Detection of Rolling-Element Bearing Faults with Different Conditions

Abstract

Bearing fault diagnosis is essential for the maintenance, durability, and reliability of rotating machines. It can minimize economic losses by removing unplanned downtime in the industry due to the failure of rotary machines. In bearing fault detection, developing intelligent techniques that can diagnose faults under different operating conditions is still a critical issue. In the current work, an intelligent fault diagnosis technique scheme is proposed for the detection of fault patterns for a given faulty bearing. A hybrid feature pool is employed in combination with artificial neural networks ANNs to perform highly effective diagnoses at various motor speeds. The hybrid feature pool is carefully extracted from both time and frequency domains. Envelope analysis is used to extract more discriminating frequency features from the raw vibration signals, while time-domain features are obtained from raw and filtered signals. With the use of 400 test samples with five bearing conditions for each rotating speed in the experiment, an accuracy rate between (97.25%) to (99.75%) was achieved based on time-frequency domain features while yielding an accuracy of (100%) based on frequency features for all rotating speeds. From the conducted tests, it can be concluded that the suggested method can accurately diagnose faults even when the network is trained at a certain speed and used to diagnose at other operating speeds.