本論文主要目的是利用倒傳遞類神經網路,結合模糊理論,利用模糊控制的穩定特性,與類神經網路的學習優點,應用在重覆性控制器於線性伺服系統之控制。 本控制方法,首先設計輸入線性伺服馬達的運動特性曲線,作為參考的命令,接著分別設計比例-積分控制器及相位超前補償器來控制線性伺服驅動系統之速度及位置迴路,藉以改善系統步階追隨命令時所產生的穩態誤差。針對可變動週期運動的輸入命令,設計糢糊類神經網路控制器,使系統能維持良好的響應及零穩態誤差,並降低系統加速度響應峰值。 本伺服系統控制以PC-Based的方式,利用數位/類比轉換卡與MATALB/SIMULINK等軟體建構一即時模糊類神經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可獲得最佳的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。
The purpose of this thesis is to improve the stability of the linear servo system by using correct repetitive controller. The back-propagation neural networks and fuzzy theorem are both used to the correct repetitive controller because of the advantages of the fuzzy stable characteristics and the neural network’s learning ability. Therefore, fuzzy-neural works was applied to the linear servo system on this thesis. For this control method, we first design a motion curve as a reference command for the input of linear servo motor, then, in order to improve the steady state error produced by system step-follow order, the proportional - integral (PI) controller and a phase-lead compensator were designed to control the speed and position loop of the linear servo drive system. The design of fuzzy-neural controller can be proved to enables the system to maintain a good response and zero steady state error, and reduce system acceleration response peak. For the servo system control, the PC-based way, digital / analog converter card with MATALB / SIMULINK software simulation are used. With the simulation and experimental results, the performance of the dynamic response of the linear servo motor will be proved. The speed responses are more accurate, the steady state error can be effectively eliminated, and reduce steady-state error and the peak acceleration response system.