Fouling Prediction using Neural Network Model for Membrane Bioreactor System

Abstract

Membrane bioreactor (MBR) technology is a new method for water and wastewater treatment due to its ability to produce better and high-quality effluent that meets water quality regulations. MBR also is an advanced way to displace the conventional activated sludge (CAS) process. Even this membrane gives better performances compared to CAS, it does have few drawbacks such as high maintenance cost and fouling problem. In order to overcome this problem, an optimal MBR plant operation need to be developed. This can be achieved through an accurate model that can predict the fouling behaviour which could optimise the membrane operation. This paper presents the application of artificial neural network technique to predict the filtration of membrane bioreactor system. The Radial Basis Function Neural Network (RBFNN) is applied to model the developed submerged MBR filtration system. RBFNN model is expected to give good prediction model of filtration system for estimating the fouling that formed during filtration process.

A Review on Membrane Fouling Prediction Using Artificial Neural Networks (ANNs)

“…Several types of ANN architectures have been used for fouling prediction, including feed-forward neural networks (FFNNs) such as radial basis functions (RBFs) [ 70 , 71 , 72 ], multilayer perception (MLP) [ 72 , 73 ], and recurrent networks (RNNs) [ 74 , 75 , 76 ]. A comprehensive comparison between MLP and RBF networks was studied by Xie et al [ 77 ].…”

“…The accuracy of an ANN model for membrane fouling prediction can be evaluated using various statistical metrics, depending on the specific problem and the desired output. The most commonly used assessment metrics include the mean squared error (MSE), root mean squared error (RMSE) (Equation ( 4)), mean absolute error (MAE), and the coefficient of determination (R 2 ) (Equation ( 5)) [51,73,90] The accuracy of an ANN model for membrane fouling prediction can be evaluated using various statistical metrics, depending on the specific problem and the desired output. The most commonly used assessment metrics include the mean squared error (MSE), root mean squared error (RMSE) (Equation ( 4)), mean absolute error (MAE), and the coefficient of determination (R 2 ) (Equation ( 5)) [51,73,90].…”

“…The most commonly used assessment metrics include the mean squared error (MSE), root mean squared error (RMSE) (Equation ( 4)), mean absolute error (MAE), and the coefficient of determination (R 2 ) (Equation ( 5)) [51,73,90] The accuracy of an ANN model for membrane fouling prediction can be evaluated using various statistical metrics, depending on the specific problem and the desired output. The most commonly used assessment metrics include the mean squared error (MSE), root mean squared error (RMSE) (Equation ( 4)), mean absolute error (MAE), and the coefficient of determination (R 2 ) (Equation ( 5)) [51,73,90]. The RMSE measures the average of the squared differences between the predicted and actual values of the output parameter; the MAE shows the average absolute difference between the predicted and actual values of the output.…”

Radial basis function neural network for 2 satisfiability programming

“…There are different techniques for training RBFNN [26], such as no-training technique and half-training technique. Calculations for the equation of the activation functions for the processing on RBFNN [27] are as follows: x is an input value in i  , Structure of RBFNN indicates the Euclidean norm on the input space as follows and Structure of RBFNN as shown in Figure 1 Accordingly, the output of RBFNN technique has the following form [31]:…”