本研究提出了一種使用區間二型遞迴模糊神經網路估測器之智慧型步階回歸控制，能夠修正非線性時變系統固有的非線性和時變控制特性。在區間二型遞迴模糊類神經網路估測器之智慧型步階回歸控制中，設計步階回歸控制法則來穩定閉環控制系統，並使用區間二型遞迴模糊神經網路來估計步階回歸設計中的總集不確定項。最初，非線性步階回歸控制的逐步設計被設定用於追蹤週期性參考軌跡，總集不確定項為保守常數。然而，實際應用通常涉及未知且難以預測的總集不確定項，為解決此問題引入區間二型遞迴模糊神經網路來即時估計總集不確定性。應用李亞普諾夫穩定性方法來確保漸近穩定性，從而製定區間二型遞迴模糊神經網路的線上學習演算法。為主動補償區間二型遞迴模糊神經網路的估計誤差亦提出自適應補償器。最後，本研究包括一個案例研究，展示具有最大每安培扭矩控制的同步磁阻馬達位置伺服驅動器的實驗結果。這些結果旨在驗證所提出的區間二型遞迴模糊神經網路智慧型步階回歸控制的有效性和穩健性。;An intelligent backstepping control with interval type-2 recurrent fuzzy neural network (IBSCIT2RFNN), which is capable of modifying the inherent nonlinear and time-varying control characteristics of a nonlinear time-varying system, is proposed in this research. In the IBSCIT2RFNN, a backstepping control (BSC) law is devised to stabilize the closed-loop control system and the lumped uncertainty in the design of BSC is estimated using an interval type-2 recurrent fuzzy neural network (IT2RFNN). Initially, a step-by-step design of a nonlinear BSC is formulated for tracking periodic reference trajectories, with uncertainties lumped by a conservative constant. However, practical applications often involve unknown and challenging-to-predict lumped uncertainty. To address this, an IT2RFNN is introduced for real-time estimation of the lumped uncertainty. The Lyapunov stability method is applied to ensure asymptotical stability, leading to the formulation of online learning algorithms for the IT2RFNN. In order to proactively compensate the estimation error of the IT2RFNN, an adaptive compensator is also presented. Finally, this research includes a case study presenting experimental results from a synchronous reluctant motor (SRM) position servo drive with maximum torque per ampere (MTPA) control. These results aim to validate the effectiveness and robust qualities of the proposed IBSCIT2RFNN.
