Optimal design of a welded beam via genetic algorithms

Introduction T HE welded beam structure is a practical design problem that is often used as a bench-mark problem in testing different optimization techniques. This problem is one of a family of structural optimization problems, which consists of a nonlinear objective function and five nonlinear constraints. There exist a number of optimization techniques that are successfully used in solving such problems. Some of these methods, like geometric programming, require an extensive problem formulation prior to the optimization procedure. Other methods, such as gradient search techniques, require derivative information that may not exist for others. This paper considers the application of a genetic algorithm (GA) in obtaining optimal design parameters for a welded beam structure. GAs are systematic search procedures—both global and efficient—based on the mechanics of natural genetics. GAs search through large spaces quickly even though they only require payoff information. Furthermore, because of the processing leverage associated with GAs, the method has a much more global perspective than many common methods in engineering optimization techniques. GAs have been applied to a variety of optimization problems—engineering, social sciences, physical sciences, computer sciences, biology, and others. In the welded beam problem described here, GAs are compared with other optimization techniques and found to have surprising speed of convergence to near-optimal solutions. Simulation results suggest that GAs can be used to solve other problems of this class with similar efficiency.