چكيده به لاتين
Considering importance and role of the large and special bridges in the transportation system and infrastructure network of each country as essential structures, they require special attention and comprehensive maintenance programs using modern and efficient methods. According to the extensive studies that have been carried out in recent decades, due to the use of modern technology in computer science and data processing, electrical science and mathematics and their use in combination with structural engineering topics, significant progress in the field of structures and specifically bridges’ health monitoring have been achieved. Currently, many dynamic health monitoring methods including the ones based on modal parameters as well as signal-based methods have been developed and used, however there are still a lot of rooms for studies with the approach of optimization, accuracy and efficiency increasing, facilitating field and computational operations and cost reduction as well. In the present study, in order to develop the subject of damage detection in bridges with the approach of simplifying and reducing field and computational complexities, some new indices in two groups based on the quasi-static component of the displacement response signals derived from the loading and analysis results of a bridge numerical model have been presented. In this regard, upon applying and then reducing the effect of noise to the recorded signals in both healthy and damaged states, the methods of Empirical Mode Decomposition (EMD) and Wavelet Transform (WT) were used to process the signals to extract quasi-static components of them with purpose of using in provided indices. At first stage, the introduced indices, by using analysis results of a bridge’s simplified model in Csi-Bridge software were evaluated from several points of view, taking into account the parameters such as severity and position of damages, different types of loading and etc. After ensuring the efficiency and accuracy of the indices, they would be examined in the final stage. In the next stage, a 3D finite element model of an existing bridge (Karoon railway bridge located in Khuzestan province) has been implemented. The bridge with general specifications of total length of 294 meters including 5 spans of 41.75 + 3x70.00 + 42.25 meters with prestressed concrete box deck, 6.0 meters wide and 4.5 meters high has been used. Mentioned bridge’s three-dimensional model was analyzed in Abaqus software, under the same live loading conditions in two healthy and damaged (simulated damages) states and once more, after going through the steps which mentioned in the previous stage, including recording and processing displacement signals, selected indices were assessed. The final investigations also indicated that the provided indices have positive and acceptable results in damage detection and determining their position along the length of the bridge with appropriate accuracy. Finally, the machine learning method was also used to quantify and optimal application of some of the provided indices, drives to achieve acceptable results with high accuracy in damage detection and determining its location as well.
