Abstract
Computational Fluid Dynamics (CFD) is used extensively in the industry and academia for analyzing the motion of solid particles and the associated solid particle erosion that may occur in various pipe components. However, CFD simulations always carry levels of inherent uncertainties due to the numerical approximations of governing equations, generated grid, and turbulence models. Also, because of the complex nature of solid particle erosion, additional uncertainties are added to erosion prediction simulations. Aspects such as particle size, number of impacts, particles’ initial condition, near-wall mesh effects, forces considered in particle tracking procedures, particle-particle interaction, and near-wall particle-fluid interactions are all possible sources of uncertainties associated with erosion prediction in CFD. Furthermore, unique problems that accompany discrete phase handling and erosion calculation needed for the industrial applications magnify the importance of uncertainty estimation in erosion calculations. Commercially available CFD codes are used with user-developed subroutines to investigate particle erosion prediction uncertainties, numerically in elbows, by considering gas and liquid flow for several pipe sizes. Moreover, different particle sizes, inlet flow velocities, turbulence models, wall functions, and erosion models are examined. According to the ASME’s Verification and Validation (V&V) standard, uncertainties are divided into 3 categories; input, numeric, and modeling. Thus, it is possible to utilize the ASME’s standard as guidance to predict uncertainty for erosion simulations. Furthermore, an extra parameter was considered for uncertainties to account for the uncertainties induced by different simulation procedures and erosion models. The current investigations resulted in developing a framework for estimating uncertainties of erosion simulation. For each simulation result, two bounds (upper and lower) were predicted for erosion. The results show that the Reynolds Stress turbulence model (RSM) and Arabnejad’s erosion model usually predict results corresponding to the lowest uncertainties.
