Fluid Noise Elimination in Centrifugal Pump Vibration Signals Using the Frequency Domain Spline Adaptive Filter

[Background]: Vibration signals of centrifugal pumps are often contaminated by complex fluid-induced noise, which complicates structural fault diagnosis.[Purpose]: This study aims to address the suppression of fluid-induced noise in centrifugal pump vibration signals by proposing and experimentally validating a noise reduction method that integrates computational fluid dynamics (CFD) with the frequency domain spline adaptive filter (FDSAF).[Methods]: Vibration acceleration signals of the pump casing were collected under rated operating conditions using a centrifugal pump vibration test rig. Pressure pulsation signals in key regions inside the pump were then calculated and extracted through CFD simulations and employed as reference inputs for the FDSAF filter to selectively eliminate fluid-induced noise from the measured vibration signals.[Results]: The proposed CFD–FDSAF-based approach effectively suppressed fluid-induced noise in vibration signals, thereby enhancing the distinguishability of structural vibration features. After denoising, the power spectral energy near the rotational frequency (49.1 Hz) increased by 1.85 times, while the energy near the blade-passing frequency (294.6 Hz) was reduced to 26.7% of its original value, with its amplitude decreased by 64.5%. [Conclusions]: This method provides a new pathway for data preprocessing in centrifugal pump fault diagnosis by enabling more reliable extraction of fault-related vibration features.