【摘要】 由于影响冷弯薄壁卷边槽钢构件弹塑性屈曲荷载的因素较多，故截面形式与承载力之间无法用精确的解析式表达。为开展截面形式优化，提高槽钢的屈曲承载力，将基因表达式编程(GEP)算法和粒子群(PSO)算法结合开展截面形式优化研究。利用Python编程对ABAQUS进行二次开发，对文献中的15个试验构件进行批量有限元计算，通过与试验值对比验证了数值计算精度；选取其中1组构件为样本开展抗屈曲截面优化，采用批量有限元计算生成包含不同截面尺寸和对应屈曲荷载的样本数据共97组；采用基因表达式编程算法对数据集样本进行数据拟合，构建截面优化目标函数的代理模型；利用粒子群优化算法得到冷弯薄壁卷边槽钢弹塑性屈曲最大承载力对应的最优截面尺寸。与优化前相比，截面优化后构件的屈曲承载力提升了30.4%。研究表明，将有限元、机器学习和传统优化方法相结合，能够有效开展薄壁槽钢的截面形式优化。更多还原

【Abstract】 There are many factors that affect the elastic-plastic buckling capacity of cold-formed thin-wall steel lipped channel members. However, the relationship between section form and bearing capacity cannot be accurately expressed analytically. In order to optimize the section form and improve the buckling capacity of channel steel, the gene expression programming(GEP) algorithm and the particle swarm optimization(PSO) algorithm were combined to optimize section form. Through calling ABAQUS by Python programming, the finite element calculation of 15 test components in the literature was carried out in batches. The accuracy of the numerical solution was verified by comparing with the experimental values. A group of components were selected as samples to carry out anti-buckling section optimization. A total of 97 ensembles of sample data including different section sizes and corresponding buckling loads were generated by the batch finite element calculation. The gene expression programming algorithm was used to fit the dataset, and the surrogate model of the objective function of cross-section optimization was constructed. The particle swarm optimization algorithm was used to obtain the optimal section size corresponding to the maximum elastoplastic buckling capacity of cold-formed thin-walled lipped channel steel. Compared with the original specimen, the buckling capacity is increased by 30.4% after section optimization. The results show that the combination of finite element, machine learning and traditional optimization methods can be adopted to effectively optimize the section form of thin-walled channel column.更多还原