| Electric motor drives applied to propulsion of Electric Vehicle (EV) have to typically traverse a wide range of speeds and power levels during their operation. It is then essential to determine and verify the adequacy of the motor design during each assumed operating condition. Also, it is necessary to determine and state the required SRM continuous power rating for the intended EV propulsion application before starting the design. Towards this end, the Thesis has developed a new sizing procedure to determine the SRM peak and continuous power rating. New equations were developed to estimate the SRM losses at any operating condition based on assumed losses at one condition.; Recent literature has identified that multi-phase operation of SRM can lead to advantages in terms of torque density, efficiency, torque ripple and acoustic noise. The Thesis introduces new coeffcients that were derived with the help of ANSYS™ based FEA studies and take the operation of each multi-phase configuration into account. These coefficients are useful when designing SRM under multi-phase operation. The design ideas are then verified by designing, building and testing the prototype of a 1-hp four-phase 8/6 SRM for EV propulsion.; Torque feedback improves the performance of the SRM drive and aids in aspects such as torque ripple minimization. Also, position sensorless SRM drives are currently being widely investigated due to their increased reliability. Under multi-phase operation, it is essential to account for the mutual flux interactions between phases in order to get high accuracy in these tasks. The Thesis proposes a new Feed-Forward Artificial Neural Network (ANN) based system to provide mapping between the SRM terminal variables' and the SRM mutual interaction function and torque.; Building upon the previous modeling research work, the Thesis proposes a simple ANN based on-line torque estimator that eliminates the need for a torque transducer, adapts to the characteristics of the individual SRM and can be easily extended to the multi-phase case.; Finally, the Thesis concludes by considering an implementation of the ANN on the Texas Instruments expressDSP™ TMS320C6701. The pipelining features of the TMS320C6701 and the Code Composer Studio features are used to obtain low execution times for the ANN. The implementation verifies the feasibility of the proposed ANN-based methods. |
