Many applications, including elbows of piping systems in pressurized water reactors, require ductile materials. However, the impingement of solid particles entrained in turbulent flows causes these materials to erode. To mitigate erosion failures, this paper proposes a low-cost method for predicting erosion in elbows carrying gas-dominated and liquid-dominated flows. The method solely relies on Computational Fluid Dynamics (CFD), drawing inspiration from experimental data of 90-degree stainless-steel elbows found in the literature. Utilizing the standard k-e turbulence model and standard wall function, penetration data of elbows operating in three different flow directions were predicted. In each flow direction, three models- DNV (2007), Oka et al. (2005), and Finnie (1960)- were evaluated. The CFD evaluation was based upon empirical data obtained in millimeters per day from four particle sizes (100, 300, 350, 450 µm), three radius-of-curvature-to-diameter ratios (1.50, 1.53, 3.25), three gas superficial velocities (25.24, 47.00, 72.00 m/s), and two liquid superficial velocities (4.00, 6.48 m/s). The DNV model exhibited a high degree of agreement with experimental data when applied to low liquid (4.00 m/s) and gas (25.24 m/s) superficial velocities. In contrast, experimental data obtained from high superficial gas (47.00, 72.00 m/s) and liquid (6.48 m/s) velocities were well correlated with the model proposed by Oka et al. At high superficial velocities, the DNV model underpredicted the data. Saffman force inclusion in particle tracking has remarkably increased the erosion prediction of all models for gas-dominated flows but resulted in less prediction for liquid-dominated flows.