Experimental study and feature identification of cavitation-induced acoustic pressure signals in a large vaned-voluted centrifugal pump
Pumps are critical mechanical devices in water conservancy and agricultural systems, yet cavitation remains a major issue affecting their operational safety and efficiency. To investigate the acoustic pressure response characteristics under cavitating conditions in a large vaned-voluted centrifugal pump, a model-scale experimental platform was developed, covering an operating range of 0.4–1.5 Q d ( Q d is the design flow rate). Acoustic pressure signals were acquired at the design flow rate under three cavitation numbers: σ = 0.2372 (non-cavitating), σ = 0.096 (initial cavitation), and σ = 0.0637 (severe cavitation). Combined with visualizations of bubble dynamics in the blade passages, a feature identification method for cavitation-induced acoustic pressure signals was established. To enhance the detection of key signal features, a denoising scheme based on variational mode decomposition (VMD) and time-synchronous averaging (TSA) was proposed. This approach effectively separates and suppresses asynchronous low-frequency noise while preserving transient mid-to-high frequency shock components associated with bubble collapses. Subsequently, a Hankel-matrix-based dynamic mode decomposition (Hankel-DMD) method was employed to solve the system’s linear operator and extract dominant modes. Results show that with increasing cavitation intensity, dominant modes with significantly increased energy emerge in the 500–1000 Hz range in the DMD spectrum of the VMD-TSA processed signals—serving as indicators for cavitation onset and evolution. Meanwhile, the amplitude of blade passing frequency and its harmonics in the low-frequency range (<400 Hz) systematically decreases, consistent with the observed attenuation of pressure pulsations in the non-bladed region of the guide vane zone. Overall, the proposed VMD-TSA denoising and Hankel-DMD feature extraction framework effectively identifies characteristic modal features of cavitation acoustic pressure signals in pumps, offering a novel technical pathway for real-time monitoring and diagnostics of cavitation in large vaned-voluted centrifugal pumps.