Single-sensor identification of multi-source vibration faults based on power spectrum estimation with application to aircraft engines

Abstract

Identifying the vibration fault of aircraft engines in lack of sufficient sensor information is a big challenge to researchers due to the restriction of sensor installation location and the nature of complicated multi-source vibrations on turbine rotors. In this paper a new method of single-sensor identification for multi-source vibration faults is proposed based on the blind source separation (BSS), the empirical mode decomposition (EMD), and power spectrum estimation. First the observed single-channel multi-source vibration signal is decomposed by EMD decomposition method, so that the intrinsic mode function (IMF) component can be obtained and new observation signal can be reconstructed. Second the source number is to estimate using the ratio matrix of power spectral density function, then the observation signals are reconstructed based on the estimated source number. By blindly separating the mixed signals matrix composed of the reconstructed observation signals and the original measurement signal, the original multi-source vibration components are obtained. The final step is to analyze the ratio of power spectral density function of signals to identify the characteristics of multi-source vibrations. Simulation and experiment results for applications in the turbine rotor system validated the proposed method on decomposing vibration signals and identifying the characteristics of vibration faults.