Challenging the multiplicative model used for gear vibration

Despite rarely being made explicit, signal models (and assumptions) for gear vibration have been fundamental in the development of both condition monitoring and operational modal analysis techniques. The analysis (for condition monitoring) or the removal (for OMA) of the dominant gear-meshing component is in fact dependent on the assumption on the number, location and patterns of the corresponding spectral harmonics. This paper discusses in detail the common modelling choices and their consequences for condition monitoring and OMA. The traditional gearmesh-carrier/shaft-modulated model is analysed and two main limitations of current models are highlighted: the additive assumption on the two gear modulating functions and the regularity of their effect on different gearmesh harmonics. The paper uses experimental gear signals to prove the validity of the newly introduced assumptions and to assess their practical significance.