The entrainment of sand particles in fluids flowing through horizontal or vertical pipes is frequently occurring during oil and gas production and transportation. Consequently, sand impingement on the internal walls of pipes and fittings leads to continuous eroding and wall thinning of these fittings. One of the practices followed to avoid erosion is by early prediction of erosion rate and controlling the flow parameters to be within a pre-specified safe operation limit. The prediction can be conducted using empirical correlations, semi-empirical models, or computational fluid dynamics (CFD) simulation. Direct impingement model (DIM) is a semi-empirical model that processes high accuracy result than the empirical correlations and involves inclusion of the effect of viscosity. DIM, however, was found more complicated than empirical correlations as it involves numerical solution of the equation of particles motion. It is desired, therefore, to develop a model that processes the features of both empirical (i.e. easy use) and semi-empirical (i.e. high accuracy) models. In this paper, a correlation has been proposed for prediction of sand erosion in elbows due to entrainment of sand in oils. The correlation has been developed by mean of comparison of results from the empirical Salama model with results from DIM for oil with different viscosities. As a result, the viscosity was incorporated into Salama model and the applicability of it is extended to oil flow. The new model was validated against results from the DIM model as well as published experimental data. The new model was found to perform better than the original Salama model for viscous oil. © 2020, Springer Nature Singapore Pte Ltd.