Research On Post-earthquake Safety Evaluation Method Of RC Frame Structure Based On Machine Learning

With the growing complexity of urban society and increasing population,structural failures caused by a strong earthquake may induce a large number of casualties and huge socioeconomic losses.Seismic performance evaluation and safety evaluation of large-scale reinforced concrete frame structure after an earthquake is still a cutting-edge problem that has not been fully solved.In response to this demand,this paper explores the use of artificial intelligence technologies such as machine learning to develop and improve post-earthquake safety assessment methods for reinforced concrete frame structures.Its core is to use machine learning to match the relationship between structural response and damage patterns that can characterize structural damage status,to establish guidelines for assessing structural post-earthquake safety,and to develop structural post-earthquake safety evaluation methods based on nonlinear time history analysis,on-site inspection and machine learning.The main research in this paper are as follows:（1）Used 16 actual ground motion records to perform amplitude modulation based on spectral acceleration（S_a（T₁,5%））to obtain ground motions with different intensity levels,an incremental dynamic analysis was performed on a 6-story RC frame structure established with the obtained ground motions of different intensity levels to obtain1044 sets of ground motion response samples.Then principal component analysis（PCA）was used to get the first principal component of base shear.The post-earthquake safety state criterion of RC frame structure was established based on the relationship between the first principal component of base shear and the maximum inter-story drift ratio.The earthquake damage database labeled with the safe state was randomly divided into the training set and the test set at random,which obtained 554 groups of safe state samples and 490 groups of dangerous state samples.（2）A post-earthquake safety state assessment method for RC frame structures based on support vector machine was proposed.The training set and the test set are randomly selected from the sample set for multiple tests based on k-fold cross validation.From the main parameters of the support vector machine algorithm and the attributes of the sample,the factors affecting the prediction performance of the support vector machine model were carried out in-depth research.The results show that the post-earthquake safety assessment method of RC frame structure based on the support vector machine can predict the post-earthquake safety status of the building structure very well,which had high accuracy in predicting the post-earthquake safety state and could be used to quickly assess the post-earthquake safety state of general RC frame structures.（3）A post-earthquake safety assessment method of RC frame structure based on random forest was proposed.The influence of the parameter of the number of decision trees in the random forest algorithm on the prediction performance of the model was studied and compared with the post-earthquake safety assessment method based on support vector machine.The results show that the overall performance of the post-earthquake safety assessment method based on random forest is better than the post-earthquake safety assessment method based on support vector machine.（4）Based on the spectral acceleration（S_a（T₁,5%））,23 actual ground motion records were amplitude-modulated,and elastoplastic time-history analysis was performed for three different RC frame structures using the obtained ground motions with different strength levels.The seismic response under the action of different intensity levels is obtained.Using machine learning algorithm to correlate structural information,earthquake motion information,and seismic response,and a seismic response prediction method based on decision tree of regression algorithm is proposed,which can predict the seismic response of the structure and the variance between the predicted and true values after inputting the given structural information and earthquake motion characteristic information.The results show that the seismic response prediction method has high accuracy.