Iterative learning control for fault-tolerance in multi-phase permanent-magnet machines

Fault-tolerant control (FTC) techniques for multi-phase permanent magnet (PM) motors are usually designed to achieve maximum ripple-free torque under fault conditions with minimum ohmic losses. A widely accepted approach is based on flux distribution or back EMF (BEM) model of the machine to calculate healthy phase currents. This is essentially an open-loop technique where currents are determined offline (based on motor fault models) for each fault scenario. Therefore, it is highly model-dependent. Since torque pulsation due to open-circuit faults are periodic, learning and repetitive control algorithms are excellent choices to minimize torque ripple. In this paper, iterative learning control (ILC) is applied as a current control technique for recovering performance in multiphase PM motor drives under open-circuit fault conditions. ILC-based FTC needs torque measurement, but avoids the need for complicated fault detection and fault diagnosis algorithms. Furthermore, a BEM+ILC based FTC is proposed that initiates the learning from a model-based approximate guess (from the BEM method). Therefore, this method combines the advantages of both model information as well as robustness to model uncertainty through learning. Hence, the proposed method is well-suited for high performance safety-critical applications. Finite element analysis (FEA) results on a five-phase PM machine are presented for verification of the proposed control schemes.