Combinatorial optimization of mistuned blade rearrangement based on reduced-order FEA model

This paper is focused on the optimization of mistuned blades assembling rearrangement under the forced response. First, in order to avoid the greatly increase of the calculation greatly by the whole circle bladed-disk finite element model, a reduced-order model is developed based on the component mode synthesis. CPU+GPU heterogeneous architecture parallel computation is used to accelerate modal analysis of the disk and blade sectors substructures. Second, a modified ant colony algorithm is applied to the combinatorial optimization to find the optimal rearrangement pattern of bladed-disk assembly. Different from classical algorithm, the individual mistuned information is used to construct heuristic function based on intentional mistuning pattern, which can avoid slow convergence of ant colony algorithm and increase the search speed efficiently. At last, a high-fidelity 3D FEM model with 43 mistuned blades is used to demonstrate the capabilities of the techniques in reducing the maximum displacement resonance response of the bladed-disk system. The numerical simulation showed that this program based on the reducedorder model proposed in this article gained 4.3 speedup compared with ANSYS full model under the scale of 500k nodes. The displacement response amplitude of the blades decreased by 32% with 60 steps (1200 times FEM calculation) by the new optimization method. The physical mechanism of reducing the bladed-disk response is explained by comparing the optimized and worst arrangement patterns. The results clearly demonstrate that the optimized rearrangement pattern of mistuned blades is able to reduce the response amplitude of the forced vibration significantly, and the algorithm proposed in this article is practical and effective.