Journal of Advances in Technology and Engineering Research
Details
Journal ISSN: 2414-4592
Article DOI: https://doi.org/10.20474/-jater1.1.1
Received: 19 October 2014
Accepted: 20 July 2015
Published: 15 October 2015
Download Article (PDF)
  • An Artificial Neural Network modelling of ginger rhizome extracted using Rapid Expansion Supercritical Solution (RESS) method

N.A. Zainuddin, I. Norhuda, I.S. Adeib, A.N. Mustapa, S.H. Sarijo

Published online: 2016

Abstract

This study explains the development of a feed forward multilayer back propagation with Levenberg-Marquardt training algorithm artificial neural network (ANN) to predict the particle size from an extraction of ginger rhizome using supercritical carbon dioxide in Rapid Expansion Supercritical Solution (RESS). The sizes of solid oil particle formation analysis are taken by using Scanning Electron Microscopy (SEM) and ImageJ, which is an image processing and analysis software. The ANN model accounts for the effects of different extraction temperatures (40, 45, 50, 55, 60, 65 and 70°C) and pressures (3000, 4000, 5000, 6000 and 7000psi) on the size of particles. A two-layer ANN with two inputs variables (extraction temperature and pressure) and one output (particle size) with 35 experimental data is taken for the modelling purpose. Different networks are trained and tested by adjusting the number of neurons within a hidden layer. Looking at validation data sets, a network has the highest (nearest to value of one) regression coefficient (R) at 0.99721 and the lowest (nearest to value of zero) mean square error (MSE) at 0.00031. Thus, it is stated as an optimum ANN model. The most suitable ANN model is found to have one hidden layer with 7 hidden neurons.