Extraction of second-order cyclostationary sources by matching instantaneous power spectrum with stochastic model - application to wind turbine gearbox

RENEWABLE ENERGY

卷:147

Article

Xin, Ge^1,2,3  Hamzaoui, Nacer²  Antoni, Jerome² 

[1] Beijing Jiaotong Univ, Sch Traff & Transportat, Beijing 100044, Peoples R China

[2] Univ Lyon, Lab Vibrat Acoust, LVA EA677, INSA Lyon, F-69621 Villeurbanne, France

[3] Beijing Jiaotong Univ, Beijing Res Ctr Urban Traff Informat Sensing & Se, Beijing 100044, Peoples R China

关键词: Wind turbine gearbox;    Machinery fault diagnosis;    Semi-blind feature extraction;    Cyclic source separation;    Cyclostationary signals;   

DOI  :  10.1016/j.renene.2019.09.087

来源: Elsevier

【 摘 要 】

The diagnosis of gearboxes plays a crucial role in the maintenance of wind turbine. Considering critical elements - i.e. gears and bearings - of gear set, the effective and exact identification of fault sources is appealing yet challenging in complex mechanical systems. Although rotating machine signals are perfectly modelled as cyclostationary (CS) processes, very few researches have so far tried to refine single CS component of interest from a mixture of multi sources; thus the efficacy of classical vibrodiagnostic tool (e.g. envelope analysis) can be greatly enhanced in a wide variety of situations, e.g. poly-cyclostationary cases in wind turbine gearboxes. As such, this paper exploits the statistical behavior of CS signals using a stochastic model based on a periodic variance to extract more specific information from the data themselves. In particular, a statistical indicator is proposed to assess the strength of CS components as well as a full-band time-dependent filter to recover the pure CS signals in the time domain. This proves very useful in many situations where the characteristic components of gears and/or bearings are embedded in heavy background noise that jeopardize their detection in practical applications. The derivation of the proposed scheme is described in detail. Its effectiveness is finally demonstrated with both synthetic and experimental examples. (c) 2019 Published by Elsevier Ltd.

【 授权许可】

Free