橋梁結構系統識別方法研究

Abstract

本研究旨在探討協方差型隨機系統識別法在唯輸出條件以及基於確定-預測性模型之遞迴式子空間系統識別法在包含輸入-輸出訊號的條件下，應用於橋梁系統識別之可行性。本文以一座混凝土拱橋為對象，輸入擾動包括行車載重以及地震波－行車載重係以等速前進之移動荷載接續在設定時距上橋的條件模擬之，由於無法量測輸入擾動歷時，只能以協方差型隨機系統識別法進行分析；地震波則採用El Centro地震，因輸入擾動函數為可量測之地表加速度歷時，因此還能選擇在確定-預測性模型下之遞迴式子空間系統識別法進行分析。分析結果顯示，在行車載重下，採用協方差型隨機系統識別法可成功識別出若干主要模態之頻率、振形及阻尼比，無論橋梁的健康狀態如何；在地震擾動下，遞迴式子空間系統識別法可以成功識別出模態頻率、模態振形以及阻尼比，協方差型隨機系統識別法的識別精度則較差，反映在振形與阻尼比的識別結果上。依本文考慮的行車載重條件下以唯輸出條件進行識別分析結果，在垂直(Z) 向振動為主的模態識別結果與遞迴式子空間系統識別法在地震擾動下相當，橫(Y)向振動模態則不如地震擾動精確。此外，橋梁在局部受損的情況下，結構主要振頻及其對應之振形等模態參數均無顯著變化，難以用於判斷橋梁結構是否受損。

The objective of this study is to explore the feasibility of using covariance-based subspace system identification method under output-only conditions as well as the recursive stochastic subspace system identification method under a deterministic-stochastic model for bridges. In this thesis, a concrete arch bridge is considered as the object. Both traffic load and earthquake are considered as the input excitations. Traffic load is simulated as point loads moving consecutively at a specified time lag in a constant speed on the bridge. As the traffic load cannot be measured in reality, only the SSI-COV method was adopted for analysis. The bi-directional El Centro earthquake records were used as the seismic inputs. Since the ground accelerations are measurable in practical applications, the R4SID method was adopted in this study in addition to the SSI-COV method. Simulation results indicate that the frequencies, mode shapes and damping ratios of the bridge for a certain predominant modes can be identified using SSI-COV under traffic load, regardless of the health conditions of the bridge. Under seismic load, the frequencies, mode shapes and damping ratios of the bridge for a certain predominant modes can also be identified using both the R4SID and SSI-COV, with SSI-COV being less accurate in predicting the mode shapes and damping. Under the considered pattern of traffic load, the identified results of the structural modes primarily in vertical (Z) direction by SSI-COV are equivalent to those under earthquake load by R4SID. The mode shapes related to the transverse (Y) directions thus identified, however, are less accurate. Moreover, the frequencies and mode shapes of predominant modes of the bridge are found barely changed when locally damaged. Therefore, they are not sufficient as indicators for damage detection of bridge structures.