This paper presents a new decoupling control of five-phase fault-tolerant permanent-magnet (FTPM) motor drives, in which support vector machine (SVM) and inverse system theory are incorporated. The inverse system is constructed to compensate the original system into a pseudo-linear system, while SVM is utilized to obtain the inverse system without knowledge of accurate motor model. The proposed FTPM motor drive is verified in Matlab/Simulink environment, showing that the d-axis current and speed of five-phase FTPM motor system are successfully decoupled. Additionally, the proposed motor drive offers fast speed response and high control accuracy.
I. INTRODUCTIONERMANENT magnet synchronous motors (PMSMs) are widely used because of their advantages such as high torque to current ratio, high efficiency, high power density, and low noise. However, in some high performance applications like electric vehicles and aerospace applications, high reliability and continued operation are required. So, fault-tolerant permanent-magnet (FTPM) motor has been investigated and used [1]. FTPM motors has not only the common features of permanent magnet motor, but also the characteristic of physical isolation, thermal isolation, magnetic separation, electrical isolation and inhibition of short circuit current. Therefore, the safety and reliability of the system has been improved. Compared with traditional three-phase one, five-phase FTPM motor has the advantages of high fault tolerance and fault isolation capabilities, high power density, and low torque ripple.Recently, many control strategies have been investigated for FTPM motor drives [2][3][4][5]. It is well known that a PMSM drive is a strong nonlinear system. So, decoupling and linearization are the key issues of control. However, due to the nonlinear nature and the coupling variables of the nonlinear system, linear control methods are inappropriate for them. Hence, nonlinear system control strategies have been presented [6].The feedback linearization method is one of the nonlinear system control methods. It contains differential geometry method and inverse system method. The differential geometry method mainly relies on precise mathematical equations. It is hard to promote in practical applications.