According to the nonstationary characteristics of vibration signals of turboprop engine rotor system

a rotor fault diagnosis method based on ensemble empirical mode decomposition (EEMD) and neighborhood rough set (NRS) was proposed. EEMD was used to decompose vibration signals. Then the time domain features and multiscale permutation entropies (MPE) of original vibration signals were calculated. Most of the fault information of rotor system was contained in the first several intrinsic mode functions (IMFs). Their time domain indicators

energy features and singular value decomposition (SVD) features were also separately calculated. NRS was applied to evaluate the attribute importance of different features

and sensitive features were selected. The sensitive feature set was fed into support vector machine (SVM) for recognizing different rotor fault states. Experimental results demonstrated that the fault diagnosis accuracy of the method reached 97.5%

and a large number of redundant features were eliminated

indicating that the method had stronger robustness.