| The concrete structure may be impacted by fire and explosion due to accidents or terrorist attacks during the service life,Therefore,based on the finite element method,this dissertation utilizes ABAQUS/explicit finite element software to simulate the impact process of reinforced concrete floor under ambient and elevated temperature due to fire.In view of the fact that numerical simulations are time-and labor-consuming,the machine learning(ML)method is adopted in this dissertation.By selecting representative working conditions and determining eigenvalues to build data sets,the ML algorithm is trained to obtain a prediction model equivalent to the finite element simulation results,so as to quickly realize the dynamic response analysis of concrete floor caused by different drop conditions under different temperature conditions It provides a reference for the study of dynamic mechanical properties of concrete floor structure under impact.This dissertation mainly covers the following works:(1)To establish different temperature fields in ABAQUS to simulate the variation rules of various physical parameters of concrete floor under actual working conditions.The research shows that with the increase of temperature,the stress of the structure itself increases,which is due to the huge thermal stress caused by the temperature change;at the same time,considering the damage of concrete materials at high temperature,the stress of the structure increases with the increase of temperature The Young’s modulus of the material will decrease with the increase of the stress,so the stress value will increase.(2)Using the finite element software ABAQUS/explicit,the dynamic impact model of concrete floor under 12 working conditions is established.The impact position,impact height,boundary condition and impact initial velocity are taken as independent variables to establish the simulation working conditions,and the influence of impact on the mechanical parameter of concrete floor,such as stress(Mises),PEEQ(equivalent plastic strain)and displacement(U)is studied.(3)A machine learning method based on finite element calculation results is used to predict the relative mechanical properties of concrete floor slabs under fire and impact.The results show that the modeling process of simulation model is complex and the calculation process is long.The accuracy of the ML method used in this paper is as high as 0.876,so the trained ML model can effectively replace the finite element simulations.In addition,compared with the time cost of finite element simulations,the MLbased algorithm can be completed in a few seconds,which greatly improves the computational efficiency. |