DIAGNOSIS OF DAMAGE IN RC STRUCTURE BASED ON STRUCTURAL RESPONSES VIA THE AI TECHNIQUE

• Main Authors: Chung-Huei Tsai, 蔡中暉
• Other Authors: Deh-Shiu Hsu
• Format: Others
• Language: en_US
• Published: 2000
• Online Access: http://ndltd.ncl.edu.tw/handle/87587896286413367945

博士 === 國立成功大學 === 土木工程學系 === 88 === This dissertation develops a feasible diagnostic model for reinforced concrete (RC) structures through the Artificial Intelligence (AI) technique, based on structural responses, to assess the severity and location of defects. Four kinds of structural response, i.e., acceleration time history (ATH), displacement time history (DTH), natural frequencies (NF), and static displacement (SD), are separately serve as the input characteristics of the neural network (NN) in the diagnostic model. A simply supported RC beam with a specified size and assumed defects is theoretically analyzed by a finite element program to produce the structural responses. The structural responses are then combined with relative damage conditions to generate training and testing numerical examples, necessary to assess the damage to the RC structure by using the NN. Two stages of diagnostic procedure are then used for the NN application to identify the damage scenarios of the relevant structures. Furthermore, several structural responses, as mentioned above, are measured from tests that also try to demonstrate the ANN base diagnostic model as presented herein, and whether it can be successfully applied to real structures. A test sample of RC beams with various extents of artificial damage is constructed and tested to diagnose the magnitude and location of damage by using well trained NNs. Finally, this study attempts to perform an objective and synthetic conclusion for various damage diagnostic results which subject to a certain location of each test RC beam from the NNs. Fuzzy logic is then applied to reduce differences between situations, especially when seemingly conflicting damage levels exist. Moreover, fuzzy logic can linguistically state the final diagnostic results that actually express or reflect the real state of the test RC beam. Therefore, this study successfully fabricates a feasible and efficient diagnostic model, which will be needed for real world damage assessment applications.