Ismail Agirman

Adaptive Torque-Ripple Minimization in Switched Reluctance Motors

Thursday, May 6, 1999
11:45 AM
422 Snell Engineering Building

Abstract

Advantages of Switched Reluctance Motor (SRM) drives include a simple structure (hence low cost), the ability to operate in harsh environments (such as at high temperatures) and under partial hardware failures, and a wide speed range. Industrial acceptance of SRM drives is presently limited by crude, almost model-independent control patterns that are easy to implement, but do not achieve the desirable smooth operation. SRMs are driven into magnetic saturation in energy efficient use, causing main control disadvantages: inherent non-linearities that present a modeling challenge, and a substantial torque ripple. The advent of power electronics and fast digital hardware, together with improved nonlinear and adaptive control methods, enables a re-evaluation of these issues.

When an accurate machine model is used for control design, both feedback linearization techniques and the use of pre-calculated optimal torque sharing functions result in good dynamic performance. However, model inaccuracies are known to degrade performance, and are often hard to avoid: adding to inaccurate measurements, the use of an accurate model entails complex on--line computations, or the use of large lookup tables that may be impractical. In this work we address this issue by combining the use of a simple and easily computable torque sharing function with adaptation. Our main goal is to reduce torque ripple. The underlying control idea is simple: assuming a position dependent -- and therefore periodic -- current excitation, model inaccuracy will result in a periodic ripple in the produced torque. Harmonic components of the produced electro-magnetic torque combined with the unknown load torque are dynamically estimated, and corresponding correction terms are added to the commanded phase currents until speed ripple is eliminated (or considerably reduced in practice). Simulations and experiments demonstrate that our algorithm substantially reduces the torque ripple.

Thesis Committee:
Prof. A.M. Stankovic (advisor)
Prof. G. Tadmor (co-advisor)
Prof. H. Lev-Ari