David Johnston (Committee Member), Gary Lamont (Committee Member), Joseph Shang (Committee Member), Scott Thomas (Committee Member), Mitch Wolff (Advisor)
Doctor of Philosophy (PhD)
Developers of aircraft gas turbine engines continually strive for greater efficiency and higher thrust-to-weight ratio designs. To meet these goals, advanced designs generally feature thin, low aspect airfoils, which offer increased performance but are highly susceptible to flow-induced vibrations. As a result, High Cycle Fatigue (HCF) has become a universal problem throughout the gas turbine industry and unsteady aeroelastic computational models are needed to predict and prevent these problems in modern turbomachinery designs. This research presents the development of a 3D unsteady aeroelastic solver for turbomachinery applications. To accomplish this, a well established turbomachinery Computational Fluid Dynamics (CFD) code called Corsair is loosely coupled to the commercial Computational Structural Solver (CSD) Ansys® through the use of a Fluid Structure Interaction (FSI) module.
Significant modifications are made to Corsair to handle the integration of the FSI module and improve overall performance. To properly account for fluid grid deformations dictated by the FSI module, temporal based coordinate transformation metrics are incorporated into Corsair. Wall functions with user specified surface roughness are also added to reduce fluid grid density requirements near solid surfaces. To increase overall performance and ease of future modifications to the source code, Corsair is rewritten in Fortran 90 with an emphasis on reducing memory usage and improving source code readability and structure. As part of this effort, the shared memory data structure of Corsair is replaced with a distributed model. Domain decomposition of individual grids in the radial direction is also incorporated into Corsair for additional parallelization, along with a utility to automate this process in an optimal manner based on user input. This additional parallelization helps offset the inability to use the fine grain mp-threads parallelization in the original code on non-distributed memory architectures such as the PC Beowulf cluster used for this research. Conversion routines and utilities are created to handle differences in grid formats between Corsair and the FSI module.
The resulting aeroelastic solver is tested using two simplified configurations. First, the well understood case of a flexible cylinder in cross flow is studied with the natural frequency of the cylinder set to the shedding frequency of the Von Karman streets. The cylinder is self excited and thus demonstrates the correct exchange of energy between the fluid and structural models. The second test case is based on the fourth standard configuration and demonstrates the ability of the solver to predict the dominant vibrational modes of an aeroelastic turbomachinery blade. For this case, a single blade from the fourth standard configuration is subjected to a step function from zero loading to the converged flow solution loading in order to excite the structural modes of the blade. These modes are then compared to those obtained from an in vacuo Ansys® analysis with good agreement between the two.
Department or Program
Ph.D. in Engineering
Year Degree Awarded
Copyright 2010, all rights reserved. This open access ETD is published by Wright State University and OhioLINK.