完整後設資料紀錄
DC 欄位語言
dc.contributor.author張家銘en_US
dc.contributor.authorChia-Ming Changen_US
dc.contributor.author李安謙en_US
dc.contributor.authorAn-Chen Leeen_US
dc.date.accessioned2014-12-12T02:23:59Z-
dc.date.available2014-12-12T02:23:59Z-
dc.date.issued1999en_US
dc.identifier.urihttp://140.113.39.130/cdrfb3/record/nctu/#NT880489093en_US
dc.identifier.urihttp://hdl.handle.net/11536/66130-
dc.description.abstract在絕大多數的工具機中,其系統存在的摩擦力現象,常常是降低工具機精度的主要原因,工業界與學界莫不以消除摩擦力為努力的目標。而且摩擦力的高度非線性特性也造成了一般在工具機上所採用的線性PID控制器補償上的困難。本論文主要之目的即為發展DSP-Based運動控制系統中,補償摩擦力影響的控制法則。 在伺服控制迴路上,我們除了採用PI-D控制器與前饋控制器作為基本的控制器架構以外,我們特別針對摩擦力現象設計了一個類神經網路補償器,其架構是以速度或誤差訊號作為類神經網路的輸入,並產生位置補償命令來克服摩擦力,以使系統能夠達到更好的精度。此類神經網路補償器的學習一方面採用離線的批次學習法,一方面使用線上的修正逆傳遞學習演算法來修正類神經網路上的加權值。 另外在系統整合方面,其實現方法是利用DSP運動控制卡當副控端(slave),配合個人電腦當主控端(master),以個人電腦上的人機界面來幫助使用者作控制器與類神經網路補償器的參數調適工作。整個伺服迴路上的設計流程可分為系統鑑別、PI-D控制器與前饋控制器設計、類神經網路補償器的設計與調適。 在隨後的電腦數值模擬與以線性馬達為對象的實驗中,驗證了本文提出的補償器架構確實能夠有效的將摩擦力所造成的誤差加以降低,證明了此類神經網路補償器實現的可行性。zh_TW
dc.description.abstractIn most of CNC machines, the friction in the system is usually the main cause of precision decline. Eliminating the friction is increasingly important in both the industry and the academia. But the linear PID controller used in general CNC machines is difficult to compensate the serious nonlinear phenomenon of friction. The purpose of this thesis is to develop a control algorithm of compensating friction phenomenon in a DSP-Based motion control system. Basically, the servo control loop of the motion system consists of a PI-D controller and a feed-forward controller. Besides, a neural compensator is also appended to the system to eliminate the friction. This compensator uses velocity or error signal as the input of the neural network compensator and generate position command to eliminate the friction. The learning in the neural network compensator is through both the offline batch learning and the online modified back-propagation learning algorithm to modify those weightings in the neural network. The motion system integrates a DSP motion control card as the slave, and a PC as the master controller. The PC also helps to tune the parameters of both the original servo controller and the new neural network compensator through a man-machine interface built in the same PC. The design procedures are as follows: system identification, PI-D and feed-forward controller design, and the design and tuning of neural network compensator. The numerical simulations and experiment results of a linear motor show that this neural compensator indeed reduces the error caused by friction. It is worthwhile to induce this kind of compensator into the servo control system.en_US
dc.language.isozh_TWen_US
dc.subject類神經網路zh_TW
dc.subject運動控制zh_TW
dc.subject摩擦力補償zh_TW
dc.subjectArtificial Neural Networksen_US
dc.subjectMotion Controlen_US
dc.subjectFriction Compensationen_US
dc.title類神經網路在DSP-Based運動控制系統上的非線性補償zh_TW
dc.titleNonlinear Compensation on DSP-Based Motion Control System by Artificial Neural Networksen_US
dc.typeThesisen_US
dc.contributor.department機械工程學系zh_TW
顯示於類別:畢業論文