標題: | 類神經網路於吊卸系統振動位移控制之應用 The Application of A.N.N. in Active Displacement Control of Suspension System |
作者: | 何佳慧 Ho, Chia-Huey 鄭復平 Cheng, Fu-Ping 土木工程學系 |
關鍵字: | 類神經網路;吊卸系統 |
公開日期: | 1997 |
摘要: | 振動是各種機械系統運轉或運動時,普遍存在的共同現象。本研究將試圖利用類神經網路於吊卸系統的振動位移控制。吊卸系統於吊卸負載物時,負載物由於受到衝擊荷重作用,將產生振動位移,於此種振動位移量的控制往往以人力輔助,但此種方式即不經濟效果亦有限。為達控制的要求,擬於負載物上加裝一控制系統。本研究嘗試以 ANSYS50A 有限元素分析軟體建立吊卸負載物的運動模型,透過分析找尋不同衝擊荷重作用下,相對應的最佳控制力與時間的組合,經由類神經網路的學習,建立衝擊荷重與控制力組合間的網路關係,於實際吊卸負載物時,以類神經網路進行離線(Off-Line)學習,即時控制,縮短時間延遲,以達較佳的控制效果。
本研究以 ANSYS50A 套裝軟體建立吊卸負載物的運動模型,分析結果經由類神經網路學習,學習效果良好,以模擬之控制力測試求得最大誤差 0.2725%,最小誤差 0.0016%,平均誤差 0.11697%;控制力作用時間測試的結果,最大誤差 0.9133%,最小誤差 0.008%,平均誤差 0.42743%。 Vibration exists in most of operating machine. In this research, active control system by ANN was used to reduce the amplitude of the displacement in suspension system. A finite element package (ANSYS) was used to simulate the dynamic suspension system dealing with the displacement variation after the hanged material was subjected an impact force. For each impact force and impact duration, one best acting force and acting duration of the active control system was found by the finite element approach. The above data establish a database for off line learning by ANN. Some data were used to test the accuracy of this ANN system. The error for test data have 0.9133% maximum error. 0.0016% minimum error and 0.11697% average error for acting force. The errors of acting duration are 0.9133% maximum. 0.008% minimum and 0.42743% average. These data show that this ANN system works well. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#NT863015038 http://hdl.handle.net/11536/63284 |
顯示於類別: | 畢業論文 |