<p>This paper focuses on optimizing the aerodynamic performance of a high-pressure turbine under multiple operation conditions. Utilizing a self-adaptive updated Kriging model, we employ the Latin hypercube sampling method and NSGA-II algorithm to find the optimum point with the maximum weighted average isentropic efficiency. The results show that compared with the original blade, the efficiency is increased by 2.35% for the optimized blade. Using the sensitivity analysis, it is indicated that the thicknesses near the leading edge and middle part in the mean section primarily influence the efficiency of turbine, in which the thickness near the leading edge is the most influential. It is also found that the proposed optimization method can greatly reduce the low-velocity regions caused by secondary flows, which thus significantly relieve the passage blockage. In addition, we notice a dramatical reduction of losses in the selected blade sections for the optimized blade. According to the entropy generation rate, the regions with high entropy generation, positioned near the pressure side, blade tip and corner regions in 50% axial blade section, and the tip regions in the section downstream the trailing edge, are also remarkably decreased.</p>