Abstract
This paper presents a new technique for online identificationof an induction motor rotor time constant. The techniqueis designed for a shaft-sensorless indirect field-oriented controlinduction motor drive with a model reference adaptive system(MRAS)-based speed estimator. The MRAS estimator is sensitiveto the changes in the rotor time constant, and online identificationof that parameter is essential. If rotor parameter error exists, itdoes not only change the achieved rotor speed, but it also changesthe dynamic behavior of the whole field control and speed estimationstructure. The proposed rotor parameter update is exactlybased on the newly introduced dynamic model of the potentiallydetuned MRAS-based speed estimator. The technique avoids theuse of test signals and rather extracts the needed information fromthe ever-present signal jitter, which is inherent to the current andspeed servo loops. This paper demonstrates that the phase angledifference between some spectral components of selected smallsignals within the speed estimator can be used for rotor parameterupdate. Computer simulations and experiments are performedunder a variety of conditions to validate the effectiveness of theproposed rotor parameter update technique.Index Terms—AC motor drives, model reference adaptivecontrol, parameter estimation.
I. INTRODUCTION
HIGH-PERFORMANCE servo applications of an inductionmotor can be made possible by implementing thevector control technology. This advance in control technology,coupled with consistent price reduction in power electronics,made the vector-controlled induction motor highly competitiveon the low-cost motor system market. Further advance in theinduction motor drive technology is also feasible, and it iscoupled with the elimination of the sensors needed for the driveto operate. In particular, the development of a shaft-sensorlessinduction drive is the best answer for the persistent demandfrom the market place for less expensive and yet more robustdrives. However, in applications where the safety regulationsapply, shaft-sensorless operation is acceptable only in caseswhere a robust reliable speed estimation is available, not beingprone to thermal drift or any other secondary effect that mayendanger correct speed estimation. Currently, there are two parallel paths toward a robust sensorlesssolution [1]. Two competing technologies are the machinemodel-based schemes and the schemes using test signal toexploit the anisotropic properties of machine. The introductionof test signal keeps machine observable even when rotorinducedvoltage approaches zero. Uses of different machineanisotropies are reported, such as magnetic saturation or rotorslotting [2] or rotor slot openings modified in sinusoidal pattern[3]. Alternatively, model-based sensorless algorithms using fullobserver approach, upgraded with online parameter identificationalgorithms, are also capable of operating at very low rotorspeeds [4]–[6]. However, vast majority of speed-sensorlessdrives are used in low-cost drive applications, without theneed to operate at standstill. For these applications, calculationintensivealgorithms associated with high-end processor usingexpensive peripherals and power supply should be avoided. In[7], Schauder investigates the rotor flux-based model referenceadaptive system (MRAS) for speed estimation in drives withindirect field-oriented control (IFOC). The method is rathersimple to implement and uses minimum processor time andmemory. Still, the major drawback of MRAS is the open-loopflux estimator sensitive to an error in terminal voltage estimationand integration, as well to an error in stator resistanceparameter Rs. Listed problems with the reference model canbe significantly reduced [8], [9]. Nevertheless, the sensitivityof the adjustable model used in MRAS to an error in the rotorcircuit parameters must also be considered. In particular, if therotor time constant parameter T∗r is not equal to its actual valueTr, an error in the estimated rotor speed will be introduced. Theproblem gets more significant in the low-rotor-speed region,where it becomes essential to upgrade the speed estimator withan online Tr identification mechanism.Variation of Tr is caused mainly by the change in rotorresistance due to temperature and also by the change in rotorinductance due to saturation. Saturation-induced variation inactual Tr value does not need to be tracked, but rather, it canbe predicted and included in the feedforward flux model [10].However, slow-tracking T∗r update algorithm is still required foronline compensation of unpredictable Tr thermal drift.The research for online Tr identification mechanism startsfor IFOC drives using shaft sensor. In those drives, an errorin T∗r greatly affects an open-loop slip estimator and leads toundesirable cross coupling and deterioration of overall driveperformance.

Download full report
http://vukosavic.etf.rs/darmar.pdf