Abstract
Uncertainties are part of any scientific measurements and calculations. As a complex phenomenon, erosion is affected by many independent parameters and consequently the uncertainties of these parameters contribute to uncertainties in erosion measurements or simulations/predictions. Erosion experiments are costly and time consuming, hence other means must be used to obtain erosion values. Computational Fluid Dynamics (CFD) codes are excellent tools in simulating flow and are used for solid particle erosion calculations for industrial applications for which measured data are unavailable. One of the most utilized frameworks for uncertainty estimation of simulations is ASME standard of Validation and Verification (ASME V&V). According to this standard, there are three main elements of uncertainties in any simulation; input, numerical, and modeling. The modeling element is heavily influenced by the results of measurements, i.e. the measured data and the corresponding uncertainties. In this research, ASME’s guideline is utilized for estimating erosion uncertainty for cases with and without data available. Two alternatives are introduced to be used as replacements for cases where data is not available. These alternatives are machine learning predictions and erosion correlations both developed based on significant number of experiments. The database used in this study includes experimental data with air-sand flows in 3 and 4-inch elbows conducted with 75, 150, and 300-μm sand particles. Finally, the estimated bounds of uncertainties are compared with those obtained by measurements. These comparisons show good agreement between the estimated and measured bounds of uncertainties.
