一適應性重複性控制器新設計方法

Abstract

本論文提出一新的設計方法來設計數位式重複性控制器，並使用適應性陷波器來估計週期性訊號週期，最後結合兩者以形成一適應性數位式重複性控制器。 重複性控制器的目的是要用來追蹤或去除一週期性的訊號，而設計此控制器的困難度在於如何使整個系統維持穩定。在文獻中的設計方法雖然系統化，但所設計之控制器實現複雜度高，而且此控制器的性能缺乏量化的分析，只能保證系統穩定，不一定是一最佳解。本論文提出一Max-Min 的量化標準來設計重複性控制器，所設計的系統在此量化標準下為最佳，且容易實現。我們並以實用的伺服馬達及磁碟機為例，設計及模擬以驗證此方法的可行性。 由於大部分實用的系統，週期性訊號的週期很難事先知道，因此本論文提出使用適應性陷波器來估計此訊號週期，並結合數位式重複性控制器，來達成適應性數位式重複性控制器，最後並模擬以驗證其可行性。

This thesis presents a new method for designing a discrete-time repetitive controller and applies the technique of the adaptive notch filter for estimating the period of a periodic signal. The resulting system after integration of the above two techniques forms an adaptive discrete-time repetitive controller. The repetitive controller serves mainly as a compensator, plugged into a control system, with the purpose to remove or track a periodic signal, which arises from the desired reference signal or the undesired disturbance. Conventional approaches design a repetitive controller focusing on how to ensure the stability of the resulting system. Hence the designed repetitive controller is usually difficult to implement, its performance is not quantified and thus is not optimum. To avoid this drawback, we define a new criterion and use this criterion to design a discrete-time repetitive controller which is not only optimum in the sense of the criterion but also simple for realization. The usefulness of the new design criterion and approach is further verified by simulations in design examples using the servo motor and disk file actuator system. In most practical systems, the period of a periodic signal is commonly unknown. We propose, in this thesis, to use an adaptive notch filter for estimating the signal period. This adaptive system integrates with the discrete-time repetitive controller, yielding an adaptive discrete-time repetitive controller. Successful results are obtained from simulation of the whole system in several examples.