Abstract: Aiming at the problems that the prediction of rotor assembly geometric accuracy is inconsistent with the measured results and the stacking optimization results are scattered greatly under deterministic conditions, the method of prediction and optimization of rotor assembly geometric accuracy under uncertainty is proposed. Considering the multi-source uncertainties in the assembly process, the source analysis and characterization of the uncertainties are firstly carried out. Then, the prediction model of rotor assembly geometric accuracy under uncertainty is established, and the random-interval mixed uncertainty model is solved based on the Monte Carlo method with a double-layer nesting approach. On this basis, a double-layer optimization model of rotor coaxiality under mixed uncertainty is constructed, and the upper bound of coaxiality overproof probability and the mean of the midpoints in all coaxiality error intervals are minimized at the same time. Finally, taking a simulated multi-stage rotor as the object, the proposed method is verified by comparing measured results, uncertainty prediction and optimization results, and results from the existing deterministic stacking method. The results show that the optimal assembly reduces the rotor coaxiality error by 57.28%, and the rotor coaxiality error values obtained by multiple assembly measurements are within the uncertainty prediction range, which indicates that the proposed method can achieve accurate prediction and reliable optimization of the rotor assembly geometric accuracy.

Key words: rotor assembly, geometric accuracy prediction, mixed uncertainty, overproof probability, coaxiality optimization