Dual-objective optimization method for aeroengine multistage rotor stacking: concentricity and perpendicularity via deep Q-network

Abstract The assembly precision of multistage rotors is crucial for the high-performance operation of aeroengines. Concentricity and perpendicularity are among the key indicators for evaluating assembly precision. Traditional trial-based assembly methods require multiple disassembly operations to meet precision requirements, which is cumbersome and may lead to component damage. This paper proposes a dual-objective optimization method to achieve concentricity and perpendicularity in aeroengine multistage rotor stacking based on a deep Q-network (DQN). First, a prediction model for concentricity and perpendicularity in the multistage rotor stacking process is established. This stacking process is subsequently formulated as a Markov decision process, and a dual-objective optimization method for multistage rotor stacking based on a DQN is proposed. Second, a six-stage rotor stacking process is simulated and an optimized stacking solution is obtained using the proposed method. Finally, a rotor stacking measurement platform is established to validate the optimization performance of the proposed method. Both simulated and experimental results show that the proposed method outperforms conventional methods (including the genetic algorithm method, the trial method, and the exhaustive method) in obtaining optimal assembly solutions that balance computation time with concentricity and perpendicularity, offering theoretical guidance and practical solutions for onetime high-precision assembly of aeroengine multistage rotors.