Prediction of red plum juice permeate flux during membrane processing with ANN optimized using RSM

Abstract

In this work, a three-layer artificial neural network (ANN) optimized by response surface methodology (RSM) was designed to predict the permeate flux of red plum juice during membrane clarification. The input parameters of the model were trans-membrane pressure (TMP), temperature, cross-flow velocity, pore size and processing time. A multi-layer feed-forward (MLFF) network using gradient descent with momentum (GDM) as learning algorithm and with one hidden layer was employed for developing predictive model. A central composite design was applied to find optimum values of number of neurons, training epoch, step size, training percentage and momentum coefficient. Also, a quadratic model was developed from training results to mean square error (MSE) of 52 developed ANNs as the response. The results showed that the training epoch had highest effect on the response of ln(MSE) and then followed by step size and momentum coefficient, respectively. Finally, the optimum values of variables to obtain minimum response were 22, 7670, 0.28, 65% and 0.85 for number of neurons, training epoch, step size, training percentage and momentum coefficient, respectively. The best ANN model for predicting permeate flux of red plum juice had a 5-22-1 topology. The MSE and coefficient of determination (R2) of the optimal topology were determined as 0.0016 and 0.986 for training, 0.0017 and 0.976 for validation and 0.0021 and 0.961 for testing data sets. The developed ANN satisfactory modeled non-linear dynamic behavior of permeate flux at different operating parameters during membrane clarification of red plum juice.