基于全局敏感性的自适应模型更新混合模拟方法

The model updating for hybrid simulation method based on global sensitivity (GS) analysis can effectively improve the accuracy of hybrid simulation with high computing efficiency. However, this always makes it fall into the local optimal value. To address this issue, momentum and adaptive step size was introduced, and an adaptive (GS-A) model updating hybrid simulation method was proposed based on global sensitivity analysis. The steel frame model was used as the analysis object, and the eight parameters including both the constitutive and model geometric ones were identified. The uniform design test was used to establish 11 model update cases and numerical simulations were performed for the hybrid simulation with no updating, the GS method, and the GS-A method, respectively. The parameter identification process, displacement time course and hysteresis curves of the 33 sets of numerical simulation cases were analyzed and compared. Results show that the GS-A method, compared with the GS method, solves the problem of the parameters falling into the local optimal value, performs higher accuracy and stability of parameter identification, and improves the model updating effect.