| In engineering science and technology, there are many computationally hard problems to which no algorithm exists to find an optimal solution CPU time effectively. These problems involve two types of difficulties: (i) multiple, conflicting objectives and (ii) a highly complex, large search space. Genetic algorithms, based on the principles of natural evolution, possess several characteristics that are desirable for this kind of problem and are preferable to classical optimization applications. In fact, genetic algorithms offer a shortcut, being able to produce good, but not perfect results much faster in terms of computer time.; This dissertation presents research in both theory and engineering applications of genetic algorithms as stochastic methods for the optimization of systems.; Based on a simple structure, procedures and operators of genetic algorithms are presented and explained. Although the classic genetic algorithm is a very powerful tool, there are ways to improve its techniques, resulting in a faster convergence and a better solution. Several advanced and newly developed techniques are introduced and investigated, including: Population Initialization, Fitness Techniques, Advanced Selection Operators, Advanced Crossover Methods, Advanced Mutation Methods, Genetic Algorithm Parameters Setting, and Multi-Parameter Representation, etc.; Three types of engineering problems are described and solved: (i) optimization of multiple objective problem, (ii) optimal controller design, and (iii) combinatorial optimization problem: Traveling Salesman Problem. All three applications are complex. Attention is paid to the incorporation of problem-based knowledge into the optimization process. Using genetic algorithms the inclusion of problem-based knowledge is possible at different levels, consequently various techniques can be developed. The incorporation of problem-based knowledge accelerates the search for solutions and improves the quality of the solutions. |
