Vibration based condition monitoring of spur gear using signal processing and machine learning

The objective of this work is to diagnose the fault of spur gear based on vibration analysis using signal processing techniques and machine learning (ML) algorithms. This paper describes two approaches of signal processing techniques, which are time-domain and frequency domain. The proposed method investigated that both approaches of signal processing are suitable for fault diagnosis effectively and has been improved by analyzing from both sides. Variation of noise level during the meshing of gears has also been measured. Statistical features extracted from recorded vibration signals using time-domain approach for healthy and faulty spur gear conditions were used as input to ML algorithms The outcome of this research validated through machine learning approaches such as the J48 algorithm, which is 97.08% classification accuracy. It has been observed that for better monitoring of the health status of the gear, both sides' signals and noise levels must be analyzed. The outcome of this work is an important consideration for fault diagnosis of spur gear as well as bearings and shaft misalignment.