123 / 2021-03-31 08:17:46

Dynamic Programming Method of Aeroengine Multi-stage Rotors Stacking Policy

Dynamic programming; Combinatorial optimization; Multi-object optimization; Rotors stacking policy; Reinforcement learning

Aeroengine high pressure compressor (HPC) multi-stage rotors assembly is stacked stage by stage by single rotor. The traditional stacking method is the trial assembly method, which usually requires multiple disassembly and assembly of rotors to ensure assembly precision, and the assembly process is cumbersome and may cause damage to parts. It is meaningful to optimize the multi-stage rotors stacking to improve the assembly quality and reduce aeroengine operational failure. The actual aeroengine HPC rotors assembly data is limited, according to the manufacture precision grade of rotors, this paper simulated the manufacture tolerance data of rotors through the Gaussian distribution and other methods, and simulated the assembly data through FEM. Aiming at the propagation and accumulation of assembly deviations in the stacking process of multi-stage rotors, based on the operational research analysis method, the bolts combination of rotors is regarded as a combinatorial optimization process, while the precision requirements of concentricity and perpendicularity of rotors are regarded as multi-object optimization, and the stacking dynamic programming model is established. With respect to the large computation to carry out the stacking dynamic programming model, the reinforcement learning algorithm is used to improve the computation speed and generate the stacking policy, and the effectiveness of stacking policy is demonstrated by FEM. This paper innovatively proposed a stacking dynamic programming method, which can generate reasonable optimal or near optimal stacking policy, and control the assembly precision of the aeroengine multi-stage rotors within the target range with single assembly. The research of this paper can be used to guide the actual stacking process of aeroengine multi-stage rotors.