Damage Diagnosis Of Sheet Pile Wharf Based On PBF Neural Networks

In the past 60 years, port construction in China has obtained the considerable development, and it has made the great contribution for the national economy of our country. But a large part of port buildings which have been built have been put into operation for a long time. Damage at various degrees may be produced in them. Damage diagnosis and repair of port buildings are the most important premise to ensure the safe use of structures.Lots of literatures about the structure damage diagnosis and neural network were consulted, and the development history and current situation about the damage diagnosis have been introduced in this paper. According to the prospect of development of the structure damage diagnosis and neural network, the RBF neural network was used in Wharf Structure damage diagnosis. Then the sheet-pile wharf was analyzed to verify the feasibility of such method.A finite element model of the sheet-pile wharf was built in ABAQUS. Only considering the damage in the sheet-pile wall and anchor tie, when the damage was produced in single element or two elements, the modal shapes and frequencies are obtained. Taking the modal shapes and frequencies as input date, four RBF neural networks, Net-R1A, Net-R1B, Net-R2A and Net-R2B, were built for damage diagnosis of the sheet-pile wharf. The analysis results show that the RBF neural network can be used in the damage diagnosis of sheet-pile wharfs.There are various errors such as noise and measurement error in the process of the structure damage diagnosis, and they cannot be avoided. Order to simulate the effect of these errors on damage identification results, the white Gaussian noise was employed in the modal shapes and frequencies, and damage identification results at various noise levels were analyzed. Then the proposal about how to simulate the model error was given.