適應性倒階類神經濾波控制器與其在伺服馬達控制上之應用

Abstract

本論文針對一未知非線性控制系統，提出一個以輻狀基底函數(radial basis functions )類神經網路(neural networks)的適應性倒階(Backstepping)控制器。在適應性倒階控制器設計中，將使用輻狀基底函數類神經網路近似未知非線性函數。一般的倒階控制器設計過程中，必須要對虛擬控制輸入微分，因而導致輻狀基底函數類神經網路在近似過程中需要執行多次微分運算。因此，為了避免輻狀基底函數類神經網路在適應性倒階控制器設計中需多次微分，本論文使用濾波器取代微分運算，以減少計算複雜度。此外，藉由李亞普諾夫函數分析整體閉迴路系統的穩定度。 最後，本文利用數個電腦模擬範例和直流伺服馬達實驗來驗証所提出方法效能與應用性，其中直流伺服馬達實驗包括具有正負電壓輸出之切換式直流電壓轉換電路設計、電壓回授電路設計與脈波寬度調變(Pulse Width Modulation)控制器設計等。

In this thesis, a radial basis function (RBF) neural adaptive backstepping controller for a class of nonlinear system with unknown nonlinearities is proposed. In backstepping design procedure, the RBF neural networks are used to approximate unknown nonlinear functions. In general, based on backstepping design technique, virtual controls must be differentiated. For this reason, differentiating the RBF neural networks is required. In order to avoid the requirement of the nth derivative of the RBF neural networks, first-order filters are added into backstepping design such that the computation burden can be effectively alleviated. In addition, the stability of the closed-loop system with first-order filters is analyzed by Lyapunov functions. Finally, simulation results and experiment results are provided to demonstrate the effectiveness and applicability of the proposed method. The experiment is composed of a DC servo motor, a switch DC-DC converter, voltage feedback circuits, and PWM (Pulse Width Modulation) controller.