本論文提出自組織式模糊控制器於磁浮控制系統。磁浮系統存在不穩定與非線性特性,因此提供一種高性能的控制器於磁浮控制系統在應用上相當重要。模糊控制器與相關人工智慧方法已被成功應用 於磁浮系統的位置控制,此類方法的缺點為設計一套最佳模糊控制器於此動態系統相當困難。本論文提出以混合粒子群演算法決定模糊控制器中每一個輸入變數之最佳模糊切割空間與相關歸屬函數的位置。由於混合粒子群演算法為一極優的最佳化工具,因此非常適合於設計最佳模糊控制模型。本論文所提出的方法應用於一自行設計、以微控制器為核心之磁浮控制系統,結果顯示,所提出的自組織式模糊控制器能較現有方法提供更佳的位置控制。
This thesis proposes a new control approach for magnetic levitation systems using self-organizing fuzzy controllers. The relevant features of the magnetic levitation control systems (MLCSs) are its inherent instability and nonlinearity in electromechanical dynamics. Development of a high-performance controller for the position control of a MLCS is thus quite important in applications. The fuzzy controller and the related artificial intelligence (AI) approaches have been successfully applied to the position control of a MLCS. The disadvantage of this sort of method is that design of an optimal fuzzy model for the dynamic system is often difficult. In this thesis, a hybrid particle swarm optimization (HPSO) algorithm is relied on to solve the problem of determining the best fuzzy partition of the input spaces and associated fuzzy membership functions for each input variable. Since HPSO is an excellent optimization tool, it is easily adequate to design the optimal fuzzy control model. The proposed approach has been verified on a microprocessor-based self-designed MLCS. The results show the proposed self-organizing fuzzy controller can provide better position control than existing methods.