Fcregram: An intuitive and effective optimal demodulation band selection method for bearing fault diagnosis

Rolling element bearings (REBs) are one of the most critical components in rotating machinery. When early-stage faults occur, the fault features are often easily submerged, making accurate and timely diagnosis very challenging. Kurtogram-type algorithms are classic and effective for optimal demodulation frequency band (ODFB) selection, possessing clear logic and a well-established theoretical foundation. However, there are two main drawbacks: First, the fixed-structure frequency band division rules do not account for the signal’s inherent structure, hindering the complete extraction of fault features. Second, the frequency band selection indicators often lack robustness and fail to adequately consider characteristics of REB faults, frequently leading to the selection of an inappropriate ODFB. To address these issues, this paper proposes a method named FCREgram, which can concisely and effectively search for the ODFB. First, taking full account of the signal’s inherent structure, a median bisection method based on resonance peak index sequence is designed for adaptive frequency band division. Then, focusing on fault characteristics, the fault characteristic relative energy indicator is devised for ODFB selection, which possesses a concise expression and outstanding performance. Finally, applying data acquired from simulation, laboratory, and actual operating conditions, it is confirmed that FCREgram exhibits excellent performance in REB fault diagnosis, outperforming benchmark algorithms such as Fast Kurtogram and Infogram.