10th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, Albany, NY

Authors: G. N. Vanderplaats
Publication Date: Aug. 30 - Sep. 1, 2004

An optimization algorithm is presented which is capable of solving nonlinear, constrained optimization tasks of well over one hundred thousand design variables. This method is an advanced exterior penalty function approach which uses very little central memory and very little computational time within the algorithm. The price paid for the large scale capability is that this method requires 3-5 times as many function and gradient evaluations to converge. An ad-hoc method for solving discrete optimization problems is included which reliably obtains a reasonable discrete solution. Examples are presented to demonstrate the method.