| Structural health monitoring system can estimate the safety status of the structures by analyzing the characteristic information collected by system,which is often used in the fields of construction structures and mechanical engineering,and the structure damage identification system is the core part of the structure health monitoring system.Using the artificial neural network to realize structure damage identification,the information collected by sensors can be processed and analyzed quickly and efficiently.With the development of related technologies,it is expected to realize the purpose of real-time and online monitoring of the structure conditions.Combined with information fusion technology,the structure damage identification system based on artificial neural network can extract effective characters from a large number of complex monitoring data in the noise environment,and get accurate results of structure state judgment.Based on BP neural network and information fusion technology,taking a cantilever reinforced concrete beam as the engineering background,the damage identification technology of the structure was studied.The main research of this thesis are as follows:(1)Based on the basic theory of artificial neural network,given the compositions of neural network and the expressions of network inputs and desired outputs,and determined the network model building process of BP neural network,deduced the formulas of network weights by gradient descent method of BP neural network based on the basic principle and formula of neural network.Also explained the classification methods of information fusion from the structure and the level of fusion information,then expounded the common information fusion algorithms and the selection basis of damage sensitive features for artificial neural network.(2)Then ABAQUS finite element analysis software was used to analyze the reinforced concrete beam,and a non-destructive damage beam and six groups of beam models of different damage locations and depths were established.Extracted some characteristic parameters of the structure and contrasted the results of modal analysis,plotted the frequency rate-of-change,the displacement mode rate-of-change and the MAC curve of the beams,studied the change rules of different parameters when the structure got a damage.The results of modal analysis showed that the mode of vibration,displacement mode and frequency change in varying degrees when the damage occurred in structure,and for different mode orders,the rates of change are quite different.(3)Based on the MATLAB software,the BP neural network model were built,using the processed feature information under different working conditions,the structural damage identification and location technology based on single type feature and information fusion technology were studied.The results showed that the accuracy of BP neural network for structural damage identification was high,and compared with only considering a single feature,the application of information fusion technology improves the accuracy of network for damage identification.When different characteristic parameters were used for damage location,only when displacement mode as network input,the neural network can accurately located the damage.The research content of this thesis can provide reference for engineering structural damage identification. |
