

Weibull distribution can be used to calculate the effective moisture diffusivity, regardless of the characteristics of moisture migration during drying process. The MR curves were fitted to the Weibull distribution [25]where MR is moisture ratio of GB seeds; is the drying time; is the scale parameter of Weibull distribution (min); is the shape parameter of Weibull distribution.

Effective moisture diffusivity () can be calculated with the following equation [25, 26]:where is the effective moisture diffusivity (m^{2}/s); is the estimate moisture diffusivity (m^{2}/s); is the volume equivalent radius of GB seeds, with 0.769 × 10^{−2} m as its value; is the scale parameter of Weibull distribution; is the physical dimension constant. For agriculture products with a shape of sphere, the value of is 18.6 [27].

##### 2.5. Estimation of Activation Energy

Activation energy () is the minimum energy that must be supplied to break water-solid and/or water-water interactions and to move water molecules from one point to another in solid [3]. The dependence of effective moisture diffusivity on drying temperature has been shown to follow an Arrhenius relationship presented as follows:where is the preexponential factor of Arrhenius equation (m^{2}/s); is the activation energy (kJ/mol);* R* is the universal gas constant (kJ/mol K);* T* is temperature (°C).

However, during the microwave drying processes, the temperature is not a directly measured variable. The Arrhenius equation was used in a modified form to illustrate the relationship between the effective moisture diffusion and the ratio of the microwave output power to sample weight (*m/P*) instead of the temperature for calculation of the activation energy. The modified Arrhenius equation (6) derived by Dadalı et al. [18] can be effectively used as follows:where is the preexponential factor of Arrhenius equation (m^{2}/s); is the effective moisture diffusivity (m^{2}/s); is the activation energy (W/g);* m* is the mass of raw sample (g);* P* is the microwave power (W).

Equation (6) can be expressed in a logarithmic form as follows:

So the activation energy can be calculated from the slope of ln() versus the ratio of the microwave output power to sample weight (*m/P*).

##### 2.6. Color Measurement

A CIE standard illuminant D65 and observer 10° were used to determine CIE color space coordinates, (whiteness or brightness), (redness/greenness), and (yellowness/blueness). GB seed samples color was measured using a colorimeter (Color Quest X, Hunter Lab, USA) before drying and prespecified time intervals during drying. Three samples were randomly selected for color measurement. , , and values of each sample were average of 6 readings.

##### 2.7. ANN Modeling

MATLAB software (Version 7.8, MathWorks, USA) was used for the design and testing of various ANN models. The ANN configuration used in this work (Figure 1) was a multilayer “feed-forward,” consisting of one input layer, one hidden layer, and one output layer with a convergence criterion for training purposes. The input variables in the input layer are microwave power and drying time and the output variables in the output layer are moisture content and the color parameters (, , and ) of GB seeds at any time. After trial and error, network unit with hyperbolic tangent sigmoid transfer function “tansig” for neurons of hidden layer, 10 neurons in the hidden layer, linear transfer function “purelin” for neuron of output layer, and Levenberg-Marquardt training algorithm “trainlm” for the training function were selected.