Loading [MathJax]/extensions/TeX/ieeemacros.js
Time series modeling of vibration signals from a gearbox under varying speed and load condition | IEEE Conference Publication | IEEE Xplore

Time series modeling of vibration signals from a gearbox under varying speed and load condition

Publisher: IEEE

Abstract:

Accurate modeling of the baseline vibration signals generated from a healthy gearbox is critical to the success of time series model-based condition monitoring approach (...View more

Abstract:

Accurate modeling of the baseline vibration signals generated from a healthy gearbox is critical to the success of time series model-based condition monitoring approach (TSMBA). Gearboxes often operate under varying rotation speed and load conditions, which makes the vibration signals non-stationary. It is challenging to accurately model such signals. Existing auto-regression models with exogenous variables (ARX) cannot model the time-varying spectral contents properly due to the limitation on its model structure. Aiming at improving the modeling accuracy, this paper proposes a functional series - operating condition dependent auto-regression (FS-OCAR) model. Legendre polynomials are used to describe the dependence between the operating condition and auto-regression parameters. FS-OCAR is validated using simulation signals from a fixed-shaft gearbox. The modeling accuracy is measured by calculating goodness-of-fit metric and mean squared error of the modeling residuals. Comparisons show that the proposed FS-OCAR outperforms the ARX.
Date of Conference: 11-13 June 2018
Date Added to IEEE Xplore: 30 August 2018
ISBN Information:
Publisher: IEEE
Conference Location: Seattle, WA, USA

I. Introduction

In the field of vibration-based condition monitoring for gearboxes [1], a promising vibration signal processing method is time series model-based approach (TSMBA) [2]. The general steps of TSMBA are as follows: (a) Identify a time series model to represent the baseline vibrations when the gearbox is healthy. (b) Construct an inverse filter based on the identified model. (c) Remove the baseline vibration components from the signals-of-interest using inverse filter. A so-called “residual signal” can be subsequently obtained. (d) Assess the health condition of the gearbox through further analyzing the residual signal, e.g., calculating statistic indicators [3].

References

References is not available for this document.