An Efficient Approach for Nonlinear Rotor Orbit Prediction Based on Artificial Neural Network Method and Fully Aeroelastic Coupling Model

Abstract Due to their low damping and nonlinear characteristics, gas foil bearings are susceptible to an imbalanced amount, so accurately predicting their dynamic response is crucial. The current time-domain-based rotor orbit prediction method has two problems. First, it is computationally expensive and time-consuming. Second, the simplified model for the complex foil structure results in low accuracy. This study combines the artificial neural network method with a novel fully aeroelastic coupling model of a multi-leaf journal foil bearing (MLJFB), which considers assembly preload, friction, and the interaction between the rotor, top foil, bump foil, and sleeve, to construct a rapid prediction model for gas film load capacity. Based on this prediction model, a nonlinear rotordynamic model is developed to enable efficient estimation of nonlinear responses, including the rotor orbit. A test rig for a high-speed MLJFB was designed and built to verify the accuracy of the theoretical model. The effects of load and rotational speed on the rotor orbits were then analyzed using a combination of theoretical and experimental approaches, with the results showing good agreement. This study provides a rapid method for predicting rotor orbits, which offer a valuable reference for the efficient optimal design and practical application of MLJFB.