The magnetic design of Switched Reluctance (SR) motors is inherently a hierarchical process. The design cycle progresses through distinct stages where the accuracy improves but computing times increase greatly, thus it often becomes impractical to furnish extensive multi-objective optimization required to accomplish the optimal design. In order to enable rapid and accurate optimization of SR motors an improved reduced-order computational method of flux tubes is implemented to complement and practically replace the time consuming 2D finite element based magnetic analysis. The paper demonstrates how the use of the improved flux tubes approach to evaluate objective functions results in substantially faster while still accurate multi-objective optimization of SR motors.
IntroductionSwitched Reluctance (SR) machines are often regarded as having the simplest mechanical design [1]- [3] compared to other conventional electric machines, such as permanent magnet dc, synchronous reluctance or induction, as they are brushless and have no permanent magnets (PM) or windings on the rotor. These design features make SR machines well suited to a wide range of applications where variable speed operation is required, such as general traction or pump drives. Moreover, the mechanical robustness of SR machines offers cheaper maintenance and better tolerance to harsh environments in which other types of machines cannot operate.Since SR motors are capable of operating in both constant torque and constant power regimes over a wide speed range [4], [5], this operating feature also makes them suitable for the automotive propulsion applications, where the absence of permanent magnets is considered necessary since this industry is very cost sensitive and insists on immunity from unforeseen material price fluctuations, as experienced by PM materials [6]. However, the absence of PM parts makes the SR machines much more difficult to control [7] due to their nonlinear torque-per-ampere characteristic. This operational nonlinearity is a result of the nonlinear magnetic characteristic as the machine's magnetic circuit becomes heavily saturated even during steady-state operation [8].Considering that SR machines are very nonlinear, they require complex analysis and robust design procedures to adequately predict their performance. Until now a rather limited success has been achieved in terms of accurate analytical design procedures for SR machines [9], [10] to provide satisfactory quantification of the machine performance. Most of the magnetic analysis and design is nowadays performed by numerical simulations based on a finite element method (FEM) in order to capture the detailed shape and allow for magnetic nonlinearity. The FEM based solutions can be accurate; however, they are time consuming and lack the intuitive insight into the cause-and-effect relationships between numerous machine design parameters. Moreover, the FEM solutions give only part of the