- 學位論文
以類神經網路模擬模糊滑動模式之主動結構控制
Simulation of Fuzzy Sliding Mode Control by Using Neural Network in Active Structural Control
摘要
本文主旨以類神經網路模擬FSMC之主動結構控制，以類神經網路自我學習式控制器為主體，以倒傳遞神經網路訓練控制規則之自我產生，期望由網路自我學習調整其控制狀態，達到最佳的控制效果。並利用單輸入多輸出控制理論推導模糊滑動模式控制（fuzzy sliding mode neural networks control, FSMC）理論，以類神經網路模擬模糊滑動模式控制，研究主動式調質阻尼器應用於樓房結構之控制效果，並進行數值模擬分析，探討可變結構滑動模式控制法(VSS)與模糊滑動模式控制(FSMC)及類神經網路模擬模糊滑動模式控制在樓房結構控制系統的可行性，並和未控制前比較，進一步探討控制過程之結構系統行為。 模擬結果證實，本文類神經網路控制可以有效應用主動式樓房結構控制系統，樓房結構系統不論在單自由度或多自由度系統在受不同外擾力作用和不同初始位移時，甚至在系統參數變動下，類神經網路控制確實可以減低結構動態反應，可得到良好的控制效果，並具有高度的強健性與容錯性。
並列摘要
The purpose of this study is to simulate Fuzzy Sliding Mode Control (FSMC) by using neural network in active structure control. The principal part of this paper is self-learning neural network controller, and produce the control rule-base by using a back-propagation neural network with self-learning the control state. The neural network was established to simulate the FSMC control theory, the FSMC controller was replaced by the neural network controller. We studied the appropriation of VSS, FSMC and neural network on structure control system, and compared to the system before controlling. The simulation results showed that the neural network control method of active control for building structures in single or multi-degree of freedom system can decrease structural dynamic responses. The results also showed good control effects and high robustness as well as wide tolerance. No matter what kinds of disturbance act or different structural initial displacement, even the system parameters change, the neural network control will still show satisfactory results.