Data Analysis of an Unsteady Cavitating Flow on a Venturi-type Profile

The instability modes and non-linear behavior of a cavitating flow have been studied based on the experimental data obtained from planar Particle Image Velocimetry (PIV). Three data-driven techniques, Proper Orthogonal Decomposition (POD), Dynamic Mode Decomposition (DMD), and Clustered-based Reduced Order Modeling (CROM), are applied to the snapshots of the fluctuating component of velocity to investigate instability modes of the cavitating flow. DMD and POD analysis yield multiple modes are corresponding to slow-varying drift flow, cloud-shedding, and Kelvin-Helmholtz (KH) instability for a fixed inlet flow condition. The high coherence measure obtained from the instabilities suggests a transfer of energy from the largest scales, fluctuating mean flow, to the smaller scales such as cloud cavitation and Kelvin-Helmholtz (KH) instability. It is demonstrated that the POD decorrelation of length scales yields inherently quasi-periodic time dynamics, e.g., incommensurate frequencies. Moreover, the eigenvalue obtained from DMD revealed multiple harmonic with different decay rates associated with the cloud cavitation. The above-mentioned intermittent transition between distinct cloud shedding regimes is investigated via Clustered-based Reduced Order Modeling (CROM). Four aperiodic shedding regimes are identified. 68% of the time, triplets of vortices are formed, while 28% of the time, a pair of vortices are formed in the near wake of the throat. Dominant mechanisms governing the momentum transport and the turbulence kinetic energy production, destruction, and redistribution in distinct regions of the flow field have been identified using Gaussian Mixture Models (GMMs). The preceding data-driven techniques and in-depth analysis of the results facilitated modeling of the cavitation inception and break-up. Accordingly, a phase transition field model is developed using the ultra-fast Time-Resolved Particle Image Velocimetry (TR-PIV) and vapor void fraction spatial and temporal data. The data acquisition is implemented in a Venturi-type test section. The approximate Reynolds number based upon the throat height is 10,000, and the approximate cavitation number is 1.95. The non-equilibrium cavitation model assumes that the phase production and destruction are correlated to the static pressure field, pressure spatial derivatives, void fraction, and divergence of the velocity field. Finally, the dependence of the model on the empirical constants has been investigated.