Object-Oriented Framework for Genetic Algorithms with Application to Space Truss Optimization

Abstract

Genetic algorithms have been shown to be very effective optimization tools for a number of engineering problems. Since the genetic processes typically operate independent of the actual problem, a core genetic algorithm library consisting of all the genetic operators having an interface to a generic objective function can serve as a very useful tool for learning as well as for solving a number of practical optimization problems. This paper discusses the object-oriented design and implementation of such a core library. Object-oriented design, apart from giving a more natural representation of information, also facilitates better memory management and code reusability. Next, it is shown how classes derived from the implemented libraries can be used for the practical size optimization of large space trusses, where several constructibility aspects have been incorporated to simulate real-world design constraints. Strategies are discussed to model the chromosome and to code genetic operators to handle such constraints. Strategies are also suggested for member grouping for reducing the problem size and implementing move-limit concepts for reducing the search space adaptively in a phased manner. The implemented libraries are tested on a number of large previously fabricated space trusses, and the results are compared with previously reported values. It is concluded that genetic algorithms implemented using efficient and flexible data structures can serve as a very useful tool in engineering design and optimization.