利用卡氏濾波器組進行感測器受損評估與結構損傷識別

Abstract

近年來，為了延長結構的使用壽命及保護結構免於遭受下次災害的破壞，結構健康監測越來越受工程界重視。大部份結構健康監測都是使用感測器量測訊號，藉由感測器所提供之資料與訊息進行結構健康監測，因此感測器的工作性會直接影響到結構健康監測的分析結果。本研究目的之一為提出一套新的方法進行感測器的故障判別。本研究考量三種訊號錯誤：訊號偏差 (additive fault)、訊號放大 (multiplicative fault)、訊號偏移 (slowly drifting fault)。首先將量測訊號建立自回歸模型 (Auto-Regressive model)，接著將此模型轉換成單一卡爾曼濾波器，由於此濾波器是利用全訊號源作為輸入，因此藉由擷取部分卡爾曼濾波器做為可變輸入輸出之多組卡爾曼濾波器。利用建立完成的卡爾曼濾波器組來估計感測器的訊號資訊，藉由計算真實量測訊號和估計訊號的差值，可觀測出破壞的模式以及發生時間，達到即時的分析處理。此外，本方法也能對於重複出現的感測器錯誤進行診斷。本研究利用模擬及於實驗資訊以人工方式加上感測器故障進行判別。從模擬與實驗結果可得知此方法能夠有效的判別出損傷的感測器、損傷種類和損傷發生的時間，使感測系統能夠順利進行結構健康監測。 在保證感測器之工作性後，結構物之感測系統可進行結構健康診斷。本研究目的之二為利用卡爾曼濾波器組進行結構的損傷識別。該方法包括四個步驟，首先利用頻率域多輸入多輸出系統識別方法或時間域唯有輸出的自回歸模型之系統識別方法，建構結構模型；然後，將該結構模型轉換成全訊號輸入之卡爾曼濾波器分離該卡爾曼濾波器成多組卡爾曼濾波器，並估計結構兩節點間的相對反應，計算真實量測結構反應與估計反應之差值；最後，利用未損傷的結構指數與事件結束的結構事件指數進行比較，將該差值透過統計表示法，建立結構損傷指標，以利判斷結構桿件損傷程度。本研究利用模擬試驗來評估方法的可靠性，並在國家地震中心進行共三組結構試驗與實驗驗證，由模擬及實驗結果皆顯示此方法能成功判斷出結構損傷之發生、損傷之位置及損傷之程度。

Structural health monitoring has drawn great attention in the field of civil engineering in past two decades. These structural health monitoring methods evaluate structural integrity through high-quality sensor measurements of structures. Due to electronic deterioration or aging problems, sensors may yield biased signals. Therefore, the first objective of this study is to develop a fault detection method that identifies malfunctioning sensors in a sensor network before the structural health monitoring begins. This method exploits the autoregressive modeling technique to generate a bank of Kalman estimators, and the faulty sensors are then recognized by comparing the measurements with these estimated signals. Three types of faults are considered in this study including the additive, multiplicative, and slowly drifting faults. To assess the effectiveness of detecting faulty sensors, a numerical example is provided, while an experimental investigation with faults added artificially is studied. As a result, the proposed method is capable of determining the faulty occurrences and types. The second objective of this study is to develop a structural health monitoring strategy for damage detection. Buildings may suffer serious damage when subjected extreme loadings such as strong winds and earthquakes. In seismic events, the error time histories between measured and estimated responses should contain the information of the structural deterioration, i.e., the locations, levels, and time of occurrences. Therefore, this study presents a new damage detection method based on prediction errors using a bank of Kalman estimators. A representative model of a building is derived from a frequency-domain multi-input, multi-output system identification method under ambient vibration prior to earthquakes. This model is then converted into a bank of estimators that calculate estimation errors. Damage is interpreted by statistical indices from these errors and allow determining the occurrence, levels, and locations of damage. A numerical example is presented to demonstrate the proposed damage detection method as well as to exhibit the damage detection performance. A series of experimental tests are carried out with this damage detection method implemented in various scenarios. The experimental verification shows that this proposed method is quite effective for seismic damage detection.