Parameter Uncertainty Quantification applied to the Duisburg Propeller Test Case

AuthorsKatsuno, E. T., Lidtke, A. K., Düz, B., Rijpkema, D., Vaz, G.
Conference/Journal22nd Numerical Towing Tank Symposium (NuTTS 2019), Tomar, Portugal
DateSep 30, 2019
At model-scale conditions (diameter-based Reynolds number below 106), laminar-to-turbulent transition plays an important role on the performance of a propeller. While at these Reynolds numbers turbulent transition and its effects on the flow are difficult to simulate, current state-of-the-art transition models (Menter et al. (2004)) provide viable means to do so. Previous studies on these new methods (Eca et al. (2016), Baltazar et al. (2018), Lopes et al. (2018)) show that, as in reality, the results are sensitive to the inlet turbulence intensity and its decay upstream of the propeller plane. In the aforementioned studies, numerical uncertainties were reported to have been small. Therefore, it is expected that variability of the user-specified inflow conditions will have a dominant effect on the solution. To better understand the associated uncertainty of model-scale open-water propeller performance predictions, Uncertainty Quantification (UQ) methods are employed in this work. Given a prescribed range of input uncertainties reflecting typical ranges found in test facilities and RANS simulation setups, a laminar-turbulent transition model is used in CFD simulations of the Duisburg Propeller Test Case. This paper then aims to use the results in order to quantify the parameter uncertainty and obtain the output-variable’s Cumulative Distribution Function (CDF), confidence interval and the Sobol indices of each input variable. Together, these quantities are used to suggest suitable ways of achieving accurate and repeatable predictions of propeller performance at intermediate Reynolds numbers.
Tags
data sciencecfdpropeller and cavitationcfd/simulation/desk studies