Internal combustion engine gearbox bearing fault prediction using J48 and random forest classifier

Abstract

Defective bearings in four-stroke engines can compromise performance and efficiency. Early detection of bearing difficulties in 4-stroke engines is critical. Four-stroke gasoline engines that vibrate or make noise can be used to diagnose issues. Using time, frequency, and time-frequency domain approaches, the vibrational features of healthy and diseased tissues are examined. Problems are only detectable by vibration or sound. The fault is identified through statistical analysis of seismic and audio data using frequency and time-frequency analysis. Vibration must be minimized prior to examination. Adaptive noise cancellation removes unwanted noise from recorded vibration signals, boosting the signal-to-noise ratio (SNR). In the first of the experiment's three phases, vibrational data are collected. To reduce noise and boost SNR, adaptive noise cancellation (ANC) is applied to vibration data from the first stage. In the second stage, ANC-filtered vibration data is subjected to three studies to detect bearing failure using J48 and random forest classifiers for online, real-time monitoring. In this experiment, one healthy and two faulty bearings are used. According to a current study, the internet of things (IoT) is a promising alternative for online monitoring of remote body health.