ACKgram: A robust approach for bearing fault diagnosis under complex interference

• Published in: Mechanical systems and signal processing Vol. 230; p. 112646
• Main Authors: Zhang, Yue, Zhao, Ming, Li, Sen, Chen, Dexin, Ma, Biao, Song, Zhihua
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.05.2025
• Subjects: Adaptive correlated kurtosis; Cyclic modulation spectrum; Fault diagnosis; Frequency band segmentation; Rolling bearing; Fault diagnosis; Frequency band segmentation; Adaptive correlated kurtosis; Rolling bearing; Cyclic modulation spectrum
• ISSN: 0888-3270
• DOI: 10.1016/j.ymssp.2025.112646

•An adaptive frequency band segmentation method based on spectral trends is proposed.•Adaptive correlated kurtosis (ACK) is introduced as a blind indicator to evaluate periodicity and impulsiveness.•A novel ACKgram framework is developed for bearing fault diagnosis under complex interference.•Simulations and experiments validated ACKgram’s superior accuracy and reliability. As a vital component in rotating machinery, bearings usually generate repetitive transients when faults occur. However, these fault-characterizing transients are easily masked by various types of interference and noise. As a benchmark method for bearing fault diagnosis solutions, narrowband filtering demodulation (NFD) often relies on fixed-mode frequency band segmentation (FBS), which could result in the omission of essential fault information. Moreover, commonly used indicators for optimal frequency band (OFB) selection either suffer from susceptibility to interference, leading to diagnostic failures, or rely on prior fault information, which limits their practical application. Therefore, a robust bearing fault diagnosis method termed ACKgram is proposed. By iteratively calculating the average envelope of the spectrum, this method enables multi-level adaptive FBS based on the specific spectrum characteristics of different signals, ensuring the completeness of fault information in narrowband signals. In addition, an indicator named adaptive correlated kurtosis (ACK) is introduced, which can be calculated independently of prior fault information. In contrast to conventional indicators, ACK simultaneously evaluates the periodicity and impulsiveness of the signal, providing a reliable criterion for OFB selection. The performance of ACKgram is validated using simulated signals with various types of interference and multiple sets of experimental signals at different rotational speeds. Compared to other typical methods, ACKgram demonstrates exceptional performance in detecting bearing faults under complex interference.