Date of Award
Spring 2001
Document Type
Dissertation - Restricted
Degree Name
Doctor of Philosophy (PhD)
Department
Electrical and Computer Engineering
Abstract
Although the first Switched Reluctance Motor (SRM) dates back to the 19th century (1838), it was not widely used in industry until very recently. The main reason for not using SRMs was their nonlinear nature and, consequently, the difficulties faced in their characterization and control. Accordingly, accurate and fast models of SRM drives are needed for the characterization and development of control algorithms for this class of drives. This dissertation presents new techniques for the fast and accurate characterization of SRM drive systems using Artificial Intelligence (AI) based model. In this work, not only the operation of the machine under normal conditions is considered, but also the machine is assumed to be operating under abnormal (fault) conditions. The work involves building two AI-based models. One model employs Artificial Neural Networks (ANNs), and the other utilizes Fuzzy Inference Systems (FIS). ANNs are used for their interpolation ability and mapping capability in highly nonlinear environment. Fuzzy Logic (FL) is applied in the modeling of the SRM drive systems because it is very suitable for problems with large degree of uncertainty but for which some knowledge is available. As the development of both models requires training data sets, this work first investigated and developed the more conventional State Space (SS) - Finite Element (FE) models. Although these models are accurate and account for magnetic material nonlinearities, they require intensive computational resources and relatively long computational time. Next the SS-FE models were validated and used to generate the information and knowledge needed for AI-based modeling. The AI-based models were used to characterize a prototype SRM drive under normal and fault operating conditions. In addition, the two AI based models were compared in order to help other future investigators in achieving the proper model for their applications.