Conservative methods for structural optimization

Abstract

A method is introduced which adds a padding term to each constraint that is proportional to the gradient of the constraint. One way to implement this approach is to determine the gradients of the constraints at the unpadded optimum design, to modify the constraint allowables, and then to restart the optimization. A second implementation is proposed which imbeds the padding calculation into the optimization process by updating the padding term periodically during the optimization process. Finally an implementation which uses nonlinear constraint approximations and a second order update method is described