Diagnosis of Reinforced Concrete Structural Damage Base on Displacement Time History using the Back-Propagation Neural Network Technique

Abstract

A state-of-the-art methodology is proposed for damage diagnosis of structures, such methodology being presented in the example of a simply supported reinforced concrete (RC) beam. The severity and location of defects within the RC structures can be assessed much more conveniently by using the back-propagation neural network technique. A simply supported RC beam with specified size (i.e., rectangular cross section and 4 m span) and assumed defects is theoretically analyzed by a finite-element program to generate training and the testing of numerical examples necessary to assess the damaged RC structure by using the neural network (NN). Numerical examples are then generated according to the displacement time history of the defected beams loaded by an impact force at the beam center. In addition, 10 sets of test beam with the assumed damage and same specified size of the numerical examples are constructed in full scale. The damage scenario of each test beam is also diagnosed by using the well-trained NN according to the displacement time history, which is the history of the responses caused by the impact loading acting at the beam centers. Based on the study and test results, the damage scenarios of the 10 sets of test beams are successfully classified.