Date of Award

2010

Publication Type

Doctoral Thesis

Degree Name

Ph.D.

Department

Civil and Environmental Engineering

Keywords

Engineering, Chemical.

Supervisor

Casey, Joseph (Psychology)

Rights

info:eu-repo/semantics/openAccess

Abstract

The densities and kinematic viscosities of the quinary regular system: benzene (1) + toluene (2) + ethylbenzene (3) + heptane (4) + cyclooctane (5) and all its corresponding quaternary, ternary, and binary sub-systems have been measured at 293.15 K, 298.15 K, 308.15 K, and 313.15 K over the entire composition range. The experimental data reported herein are considered valuable additions to the literature. The experimental data gathered in the present study were utilized in testing the predictive capabilities of some well know viscosity models available in the literature. In addition, a new multi-layer artificial neural network (ANN) has been developed for the prediction of the kinematic viscosities of multi-component regular liquid mixtures. The concept of modular neural networks has been successfully applied to the design of the current network. Only a part of the experimental binary data was required for the training of the developed network. The remaining data on the binary systems were used for testing the ANN-based model. The developed neural network resulted in excellent viscosity predictions for the cyclooctane-containing systems. The predictive capability of the ANN in the case of the cyclooctane-containing systems was superior to the predictive capabilities of the other tested models for all systems. The predictive version of the McAllister three-body interaction model was the best to predict the kinematic viscosities of non-cyclooctane-containing systems. The predictive version of the McAllister three-body model worked very well when the molecular diameter ratio between system components was less than 1.5. The reliable and accurate data resulting from the present study helped in both critically testing existing viscosity models and in developing a new model that is based on the ANN. Results of the present study are promising for continued work in the same area.

Share

COinS