Research On Structural Damage Based On Convolutional Neural Network And Recurrent Neural Network

With the use of time and the effect of the external environment,the structure of building engineering will definitely undergo corresponding aging,with different degrees and different forms of structural damage.The identification,detection,repair and maintenance of structural damage play an important role in structural safety.At the same time,the earlier the damage location and extent of the structure is found,the lower the cost of maintenance.Therefore,it is of great significance to identify the damage location and damage degree of engineering structures.Convolutional neural network and recurrent neural network are important research achievements in the field of artificial intelligence in recent years.This method has its own special features compared with traditional pattern recognition methods.Through the detection of data and the use of this method,it can learn and train itself,so as to obtain the relevant damage of the structure.Using convolutional neural network and long and short term memory network(a kind of recurrent neural network),based on data analysis,the scientific laws behind these data are searched,and the damage and damage degree are found,which is what makes this paper special.However,researches on the application of convolutional neural network,especially recurrent neural network,in the field of structural damage identification are limited.In this paper,relevant researches are carried out on the application of these two kinds of neural networks in structural damage identification.The main work includes the following contents:(1)The research on structural damage identification in the field of artificial intelligence is summarized,and several deep learning methods in structural damage identification are introduced.(2)The simple supported beam is used to simulate the bridge structure.Under the action of uniform moving concentrated load,the vertical displacement at the measuring point is extracted through the simulation analysis of ANSYS,and the network framework is built based on MATLAB.Two kinds of convolution neural network and long-term memory network are used to identify the damage of the simply supported beam.By comparing the damage identification effect of the three kinds of neural network,the preliminary exploration of which neural network is made It is more suitable for the damage identification of this type of structure and the main factors affecting the accuracy of damage identification.(3)Taking a typical frame structure as an example,under the excitation of El Centro seismic wave including dead load,the acceleration of column joints is extracted through ANSYS simulation analysis,and the extracted data are preprocessed.The feasibility of longterm and short-term memory network technology in damage identification of this kind of structure is preliminarily discussed.