Application of Artificial Neural Network to predict phosphoric acid slurry viscosity

Abstract

At the level of the attack unit of the phosphoric acid production process, the control and monitoring of the dynamic viscosity of phosphoric acid slurry is crucial to understand its rheological behavior. This information contributes to the resolution of problems that may be encountered in the flow. The objective of this work is to obtain a reliable artificial neural network model to predict this rheological property. First, experimental data of phosphoric acid slurry are analyzed. The dataset is composed of 468 samples with three explanatory variables namely: temperature, shear rate and solid content and a target variable which is dynamic viscosity of the phosphoric acid slurry. Results have shown that solid content has the greatest effect on the dynamic viscosity of the slurry, followed by shear rate and then comes temperature. Then, a neural network model with the topology (3-5-1) have been developed and have shown accurate predictions of the slurry viscosity.