Data-driven modal decomposition of R134a refrigerant cavitating flow in Venturi tube

Abstract

This study utilized high-speed camera and large eddy simulation methods to explore the cavitating flow mechanisms and turbulence structures of R134a refrigerant inside a Venturi tube under varying cavitation numbers (CNs). Data-driven modal analysis approaches, proper orthogonal decomposition (POD) and dynamic mode decomposition (DMD), were introduced to identify and extract the energy hierarchy and transient characteristics within the cavitating flow. The analysis of grayscale images indicated that the cavitating flow gradually transitioned from quasi-periodic to unsteady flow as the CN decreased, and the severity of cavitation correlates with lower peak frequencies. The POD analysis facilitated the extraction of coherent structures in the cavity's temporal evolution, and the results indicate that the quasi-ordering shedding and collapse of large-scale cavity clouds predominantly occur under low cavitation intensity conditions. As the CN increases, the influence of small-scale cavity shedding becomes more significant. The first 30 most energetic modes occupied over 75% of the entire energy, and they were used to reconstruct the cavitating flow, achieving good consistency with transient flow snapshots. Additionally, the DMD results of the cavitating flow yield three frequency spans, including several prominent characteristic frequencies. These spans are closely linked to the cavity cloud structures of varying scales, unveiling the structural characteristics of unsteady cavitating flow.