The paper presents a fast, accurate and simple systematic approach for online condition monitoring and severity identification of ball bearings. This approach utilizes compact 1D convolutional neural networks (CNNs) to identify, quantify, and localize bearing damage. The proposed approach is verified experimentally under several single and multiple damage scenarios. The experimental results demonstrated that the proposed approach can achieve a high level of accuracy for damage detection, localization and quantification. Besides its real-time processing ability and superior robustness against the high-level noise presence, the compact and minimally-trained 1D CNNs in the core of the proposed approach can handle new damage scenarios with utmost accuracy.

This research aims to analyze the vibration response of damaged rolling element bearings experimentally and to assess their degree of degradation by examining parameters extracted from the time domain. This task was accomplished in three phases. In the first phase, a test rig was carefully designed and precisely manufactured. In particular, an innovative solution for rapidly mounting and dismounting bearings on the supporting shaft was tested and used successfully. In the second phase, a specific technique of seeding defects inside the ball bearings was developed. In the last phase, damaged bearings (and healthy ones serving as a reference) were installed on the test rig, and different vibration measurements were taken. The results obtained from this work show that different parameters could be extracted from the time domain. In addition to the six common indicators (peak, root mean square, crest factor, kurtosis value, impulse factor, and shape factor), four hybrid new ones have been proposed (Talaf, Thikat, Siana and, Inthar). The experimental results confirm the well-known efficiency of kurtosis in the detection of bearing defects. However, the newly proposed parameters were found to be more responsive to defect growth.