參數最佳化之適應性類神經網路控制器於主動式氣壓隔振系統之控制

Pneumatic Vibration Isolator Control by Using Adaptive Neural Network Controller

梁晶煒(Jin-Wei Liang) ； 陳宏毅(Hung-Yi Chen) ； 周彥騰(Yan-Teng Chou) ； 王敬智(Ching-Chih Wang)

《應用聲學與振動學刊》 7卷2期 (2015/12) Pp. 27-35

https://doi.org/10.6282/JASV.2015.7.2.03

摘要

氣體的可壓縮性及氣壓系統孔口流等特性，使得氣壓隔振系統具有高度的非線性，若要建立系統正確之數學模式是非常不容易的。本研究嘗試以函數近似法為基礎之適應性滑動控制器，結合適應性類神經網路控制器，來針對主動式單缸膜片氣壓隔振系統進行控制。類神經網路具有強大的學習能力，高度的容忍誤差、平行運算之學習能力等特性，故可用於非線性動態函數之近似。本研究以適應性小波類神經網路與適應性輻射函數神經網路控制法則結合函數近似法則來進行控制器之設計，並以田口法來獲得系統之最佳控制參數，針對主動式單缸膜片氣壓隔振系統進行控制。從實驗結果可以得知，本研究所設計之控制器在單缸膜片式氣壓隔振系統之控制中，可呈現明顯之隔振成效。

關鍵字

單缸膜片氣壓隔振系統 ； 適應性小波類神經網路 ； 適應性輻射類神經網路 ； 田口法

並列摘要

It is well known that a pneumatic actuating system has nonlinear uncertainty and time-varying characteristics. It is difficult to establish an accurate process model for designing a model-based controller to monitor the pneumatic actuating force. An intelligent control strategy for a pneumatic vibration isolation system is developed in this research. In this paper, a model-free adaptive wavelet neural network (AWNN) controller and radial basis function neural network (ARBFNN) controller is proposed to control a diaphragm-type pneumatic vibration isolator. This approach has online learning ability and the advantage to achieve the controller design without knowledge of the system dynamic model. In order to validate the proposed method, a composite control scheme using pressure and velocity measurements as feedback signals is implemented. In addition, Taguchi method had been utilized to obtain the optimal control gain values for this control system. Experimental results are executed to show the control performance of the proposed intelligent controller.

並列關鍵字

Diaphragm-Type Pneumatic Vibration Isolator ； Adaptive Wavelet Neural Network Controller ； Adaptive Radial Basis Function Neural Network ； Taguchi Method