Optimal Sizing/Shape Design of Steel Portal Frames Using Genetic Algorithms

Abstract

This chapter presents a practical optimization approach for steel portal frames. The Genetic Algorithms (GAs), which are based on the principles of survival-of-the-fittest, are employed to generate a pool of solutions (population), represented by a string structure (coding structure). The selection operator is used to generate the new population of the next generation. The selection probability of a string surviving in the next generation depends on the objective function value of the solution and the selection method such as proportionate selection. The crossover operator exchanges the string bits after the crossover point between two selected parent strings to form two offspring strings. Crossover point is randomly found along the strings. The mutation operator occurs with some specified probability for each bit in the strings that have undergone crossover. The nonuniform mutation is selected for its ability to uniformly search the design space at the beginning and very locally at later stages to improve GAs' fine-tuning capability. Penalty function is used to transform the constrained optimization to a structure evaluation function for GAs to evaluate the performance of candidate solution.