柯文俊臺灣大學：工程科學及海洋工程學研究所李文仁Lee, Wan-YanWan-YanLee2007-11-262018-06-282007-11-262018-06-282007http://ntur.lib.ntu.edu.tw//handle/246246/51097本文提出一套前處理技術，藉以輔助各種結構系統識別方法，簡化整個系統識別的作業流程。在實測振動資料中常因無法避免的量測通道雜訊干擾或是對結構特徵不清楚之下造成過採樣，本文所提之前處理技術以組合了降頻採樣濾波器組、時譜圖及模態穩定圖等三種方法來有效評估結構系統之自然振頻。透過前處理的技術，可以減少試誤的過程，增加整個識別流程的效率。 為了驗證本文提出之前處理技術可有效地應用於真實世界資料，文中透過一個十自由度系統，在加速度的量測條件下，加入不同的雜訊用以測試前處理技術抵抗雜訊的能力，並針對其識別結果之準確性做一綜合性的比較。最後探討三組不同結構的實驗例，分別為懸臂鋼樑、鋁合金板以及機車車架結構。懸臂鋼樑主要用於檢驗自由響應資訊之結構系統識別；鋁合金板為驗證結構在受迫響應之下，文中所提出之前處理技術之可行性；機車車架結構則為展現前處理技術在未知雜訊干擾下，對於結構系統識別之必要性及其優越性。由分析結果顯示本文提出之前處理技術可提高整個結構系統識別的效率及準確性。In this report the prepossessing techniques has been proposed to assist various kind methods of structural system identification in simplifying the whole system identification procedure. Because of the noisy vibrating data measured from unavoidable data channels or the over determined condition had effected up on the modal parameters, the identification procedure firstly employs the prepossessing techniques to reduce the affection of noise interference. By the way, it can reduce the course tried by mistake and increase the efficiency of the whole system identification procedure. The prepossessing techniques combine the effective capabilities of decimator filter banks, time-frequency analysis and modal stabilization diagram analysis. In order to prove the prepossessing techniques can be applied to analyze the real-world data effectively, a numerical example of simulation for a ten-degree-　of-freedom system was employed to test the identification capability of prepossessing techniques under different noise levels. Finally, we probed into the experiment examples of three different structures, cantilever, aluminum alloy board, and a framed structure of motorcycle. The experiment of cantilever was proved the identification results of free vibration, the experiment of aluminum alloy board was proved the prepossessing techniques is usefulness at force vibration, and the experiment of motorcycle was demonstrated the necessity and superiority of prepossessing techniques when the data is under unknown affection of noise interference. The results obtained from the prepossessing techniques reveal that the analysis can raise efficiency and accuracy.中文摘要 I 英文摘要 II 簡稱術語對照表 III 目錄 IV 圖目錄 VI 表目錄 X 符號說明 XII 第一章：導論…………………………………………………………1 1.1 研究目的…………………………………………………………1 1.2 文獻回顧…………………………………………………………2 1.3 報告架構…………………………………………………………5 第二章：狀態空間系統………………………………………………8 2.1狀態運動方程式…………………………………………8 2.2連續與離散時間狀態空間轉換……………………… 12 2.3離散時間等效狀態方程組…………………………… 14 第三章：前處理技術在模態識別上的應用 18 3.1 降採樣………………………………………………………… 18 3.1.1降採樣的頻域特性…………………………………………20 3.2時譜分析…………………………………………………………23 3.3模態穩定圖分析…………………………………………………31 3.3.1 前向型自我回歸模型(VAR) ………………………… 32 3.3.2 隨機次空間運算識別(SSA) ………………………… 34 第四章：數值模擬例 43 4.1十自由度系統之描述…………………………………43 4.2不同雜訊水平下前處理技術之評估能力……………48 4.3過採樣資料分析之模擬…………………………………………61 4.4模擬識別結果探討……………………………………74 第五章：結構識別應用…………………………………………… 76 5.1懸臂鋼樑………………………………………………76 5.1.1懸臂樑理論分析………………………………… 77 5.1.2懸臂鋼樑識別結果探討………………………… 80 5.2以強迫振動儀激振之鋁合金板………………………91 5.2.1鋁合金板識別結果探討…………………………92 5.3 機車車架結構……………………………………… 99 5.3.1機車車架結構識別結果探討………………… 102 第六章：結論與展望 114 6.1結論…………………………………………………………… 114 6.2展望…………………………………………………………… 116 參考文獻………………………………………………………… 1179085507 bytesapplication/pdfen-US模態參數時譜分析模態穩定圖前向型自我迴歸模型隨機次空間運算特徵根濾波器modal parameterstime-frequency analysismodal stabilization diagramvector forward auto-regressivestochastic subspace algorithmeigenvalue filterthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/51097/1/ntu-96-R93525032-1.pdf