Study on Dynamic Adsorption of Complex System In Solid-Liquid Phase Modelling Using Artificial Neural Networks

Abstract

This work aims to develop an ANN model to predict the dynamic adsorption of complex system of adsorbent- adsorbate in solid-liquid phase on different parameters through an adsorption column. Nine neurons were used in the input layer, fourteen neurons and ten neurons were used respectively in the first and the second hidden layer. One neuron was used in the output layer. A set of 2007 data points were used for testing the neural network. Levenberg Marquardt learning (LM) algorithm, logarithmic sigmoid transfer function and linear transfer function were used for the hidden and output layer respectively. Results with the ANN showed a correlation coefficient R² = 0.9976 and 0.9969 respectively for total database and for validation phase between simulated data and those obtained from the literature with a root mean square error RMSE = 0.0268 and 0.0305for total database and for validation phase respectively.

Moreover, to determine the most suitable model, Thomas and Bohart-Adams models were applied. The comparison between root mean square error (RMSE), sum of the absolute error (SAE), Chi-square statistic test (X²) and correlation coefficient (R²) showed that the neural network model gave far better. In general, the developed model provides the highest agreement vector values of [R²,\(\begin{equation}\label{eq1}
\alpha, \beta
\end{equation}\)] with a root mean square error value (RMSE) closed to zero.