| 吕小龙,黄丹,姜冬菊.基于POD-RBF代理模型的迭代更新反演方法[J].计算力学学报,2022,39(4):506~511 |
|
| 
码上扫一扫！ |
| 基于POD-RBF代理模型的迭代更新反演方法 |
| Iterative updating inversion method based on POD-RBF surrogate model |
| 投稿时间：2020-12-01 修订日期：2021-04-09 |
| DOI：10.7511/jslx20201201001 |
| 中文关键词: 代理模型 POD-RBF 参数反演 粒子群算法 高斯-牛顿法 |
| 英文关键词:surrogate model POD-RBF parameter inversion PSO Gauss-Newton method |
| 基金项目:国家重点研发计划(2018YFC0406703);国家自然科学基金(12072104;51679077;11932006)资助项目. |
|
| 摘要点击次数: 393 |
| 全文下载次数: 204 |
| 中文摘要: |
| 针对常规的水工大坝等大型工程结构参数反演需要耗费大量有限元正分析机时的问题,建立了具有较好反演精度和泛化性能的POD-RBF代理模型和快速迭代更新反演算法。基于有限元分析获得足量数据样本,利用POD提取本征向量,并使用RBF方法进行插值得到有限元模型的代理模型;同时结合粒子群算法的全局寻优能力和高斯-牛顿法的快速局部收敛优势,建立了一种新的高效率迭代反演方法,并应用于混凝土大坝分区弹性模量反演。结果表明,该方法适用于大坝等大体积混凝土结构的力学参数反演。同时,相较于传统的单一反演方法,该方法在反演效率和反演精度两方面均显示出优势。 |
| 英文摘要: |
| To address the high computation workload in finite element analysis in the general inverse analysis of large engineering structures like hydraulic dams,a precise and robust POD-RBF surrogate model and an iterative updating inversion algorithm were constructed.Based on the data samples obtained from the finite element method,we can interpolate eigenvectors abstracted by POD method using RBF to get a surrogate model of the finite element model.Combining the global optimization ability of particle swarm optimization algorithm and the fast local convergence advantage of Gauss-Newton method,a new efficient iterative inversion method was established and applied to the zonal elastic modulus inversion of a concrete dam.The results shown the applicability of this method for mechanical parameter inversion of large concrete structures such as dams.Meanwhile,compared with traditional methods,this inverse framework consisting of the POD-RBF surrogate model and the iterative updating method can get good performance both on the inversion accuracy and the efficiency of computation. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
