Background
The objective of this proposed work was to determine the main contributing conditions that influence erosion and to improve the accuracy of erosion prediction computationally. Material erosion can occur from solid particles impacting a surface. In the oil and gas industry, this can occur during sand control for reservoir production or injection. Material erosion from solid particle impact is a complex process and can be predicted either experimentally or computationally, where each has its own advantages and limitations. Many studies have been accomplished to accurately determine the material penetration depth due to erosion; however, all of the relevant reference parameters at the surface of interest are not typically measured and accounted for when determining material erosion. These conditions are difficult to obtain using traditional experimental methods.
Material erosion prediction from computational modeling is desirable as it is a low-cost alternative to testing and can be used to simulate scenarios that cannot be replicated with current testing. However, current erosion models are semi-empirical and were developed based on a specific experimental test. The research element of this project was to advance the understanding and accuracy of how erosion is predicted both experimentally and computationally by measuring all reference parameters that are said to strongly influence erosion at the material surface for various flow conditions and material properties. Additionally, this project aims at demonstrating the ability to model erosion with an accuracy of ±50% when compared to the experimental test results based on varying flowing and erosive environmental conditions.
Approach
To meet this objective and to fill in the gaps with our current state of knowledge for erosion prediction, the particle and material breakdown effects were determined through a combination of material erosion impingement testing, particle image velocimetry (PIV) tests, and a computational fluid dynamics (CFD) simulation effort. Correlations were developed from both the impingement coupon tests and the PIV tests and were used to update and refine a computational erosion submodel. These tests were conducted for a sweep of particle concentrations, material hardness, and flow rates with both silicon carbide and quartz solid particles.
Figure 1: Through a sensitivity study on the new SwRI erosion submodel, the erosion rate was found to be most sensitive to particle impact angle and particle impact velocity in comparison to the other parameters that affect specific erosion. All simulation erosion rate results using the new SwRI erosion submodel were within the project’s objective accuracy of ±50%, where the obtained accuracy of the simulation results was within ±28%. Some of the experimental and computational results from this effort are depicted here.
Accomplishments
The mass loss for all test materials over time linearly increased with increasing test duration. This material mass loss increased with increasing flow rate as well as decreasing material hardness. The mass loss results obtained with the more fragile quartz particles were less in value compared against those obtained from the silicon carbide tests with the same corresponding flow conditions. The reduction in mass loss between the two particles is attributed to the particles breaking down greater than 10% of their initial value.
A new SwRI erosion submodel was developed that was based on regressions performed on each independent parameter that included inlet and slip velocity, particle diameter, Brinell material hardness, impact angle, and particle concentration. Through a sensitivity study on the new SwRI erosion submodel, the erosion rate was found to be most sensitive to particle impact angle and particle impact velocity in comparison to the other parameters that affect specific erosion. All simulation erosion rate results using the new SwRI erosion submodel were within the project’s objective accuracy of ±50%, where the obtained accuracy of the simulation results was within ±28%. Some of the experimental and computational results from this effort are depicted in Figure 1.