Date of Award
Summer 2016
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Electrical and Computer Engineering
First Advisor
Demerdash, Nabeel A. O.
Second Advisor
Ionel, Dan M.
Third Advisor
Nehl, Thomas W.
Fourth Advisor
Jahns, Thomas M.
Abstract
The common practices of large-scale finite element (FE) model-based design optimization of permanent magnet synchronous machines (PMSMs) oftentimes aim at improving the machine performance at the rated operating conditions, thus overlooking the performance treatment over the entire range of operation in the constant torque and extended speed regions. This is mainly due to the computational complexities associated with several aspects of such large-scale design optimization problems, including the FE-based modeling techniques, large number of load operating points for load-cycle evaluation of the design candidates, and large number of function evaluations required for identification of the globally optimal design solutions. In this dissertation, the necessity of accommodating the entire range of operation in the design optimization of PMSMs is demonstrated through joint application of numerical techniques and mathematical or statistical analyses. For this purpose, concepts such as FE analysis (FEA), design of experiments (DOE), sensitivity analysis, response surface methodology (RSM), and regression analysis are extensively used throughout this work to unscramble the correlations between various factors influencing the design of PMSMs. Also in this dissertation, computationally efficient methodologies are developed and employed to render unprohibitive the problems associated with large-scale design optimization of PMSMs over the entire range of operation of such machines. These include upgrading an existing computationally efficient FEA to solve the electromagnetic field problem at any load operating point residing anywhere in the torque-speed plane, developing a new stochastic search algorithm for effectively handling the constrained optimization problem (COP) of design of electric machines so as to reduce the number of function evaluations required for identifying the global optimum, implementing a k-means clustering algorithm for efficient modeling of the motor load profile, and devising alternative computationally efficient techniques for calculation of strand eddy current losses or characterization of the mechanical stress due to the centrifugal forces on the rotor bridges. The developed methodologies in this dissertation are applicable to the wide class of sine-wave driven PM and synchronous reluctance machines. Here, they were successfully utilized for optimization of two existing propulsion traction motors over predefined operating cycles. Particularly, the well-established benchmark design provided by the Toyota Prius Gen. 2 V-type interior PM (IPM) motor, and a challenging high power density spoke-type IPM for a formula E racing car are treated.