Mathematical Programming Procedures for Mixed Discrete-Continuous Design Problems

Abstract

Two mathematical programming procedures for treating nonlinear problems involving mixed variables are presented. One involves a relatively simple concept. First an optimum is located treating all variables as continuous. Adjacent discrete points are then evaluated in order of increasing distance from the all-continuous optimum, each evaluation requiring an optimization of the continuous variables, if any, until a satisfactory design is found. The other method utilizes an optimal discrete search to locate the optimum. These procedures are applied to the minimum weight design of stiffened, cylindrical shells where they prove to be effective.