Research on Damage Identification of Frame Structure Based on Deep Learning
Convolutional neural network model and long-term and short-term memory network model were two widely used deep learning network models.In order to explore the application effect of the two models in structural damage identification,two network models were used to study the damage identification of steel frame structures.Three-story frame structure was taken as an example,the modal strain energy difference after reducing the dynamic characteristics of the unit was selected as the damage index,and inputted into the two neural network models respectively.The damage degree identification and damage location identification of beam-column elements were analyzed.The results show that the two network models can quickly grasp the dynamic characteristics of the structural unit.After learning the modal characteristics of the frame structure,they can accurately identify the location of the damage unit,and can accurately predict the damage degree of the unit.It verified that the two network models have good applicability in the damage identification with modal strain energy difference as the index.By comparing the performance of the two network models,it was found that the convolutional neural network has higher training efficiency and better generalization performance.
NETL
NSTL
万方数据
