Abstract
Phase behaviours are essential in studying chemical systems, simulation, process design, control, and optimisation. It is a crucial concept of process integration and intensification of separation and reaction processes. The study of phase behaviour includes the concept of phase stability, phase equilibrium for both reactive and nonreactive systems, and parameter estimation for thermodynamic models. The problem can be solved by formulating them into optimisation problems, which are complex, nonlinear, and with several local minima, and therefore need robust and reliable algorithms. Amidst several methods of solving optimisation problems, including engineering and real-world problems, stochastic algorithms seem promising in overcoming computational challenges such as initialisation and others.
This study covered the application of newly developed stochastic optimisation algorithms to solve the myriad of phase stability and equilibrium problems for both reactive and nonreactive systems. Several performance indices were used in evaluating the ability of these algorithms to solve this problem of different computational difficulties, with a view of elucidating their strengths and weaknesses. The need to enhance the algorithms for better performance was carried out by intensification with local optimisers. Novel hybrid algorithms that combined desirable characteristics of parent algorithms were also developed, and their performances were compared with several other algorithms. Problem formulation and algorithms were developed to estimate reliable binary interaction parameters that satisfied thermodynamics stability criteria and predicted the actual empirical behaviour of chemical systems with high accuracy.
The composition of this thesis includes the following:
•
24 different phase stability and equilibrium problems with various degrees of complexity and thermodynamic models such as Equations of state, Wilson, NRTL, and UNIQUAC models.
•
Stochastic algorithms applied and enhanced are the pathfinder, wild horse, honey badger, red fox, gorilla, and pelican optimisation algorithms.
•
New hybrids of pelican and gorilla, driving test and honey badger optimisation algorithms were developed.
Overall, this thesis presented robust and reliable algorithms for solving phase problems with high-performance profiles. Honey Badger Algorithms (HBA), Wild Horse Optimisation (WHO), Pelican Gorilla optimisation algorithms 2(PGOA2) all have high performance profiles which are close to 1, overall success rates of 99.7%, 99.2%, and 99.9% respectively and their
5
performances surpassed most of the algorithms presented in the literature regarding reliability, success rate, performance profile, and computational cost.