系統參數變動對控制性能有極大影響,必須即時鑑別出系統參數之變動,方能有效地設計控制器與提升系統性能,因此系統參數鑑別之準確性是十分重要的工作,所以本文提出離線參數鑑別及在線參數鑑別之法則來精確地估測出受控体系統參數。 利用參數鑑別之結果來設計I-PD迴授控制器與前饋控制器,其中I-PD迴授控制器相較於傳統串級式控制架構而言,它可提升系統頻寬並同時達到速度與位置之控制。前饋控制器則用來提升軌跡追蹤性能,但前饋控制器會造成系統飽和現像發生,因此本文於前饋控制器中加入一比例因子,並利用工業上之經驗法則來調適此因子。 上述之系統參數鑑別法則、迴授控制器及前饋控制器將透過模擬與實驗來驗証其性能,針對實驗部份,本文將建構即時多工的環境來實現控制系統與相關之演算法。
The mechanical systems, such as machine tools, semiconductor manufacture equipment, mechanical manipulators and automatic inspection machines, must be supported by motion controllers, which ensure robust, high speed and high accuracy tracking performance. To control the system adequately, the system parameters must be known and the parameters of controller has to be adjusted accordingly. This paper presented a digital servo driver that realizes an auto-tuning feedback and feedforward controller design using off-line and on-line parameters identification. Firstly, the inertia constant, damping constant and the disturbed load torque of the servo motor are estimated by off-line identification. In manufacture processing, the system parameters are varied according to different working conditions. Therefore, two algorithms of on-line parameters identification are proposed to identify the variant inertia constant, damping constant and the disturbed load torque. The controller proposed here uses an I-PD feedback controller, which has larger bandwidth than cascade control of velocity and position loop, and a direct velocity and acceleration feedforward (DVAFF) controller to improve the tracking performance. By using the estimated inertia constant and damping constant, the I-PD controller and DVAFF controller can adaptively tune itself to compensate the parameter variations. The proposed auto-tuning digital servo controllers are evaluated and compared experimentally with a traditional controller on a microcomputer-controlled servo motor positioning system. The simulated and experimental results show these controllers robustly sustain the control performance and dramatically reduce the tracking error even if the plant parameters are varied.