Zhang, Min (2017-12). Experimental Study of the Static and Dynamic Characteristics of a Long (L/D=0.65) Smooth Annular Seal Operating Under Two-Phase (Liquid/Gas) Conditions. Doctoral Dissertation.
This research documents the development of a 2-phase annular-seal stand (2PASS) in the Turbomachinery Laboratory at Texas A&M University to investigate the static and dynamic characteristics of annular seals operating under 2-phase flow conditions. The 2PASS is modified from an existing air-annular seal test stand. It uses either spargers or a specially designed mixer to make a 2-phase flow, comprised of air and silicone oil (PSF-5cSt). Smooth annular seals with (length L=57.785 mm and length-to-diameter ratio L/D=0.65) are tested in the centered position for three radial clearances (Cvr=0.140, 0.163, and 0.188 mm) with zero intentional pre-swirl. Due to the difficulties in making homogeneous mixtures over the gas-volume fraction (GVF) range from 10% to 92%, tests are performed at pure- and mainly-oil conditions (GVF=10%) and pure- and mainly-air conditions (92%=GVF=100%), respectively. Under pure- and mainly-oil conditions, tests are conducted at exit pressure Pve=6.9 bars, rotor speed ?=5, 7.5, 10, and 15 krpm, pressure drop PD=31, 37.9, and 48.3 bars, and inlet GVF=0%, 2%, 4%, 6%, 8%, and 10%. At pure- and mainly-air conditions, tests are performed at inlet pressure Pvi=62 bars, ?=10, 15, and 20 krpm, pressure ratio PR=0.57, 0.5, and 0.43, and inlet GVF=100%, 98%, 95%, and 92%. Leakage mass flow rate ??? and rotordynamic coefficients are measured, and the effects of changing inlet GVF, Cvr, PD (or PR), and ? are studied. Test results show that adding fine air bubbles into the oil flow does not significantly change ???, but remarkably influences the seal's rotordynamic performance. Increasing inlet GVF from zero to 10% generally decreases direct stiffness K when Cr=0.163 and 0.140 mm, but increases K when PD=31 and 37.9 bars and Cr=0.188 mm. As K drops to a large enough negative value, the seal stator's 1st natural frequency drops significantly, causing a sub-synchronous vibration (instability at the stator's 1st damped natural frequency), preventing further tests. Increasing inlet GVF from zero to 10% has little effect on effective damping Cveff and does not change the seal's stabilizing force when Cvr=0.188 mm, but generally increases Cveff and makes the seal more stabilizing when Cvr=0.163 and 0.140 mm. For all three clearances under pure- and mainly-air conditions, ??? drops slightly (by less than 6%) as inlet GVF decreases from 100% to 98%, and then increases (by about 45%) with further decreasing inlet GVF to 92%. The oil presence in the air stream significantly impacts the seal's rotordynamic characteristics. Direct dynamic stiffness KvO is frequency dependent and generally increases as O increases, especially after injecting oil. As inlet GVF decreases from 100% to 92%, KO generally decreases except that it increases as inlet GVF decreases from 100% to 98% in the following circumstances: (1) cases with PR=0.43 and Cr=0.188 mm, and (2) the case with PR=0.43, ?=20 krpm, and Cr=0.163 mm. Decreasing the seal's KO would decrease the critical speed of the rotor in a centrifugal compressor. As O increases, vCeff changes from negative (destabilizing) to positive (stabilizing) at Ovc (the cross-over frequency). Oc is of great interest to the system stability. Injecting oil into the air stream increases Ovc, making the seal less stabilizing. Increasing Cvr from 0.140 to 0.188 mm decreases Oc, making the seal more stabilizing. However, an immediate disadvantage is that ??? increases significantly (by about 58%) as Cr increases from 0.140 to 0.188 mm. A program developed by San Andr?s based on a bulk-flow model and the Moody friction model with isothermal and homogeneous-mixture assumptions produces predictions to compare with test results. For all three clearances at pure- and mainly-oil conditions, the program reasonably predicts ??? and Cveff. For most cases, K predictions are not close to test results. In a centrifugal pump, the poor agreement in K leads to prediction inaccuracy in the rotor's critical speed. For