The effect of environmental noises on the acoustic detectability of damaged bearings in rotating machines

Bearing condition monitoring in real environments can be confronted with the presence of disturbing ambient acoustic noise, which can penalise the acoustic modality for fault diagnosis. In this context, performance of signal processing methods for detecting bearing faults, developed without added noise, can be greatly reduced depending on the type of disturbing ambient noise added. The purpose of this study is to examine the bearing failure diagnosis by combining the signal measurement with two different types of ambient noise, including automobile and train noise. The measurement signal is measured by a microphone located near the rotating machine, which rotates at 1000 and 3000 rpm. A healthy bearing and a damaged bearing will be considered. A sound signal processing method that combines filtering and detecting steps is used to diagnose the bearing's condition. The first step will evaluate autogram pre-processing versus kurtogram pre-processing. The second step will evaluate the power after filtering of the fault frequency identified. ROC performance curves will be obtained by varying the signal-to-noise ratio of the added noise.