基于全局敏感性的模型更新混合模拟方法

To address the uncertainty of numerical structural parameters in hybrid simulation, a model updating method for hybrid simulation based on Sobol global sensitivity analysis was proposed, and the test substructure was used as the comparison target for parameter updating. The common steel frame and reinforced concrete frame models were established, their independent constitutive parameters were derived as update objects, and the variation of global sensitivity coefficient with each parameter during the whole process of earthquake was obtained. Then the parameters which affect results significantly were selected as key parameters. Two update strategies, key parameters updating and all parameters updating, and parameters' search step size were determined. The hybrid simulations were conducted using nine groups of different initial parameter sets. The results of the study show that the parameter sensitivity changes continuously throughout the earthquake. Based on this, the multi-parameters are updated in different stages with unequal search scope, which effectively reduces the impact of parameters interaction on the model updating and balances the influence of each parameter change on the system. At the same time, the uniform design ensures that the parameter updating points are representative. Compared with the hybrid simulation without updating, this method effectively reduces the relative cumulative error of the two types of structures to within 4%. As for the results of multi-parameter identification, the relative error of the steel frame is within 1.5%, and the reinforced concrete frame is basically within 10%. This method is feasible and effective in improving the accuracy of hybrid simulation and multi-parameter identification.