Bearing failure detection in farm machinery using low-cost acoustic techniques

Unexpected bearing failures in agricultural equipment can result in considerable inconvenience, potential hazard, and monetary loss. Continuous bearing condition monitoring using vibration and audio spectrum analysis can detect imminent bearing failures before dangerous situations develop. Current application of bearing condition monitoring using vibration signature analysis has focused on fixed industrial applications involving high cost equipment and high shaft speeds. The feasible application of current technology on mobile agricultural equipment requires a lower-cost method of signal sensing and processing. Lower shaft speeds and the use of neural-net based pattern recognition techniques can allow the use of lower-cost transducers necessary for practical adoption on mobile equipment.

A test apparatus was developed and constructed to test the feasibility of using a electret microphone as a sensor for monitoring bearing condition through vibration signature analysis. Laboratory experiments designed to evaluate the sensors with test bearings at both an advanced and an early stage of wear were developed. A field-worn bearing that had been removed from a unit of farm machinery was obtained and examined. Twelve new bearings identical in type to the field worn bearing were also examined, altered to simulate an early stage of wear, and re-examined. Identical experiments were conducted on both the field-worn and altered bearings. The signal acquired from a field-worn bearing was compared with the signal acquired from a new bearing using a two-sample, paired t-test for means at an alpha level of 0.05 and a graphical defect frequency analysis. The signals acquired from the altered bearings were compared with the signals acquired from the new bearings using the same statistical and graphical test performed on the field-worn bearings.

The goal of the analysis was to evaluate the performance of the microphone by attempting to identify significant frequency patterns that could be used to determine the condition of the test bearing while in operation. Duplicate experiments conducted using the accelerometer provided a comparison of sensor performance. The microphone performed well for bearings at an advanced stage of wear. The sensitivity of the accelerometer resulted in better performance when used with bearings at an early stage of wear.