Research On Pile-Cap Damage Localization Using Probabilistic Neural Network

As we know, building is an important component of people's daily life, and any building must be constructed in the ground below a certain depth of soil or rock layer. As a result, building transmits its load to the ground through this foundation. Along with the rapid development of the factory production and mechanized construction level, pile group thick slab foundation is widely used in all types of modern buildings. When the building is completed, the building's foundation failure will lead to building expire and result in heavy economic losses. Therefore, the damage identification for building's foundation and detecting the foundation's health status are of great significance. For these, the research work of this paper is based on the following aspects.Based on the domestic and international related works of identification of the structural damage and artificial neural network (ANN) data analysis, we have summarized three methods for damage identification: Static Diagnostic Method, Dynamic Diagnostic Method, and Neural Network Intelligent Diagnosis Method. As noise is inevitable in civil engineering, we have proposed using the Probabilistic Neural Networks to localize the damage of pile cap, and systematically elaborated its principles, the formulas and its applications.In Southeast University, they have done a lot of things based on the test and 3-dimensional structural nonlinear analysis by Finite Element Method. With further, this paper established the Finite Element model of six pile cap and pointed out that the mode of load transfer is space truss model.And then we have established the Finite Element damage modal of six pile cap and simulated many kinds of damages, identified the damages using the damage signal index which was polluted by noise, and trained BP network with the same parameter. The results of the two networks are proving that the probability neural network has a better performance of recognition and the extrapolation without much training time. The result showed that PNN has a better classified ability. It is feasible that using PNN to identify damage of pile cap which is based on the space truss load transfer mechanism.