中文摘要：

根据环境振动测试的模态数据,采用复形法对广州国际体育演艺中心屋盖结构进行有限元模型修正研究,并与ANSYS一阶算法、改进BP神经网络算法和粒子群算法（PSO）修正有限元模型进行对比分析。研究结果表明：修正后的有限元模型的模态频率与实测的模态数据吻合较好,复形法可以作为大跨屋盖结构有限元模型修正的一种有效方法。

英文摘要：

According to the identified model parameters from full-scale measurements of structural vibrations,the complex method was employed to update the finite element model（FEM） of the roof of Guangzhou International Sports Arena.For the purpose of comparison,the results of some other conventional algorithms including ANSYS first-order algorithm,improved BP neural network algorithm and PSO algorithm were also presented.The results show that the modal frequencies of the updated finite element model（FEM） agree well with that from full-scale measurements,and the complex method can be used as an efficient way of FEM updating for large span roof structures.