CRACK DEPTH DETECTION OF MASSIVE STRUCTURES BASED ON DATA-DRIVEN ALGORITHM

Cracks are the main diseases of concrete structures. Finding out the depth of cracks can provide reliable information for the durability and safety evaluation of structures, but it is also one of the difficulties in the detection of concrete structures. A data-driven learning algorithm is proposed to predict the crack depth by investigating the signal of wave propagation through the cracked structure. The wave propagation process in the cracked massive concrete structure is simulated by using the extended finite element methods (XFEM) and the boundary absorbing layer model, and the observation signal of the receiving point is paired with the crack information. Based on the machine learning model of artificial neural network, a fracture depth prediction model based on XFEM datasets is established. For the concrete structure with unknown crack information, through the measured observation point signals, the established machine learning model is utilized to realize the real-time prediction of crack depth. The performance of the algorithm is verified by two numerical examples. The results show that the proposed algorithm can accurately predict the crack depth.