Development of a Human Centric Multidisciplinary Design Optimization Method Using Fuzzy Logic Systems and Controllers.

Abstract

During early stages of design of large scale engineering systems, such as ships, competing objectives of multiple discipline analyzers must be taken into account. Currently, tools such as Multidisciplinary Design Optimization (MDO) provide an automated method for analyzing trade-offs between competing discipline design objectives. While automated design methods are useful in the design world, automation removes the human ability to interpret results of analysis. Current design tools lack the ability to incorporate the intent of experienced designers in the communication between disciplines, so MDO results are dependent on the fidelity of the analysis models being used rather than human input. When evaluating designs, a team of experienced engineers communicates information about the design space to each other with respect to their own disciple's objectives and design constraints.

This thesis presents a new method of MDO which provides a framework to emulate the intent of an human designer in a MDO optimizer. The method uses the ideas of managing trade-offs and modeling discipline interactions from MDO. Typical MDO algorithms utilize a system level optimizer, which receives information from discipline optimizers and moves toward a globally optimal solution. Generally, the information communicated to the system level optimizer is point data about the individual discipline's optimal point, which is then interpreted as the potential improvement from the current system design state. The proposed algorithm would communicate the discipline's preferred areas of the design space to be combined with the other discipline's preferences in an attempt to find a globally optimal solution.

This design intent modeling is introduced through fuzzy logic systems and fuzzy logic controllers. Fuzzy logic systems provide a method to interpret design analysis tools and glean more information to use in the decision making process. Fuzzy logic controllers are often used to emulate human decision models in the control of physical systems. These methods are extended to evaluation of designs to combine data from multiple disciplines to find the optimal system design while considering trade-offs between multiple disciplines. This dissertation will present the development of a novel method of MDO which incorporates fuzzy logic systems and controllers in the MDO optimizer.