臺灣大學: 機械工程學研究所陳明新林逸哲Lin, Yi-JeYi-JeLin2013-04-012018-06-282013-04-012018-06-282010http://ntur.lib.ntu.edu.tw//handle/246246/256256本論文針對受到決定性干擾的線性迴歸系統提出一個線上強韌系統識別演算法。相較於傳統的強韌識別方法，本論文所使用的演算法不但能夠在沒有任何干擾相關資訊的情況下有效估測出系統參數，更能進一步的獲得干擾的變化。藉由線上時變的多項式做為干擾模型，將干擾以係數為未知常數的多項式函數模型來表示，並整合系統與干擾的未知參數形成線性迴歸系統，最後配合Kalman Filter估測器同時估測出系統參數與未知的干擾。在估測的過程中，所選取的估測器參數將影響估測的結果，甚至某些估測器參數的設定將導致估測系統中Riccati 方程式的解趨近無窮大，而達不到預期的參數識別效果，為了解決這方面的問題，本論文將設計一個上限來限制Riccati 方程式的解，確使整個設計能夠順利運作並保證估測結果能夠擁有一定的準確性。This thesis proposes an on-line robust identification algorithm to estimate unknown parameters in a linear regression form that is contaminated by a deterministic disturbance signal. In this thesis, not only the parameters will be obtained by this algorithm, but also the disturbance will be estimated by an on-line polynomial fitting model without any disturbance information. In this algorithm, we use a polynomial fitting model with unknown coefficients to represent the disturbance in the system. Both unknown parameters of system and the time-varying disturbance will be estimated accurately by a Kalman filter observer. However, the results of the estimation will be affected by the choice of observer parameters. Under certain circumstances, the singular values of the solution in the Riccati equation will approach infinity. To prevent the singular values from increasing unlimitedly, a threshold for reset is set in order to ensure theis algorithm is capable of successfully estimating the real parameters.826543 bytesapplication/pdfen-US強韌系統識別參數識別線上系統識別Kalman FilterRiccati 方程式上限robust identificationparameter estimationon-line system identificationRiccati Equationthresholdthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/256256/1/ntu-99-R97522824-1.pdf