Skip to main content

Full text of "NASA Technical Reports Server (NTRS) 19910011176: A model for the space shuttle main engine high pressure oxidizer turbopump shaft seal system"

See other formats


f7 




NASA Technical Memorandum 103697 .. . 


///'3 7 

¥3 i 


A Model for the Space Shuttle Main Engine 
High Pressure Oxidizer Turbopump 
Shaft Seal System 


Daniel E. Paxson 
Lewis Research Center 
Cleveland, Ohio 


i 

i 




Prepared for the 

Second Annual Conference on Health Monitoring for 

Space Propulsion Systems 

sponsored by the University of Cincinnati 
Cincinnati, Ohio, November 14—15, 1990 


NASA 

v= (NASA-TM-101697) a MODEL FOR THfc SPACfc 

SHUTTl E MAIN ENGINE HI GH PRESSURE QXlDlZtR 
TilK^O D UMP SHAFT S r AL SYSTEM (NASA) 31 p 

CSCL 131 

G 3/ 3 7 


N9 1-20489 


Unci u s 
0001^09 


• - • \ 


? 



A Model for the Space Shuttle Main Engine 
High Pressure Oxidizer Turbopump Shaft Seal System 


Daniel E. Paxson 

National Aeronautics and Space Administration 
Lewis Research Center 
Cleveland, Ohio 44135 


Abstract 

A simple static model is presented which solves for the flow properties of pressure, temperature, and 
mass flow in the Space Shuttle Main Engine High Pressure Oxidizer Turbopump Shaft Seal System. 
This system includes the Primary and Secondary Turbine Seals, the Primary and Secondary Turbine 
Drains, the Helium Purge Seals and Feed Line, the Primary Oxygen Drain, and the Slinger/Labyrinth 
Oxygen Seal Pair. The model predicts the changes in flow variables that occur during and after 
failures of the various seals. Such information would be particularly useful in a post flight situation 
where processing of sensor information using this model could identify a particular seal that had 
experienced excessive wear. Most of the seals in the system are modeled using simple one 
dimensional equations which can be applied to almost any seal provided that the fluid is gaseous. A 
failures is modeled as an increase in the clearance between the shaft and the seal. Thus, the model 
does not attempt to predict how the failure process actually occurs (e.g. wear, seal crack initiation, 
etc.). The results presented herein were obtained using a FORTRAN implementation of the model 
running on a VAX computer. Solution for the seal system properties is obtained iteratively; however, 
a further simplified implementation (which does not include the Slinger/Labyrinth combination) has 
also been developed which provides fast and reasonable results for most engine operating conditions. 
Results from the model compare favorably with the limited redline data available. 


1 



Nomenclature 


C D 

Ci 
d . 
G. 
h . 

K 

K 

L . 
M, 
M 2 

iti 

N . 

P o 

P e 

R . 


31 

T 

1 o 

Y 

P 

P 


clearance between shaft and seal (radial) in inches 
discharge coefficient 
experimental coefficient 
diameter of shaft in inches 

geometric factor depending upon slinger dimensions 

fluid enthalpy in btu/lbm 

smooth side slip coefficient 

vaned side slip coefficient 

seal length in inches 

mach number at the end of the inlet nozzle 
mach number at the end of the channel 
flow rate in lbm/sec 
number of teeth in labyrinth seal 
pressure at origin or inlet of a seal in psi 
pressure seal exit in psi 

real gas constant for a specific gas in ft-lbf/lbm*°R 
radial position in inches 
flow resistance in sec 2 /in 5 

temperature at origin or inlet of a seal in degrees Rankine 

ratio of specific heats 

fluid viscocity in Reyns 

fluid density in Ibm/ft 3 

shaft rotational speed in rpm 


2 



Introduction 


Seal failures have been recognized as an important potential failure mode in reusable rocket engines 
[1,23]. Seals are critical for the operation of many major components and appear in many forms 
throughout the engine. However, despite their varied appearance many of these seals function in a 
similar fashion and thus lend themselves to a generalized representation. Unfortunately, to the 
author’s knowledge, no reliable model is readily available in the literature today. In an effort to 
eliminate this void, the present investigation was undertaken and resulted in the model to be 
described below. 

