Pareto Artificial Life Optimization Algorithms for Multi-objective Optimization Problems

Abstract

The design of complex machinery is an iterative process. Components are redesigned iteratively until acceptable performance and reliability is obtained. In preliminary design of machinery, the optimum design is carried out to reduce the iterative redesign process. Therefore, the optimization of the system has become an important part of design process. There are very efficient optimization methods called the hill-climbing methods. However, it is clear that these methods provide local optimum values only and these values depend on the selection of the starting point. Therefore this research concerns with global optimization methods to overcome these problems. One of the global optimization methods is the artificial life algorithm (ALA) for function optimization. This study proposed a hybrid ALA called the enhanced artificial life optimization algorithm (EALA) to overcome the demerits of the ALA which are low speed of convergence and low accuracy after generating colony. Most of engineering optimization problems often consists of several objective functions rather than a single objective function. Basically, there are two kinds of approaches to solve the multi-objective optimization problems (MOP). The first approach transforms a given multi- objective optimization problem into a single objective optimization problem (SOP). In order to provide possible solutions for the final decision maker, this approach has limitation that is only one solution is provided. The second approach is based on the concept of Pareto optimality to avoid this difficulty and to explore various possibilities. In order to apply artificial life algorithm to MOP in engineering problems, it is necessary to solve the Pareto optimization problem. Therefore, in this study, artificial life optimization algorithm has been expanded to enable the application of Pareto optimization to solve the MOPs.