FORECASTING THE MOISTURE RATIO OF WILD SPINACH (GNETUM AFRICANUM) USING BIO-INSPIRED ALGORITHMS.pptx

FORECASTING MOISTURE RATIO

1. FORECASTING THE MOISTURE RATIO OF WILD SPINACH ( GNETUM AFRICANUM ) USING BIO-INSPIRED ALGORITHMS

2. OUTLINE Background of study Statement of the problem Aim and Objective of the study Literature Review Materials and methods Results and Discussion Conclusion and Recommendation

3. BACKGROUND OF STUDY Drying as the name implies is a moisture removal technique used in the preservation of high moisture content vegetables and fruits. It is a chemical engineering unit operation/process. The objective associated with drying of food products is the reduction of moisture content to prolong the life span of the products by reducing water activity to a level low enough where microbial growth, enzyme reactions and other deteriorative reactions are inhibited ( Khawas et al., 2013). Neural networks are recognized as good tools for dynamic modeling, and have been extensively studied since the publication of the perceptron identification methods ( Rumelhart et al ., 1986). The interest of such model includes the modeling without any assumptions about the nature of underlying mechanism and the ability to consider non-linearities and interactions between variables (Bishop, 1994). Recent results establish that it is always possible to identify a neural model based on the perceptron structure, with only one hidden layer, for either steady state or dynamic operations. An outstanding feature of a neural network is the ability to learn the solutions of problems from a set of examples, and to provide smooth and reasonable interpolations for new data.

4. STATEMENT OF PROBLEM According to Majid et al ., 2013, Models that predict the drying phenomenon of agricultural products fall mainly into three categories: Theoretical, Semi-Theoretical and Empirical. Although empirical mathematical models give accurate result for a specific experiment, it cannot predict drying trends over a range of drying parameters. An attempt would result in complexity and long computing time ( Tripathy and Kurmar , 2009). This Problem may be avoided by use of analytical drying models. But these models are generally solutions of simultaneous heat and mass transfer differential equations and the final result may be very complicated and difficult to use in actual drying systems. 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 ., observed that ANN model was more productive and precise than mathematical modeling methods for predicting changes in the moisture ratio of green bell pepper during hot air fluidized bed drying.

5. AIM AND OBJECTIVES OF STUDY The major aim of this work is to use a neural network tool to predict the removal of moisture content from Wild spinach and also compare the predicted data with the experimental data generated. The objectives of this study are listed below: To model experimental drying data of Wild spinach during Oven drying process using ANN methodology. Compare the performance of the neural network in relation to the data predicted with the experimental data generated. To show the ability of ANN models to describe the drying behavior of Wild spinach in different experimental conditions.

6. SCOPE AND LIMITATION Drying of wild spinach using DHG-9101 Laboratory Drying Oven Training of experimental data using neural network toolbox. Due to unavailability of data, the sturdiness of the developed model will not be tested with new sets of data.

7. LITERATURE REVIEW Author(s) Aim(s) Mode input(s) Model Output (s) Best Model(s) Aktas et al., 2015 Predict moisture content and total energy consumption of bay leaves drying process in a closed-loop heat pump dryer. Drying time, drying air temperature, relative humidity, volumetric flow rate. Moisture content and total energy consumption A four-hidden layer MLP ANN with 6,3,6 and 2 neurons in the hidden layers respectively. Khawas et al., 2016 Predict and optimize quality parameters of culinary bananas during vacuum drying. Drying temperature, sample slice thickness and pre-treatment. Rehydration ratio, scavenging activity, non-enzymatic browning and hardness. A one-hidden layer MLP ANN model with 5 hidden neurons and sigmoid transfer function. Tarafdar et al., 2017 Modeled water activity of button mushrooms during freeze drying. Time, Initial moisture content, vacuum pressure, sample thickness and primary and secondary drying temperatures. Water activity. A one-hidden layer MLP ANN model with 13 neurons, LM error minimization algorithm and hyperbolic tangent transfer function. Li et al., 2016 Developed and evaluate a recurrent self-evolving fuzzy neural network predictive control scheme for microwave drying process. Current temperature, Current moisture and current input power. Current temperature, current moisture and current input power. Recurrent self-evolving fuzzy neural network.

8. MATERIALS AND METHOD Materials: Wild Spinach Drying Pan Electrical weighing balance Electrical Drying Oven Knife etc. Methods: Sample preparation Calibration of Drying Oven Drying of sample Training of Experimental Data’s (Simulation)

9. RESULTS AND DISCUSSION The type of network used in this work was the multi-layer perceptron network. Multi-layer perceptron network is one of the most popular and successful neural network architectures, which are suited to a wide range of applications such as prediction and process modelling. MLP network comprises a number of identical units organized in layers, with those on one layer connected to those on the next layer, so that the outputs of one layer are regarded as inputs to the next layer. MLP neural networks are normally trained using a supervised training algorithm. Different types of activation functions can be utilized for this network; however, the common ones, which are sufficient for most applications, are the sigmoidal and hyperbolic tangent functions. Experimental data from this study were used to train and test an Artificial Neural Networks model for prediction of wild spinach moisture ratio during the drying process. A total of 162 data points was collected for the four drying temperature conditions. The experimental data were randomly divided into three sets. One set was used as training data and the rest for testing and validating the model. The training data was by default setting, randomly divided into 70% for training, 15% for both validation and testing. In this work, a one hidden layer with 10 neurons with network topology 3-1-1 was used after using different ANN model configuration for the training process.

10. RESULT AND DISCUSSION CONT’D

11. RESULT AND DISCUSSION CONT’D

12. CONCLUSION AND RECOMMENDATIONS Overview of Research Research Process and Results Deduction Recommendation

13. THANK YOU

14. Additional Information No of Hidden layer Neuron in hidden layer   MSE   R 2   Epoch Transfer function   1   10   0.00017506   0.99935   98   Tansig   1   10   0.027978   0.88896   4   Purelin   1   10   0.00014851   0.99921   17   Sigmoid   1   8   0.00065009   0.99786   29   Tansig   1   8   0.006968   0.86861   2   Purelin   1   8   0.00062686   0.9960   9   Sigmoid