Sensorless position estimation in fault-tolerant permanent magnet AC motor drives with redundancy.

摘要

Safety critical applications are heavily dependent on fault-tolerant motor drives being capable of continuing to operate satisfactorily under faults. This research utilizes a fault-tolerant PMAC motor drive with redundancy involving dual drives to provide parallel redundancy where each drive has electrically, magnetically, thermally and physically independent phases to improve its fault-tolerant capabilities. PMAC motor drives can offer high power and torque densities which are essential in high performance applications, for example, more-electric airplanes. In this thesis, two sensorless algorithms are proposed to estimate the rotor position in a fault-tolerant three-phase surface-mounted sinusoidal PMAC motor drive with redundancy under normal and faulted operating conditions. The key aims are to improve the reliability by eliminating the use of a position sensor which is one of major sources of failures, as well as by offering fault-tolerant position estimation. The algorithms utilize measurements of the winding currents and phase voltages, to compute flux linkage increments without integration, hence producing the predicted position values. Estimation errors due measurements are compensated for by a modified phase-locked loop technique which forces the predicted positions to track the flux linkage increments, finally generating the rotor position estimate. The fault-tolerant three-phase sensorless position estimation method utilizes the measured data from the three phase windings in each drive, consequently obtaining a total of two position estimates. However, the fault-tolerant two-phase sensorless position estimation method uses measurements from pairs of phases and produces three position estimates for each drive. Therefore, six position estimates are available in the dual drive system. In normal operation, all of these position estimates can be averaged to achieve a final rotor angle estimate in both schemes. Under faulted operating conditions, on the other hand, a final position estimate should be achieved by averaging position estimates obtained with measurements from healthy phases since unacceptable estimation errors can be created by making use of measured values from phases with failures. In order to validate the effectiveness of the proposed fault-tolerant sensorless position estimation schemes, the algorithms were tested using both simulated data and offline measured data from an experimental fault-tolerant PMAC motor drive system. In the healthy condition, both techniques presented good performance with acceptable accuracies under low and high steady-state speeds, starting from standstill and step load changes. In addition, they had robustness against parameter variations and measurement errors, as well as the ability to recover quickly from large incorrect initial position information. Under faulted operating conditions such as sensor failures, however, the two-phase sensorless method was more reliable than the threephase sensorless method since it could operate even with a faulty phase.