Diagnosing of rolling-element bearings using amplitude level-based decomposition of machine vibration signal

•A new automatic approach for rolling-element bearing diagnostics is proposed.•A signal is divided into three amplitude-level-dependent components.•Extraction of a bearing signal from a machine vibration signal was performed.•Spectral analysis of the empirically determined local amplitude of a signal is used.•The proposed approach enables early detection and identification of bearing faults. In the last few decades there has been a significant development in the use of vibration measurement and analysis for monitoring the condition of rolling-element bearings. Although a lot of vibration diagnostic techniques have been developed, in many cases these methods are quite complicated to use and are time consuming. They are even impractical in real-world applications or are only effective at later stages of damage development. One of the main reasons for the ineffectiveness of many diagnostic approaches is the fact that in complex industrial environments the vibration signal of the rolling-element bearing may be covered or concealed by other vibration sources, such as gears. As a result, the development of methods for extracting an informative bearing signal from a machine vibration signal is one of the most important topics in diagnosing rolling-element bearings operating in complex industrial environments. This paper presents the diagnostic approach enabling early detection of a rolling-element bearing fault at the low-energy stage of its development. A key element of this approach is the completely automatic method of amplitude level-based signal decomposition, which enables an extraction of an informative bearing signal from a machine vibration signal. In order to perform a bearing fault-related feature extraction from a low-energy component of a vibration signal, the spectral analysis of the empirically determined local amplitude is used. The practicability and the effectiveness of the proposed approach have been tested on simulated and real-world vibration data. Tests of the devised approach give better results than the classical method and show that this approach is appropriate and effective at identifying bearing damages at early stages of their development.