Non-Contact Fault Diagnosis of Bearings using Ensemble and NaÏve Bayes Classifiers

Condition-based maintenance involves maintaining equipment based on the real-time condition of its subsystems. Machine health monitoring typically involves analyzing signals such as vibration, temperature and acoustics, with vibration signature being the most commonly used parameter. Vibration analysis is considered to be a highly effective method for monitoring a machine's operating conditions. Its non-destructive nature makes it increasingly popular in industry, as it enables sustainable monitoring without disrupting the ongoing process. In some cases, the source of the vibration signal may be visible, allowing for corrective actions to be taken. However, in the majority of cases, components are hidden within the machine's structure, making it difficult to identify and locate the source of the signal. To address this challenge, sensors are strategically placed to detect signals produced by components that may need attention to prevent failures. In this article, non-contact fault diagnosis of bearings has been presented using Ensemble and Naïve Bayes classifiers. The findings indicate that the vibration signatures acquired from the developed non-contact sensor are consistent with the accelerometer under the same operational conditions.