Wear and Dynamics Study of Interstage Seals in Electric Submersible Pumps

Abstract

Based on field data and lab test results, interstage seal wear in electric submersible pumps used in oil production is a major cause for system failures. Due to their unique wear resistance, tungsten carbide (WC) seals are the industry standard; however, due to the brittleness of WC, interstage seals may fracture during pump installation, transportation, or operation. Reinforced thermoplastics have been proposed as alternative material to avoid fracture failures. A component-level vertical test rig was built to evaluate the dynamic performance and wear progress of interstage seals. The test rig simulates ESP operating conditions. Seals of different designs and materials were tested under various working conditions, including different gas volume fractions (GVF). Test results for reinforced thermoplastic seals indicate that their wear rate is much higher than that observed for WC seals. Axial grooves did not decrease wear rate but did increase leakage flow rate. The component-level test results allowed comparison of the performance of seal materials and geometries, but extrapolating those results to predict seal wear rates over the working lifetime of ESPs was not possible. Thus, we introduced a numerical method that employs semi-empirical wear fluid models and a structure interaction approach to model the seal, shaft, and fluid system, considering that the seal response is dominated by both fluid forces from the thin film and structural forces from the shaft. The transient seal orbits and eccentricity are calculated from the unbalance response, while the wear rate induced by the sand particles and system dynamic response was modeled according to the dominant wear mechanisms, such as three-body abrasion and erosion. When benchmarking numerical predictions against experimental data, the wear rate numerical predictions showed similar trends when compared to experimental data for the 0% and 30% GVF cases. Adjusting the empirical coefficients included in the wear models by a factor of ~2x yields a good correlation between the experimental results and the predictive model. On the other hand, the 60% GVF predictions and experimental results do not follow a similar trend and cannot be reconciled by adjusting the wear models’ empirical coefficients, a discrepancy indicating that a phenomenon affecting the wear mechanism not included in the current modeling approach may be dominant, thus limiting the applicability of this methodology to low (~30%) GVF values.