2019-01-012024-05-18https://scholars.lib.ntu.edu.tw/handle/123456789/700605摘要：結構健康監測利用量測結構反應之手段，通過結構動力特性與量測反應之關係，建立數值與實測之資料庫，應用相關之結構健康監測與損傷診斷之演算法，進而了解結構a)是否發生損傷、b)損傷之位置、c)損傷之程度及d)殘餘性能。該健康監測手段，如同民眾前往醫院健康檢查，藉由精密儀器進行生理數據量測，與過往病歷比較，加上專業醫師學理與經驗判斷，最後斷定該民眾目前生理狀況、決定病灶之位置與程度，並告知可恢復之方式與時間，或提供醫療處置方式。在健康檢查過程中，首要步驟為量測生理數據，該量測必須能夠正確及準確提供病人之各種生理反應，進而推估病人之狀況。結構健康監測也是如此，量測的方式必須準確與全面，且須利用多重物理量之量測，達到可將結構反應、結構行為與數值模型相比對，最終可以確定結構之病灶，並推估結構之殘餘性能。 本計畫將利用影像量測之方式，整合傳統加速規量測，進行結構動力特徵萃取，或結構桿件變異性之量測(如混凝土裂縫)，並藉由該動力特徵、桿件變異性與數值模型之比對，了解結構之殘餘性能與目前之耐震能力。就動力特徵萃取而言，將透過特殊幾何位置、多相機之配置，影像量測結構受微振、地震之反應，搭配結構加速度反應，藉由電腦視覺之方式與反應之時頻域分析，最後萃取出結構重要之動力特徵參數，如結構各模態之自然頻率、阻尼比及振形；就結構桿件變異性而言，主要著重於混凝土裂縫之識別與幾何判定，將透過相同之特殊幾何、多相機配置，與卷積式類神經網路模型，自動擷取混凝土裂縫於重要桿件中之位置，運用電腦視覺判定裂縫之尺度(如長度、寬度)，經由長期之觀察，了解該桿件劣化之程度，及早預警該桿件之損傷。<br> Abstract: Structural health monitoring (SHM) is a strategy which exploits sensing technologies to evaluate the relationship between dynamic characteristics and responses of structures, to build the numerical and measured database, and to assess health conditions of structures by sophisticated algorithms. Then, these algorithms provide the understanding of structures for a) the occurrence of damage, b) damage locations, c) damage severity, and d) remaining performance. SHM is quite similar with health exams for people. Doctors use professional devices to measure physical data from people, compare these data with previous ones in a health profile, evaluate health by knowledge-based or empirical judgements, and consequently determine health conditions, locations and severity of sickness, time and methods for recovery, and required medical operations. In the process of health exams, the most essential step is the professional devices to correctly and accurately measure physical reactions. Likewise, SHM relies on accurate and versatile sensing technologies to acquire multi-metric measurements of structures. These measurements can be finally interpreted by the SHM algorithms to ensure the damage locations as well as to estimate the remaining performance of structures. In this research project, an image-based structural health monitoring system, in conjunction with conventional sensing technology (i.e., accelerometers), will be developed to detect damage in an element-, member-, and system-level of structures. This image-based SHM system consists of multiple cameras with a specifically geometric configuration, allowing responses and deformations of structures to be measured under ambient vibrations and seismic excitation. For the element-level measurements, the development emphasizes the autonomous concrete crack detection that can determine number and geometric dimensions (e.g., lengths and widths) of cracks through the artificial intelligence, image processing, and computer vision techniques. This concrete crack detection requires a long-term implementation to observe the deterioration of concrete elements and to provide early warning of damage. For the member-level measurements, the static and dynamic deformations over a critical member in structures will be acquired to evaluate the degradation of this member. These deformations can indicate damage levels of critical members in a structure. For the system-level measurements, point displacements will be measured along with acceleration responses from conventional sensors, and these multi-metric data will be turned into dynamic characteristics (i.e., natural frequencies, damping ratios, and mode shapes) of the whole structural system. The variations of these dynamic characteristics will be the source to assess the integrity of structures.影像量測結構動力特徵萃取結構桿件變異性電腦視覺卷積式類神經網路模型自動擷取混凝土裂縫Optical measurementdynamic characteristics extractionelement-level deteriorationcomputer visionconvolution neural network modelautonomous concrete crack detection