Automated Optimum Design from Discrete Components

Optimization in structural design using discrete variable can exploit the discrete characteristics of the design problem and avoid artificial and often self-defeating discretation from problem solutions in continuous variables. A method for solving member-sizing problems using wide flange tables, a current design code, and linear elastic behavior due to arbitrary loading conditions is shown to be within reasonable economic bounds. The method consists of a combinatorial algorithm classified as a branch-and-bound type. Starting with a lower-bound infeasible solution the algorithm detects a feasible solution for an upper bound and explores the region for a local optimum, bouncing along the constraints if the value of the objective function can be improved. Reanalysis is performed at each step with an exact method which performs only part of normal analysis steps. Included as part of the STRUDL information system, the programmed algorithm allows an engineer to assist in improving the optimization process and perform optimization on a structure on a piece-wise basis.