Experimental Study of the Static and Dynamic Characteristics of a Long Smooth Seal With Two-Phase, Mainly Air Mixtures

Abstract

Listen

Translate article

A two-phase annular seal stand (2PASS) has been developed at the Turbomachinery Laboratory of Texas A&M University to measure the leakage and rotordynamic coefficients of division wall or balance-piston annular seals in centrifugal compressors. 2PASS was modified from an existing pure-air annular seal test rig. A special mixer has been designed to inject the oil into the compressed air, aiming to make a homogenous air-rich mixture. Test results are presented for a smooth seal with an inner diameter D of 89.306 mm, a radial clearance Cr of 0.188 mm, and a length-to-diameter ratio (L/D) of 0.65. The test fluid is a mixture of air and silicone oil (PSF-5cSt). Tests are conducted with inlet liquid volume fraction (LVF) = 0%, 2%, 5%, and 8%, shaft speed ω = 10, 15, and 20 krpm, and pressure ratio (PR) = 0.43, 0.5, and 0.57. The test seal is concentric with the shaft (centered), and the inlet pressure is 62.1 bar. Complex dynamic-stiffness coefficients are measured for the seal. The real parts are generally too dependent on excitation frequency Ω to be modeled by constant stiffness and virtual-mass coefficients. The direct real dynamic-stiffness coefficients are denoted as KΩ; the cross-coupled real dynamic-stiffness coefficients are denoted as kΩ. The imaginary parts of the dynamic-stiffness coefficients are modeled by frequency-independent direct C and cross-coupled c damping coefficients. Test results show that the leakage and rotordynamic coefficients are remarkable impacted by changes in inlet LVF. Leakage mass flow rate m˙ drops slightly as inlet LVF increases from zero to 2% and then increases with further increasing inlet LVF to 8%. As inlet LVF increases from zero to 8%, KΩ generally decreases except it increases as inlet LVF increases from zero to 2% when PR = 0.43. kΩ increases virtually with increasing inlet LVF from zero to 2%. As inlet LVF further increases to 8%, kΩ decreases or remains unchanged. C increases as inlet LVF increases; however, its rate of increase drops significantly at inlet LVF = 2%. Effective damping Ceff combines the stabilizing impact of C and the destabilizing impact of kΩ. Ceff is negative (destabilizing) for lower Ω values and becomes more destabilizing as inlet LVF increases from zero to 2%. It then becomes less destabilizing as inlet LVF is further increased to 8%. Measured m˙ and rotordynamic coefficients are compared with predictions from XLHseal_mix, a program developed by San Andrés (2011, “Rotordynamic Force Coefficients of Bubbly Mixture Annular Pressure Seals,” ASME J. Eng. Gas Turbines Power, 134(2), p. 022503) based on a bulk-flow model, using the Moody wall-friction model while assuming constant temperature and a homogenous mixture. Predicted m˙ values are close to measurements when inlet LVF = 0% and 2% and are smaller than test results by about 17% when inlet LVF = 5% and 8%. As with measurements, predicted m˙ drops slightly as inlet LVF increases from zero to 2% and then increases with increasing inlet LVF further to 8%. However, in the inlet LVF range of 2–8%, the predicted effects of inlet LVF on m˙ are weaker than measurements. XLHseal_mix poorly predicts KΩ in most test cases. For all test cases, predicted KΩ decreases as inlet LVF increases from zero to 8%. The increase of KΩ induced by increasing inlet LVF from zero to 2% at PR = 0.43 is not predicted. C is reasonably predicted, and predicted C values are consistently smaller than measured results by 14–34%. Both predicted and measured C increase as inlet LVF increases. kΩ and Ceff are predicted adequately at pure-air conditions, but not at most mainly air conditions. The significant increase of kΩ induced by changing inlet LVF from zero to 2% is predicted. As inlet LVF increases from 2% to 8%, predicted kΩ continues increasing versus that measured kΩ typically decreases. As with measurements, increasing inlet LVF from zero to 2% decreases the predicted negative values of Ceff, making the test seal more destabilizing. However, as inlet LVF increases further to 8%, the predicted negative values of Ceff drop versus measured values increase. For high inlet LVF values (5% and 8%), the predicted negative values of Ceff are smaller than measurements. So, the seal is more stabilizing than predicted for high inlet LVF cases.