Dependency-tracking object-oriented multidisciplinary design optimization (MDO) formulation on a large-scale system

Advances in computer technology and analysis software are making optimization of engineering systems more attractive and affordable than ever before. Optimization is becoming a necessary tool in order for companies to stay competitive. While the concept of optimization has been known almost as long as mankind, specific procedures on how to optimize engineering systems are younger. Currently, efforts are being made to reduce the computational time and simplify the organizational complexity involved with solving multidisciplinary systems.; The work presented in this dissertation deals with how an object-oriented, dependency-tracking, demand-driven language can be used in reducing the computational time in performing multidisciplinary design optimizations. The work also discusses how the object-oriented language was used in integrating optimization functionality with a missile design system.; The object-oriented dependency-tracking demand-driven language is applied to a large-scale multidisciplinary missile system involving disciplines such as a geometry engine, weight analysis, propulsion, aerodynamics, trajectory analysis, and cost analysis.; Also discussed is the need for using approximations in optimizing a large-scale system. Designed experiments and response surface techniques were employed in creating approximation models for the problem at hand. Using these approximations to evaluate the responses was found to be useful at points in the design space where one or more responses could not otherwise be evaluated.; Different optimization schemes were studied including response surface analysis of different resolutions in conjunction with higher fidelity optimization and higher fidelity optimization without approximation models.; The contributions of this work are the application of MDO capabilities to a large-scale missile design system modeled in an object-oriented dependency-tracking environment, the use of response surface approximations to fit areas in the design space that could otherwise not be determined, and the use of a dependency-tracking demand-driven language to reduce computational effort.