Film cooling is a technology used in the field of turbomachinery, such as in land-based power generation and in aircraft engines, which serves to help provide a safe thermal condition in key hot hardware components. With a significant presence of turbulence present in the flow-field, these film cooling jets emanate into a crossflow environment in such a way that is very challenging for computational fluid dynamics to accurately predict. One impact of such is significant uncertainty for those in industry when conducting both aerodynamic and thermal designs. This study presents numerical approaches which are demonstrated on film cooling flow-fields and are underpinned by experimental measurements on identical geometrical specimen and testing conditions. Extensive taxonomy and modifications are given to eddy viscosity-based turbulence models, and their available constituent components, which will directly benefit those in industry designing hardware with computational fluid dynamics. These computational fluid dynamics simulations and film cooling effectiveness experimental measurements utilize an array of laid-back diffuser film cooling holes which are laterally expanded at ten degrees, for blowing ratios between 0.5 and 2.5. Further investigation is executed with similar computational fluid dynamics methodologies for film cooling holes laterally diffused film holes at fourteen degrees, for which flow and turbulence experimental measurements are provided. Additionally, unique eddy-viscosity based model extensions are presented which have previously not been evaluated on similar industrial flow configurations. These models account for the anisotropic state of the turbulence from within the turbulent dissipation rate equation and prove viable for naturally mitigating some of the typical sources of error to be expected in eddy-viscosity models (such as length scale predictions). The lag elliptic blending k-epsilon model and elliptic blending Reynolds stress models serve as advanced RANS models for which these extensions are tested on. Large eddy simulation study is also conducted for thorough comparisons of such models, and as well to support ideas of an uncertainty parameter which can be useful when understanding failures of typical turbulence models in such flow-fields. Lastly, these assertions regarding turbulence modeling are complemented with a random forest regression machine learning algorithm focused on predicting the anisotropic nature of the turbulence. While thermal (effectiveness) performance results are presented surrounding experimental measurements, the sole focus of the machine learning study is the prediction of the turbulence. The algorithm is trained on combinations of data from six high quality non-film cooling simulations, whereby making successful predictions on a film cooling flow-field. From this evidence that machine learning algorithms can predict the turbulence anisotropy in this class of flow-fields, such turbulence models presented can be sensitized and improved with this new formulation. Ultimately, the objective is to provide a validated machine learning approach for film cooling flows which can be used to improve turbulence predictions from eddy viscosity-based models.
If this is your thesis or dissertation, and want to learn how to access it or for more information about readership statistics, contact us at STARS@ucf.edu
Doctor of Philosophy (Ph.D.)
College of Engineering and Computer Science
Mechanical and Aerospace Engineering
Length of Campus-only Access
Doctoral Dissertation (Campus-only Access)
Hodges, Justin, "Investigating Film Cooling Flows with Advanced Turbulence Modeling, Machine Learning, and Experimental Methods" (2020). Electronic Theses and Dissertations, 2020-. 441.
Restricted to the UCF community until November 2025; it will then be open access.