Prediction And Optimization Of Impact Resistance Of Steel Plate-Concrete Composite Structures Based On Machine Learning

Steel plate-concrete composite structures(SC structures)are mainly composed of external steel plates,concrete and internal connectors,and can be divided into Half-SC structures and Full-SC structures according to the arrangement of steel plates and connectors.With the gradual research on SC structures by scholars from various countries,many countries have formulated relevant codes for static performance and seismic performance of SC structures,but the experimental and theoretical studies on the impact resistance of SC structures are still relatively few.Therefore,how to efficiently design the impact resistance of SC structures has a very important engineering application value.The main contents and achievements of this paper are as follows:(1)The sensitivity analysis of Half-SC and Full-SC structures under low-velocity impact was carried out using the Elastic net method,respectively.The results show that the parameters that have the greatest influence on the impact resistance of Half-SC structures are concrete thickness and distributed steel diameter,and the parameters that have the greatest influence on the impact resistance of Full-SC structures are concrete thickness and steel plate thickness.The above parameters should be given priority in the structural design.(2)In order to reduce the computational cost of SC structures dynamic response under impact,three machine learning models of BP neural network,support vector regression(SVR)and Gaussian process regression(GPR)are established to predict the maximum deformation of SC structures under impact,and the hyperparameters of the machine learning models are determined by using cross-validation and Bayesian optimization.The accuracy and computational efficiency of the three machine learning algorithms are compared on this basis,and the results show that all three machine learning models can predict the maximum deformation of the SC structures well,and the time to calculate the maximum deformation can be significantly reduced compared with the traditional methods,which can replace the traditional calculation methods and effectively improve the computational efficiency of the target equation.(3)A Half-SC roof plate in a mountainous area impacted by falling rocks and a FullSC wall in a CA20 module of an AP1000 nuclear power plant impacted by tornado wracking materials are used as examples for impact resistance optimization design.First of all,different optimization variables,optimization objectives and optimization scope are set according to different cases,and a genetic algorithm(GA)is used to optimize the target case.The final optimization results give a variety of optimization schemes,which can greatly reduce the deformation of SC structures under impact,which provides a reference for engineering design.