- 學位論文
以類神經網路為基礎之線性伺服系統重覆性控制
Repetitive Control of Linear Servo System Based on Artificial Neural Network
摘要
本論文主要目的是將類神經網路學習的特色，應用於線性伺服系統之重覆性控制。首先，根據線性伺服系統運動特性曲線設計一相位超前補償器，藉以改善系統追隨位置命令時所產生的相位誤差。其次，針對一般控制器在追隨週期信號時，所產生的週期性誤差，而設計一重覆性控制消除週期性的穩態誤差。另外，為了降低線性伺服系統的加速度響應峰值，避免機具運作時，產生較大的震動，利用類神經網路針對不同運動特性曲線進行學習訓練，將其結果運用於重覆性控制器中，藉以改善傳統重覆性控制器中固定的二階低通濾波器的值，使得重覆性控制在不同運動特性曲線的連續週期命令下，能有良好的運動特性響應與穩態誤差的消除。 線性伺服驅動系統控制以PC-Base的方式，利用數位/類比轉換卡與MATLAB/ Simulink等軟體建構一即時重覆性位置運動控制系統，配合電腦模擬及實驗驗證本控制法則確實改善系統輸出響應和穩消除態誤差，並降低系統加速度響應峰值。
並列摘要
The objective of this thesis is to learn of artificial neural network characteristic, Application of linear servo system controller of repetitive control. First, according to the linear servo motion characteristic curve system design the phase lead compensator to improve the system to follow the position command generated by the phase error. Secondly, Conventional controller for tracking period signal generated periodic error. And the design repetitive controller reduce periodic steady-state error. In addition, linear servo system in order to reduce the peak acceleration response and avoid the operation of machinery, have a greater shock, the utilizes of neural networks for different motion characteristic curve training. The result will be applied to repetitive controller to improve the traditional fixed repetitive controller of the second-order low-pass filter value, Makes control of repetitive in different characteristic curves of the continuous cycle of the order to the motion characteristics of a good response and the reduce of steady-state error. Linear servo-drive system to control the PC-Base approach, using digital / analog interface and MATLAB/Simulink software to build a real-time position of repetitive motion control system, With computer simulation and experimental verification of this control law is indeed to improve the system output response and reduce steady-state error and the peak acceleration response system.