Research On Steel Structure Damage Detection Method Based On The Fusion Of BP Neural Network And Transmissibility Function

With the rapid development of urban modernization,the steel structure has been widely applied in the field of civil engineering construction,however,a series of questions caused by steel structure damage are also followed.When the structure is tested for damage,the transmissibility function is concerned because it is not affected by external forces.However,the existing transmissibility function method is not comprehensive and accurate for the damage detection of steel structure.In this paper,the damage detection method of steel structure based on the combination of the BP neural network model and the transmissibility function is studied.In this paper,the damage detection method based on neural network and transmissibility function is discussed in detail,and the feasibility of this method is verified from three aspects: theoretical derivation,simulation and experimental verification.The main work is as follows:Firstly,the frequency response characteristics of vibration acceleration are studied based on the differential equation of motion with multiple degrees of freedom,and the damage transmissibility function model is established.Based on this,the sensitivity of damage detection with the change of transmissibility as damage index and the recurrence law of symmetric tridiagonal matrix is studied.As a damage index,the transfer rate is not only simple and accurate,but also not affected by external excitation.Then,ANSYS software is used to model,load and solve the steel structure frame,and the node acceleration perpendicular to the detection surface is obtained.Then MATLAB software is used to Fourier transform the node acceleration,divide the frequency spectrum and subtract the transmissibility function to get the change of the transmissibility function.After that,a method of steel structure damage detection is studied,which combines the BP neural network model with the transmissibility function.The main idea of this method is to take the transmissibility function of the structure as the input parameter of the neuralnetwork.Based on the sensitivity analysis method of the transmissibility index,the feasibility of the algorithm is theoretically analyzed.Then,the change of the transmissibility function obtained by ANSYS and MATLAB is used As the input parameter of the BP neural network,the error is corrected by back propagation,and the output is 1 or 0 to indicate whether there is damage or not,so as to simulate the damage location of the steel structure.Finally,through the establishment of the portal frame experimental platform,three models of undamaged,mass change and stiffness change are tested to get the node acceleration perpendicular to the detection surface,and then the MATLAB software is used to Fourier transform,spectrum division and transmissibility function subtraction of each node acceleration to get the change of transmissibility function.The results show that compared with the traditional transmissibility function method,this method can identify the damage of each position of the steel structure,and the recognition rate has been greatly improved,which can effectively and accurately locate the damage of the steel structure.