Abstract: An improved trust region hybrid simulation model updating method is proposed to address the high computational cost and slow convergence in multi-parameter model optimization. This method integrates the trust region algorithm with classification stage sensitivity analysis and uniform design. The trust region is established with the core parameters as input variables, while the relative cumulative restoring force error between the checking and experimental substructures serves as the objective function. The trust region algorithm is employed to determine the optimization center and to calculate the trust region radius. The uniform design is applied to generate the sample space and to perform the search of the current optimal parameter set. Taking the steel frame as an example, four scenarios are set up and ten random conditions are generated for numerical verification. Simulation results indicate that the improved trust region hybrid simulation model updating method reduces errors and keeps them within 1.5%, demonstrating high accuracy. The computational effort for identification is reduced, with calculation time cut by more than a half and a significant reduction in the number of convergence steps. The feasibility and effectiveness of the method proposed in improving the recognition accuracy and speed of hybrid simulation model updating are verified.