Vortex Characterization of Rotor-Stator Cavities Dominated With Radial Flows

Abstract

Abstract Rotor-Stator cavities are quite common and an important feature in the turbomachinery application. The static pressure and total temperature change in the rotor-stator cavities play a critical role in the design of the secondary air system, bearing load and component life. Hence understanding the vortex characteristics and windage heating is vital to accurately estimate the swirl velocity, static pressure, and air temperature surrounding rotor/stator structures. This paper describes the development of a numerical methodology aimed at studying the vortex dynamics and windage heating in rotor-stator cavities. Specifically, a 2D axisymmetric validated CFD model is developed utilizing commercial finite volume-based software incorporating the Reynolds Stress turbulence model. Using the validated model, a broad parametric study is conducted by varying the aspect ratio of the cavity, inlet swirl fraction and mass flow rate at a constant rotor speed. Finally, the radial variation of swirl velocity, static pressure and total temperature are studied for the radial inward and outward flows. The nature of the vortices is found to be very complex and they cannot be categorized into either free or forced vortex. The vortices change radially in the cavity with change in the flow, inlet swirl and cavity aspect ratio. Practically it is difficult and cumbersome to arrive at a universal characteristic equation for the range of flow conditions for a cavity of interest. In this paper an attempt is made to arrive at the universal characteristic equation using machine learning based surrogate model for the rotor-stator cavity. A polynomial regression is developed using the CFD generated data sets by suitably training and testing the model. Feature engineering and hyperparameter tuning techniques are used to select the important features from the input parameters to make the model simple without significant accuracy loss. The results from the surrogate model show accuracy of 97% with an independent set of CFD data used in validation. Understanding of the vortex dynamics from CFD and the polynomial vortex characteristic equations will be instrumental in the design of secondary air system, bearing load and component life in turbomachinery.