State Space Formulation of Nonlinear Vibration Responses Collected from a Dynamic Rotor-Bearing System: An Extension of Bearing Diagnostics to Bearing Prognostics

Abstract

Bearings are widely used in various industries to support rotating shafts. Their failures accelerate failures of other adjacent components and may cause unexpected machine breakdowns. In recent years, nonlinear vibration responses collected from a dynamic rotor-bearing system have been widely analyzed for bearing diagnostics. Numerous methods have been proposed to identify different bearing faults. However, these methods are unable to predict the future health conditions of bearings. To extend bearing diagnostics to bearing prognostics, this paper reports the design of a state space formulation of nonlinear vibration responses collected from a dynamic rotor-bearing system in order to intelligently predict bearing remaining useful life (RUL). Firstly, analyses of nonlinear vibration responses were conducted to construct a bearing health indicator (BHI) so as to assess the current bearing health condition. Secondly, a state space model of the BHI was developed to mathematically track the health evolution of the BHI. Thirdly, unscented particle filtering was used to predict bearing RUL. Lastly, a new bearing acceleration life testing setup was designed to collect natural bearing degradation data, which were used to validate the effectiveness of the proposed bearing prognostic method. Results show that the prediction accuracy of the proposed bearing prognostic method is promising and the proposed bearing prognostic method is able to reflect future bearing health conditions.

Keywords: acceleration life testing; bearing prognostics; dynamic rotor-bearing system; non-linear vibration responses; remaining useful life.

Figure 1

A flowchart of the proposed bearing prognostic method for prediction of bearing remaining useful life (RUL).

Figure 2

The experiment setup: (a) the locations of four accelerometers; (b) the application of the hydraulic jack to the outer race of the bearing; (c) the hydraulic jack without touching the surface of the bearing; and (d) the hydraulic jack applied to the surface of the outer race.

Figure 3

A vibration measurement collected from the vertical direction of the casing of the right bearing at measurement number 50: (a) its temporal signal and (b) its corresponding frequency spectrum.

Figure 4

Bearing degradation assessment: (a) the RMS after applying the high-pass filter; (b) the degradation trend generated by the BHI.

Figure 5

Bearing RUL prediction by using the proposed bearing prognostic method at measurement number 30. (a) The degradation trend; (b) the probability density function (PDF) of the RUL.

Figure 6

Bearing RUL prediction by using the proposed bearing prognostic method at measurement number 50. (a) The degradation trend; (b) the probability density function (PDF) of the RUL.