Prediction of pore size of ultrafiltration membrane by using artificial neural network (ANN)
The objective of this research is to develop a neural network model to predict the pore size of ultrafiltration membrane. Usually, the pore size of ultrafiltration membrane was determined experimentally using permeation and rejection rate experiments, followed by empirical equations. Therefore, in t...
Saved in:
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Penerbit UTM Press
2008
|
Subjects: | |
Online Access: | http://eprints.utm.my/id/eprint/8721/1/UTMjurnalTEK_49F_DIS%5B23%5D.pdf http://eprints.utm.my/id/eprint/8721/ |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | The objective of this research is to develop a neural network model to predict the pore size of ultrafiltration membrane. Usually, the pore size of ultrafiltration membrane was determined experimentally using permeation and rejection rate experiments, followed by empirical equations. Therefore, in this study, Artificial Neural Network (ANN) has been proposed as an alternative method to predict the pore size of flat sheet ultrafiltration membranes. Experimental data were collected from the previous research whereby the polyethersulfone (PES) polymeric membranes were fabricated with lithium bromide (LiBr) additive. The membranes were tested by using various polyethylene glycol PEG molecular weights solution. The neural network has a pyramidal architecture with three different layers which consists of an input layer, hidden layer and output layer. Feed-forward Backpropagation (FFBP) network was constructed in MATLAB version 7.2 environment by using Levenberg-Marquardt algorithm (trainlm) training method. In addition, Bayesian regularization method was introduced to improve the neural network generalization. The simulated results obtained from this study were then compared to the experiment results so as to obtain the best model with the smallest Root-Mean Square (RMS) error. The results revealed that the constructed networks were able to accurately estimate the pore size of ultrafiltration membrane. |
---|