Improved Backward Ridge Tracking for Rolling Bearing Fault Diagnosis Under Strong Noise

To avoid the operational interference and increased cost caused by encoder installation in rotating machinery, tacholess order tracking (TLOT) based on vibration signals plays an important role in instantaneous frequency (IF) estimation. However, existing TLOT methods are susceptible to noise and overly reliant on prior information, which limits their applicability. To address these issues, this paper proposes an improved backward ridge-tracking (IBRT) algorithm with an adaptive dynamic search range. The proposed method incorporates the chirp rate to enhance the backward ridge-tracking strategy and constructs an adaptive cost function, which effectively reduces the penalty on true ridge frequency variations under low signal-to-noise ratio (SNR) conditions, automatically balances peak amplitude and IF continuity, and requires no manual parameter tuning. In addition, an adaptive ridge search boundary is designed based on the improved tracking strategy to enhance noise robustness. Comparative experiments with four existing IF extraction methods demonstrate that IBRT exhibits significant advantages in noise suppression and parameter adaptivity.