Intelligent Method For Crack Diagnosis Of Reinforced Concrete Beams

The structural crack diagnosis has been an important research subject in the field of civil engineering. The cracks exist widely in concrete structures will decrease the strength and stiffness of the structures greatly or even cause serious damage when the width and depth of the cracks reach some extent. Therefore, to avoid the further development of cracks and subsequent damage, the location and depth of cracks must be diagnosed in real-time and some necessary protecting measures should be adopted.In this dissertation, the statistical property of structural dynamic response is employed as the damage index. The artificial neural network (ANN) method based on the damage index is proposed and implemented in the crack diagnosis of simple supported RC beam. The ANN method application is also extended to more fields. The primary work and innovations are as follows: (1) Basing on the sensitivity analysis of the variance and variance variation of the structural dynamic response to the change of crack's width and depth, the variance variation is selected as a damage index for the ANN input. (2) The three-dimension modeling and element killing method in ANSYS software are applied to simulate the crack location and depth, which is approximate to the real structure damage. The crack diagnosing numerical simulation of the simply supported RC beam is done, considering three damage cases: single crack, double cracks and tri-cracks. Each case is also diagnosed from the crack location and depth.Some conclusions are given as: (1) The well-trained ANN can identify the location and depth of the crack in the simply supported RC beam with good accuracy under single and double crack damage. (2) In the multi-crack damage case, for the location identification, the ANN can only identify one crack accurately and have a bad validity in the identification of the other two cracks; for the depth identification, opposite results are acquired: the two crack depth are identify accurately but another one can not be determined effectively. Therefore, it can be seen that the ANN is more effective in the identification of crack location than that of crack depth. (3) A modified method that increasing the dimension of ANN input vectors through adding the gauging point of dynamic displacement response is used in multi-crack damage diagnosis. The identifying efficiency is improved significantly. Thus, the efficiency and applicability of the ANN structural damage identification method based on the statistical property of structural dynamic response in crak diagnosis is verified.