Adaptive state filtering with neural networks for sensorless induction motor speed estimation /
Effective sensorless dynamic speed estimation is desirable for both on-line induction motor condition monitoring and speed sensorless motor-drive applications. In this research a "semi-parametric" adaptive speed filter is used for induction motor speed estimation. The proposed method of...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
2000.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=728408401&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | Effective sensorless dynamic speed estimation is desirable for both on-line induction motor condition monitoring and speed sensorless motor-drive applications. In this research a "semi-parametric" adaptive speed filter is used for induction motor speed estimation. The proposed method of speed filtering is based on recurrent neural networks and it is an application of a recently developed adaptive state filtering algorithm. The adaptive speed filter is formulated using the two-step prediction-update approach, based on the fundamental principles of the Kalman Filter. The proposed adaptive speed filter is constructed using neural networks, which are trained using measured terminal voltages, currents, and inferred motor speed. The rotor slot harmonic detection-based inferred speed estimates are used as targets for the adaptive speed filter training. In addition to the measured motor terminal quantities and inferred speed, only motor nameplate information is required for the adaptive speed filter setup and operation. Two adaptive speed filters are developed for motor speed estimation: one for motors running from the power supply mains, and a second for motors running off a voltage source inverter. The speed filter developed for induction motor running from power supply main gives good speed estimates under varying load conditions. The developed speed filter is also shown to give good speed estimates under power supply imbalance and faulty machine conditions. Further, the developed speed filter is scalable, and after incremental tuning, it is successfully used for speed estimation of induction motors with higher power rating. In the second case, the performance of the adaptive speed filter developed for induction motors running from a voltage source inverter has been presented. Even though the adaptive speed filters are developed using motor terminal measurements, the filtering accuracy is comparable to speed estimation techniques discussed in the literature. This research demonstrates feasibility of neural networks-based adaptive speed filtering for effective induction motor speed estimation without explicit use of the motor parameters and dynamics. The use of only standard motor electrical measurements and the scalability of the developed filter to motors with different power ratings enhance the applicability of this development. This work can be used as a stepping-stone for the development of load independent fault detection schemes and fault tolerant control algorithms for induction motors. |
|---|---|
| Item Description: | Vita. "Major Subject: Electrical Engineering". |
| Physical Description: | xv, 134 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilm Inc. |
| Bibliography: | Includes bibliographical references (leaves 117-125). |