Artificial Neural Network (ANN) model development for predicting just suspension speed in solid-liquid mixing system

Abstract

Just-suspension speed (Njs) is one of the important criteria for the design of agitators for solid-liquid mixing processes. In this manuscript a novel approach on using Artificial Neural Network (ANN) modeling for of just-suspension speed prediction is developed based previous published work that contains 950 datasets including various solid loading, solid density, solid diameter, tank diameter, solution density, impeller diameter, number of impeller blade, blade hub angle, blade tip angle, the width of blade and the ratio of clearance between an impeller and the bottom of the tank with the tank diameter whereas the corresponding to just-suspension speed as an output. Multilayer perceptron type of feed-forward back-propagation neural network was employed for building the ANN model. It found that the configuration of 8 neurons in 1 hidden layer using tangent sigmoid as transfer function presented as the optimum ANN model (11-8-1). Results show the proposed ANN model could provide the desired accuracy on predicting just-suspension speed by achieves 0.96 of correlation coefficient and 0.0059 of mean square error. In addition, the results showed that the integrated Genetic Algorithm-Artificial Neural Network (GA-ANN) model enhanced the accuracy for predicting the just-suspension speed compare ANN model. This novel approach showed the high potential to be applied in chemical process industrial design system.