基于贝叶斯法的复杂有限元模型修正研究

从概率思维角度出发，证明基于最大熵原理的贝叶斯反分析准则函数法和解不适定问题的正则化方法是一致的，提出一种基于信息融合和贝叶斯理论的模型修正方法。该方法采用试验设计构造样本，采用二次响应面作为快速运行模型，通过响应面自身的特性和精度要求进行收敛判断，在响应面迭代中确定信息融合系数（设计规范、有限元计算信息、实测信息）和待修正的设计参数值。该方法充分利用先验信息，迭代计算量较小，可推广至大型复杂非线性结构。某抽水蓄能电站地下厂房结构的有限元模型修正结果验证了该方法的有效性

Complicated finite element model updating based on Bayesian method

It was proved from the probabilistic perspective that Bayes criterion function method based on maximum entropy principle is in accordance with the regularization method solving ill-posed problems.A new method based on information fusion and Bayes theory for model modification was established.Samples were designed with tests and the second-order response surface was adopted as the fast operation model easily modified and perfected by fully use of prior information(design code information,finite element calculation information and test information).This method could be extended to large and complicated nonlinear structures for its good computational efficiency and solution convergence.The underground powerhouse of a pumped-storage power station was employed to verify the effectiveness of the proposed method.