Advanced Turbulence Modelling and Near-Wall Treatments for Rotating Ducts and Centrifugal Compressor Flows

Abstract

The main aim of the thesis is to report numerical investigations ofdeveloping flows through a centrifugal blade passage. Theinvestigation focused on exploring alternative ways of predictingthe near wall region and rotation using advanced stress transportmodels and wall functions within such geometries.In the initial stages of this study, the flow fields withinstationary and rotating ducts have been reproduced through the useof various turbulence models and wall treatments. The aim ofexploring such geometries was to gain insight into the flow physicswithin them as these flows share similar features found in the flowwithin centrifugal impellers. Initial refinements of the advancedwall treatment and the approaches used to couple the wall functionsand the turbulence models were also performed on the ducts. Thisstudy confirmed that linear eddy-viscosity schemes fail to capturemany features associated with such flows. The two stress transportschemes tested (a basic linear one, and more advanced non-linearform) performed similarly to each other, and gave reasonableagreement with data, although they both failed to faithfullyrepresent the effects of secondary flows on rotation on the suctionside. It was noted that the advanced wall function tested (theAnalytical Wall Function, or AWF, based on solving a simplified formof the momentum equation across the near-wall cell) in its originalform failed to reproduce the effects of rotation and that it washighly sensitive to the approach used to approximate convectiveterms within it. Reasonable results for the rotating ducts wereobtained with the AWF using a refined approximation of theseconvective terms, although turbulence levels were generallyunderpredicted close to the wall.A stationary impeller blade passage of the NASA Low speedcentrifugal compressor (LSCC) has then been simulated. The standard$k-\varepsilon$ and $k-\omega$ models, and two stress transportschemes were used to model the core flow region, while the standardand analytical wall functions were used to predict the near-wallregion. Although there is no corresponding available experimentaldata, the study was aimed at exploring the behaviour of thedifferent turbulence models and wall treatments and gaining insightinto the flow field in the absence of rotation. It was found that,in general, both stress transport models perform similarly in termsof reproducing the flow features within the blade passage and thatin the first part of the passage, where there is substantialboundary layer growth and flow development, both the velocity andturbulence fields are mainly affected by the choice of the walltreatment, while in the second half, the predictions of the flowfield are mainly dependent on the choice of the outer turbulencemodel. The turbulence levels predicted by the AWF were generallylower than those returned by the standard wall function.Finally the flow within the rotating NASA Low speed centrifugalimpeller was modelled using the previously described turbulencemodels and wall treatments and comparisons were made with existingexperimental data. It was noted that the various combinations ofturbulence models and wall treatments produced generally similarresults for the various velocity components, although the meridionalvelocity component was slightly underpredicted in locations. Thespanwise velocity was not, in general, faithfully produced by thevarious turbulence models, especially in the second half of theimpeller passage. Additionally, similar values for turbulencequantities were produced via the different turbulence models andwall functions, with the main difference being that the use of theAWF often resulted in lower levels of turbulent kinetic energy thanthose computed via the standard wall function.

Date of Award	1 Jan 1824
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Timothy Craft (Supervisor) & Ali Turan (Supervisor)