【作者】 孙鸣； 黄斌；

【Author】 Sun Ming;Huang Bin;School of Civil Engineering & Architecture,Wuhan University of Technology;

【机构】 武汉理工大学；

【摘要】 通过基于贝叶斯方法的模型修正方法结合先验信息与样本信息对模型进行修正从而使模型能更好反映结构力学性能,为工程决策服务,是目前较为主流的一种方法。然而基于抽样的各类贝叶斯有限元模型修正方法涉及到多次模型的正向计算,计算耗时较大,这一问题在实际大型结构中更为明显。基于此,本文提出了同伦代理模型的大型结构贝叶斯随机模型修正方法,该方法利用用同伦代理模型替代FEM模型进行计算,同时在马尔科夫(Markov)链抽样中,采用延迟拒绝自适应抽样算法(DRAM),保证抽样效率,用受偏心荷载的单向板作为数值算例验证了该方法的高效与精确性。最后,利用FEM模型交互有限元模型信息提取技术,构筑了某实桥结构的同伦代理模型,基于桥梁静载试验数据,利用该方法对位于宜昌市的某城市桥梁进行了模型修正。计算结果与实测数据较为吻合,模型修正效果较好,表明该方法能较好的应用于实际工程中。更多还原

【Abstract】 The model updating based on Bayesian method combines prior information and the information from samples,and the updated model can reflect the mechanical properties of the structure and serve for engineering decision-making,which is a relatively mainstream method at present.However,Tens of thousands of model calculations are inevitable in all kinds of bayesian model updating based on sampling,which is time-consuming.And this problem is more serious in actual large structures.Based on this,A new model updating method is proposed by combining Bayesian and homotopy-based response Meta-model,which uses homotopy-based response Meta-model instead of FEM model to calculate,and uses delayed rejection adaptive sampling algorithm(DRAM)in Markov chain sampling to ensure sampling efficiency.The efficiency and accuracy of the method are verified by a numerical example of a plate subjected to eccentric load.Based on the static load test data of bridge girder,the homotopy-based response Meta-model of a real bridge structure in Yichang city was constructed using the interactive finite element model information extraction technology.The calculated results are in good agreement with the measured data,and the model modification effect is good,which indicates that the method can be applied to practical engineering.更多还原