Online quantitative detection for rotor misalignment based on data-driven modeling under nonstationary harmonic excitations

Misalignment is a commonly encountered malfunction that can generate reaction forces and moments in the coupling leading to excessive vibration in the rotor. Generally, the vibration responses of a rotor with shaft misalignment comprised of nonlinear components. Some widely focused linear detection methods are incapable of extracting the key nonlinear information. Nonlinear output frequency response functions weighted contribution rate (WNOFRFs), as a convenient nonlinear detection method, fails in online fault detection due to the requirement for more than one-time tests. Although the nonlinear autoregressive with eXogenous inputs (NARX) model can address this challenge, it encounters difficulties associated with its instability and inaccuracy when applied to nonlinear systems under nonstationary harmonic excitations. This research proposed an innovative online quantitative detection method based on data-driven modeling under nonstationary harmonic excitations. Additionally, a novel index related to WNOFRFs for representing nonlinear features is put forward. Simulation and experiments conducted on rotor shaft misalignment validate the efficacy of this method. This approach is not only proficient in the online detection of rotor misalignment but also in quantitatively representing detailed angles of rotor shaft misalignment.