Recurrent Neural Fuzzy Network for High -precision Motion Controlof PMLSM Drives
碩士 === 大葉大學 === 機械與自動化工程學系 === 98 === This thesis proposes a recurrent neural fuzzy network (RNFN) for the high-precision motion control of permanent magnet linear synchronous motor (PMLSM) drives. The RNFN Control system consists of two network structures; namely, RNFN identifier (RNFI) and RNFN C...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | zh-TW |
| Published: |
2010
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/40180563405016528550 |
| Summary: | 碩士 === 大葉大學 === 機械與自動化工程學系 === 98 === This thesis proposes a recurrent neural fuzzy network (RNFN) for the high-precision motion control of permanent magnet linear synchronous motor (PMLSM) drives. The RNFN Control system consists of two network structures; namely, RNFN identifier (RNFI) and RNFN Controller (RNFC). The RNFI is first trained to capture the inverse dynamics of the PMLSM drive and then is used as a feedforward controller to calculate the desired control force of the PMLSM along a desired trajectory; while RNFC is used as an error-feedback Compensator to minimize the trajectory tracking error resulted from system uncertainties. Structure and Parameter learning algorithms are concurrently preformed is RNFN online. A recursive recurrent learning algorithm based on the gradient descent method is derived for the parameter learning. An analytical method based on a discrete-type Lyapunov function is proposed to guarantee the convergence of RNFN by choosing varied rates. The experimental setup is comprised by a host computer, a servo controller board, a motor driver and a PMLSM. Simulation sand experiments performed on a PMLSM drive demonstrate the effectiveness off the proposed control system.
|
|---|