Planetary gearbox condition monitoring based on modulation analysis

The epicycle gearbox or planetary gearbox (PG) is a central power transmission systems of important machines such as helicopters and wind turbines which are mission critical and high cost systems. Condition monitoring (CM) has been explored extensively in recent years to avoid any unexpected interruptions and severe accidences caused by faults PGs. Although, considerable advancements in CM techniques, there still existed significant deficiency such as insensitivity, false diagnosis and high costs in implementing such techniques in industries. To improve CM techniques, therefore, this thesis focuses on an investigation of advanced signal analysis techniques such as higher order spectra (HOS) in order to achieve full characterisation of the nonlinear modulation processes of PG dynamics and thereby develop accurate diagnostic techniques.

The lumped mass model is established for modelling the dynamic behaviour of the PG under investigation, which allows the vibration behaviours to be understood for analysing different abnormalities such as tooth breakages and gear errors. This paves the way for subsequent data analytics and fault diagnostics using modulation signal bispectrum (MSB) that allows the vibration data to be examined through HOS, but it is significantly efficient in characterising the multiple and nonlinear modulations of PG dynamics alongside superior noise reduction performance.

Different degrees of misalignments in the PG drive system has been investigated and successfully diagnosed using MSB analysis of vibration measurements.. Moreover, the investigation included detection of tooth breakage faults of different severities in both the sun and a planet gear. The tooth faults were diagnosed using the recently developed MSB through accurately representation and estimate of residual sidebands induced by these faults. Consequently, MSB analysis produces an accurate and reliable diagnosis in that it gives correct indication of the fault severity and location for wide operating conditions.

Furthermore, these fault diagnosis practices allows the establishment of residual sideband analysis approach. These residual sidebands resulting from the out-of-phase superposition of vibration waves due to asymmetric, multiple meshing sources are much less influenced by gear errors than the in-phase sidebands due to faults or new occurrences of the symmetricity. MSB can provide an accurate characterisation of the residual sidebands and consequently produces consistent diagnosis as confirmed by both simulation and experiment.