Abstract
The construction of a model for the prediction of process outputs is a valuable tool in the field of engineering. The models play an important role in the simulation and optimization of systems leading to the design of efficient and economical processes.
Since 1943 neural network (NN) techniques have been considered as promising tools for use in simulation, prediction and modelling because of their simplicity.
In this thesis a feed-forward neural network (FFNN) with back-propagation (BP) is used to test its effectiveness in modelling the ion-exchange process.
The ion-exchange process has been widely employed in the removal of heavy metals from industrial wastewater. This process is a complex non-linear process involving many factors influencing the chemical process which is not well understood (the ions uptake mechanisms from the pregnant solution, the subsequent step being the elution). In order to improve the performance of the ion-exchange process, optimization and analysis of the process should be accomplished. Modelling and simulation are tools which can be used to achieve the objectives.
The experimental design using analysis of variance (ANOVA) was chosen to compare to the NN techniques and for optimizing the effective input parameters (pH, temperature and initial concentration). The FFNN successfully tracked the non-linear behaviour of the ion-exchange process versus the input parameters with a mean square error (MSE), correlation coefficient (R) and mean square relative error (MSRE) of 0.102, 0.998 and 0.004, respectively. The results showed that the FFNN modelling techniques could effectively predict and simulate the highly complex system and non-linear process such as the ion exchange using activated zeolite...
D.Tech. (Chemical Engineering)