面向桥梁数字孪生模型更新的多策略集成学习代理优化方法OA

To improve the efficiency and accuracy of high-frequency updating of bridge digital twins based on finite element model updating,a Kriging-based clustered parallel optimization-driven surrogate updating method was proposed.First,by embedding multiple active learning functions,a clustered parallel surrogate optimization algorithm was designed,and its optimization efficiency and accuracy were analyzed using com-plex test functions.Then,an efficient updating framework for bridge digital twins was constructed.This method was applied to a metro viaduct bridge,so as to reveal its advantages and practicality in representing the variation of bridge performance with the parameters.The results show that in complex function examples,compared with the traditional optimization algorithms,the proposed algorithm can significantly reduce the computational cost of numerical simulations while maintaining high optimization accuracy.In the bridge model updating based on measured frequencies,the convergence accuracy of the proposed method is improved by one order of magnitude compared with that of the standard Kriging surrogate-based updating method,and the com-putational time is reduced by about 58%compared with that of the particle swarm optimization-based updating method.The updated twin model can accurately predict the true dynamic responses of in-service bridges.