Vortex energy conversion and dynamic variation of excitation force under cavitation conditions in a centrifugal pump

This study systematically analyzes the cavitation–vortex evolution and vortex dynamic energy conversion in a single-stage centrifugal pump. It integrates numerical simulations, experimental measurements, the Omega-Liutex method, the vorticity transport equation, the kinetic energy transport equation, and deep learning techniques, and further develops a vortex–axial/radial force correlation model. The results demonstrate that, during cavitation development, bubbles at the blade inlet interact with strong vortices. The attached vortices on the suction surface tear and disintegrate while the bubbles expand toward the inlet, weakening the weaker vortex structures. The pressure gradient primarily governs the kinetic energy conversion of vortices, whereas the vorticity-bearing movement contributes minimally. The cross-power spectral density between pressure propulsion power and Lamb vector divergence reaches its maximum, and the dominant frequency components of kinetic energy transport terms and fluctuating pressure signals correspond to the rotational frequency. As the net positive suction head decreases, both signal correlation and phase difference increase. The axial force is sensitive to cavitation onset but exhibits limited variation, whereas radial forces show significant changes. Convolutional neural network outperforms traditional machine learning models in predicting vortex-induced excitation forces.