In this paper, the two methods of CO and BLISS 2000 are compared by using a classical numerical example firstly. The results show that the BLISS 2000 performs better in both computational accuracy and efficiency, because the scale of the optimization problem, such as the number of variables and constraints, in the BLISS 2000 is less than that in the CO. Moreover, the BLISS 2000 optimizes directly the system objectives while the CO focuses on decreasing the discrepancy between the coupled variables in different fields.
The two methods are then applied to an aerodynamics-thermal-structure coupled design problem for the turbine blades of an aircraft engine. Based on the results of sensitivity analysis, the number of design variables is reduced from 34 to 12. To eliminate the impact of the surrogate models on the different MDO algorithms as much as possible, the same initial Kriging surrogate models are not refreshed during the MDO procedures. Without the high fidelity simulation such as CFD and FEM analysis in the MDO processes, the BLISS 2000 method shows the more powerful capability of the convergence than that of the CO method, as shown in the numerical example. The optimization steps of the system level for BLISS 2000 are approximate 1/5 of those for CO, and the iteration number of step in the sub-system level with BLISS 2000 is only about 1/8 of that with CO.
Based on the weight sum of two objectives, the BLISS 2000 shows more robust than the CO because both weight reduction and aerodynamic efficiency are improved in the BLISS 2000 but only the blade weight reduction is gained in the CO. For other multi-objective optimization approaches, however, it still needs to be demonstrated through more studies.
Structural Analysis and Optimization A Must in Spacecraft Projects
