Evaluation and optimization of axial air gap propulsion motors for naval vessels.
Loading...
Authors
Thomas, Mark W.
Subjects
Advisors
Kirtley, James L., Jr.
Date of Issue
1996-05-10
Date
Publisher
Massachusetts Institute of Technology
Language
en_US
Abstract
A unique method is used to optimize a design to multi-objective criteria. While the method is potentially applicable to any optimization where the cost function is not well defined, the products considered here are synchronous axial gap electric motors (both wound rotor and permanent magnet) and the application for which the motors are optimized is warship propulsion. All motors are rated at 40,000 hp, or approximately 25 megawatts. A preliminary design of an axial gap motor in this power range was completed as part of doctoral research by T. J. McCoy [1]. All wound rotor designs in this study are based on his work. However, the McCoy motor includes a rotating thermosyphon cooling system, which is omitted here in favor of a simple heat density calculation. A permanent magnet machine design is presented and the resulting motors are optimized simultaneously with wound rotor types based on Naval propulsion criteria. This optimization method was originated by J. A. Moses et al. [2] and is termed the Novice Design Assistant. It involves the repetitive computer generation of designs through random combinations of design parameters. The results are compared to a database of previous designs. Any design found to be dominated in all desired attributes by another is discarded: otherwise it is added to the database. Dominance is determined by an evaluator module, tailored to the application, which compares motor attributes such as physical dimensions, weight and efficiency. The result of many iterations is an ndimensional "frontier" of non-dominated designs, where n is the number of attributes considered. Since the number of feasible possibilities is large given that some design parameters may vary continuously, a low "hit" rate is avoided by mapping successful parameter combinations back to the design module using a Gaussian distribution. This mapping process is similar to that used by U. Sinha in applying the method to commercial induction motors [3], and preserves the creativity of the method while decreasing computer overhead time.
Type
Thesis
Description
Series/Report No
Department
Ocean Engineering
Organization
Identifiers
NPS Report Number
Sponsors
Funder
Format
83 leaves