基于Kriging模型的钢管混凝土连续梁拱桥有限元模型修正

CFST arch/continuous beam bridge FEM model updating based on Kriging model

提出基于Kriging模型的有限元模型修正方法。Kriging模型为据区域内若干信息样品某种特征数据对该区域同类特征未知数作线性无偏、最小方差估计方法,其只用少量样本即可获得较高精度预测结果。用Kriging模型对平面桁架进行有限元模型修正,验证该方法的可行性与准确性;对一连续梁拱桥进行模型修正,并与GA算法、BP神经网络方法模型修正结果比较分析。Kriging模型仅需一定量测量频率信息即可完成模型修正,能避免修正过程中进行有限元模型迭代计算。结果表明,该方法能准确预测有效频率范围(active frequency range)外模态信息,计算效率、精度较高,可用于工程实践。

A new method for FEM updating based on Kriging model was developed.The Kriging model is a linear unbiased minimum variance estimation to the unknown data in a region according to some characteristic information of region’s samples which have similar features with the unknown data.This method can obtain higher accuracy predicted results based on a small number of samples.Through a planar truss FEMupdating example,the feasibility and accuracy of the Kriging model were verified.And then the Kriging model was applied to a concrete-filled-steel-tubular （CFST）arch/continuous beam bridge FEM updating and the results were compared with those by the method of genetic algorithm （GA） combined with BP neural network.The analysis results show that the Kriging model needs only a certain amount of measured frequency data for FEM updating.There is no iterative calculation like in the FEM,which will exhaust much calculation time in updating program.This method can accurately predict the modal information outside the active frequency range.The results testify the high computational efficiency,accuracy and feasibility of the method applied in actual engineering.