A new genetic algorithm for multiobjective optimization

In recent years, much interest has been generated for using genetic algorithms to optimize ccrtain classes of problems. Genetic algorithms have definite advantages over some other types of optimization algorithms in that they are quite robust, and do not require knowledge of the gradients of the objective function. Because of their unique breeding and selection processes, genetic algorithms can be used with equal ease on linear and nonlinear optimization problems, which makes them an excellent all-purpose optimization algorithm. This paper will discuss a new genetic algorithm (GENMO) that can be used to simultaneously optimize multiple objectives. Multiobjective optimization is necessary in design of turbine blades, aircraft and other complex real world structures. The GENMO algorithm can be used to generate Pareto sets — which contain the trade-off information - for two or more conflicting objective functions. This is demonstrated through applications to composites and to turbomachinery airfoils.