A comparative study of adaptive filters in detecting a naturally degraded bearing within a gearbox

• The problem of bearing fault diagnosis in gearboxes has been investigated.
• Vibration analysis was used for failure detection.
• Linear Prediction, SANC, LMS and FBLMS algorithms were applied to separate the bearing signal.
• The LMS technique demonstrated the ability to identify the defect earlier than all other methods.

The diagnosis of bearing faults at the earliest stage is critical in avoiding future catastrophic failures. Many diagnostic techniques have been developed and applied in for such purposes, however, these traditional diagnostic techniques are not always successful when the bearing fault occurs within a gearbox where the vibration response is complex; under such circumstances it may be necessary to separate the bearing vibration signature.This paper presents a comparative study of four different techniques for bearing signature separation within a gearbox. The effectiveness of these individual techniques were compared in diagnosing a bearing defect within a gearbox employed for endurance tests of an aircraft control system. The techniques investigated include the least mean square (LMS), self-adaptive noise cancellation (SANC) and the fast block LMS (FBLMS). All three techniques were applied to measured vibration signals taken throughout the endurance test. In conclusion it is shown that the LMS technique detected the bearing fault earliest.