Abstract: In view of the complexity and the time-consuming problem of the double-loop method of the reliability-based design optimization(RBDO) for crane metal structures(CMSs) based on static surrogate model, a reliability decoupling design optimization method for CMSs based on dynamic Kriging agent model with multi-point addition strategy is proposed. In this work, the improved mean value method of the performance measurement method is adopted for inverse reliability analyses, and the reliability decoupling is performed by calculating the offset vector, which render the complex RBDO problem transform into a deterministic optimization one. At the same time, during the optimization iteration process, the multi-point addition strategy of U learning function and optimal point addition and the multi-point addition strategy of constraint boundary and optimal point addition are constructed to extract more effective high-fidelity sample points to dynamically update the Kriging agent model, so that the optimization is built on the continuously updated Kriging agent model, and the optimization ends while the model accuracy meets the requirements. Finally, verification by example shows that the proposed method simplifies the procedure of RBDO, improves the optimization efficiency and provides a new effective method for the RBDO for CMSs based on the agent models.

Key words: crane metal structure, offset vector, reliability decoupling, dynamic Kriging agent model, multi-point adding strategy