An adaptive dual-Kriging method based on parameter sensitivity analysis and application to vibration reduction optimization of helicopter rotor test-bed

The surrogate-based model optimization method has been successfully applied in various engineering optimization problems across multiple fields. However, when facing high-dimensional optimization problems with multi-design variables, the efficiency of surrogate-based model optimization is often hindered due to the proliferation of sample points and the computational complexity associated with high-dimensional matrices. In this work, an optimization method of adaptive dual-Kriging method based on parameter sensitivity analysis (PSAD-Kriging) is proposed to solve the inefficient problem of the surrogate-based model in high-dimensional optimization problem. In the PSAD-Kriging method, Kriging models with different accuracy are introduced to ensure the accuracy of optimization and improve optimization efficiency. The low-accuracy Kriging model is used for parameter sensitivity analysis to compute the sensitivity of design variables, which can reduce the dimensions of design space and improve optimization efficiency. The high-accuracy Kriging model is adopted to complete the adaptive filling point process to obtain the optimal solution of the optimization problem. The PSAD-Kriging method is applied to the vibration reduction optimization of the helicopter rotor test-bed, and compared with the other three traditional Kriging model method to verify the high efficiency and high accuracy of the PSAD-Kriging method proposed in this work. The results indicate that the PSAD-Kriging method improves the optimization efficiency by 30.02% on the premise of ensuring good prediction accuracy. Moreover, the maximum displacement response of the rotor test-bed is decreased by 31.32% after the vibration reduction optimization by PSAD-Kriging method. Therefore, the PSAD-Kriging method proposed in this work provides a novel solution for high-dimensional optimization problems with multi-design variables and can be effectively applied to engineering applications.