基于机器学习的模块间节点受剪性能多目标优化OA

Modular steel structure buildings have become an important development trend for green and low-carbon construction due to their advantages of a high degree of industrialization and high construction efficiency.The Inter-module connections have significant effects on the mechanical performance of modular buildings.However,the shear performance of the inter-module connections has not been fully investigated in existing studies.Based on the previous experimental study on fully prefabricated liftable connection(FPLC)of modular steel structures,this paper established a refined finite element model and conducted parametric analysis to obtain a shear performance database of FPLC with 1000 different parameters.Six mainstream machine learning algorithms were utilized to predict the shear performance of the FPLC.The results indicated that the neural network optimized by genetic algorithm(GANN)provides better prediction accuracy for the shear load bearing capacity,and the support vector machine stacking algorithm(SVR)shows higher prediction accuracy for the ultimate displacement.By stacking the two algorithms with higher prediction accuracy as a proxy model and linking this model with the non-dominated sorting genetic algorithm II(NSGA-Ⅱ),a multi-objective optimization method for the shear performance of the FPLC was established,and the optimized joint configuration was proposed.The finite element model of a four-story modular steel structure was established,and the static analysis was carried out under vertical load and wind load.The shear performance and inter-story drift rations of the four-story modular steel structure before and after optimization were compared to verify the reliability of the optimization method.