Using Cepstrum and Historical Data to Detect Planetary Stage Fault

摘要

This work shows how cepstrum helps identify families of sidebands related to gear faults in the planetary stage of a wind turbine gearbox. Quefrency peaks (representation of those sidebands families) identified can be trended over time, and early stage fault detection can be achieved. Normally the kinematical data of a specific gearbox, such as gear mesh frequency and shaft rotational speed, must be known in order to trend the quefrency peaks. Compiling cepstrum signatures from hundreds of turbines with known and unknown faults shows that turbines with similar gearbox ratios can have common increasing quefrency peaks related to certain faults. Information regarding these quefrency ranges can then be used for trending. In this way, it is not necessary to know the details of gear counting for each turbine.