以直接反應量測為主之結構損傷偵測評估

Abstract

利用結構物的振動資料進行損壞偵測(Vibration-based Damage Detection, VBDD)為損壞偵測其中一項重要的課題，不僅能偵測出損壞的發生且能偵測出損壞的位置並量化損壞的程度。本研究應用唯輸出(Output-only)的量測資料進行分析，並提出一套系統性的損壞評估架構，包含識別損壞位置與損壞程度的量化。本研究提出四種程度的損壞識別方法，首先，為了識別是否有損傷發生，使用(1)主子空間(Subspace)與零子空間(Null-space)損壞因子，(2)奇異譜分析法(Singular Spectrum Analysis, SSA)之奇異值差異，(3)互相關函數之振幅向量確信準則(Cross Correlation Function Amplitude Vector Assurance Criterion, CVAC)、(4)功率譜密度函數之振幅向量確信準則(Power Spectral Density Function Amplitude Vector Assurance Criterion, PSDAC)與(5)兩階段式AR-ARX模型偵測損壞。接著使用隨機子空間識別法(Stochastic Subspace Identification, SSI)觀察具有物理意義的系統參數的變化。為了找出損壞位置，使用(1)奇異譜分析法之重建訊號與原訊號之差異與(2)由小波封包轉換(Wavelet Packet Transform, WPT)之Novelty Index。最後為了量化損壞程度，(1)依據識別的模態參數配合Model Updating的技術與(2)使用正規化之勁度矩陣計算樓層間的勁度折減率(Stiffness Reduction Ratio)。除了使用從水工試驗廠進行的水工沖刷試驗與於國家地震中心進行的六層樓鋼構架切割鋼柱的振動台試驗記錄到的實驗資料外，於宜蘭牛鬥橋進行現地量測以驗證所提出的方法之適用性。最後討論使用方法之計算效率並探討進行即時損壞偵測的可行性。

One of the important issues to conduct the damage detection of a structure using vibration-based damage detection (VBDD) is not only to detect the damage but also to locate and quantify the damage. In this paper a systematic way of damage assessment, including identification of damage location and damage quantification, is proposed by using output-only measurement. Four level of damage identification algorithms are proposed. First, to identify the damage occurrence, (1) Subspace and Null-space damage index. (2) from Singular Spectrum Analysis (SSA) compute the eigenvalue difference ratio. (3) Cross Correlation Function Amplitude Vector Assurance Criterion (CVAC). (4) Power Spectral Density Function Amplitude Vector Assurance Criterion (PSDAC). (5) Two Stage AR-ARX are discussed for detecting the damage. Second, use Stochastic Subspace Identification (SSI) to detect the change of dynamic characteristics. Thirdly, to locate the damage, (1) from SSA we can compute the difference between original signal and reconstruct signal. (2) Novelty Index, defined as the Euclidean norm of the time-frequency Hilbert amplitude spectrum of measurement between the intact and the damaged structure, is applied to locate the damage. Finally, to quantify the damage (1) combine modal parameters with the model updating technique. (2) from normalized stiffness matrix identify the inter-story stiffness reduction ratio. Experimental data collected except from the bridge foundation scouring in hydraulic lab and a series of shaking table test of a 6-story steel structure with the cut in column member, in situ structure like Niu-Dou Bridge are used to demonstrate the applicability of the proposed methods. The compotation efficiency of each method is also discussed so as to accommodate the online damage detection.