Erosion prediction models are important tools for preventing costly failures that can happen due to solid particle erosion. Various models such as computational models, mechanistic models and correlations, and machine learning approaches have been developed and utilized by researchers for predicting the erosion rate. The mentioned models have been developed under certain conditions, therefore, there are limitations on the application of each model. In the present study, mechanistic models, a two-dimensional CFD-based model developed by the Erosion/Corrosion Research Center are compared to other models from the literature including DNV RP O501 correlation. Experimental erosion data for standard elbows obtained by ultrasonic measurements in literature are used to evaluate different models. Experiments cover a wide range of flow conditions with air, water, and sand, 50.8 mm, 76.2 mm, and 101.6 mm pipe sizes, and particle sizes from 20 μm to 300 μm. The accuracy of the models for various flow regimes is examined and their limitations are discussed. Moreover, to expand the examined flow conditions, CFD data for high-pressure gas-sand flows, high-viscosity liquid-solid flows, and large-diameter pipes are used to evaluate a modification to the mechanistic model.


Erosion, a mechanical process during which material is removed from the pipelines and other flow-containing equipment, can occur when solid particles such as sand are carried by the flow. Erosion is more critical when there is a change in the flow direction, such as particle-laden flows in elbows and tees.

While experimentation is a possible approach to obtain erosion rates, the conditions under which tests could be performed are limited in some respects. In the oil and gas applications, the pressure, temperature, and fluid properties commonly found in the field may not be achieved with the experimental facilities. Moreover, there is a restriction regarding the dimensions of the domain such as pipe size for achieving the desired velocities and measurable erosion rates. In addition, performing experiments for all possible conditions is not feasible. Under these circumstances, to avoid failure in the system, modeling and numerical approaches are required to predict the erosion rate for the current and future flow conditions. Various types of models are introduced in the literature for predicting erosion. Such models include correlations, mechanistic models, Computational Fluid Dynamics (CFD)-based models, and Machine Learning approaches.

This content is only available via PDF.
You can access this article if you purchase or spend a download.