In any rotating machinery, the geometry of the seal influences the extent of systemleakage. The path taken by the flow in the clearance volume is dependent on the seal and rotor profile. The clearance between a new "seal-rotor" combination is uniform except for small variations during manufacturing and assembly. With time this annular cross section undergoes further physical changes causing non-uniform flow in the annular volume. This azimuthally varying leakage through the seal-rotor annulus creates unbalanced forces on the rotor causing it to whirl. It is essential to identify the reasons for these unwanted forces. Velocity profiling of the clearance volume flow was performed by Morrison et al. (1992) using 3-D LDA measurements on annular and labyrinth seals operating with 50% dynamic eccentricities and a whirl ratio of one. However, this alone does not provide a complete matrix of data for the conditions prevailing in the clearance zone. Additional information of mean and instantaneous wall pressure distributions for 0%, 10%, 25% and 50% rotor dynamic eccentricity for whirl ratios of zero and one, with positive pre-swirl, no pre-swirl and negative pre-swirl conditions were measured by Robic (1999). The data collected showed that the pressure field on the seal walls reversed itself between the whirling and non-whirling conditions. As a continuance of the earlier works, the present effort investigates the effect of whirl ratio variation for a 50% eccentric smooth annular seal at a leakage Reynolds number of 24000. An attempt has been made to collect pressure data for negative whirl ratios also under similar test conditions. A seal test rig capable of handling different eccentricities and whirl ratios simultaneously was designed and constructed for this purpose. Mean and instantaneous wall pressure data were recorded for 50% eccentricity with whirl ratios between ? 1 for a rotor speed of 1800. For a rotor speed of 2700, whirl ratios tested were between ? 0.6 and for 3600 rotor speed, whirl ratios ranging between ? 0.5 were tested. From the collected data a detailed analysis of wall pressures along the seal surface is performed following the technique described by Winslow (1994) and Robic (1999).

Windback seals work similarly to labyrinth seals except for the effect of helical groove. These seals are essentially a tooth on stator or tooth on rotor labyrinth seal where the grooves are a continuous helical cut like a thread. Windback seals are used in centrifugal gas compressor to keep oil out of the gas face seal area. These face seals cannot be contaminated by oil. A purge gas is applied to the seal to help force the oil back into the bearing area. The windback seal should be designed to prevent any oil contamination into the supply plenum and to reduce purge gas leakage. The CFD simulations have been performed with the effect of clearance, tooth width, cavity shape, shaft rotation, eccentricity, and tooth location on the seal leakage performance and the flow field inside the seal. The leakage flow rate increases with increasing the pressure differential, rotor speed, radial clearance, cavity size, and shaft diameter and with decreasing the tooth width. The eccentricity has a minimal effect for the windback seal. From oil simulations, the windback seal with 25% rotor eccentricity has some of the journal bearing action and drives back flow into the gas plenum. However the windback seal can be used to force the oil back into the bearing side before starting the compressor by applying a purge gas flow since the positive axial velocity inside the cavity is larger than the negative axial velocity. m A Rw cav & / ? is constant for varying shaft rotation since the leakage flow rate for the windback seal increases linearly as the the rotor speed increases. The leakage flow rate for the windback seal increases as the groove size increases due to the pumping action of the windback seal. A windback seal design based upon the numerical simulations that minimize gas leakage and help prevent gas face seal oil contamination was optimized. The windback seal has two leakage flow paths. Since the leakage flow rate under teeth of windback seals is the same as for a similar geometry labyrinth seal, the flow under the teeth can be predicted by two-dimensional labyrinth seal analysis. An empirical model for the leakage rate through the cavity has been developed which fits the data with a standard deviation of 0.12.

Friction factor data are important for better prediction of leakage and rotordynamic coefficients of gas annular seals. A flat-plate test rig is used to determine friction factor of hole-pattern/honeycomb flat-plate surfaces representing annular seals. Three flat-plates, having a hole-pattern with hole diameter of 12.15 mm and hole depths of 0.9 mm, 1.9 mm, and 2.9 mm, are tested with air as the working medium. Air flow is produced between two surfaces, one having the hole-pattern roughness representing the hole-pattern seal and the other smooth, at the following three clearances of 0.254, 0.381, and 0.635 mm and three inlet pressures of 56, 70, and 84 bar with all possible pressure ratios at each configuration. The friction factor data are presented for all tested configurations, with description of the test rig and theory behind the calculations. The effect of hole diameter, hole depth, clearance, Reynolds number, and inlet pressure are analyzed, and friction factor models based on these parameters are calculated. Friction factor upset (an undesirable phenomenon making the test data non repeatable) is also explained. Dynamic pressure data are presented, measured from dynamic pressure probes located at both the hole-pattern plate and the smooth plates at different locations.