Abstract
Hydraulic seals are used as intershaft seals in aero-engines and consist of an oil filled trough on the higher speed shaft and a fin on the lower speed shaft that dips into the oil forming the seal. Rotation is imparted to the sealing fluid within the trough and, similar to a manometer in operation the liquid either side of the fin can be at different heights allowing the seal to withstand differential pressure. In normal operation hydraulic seals do not leak air but if the differential pressure becomes too high the seal will break down and leakage will occur.
There is limited published research relating to hydraulic seals and the accuracy and reliability of the existing design approaches based on analytical derivations is not fully known. This acknowledged need to improve the ability to develop accurate computational models of hydraulic seals provides context for the current study.
An approach to evaluate the maximum pressure capacity of a hydraulic seal is therefore introduced in this work. Building on previously published studies, this paper presents results of a 2D numerical study into the performance of a simplified hydraulic seal geometry. This paper reports a numerical CFD methodology based on an axisymmetric Volume-of-Fluid (VOF) method. In this study there is no oil feed into the trough.
Results are presented for a range of shaft speeds of 2000–8000 rpm for the high speed shaft and 1000–4000 rpm for the low speed shaft. Fin position within the trough was varied. A criteria for broken seal was developed. The CFD data shows that the seal can withstand higher pressure at higher shaft speed with the characteristic following the expected linear relationship between differential pressure and shaft speed squared. The seal could withstand a higher differential pressure if the fin was closer to the housing on the high pressure side with this being attributed to the secondary air flow in the cavity.
The average core velocity was compared to values obtained using different analytical approaches and it was found that one where core angular velocity is proportional to the area averaged rotational velocities of the housing and fin was the best match to CFD data.