Study on nonlinear regression modeling methods of the Permanent Magnet Drive

The Permanent Magnet Drive is a new kind of advanced energy-saving products with disc structure. This paper mainly aims at nonlinear regression modeling of the Permanent Magnet Drive with BP neural network, adaptive network-based fuzzy inference system and support vector machine. The sample points for training are the electromagnetic characteristics of the Permanent Magnet Drive and obtained from finite element method simulation. The accuracy, validity and prediction capability of these three modeling methods are verified and compared. Simulation results prove that, all approaches have highly precise, little time consuming for convergence and strong prediction performance, and support vector machine reflects the characteristics of the Permanent Magnet Drive best. The nonlinear models built in this paper provide effective tools for the performance analysis, the control system establishment and the optimization design of the Permanent Magnet Drive.