LEE- Gemi ve Deniz Teknolojisi Mühendisliği Lisansüstü Programı

Bu topluluk için Kalıcı Uri

http://hdl.handle.net/11527/19286

Gözat

Konu "Belirsizlik analizi" ile LEE- Gemi ve Deniz Teknolojisi Mühendisliği Lisansüstü Programı'a göz atma
  • Öge
    Quantifying uncertainties in numerical predictions of dynamic cavitation
    (Graduate School, 2023-06-15) Kara, Erdinç ; Kınacı, Ömer Kemal ; Lidtke, Artur K. ; 508201107 ; Shipbuilding and Ocean Engineering
    In engineering and physics applications, cavitation is a physical phenomenon can be encountered in various fluid systems including pumps, turbines, and propellers, when the pressure of the fluid drops below the vapor pressure, and causes the fluid to evaporate and form small bubbles or voids in the fluid. Cavitation in marine propellers is a critical issue due to its adverse effects on many aspects of operations such as causing noise and vibration, damaging the propeller structural integrity and reducing propulsion efficiency. Therefore, accurate estimation of cavitation from the early stages of design is important to provide better propeller design. However, numerically predicting the cavitation behavior is a difficult task due to the complexity of the problems, the high computational cost of simulations, and various numerical uncertainties. Considering the numerical methods used in the determination of propeller cavitation, it is seen that the use of computational fluid dynamics (CFD) has increased simultaneously with the rapidly advancing computer technology. However, there are various uncertainties that can have important effects on the results such as the input parameters used in the simulations, the quality of the network, the physics models used, the modeling of the computational space, and the definition of the boundary and initial conditions. In the literature, there are various studies in many different areas related to uncertainty quantification and sensitivity analysis. Simultaneously with the increase in the use of CFD tools, studies on the application of uncertainty quantification approaches to CFD problems are gaining momentum. Although there are many studies on numerical analysis of cavitation or uncertainty analysis studies applied on CFD problems in the literature, no studies were found in which numerical uncertainty analysis was performed on the cavitation phenomenon. For this reason, in the current thesis study, the analysis of input parameters and discretization uncertainties separately and together in the numerical estimation of cavitation is emphasized. Since three-dimensional numerical simulation of propeller cavitation is espensive in computational cost, in this study, cavitation behavior on a two-dimensional hydrofoil was investigated as a test case to estimate parameter and discretization uncertainties and combine them into a single value. The test case is selecected as the NACA 66 foil, which has been experimentally studied by many researchers before. Angle of attack and cavitation number were chosen as input parameters and their effects on properties such as force coefficients (lift and drag) and cavitation properties were investigated. ReFRESCO, the internal code of the Maritime Research Institute Netherlands (MARIN), was used in all simulations in the study. Sobol indices were calculated to measure the relative importance of the input parameter and discretization uncertainties. In addition, discretization uncertainty is considered as the third input parameter to investigate the effect of discretization uncertainty on output parameters. Instead of sampling the large number of samples needed for uncertainty quantification analysis, a suitable surface is defined with data generated with only 25 sampling points. Only CFD simulations of these points were performed. Then, using the surrogate model approach from the results obtained, separate response surfaces were created for each of the output parameters; Sobol indices were calculated precisely by taking $2^{15}$ points from these response surfaces. The results show that confidence intervals and Sobol indices for input parameters such as drag coefficient ($C_D$) and cavitation length ($L_{cav}$) remain mostly unchanged despite changes in grid refinement; shows that parameter uncertainties are the dominant factor for these output parameters. Regarding the lift coefficient ($C_L$), it was observed that the effect of discretization uncertainty was larger than $C_D$ and $V_{cav}$, but still parameter uncertainty had a more significant effect than discretization uncertainty. Contrary to other output parameters, the results for the cavitation volume ($V_{cav}$) parameter show that discretization uncertainty has a greater effect on this parameter than the parameter uncertainty. However, the study still showed that parameter uncertainty plays an important role in determining this output parameter. For future research, the scope of this study can be expanded to develop a formulation can be that considers the effects of iterative uncertainty or time step uncertainty as well as measuring discretization and input uncertainty for time-unsteady-flow applications. Other parameters, such as coefficients related to turbulence or cavitation models, can be included in the analyzes to account for additional degrees of modeling uncertainties in addition to grid and operating point conditions.