Research On The Fragility Of Prestressed Prefabricated Frame Structures Based On Modal Analysis And Machine Learning

With the accelerating process of building industrialization,prestressed prefabricated concrete structures have been used more and more widely.Prestressed prefabricated concrete structure is a kind of building structure system with prestressed concrete components as the main body and manufactured by industrial assembly.As one of the prestressed prefabricated concrete structures,the unbonded prestressed prefabricated concrete frame structure has obvious advantages over the traditional wet bonding mode,such as fast construction,no wet operation,not affected by weather and so on.At the same time,it is better than the cast-in-place structure in ductility and deformation recovery ability.At present,the research on the connection mode,energy dissipation capacity and self-reset of the prestressed prefabricated frame joints has become mature.However,this structure was born in the late 20 th century,and its seismic performance has not been tested by actual earthquakes.Therefore,seismic fragility analysis and seismic risk assessment of this structure are helpful for its popularization and application.This thesis mainly takes the unbonded prestressed prefabricated frame as the research object,and uses Open SEES finite element software to analyze the structural fragility and assess the seismic risk of the structure.The specific work is as follows:(1)The flexural and shear coupling effect of reinforced concrete columns is studied,and the finite element model is established to compare and verify the pseudostatic test of two reinforced concrete columns.The model can simulate the stress situation of reinforced concrete columns well.Based on the stress analysis of unbonded prestressed prefabricated structural beam-column joints,the calculation formula of four limit state points on the skeleton curve is obtained,and the relationship model between the bending moment and the Angle of the beam end is established.Then,the model is verified by two node experiments.(2)The flexural-shear coupling effect of reinforced concrete column and the moment-Angle relation of beam end are introduced into the finite element model of frame structure.The shaking table test of cast-in-place reinforced concrete frame and the pseudo-static test of unbonded prestressed prefabricated frame are compared and verified.The results show that the model considering the flexural shear effect can better simulate the shear failure of the bottom frame column.The finite element model with the bending moment-angle relation of beam end can simulate the structural response of unbonded prestressed prefabricated frame more accurately,which verifies the accuracy of the model.(3)A damage index(DI)based on structural mode was established,which fully considered the changes of natural vibration frequency and mode before and after damage.The principle of mode recognition based on covariance driven random subspace method(COV-SSI)is explored.The method is used to identify the modal of steel frame and reinforced concrete frame to verify the accuracy of the method.(4)According to the principle of structural seismic fragility analysis,90 ground motion records that meet the requirements are selected from PEER database,and incremental dynamic analysis is carried out on the finite element models of two kinds of frame structures by using the equal-step method.IDA curves are drawn according to the analysis results,and the probability-seismic demand models of two kinds of frame structures under different damage indexes are fitted.According to this model,the seismic fragility curve is calculated.The seismic risk curves of two kinds of frame structures are calculated based on the seismic risk analysis principle.By integrating the seismic fragility curve and the seismic hazard curve,the annual exceedence probability of structural collapse is obtained.Based on this,the earthquake risk assessment of the two frame structures is carried out.The evaluation results show that the exceedence probability of the two structures in 50 years is much lower than the seismic design code of China.For different damage indexes,the annual exceedance probability and 50-year exceedance probability of prestressed prefabricated frame structure are slightly higher than those of cast-in-place frame structure.(5)Select 30 ground motions from 90 ground motions as the input time series of physical guided convolutional neural network(Phy CNN)model,and take the response data of two kinds of frame structures under the action of these 30 ground motions as the output time series.Train the input and output time series.The trained model is used to predict the time series of the remaining 60 structures under the action of ground motion.The predicted results are in good agreement with the actual results.The response time series predicted by Phy CNN were used to fit the probabilistic seismic demand models of two kinds of frame structures,and the structural Fragility curve,the annual failure probability and the 50-year failure probability of structural damage were calculated.These results are compared with the actual results.The verification results show that there is little difference between the predicted results and the actual results.The use of Phy CNN to predict structural response,seismic Fragility analysis and seismic risk assessment has a certain reliability.There are 86 figures,17 tables and 100 references in this thesis.