Before describing the model itself it is worthwhile to discuss the nature of failure modelling in general 
and its relation to the approach taken in this investigation. There are two basic approaches which 
seem to be espoused in the field. The first is a true first principles approach in which one asks, "what 
are the dynamics of the failure?" An attempt is made to actually link together the fluid mechanical 
and structural interactions which lead to a given component failure. This is a bold and difficult 
approach which has yet to yield fruitful results, however; if successful, it will give the most valuable 
information. The second approach proceeds along the lines of asking, "suppose a certain failure does 
occur, what are the effects on the rest of the engine?" This is a purely pragmatic line of thinking in 
which the only goal is detection and isolation of a failure. No consideration is given to its cause. It 
is this approach which was taken in the present work and which will now be discussed. 

A schematic of the High Pressure Oxidizer Turbopump (HPOTP) seal system on the Space Shuttle 
Main Engine (SSME) is shown in Figure 1. Generally, its purpose is to prevent the leakage of any 
fluids in down the shaft and in particular, to prevent the mixing of hot Hydrogen rich turbine gases 
on one side of the pump from combining with Oxygen from the other side. Also shown in the figure 
are the points at which pressure and temperature sensors reside in the actual HPOTP. There are 
five static seals , that is seals which have no moving parts, and a dynamic seal, often referred to as 
a slinger, which depends upon shaft rotation in order to function. Proceeding from right to left, the 
first two seals are referred to as the primary and secondary turbine seals. These separate the hot 
turbine gases from the Helium purge gases. The second set of seals are referred to collectively as 
the Purge seals. These serve to separate hot gas which has leaked through the turbine seals from 
gaseous oxygen which has leaked through the slinger/labyrinth seal combination on the left of the 
figure (often called the Primary Oxidizer Seal). Gaseous, inert Helium is forced through the purge 
seal pair from high pressure tanks on board the shuttle. This gas then mixes with the hot turbine gas 
and gaseous oxygen on either side of the seal pair and exits through separate drain lines which are 
also shown in Figure 1. The constant flow of Helium thus provides a barrier between the two gases 
(Hydrogen rich turbine gas and gaseous oxygen) which if combined would be extremely volatile. The 
labyrinth seal, in combination with the slinger, serves to stem the flow of oxygen leaking from the 
high pressure LOX pump. Like the other seals mentioned, it provides high resistance to through flow 
by maintaining a small clearance between the seal face and the shaft. The labyrinth differs from the 


3 



others however, in that it is composed of many thinner seal faces, or teeth, instead of just one thick 
face. Thus, there is a relatively small drop in pressure across each of the labyrinth teeth, but the 
accumulated pressure drop across the entire seal is quite large. Finally, the slinger seal is a dynamic 
seal (e.g. has moving parts) which serves two purposes. First, it restricts the flow. Secondly, it 
gasifies the liquid oxygen which does leak through by doing work on it. Many papers have been 
written on the slinger seal [4, 5, 6, 7] and details of the principles of operation will not be presented 
here. Suffice it to say that generally speaking, a slinger is conceptually the same as a pump in which 
the pressure gradient that it must overcome is too great and the fluid flows backward. The 
shortcoming of this analogy is that unlike most pumps, the slinger has the added complexity of a 
gas/liquid interface. 

Model Description 


Annular Seals 


The Turbine seals and the two seals which compose the Purge seal are all simple annular clearance 
seals thus, a general description of them was sought. It was found that they could be well modeled 
by considering each to be composed of a nearly isentropic entrance nozzle attached to a one 
dimensional channel with frictional losses through which perfect gases flowed. This is shown in 
Figure 2. Several implications are embedded in this model. First, the clearance of the seal is 
presumed to be much less than the radius of the shaft. Second, the fluid velocities associated with 
axial flow through the seal are assumed much greater than those in the circumferential direction 
caused by the rotation of the shaft. This is a reasonable assumption since seals typically operate at 
or near a choked condition, and the speed of sound in the gases is significantly larger than the 
characteristic circumferential speed which is the shaft angular rotation rate multiplied by the shaft 
radius. Thirdly, it is supposed that fluid velocities inside the seal proper are much greater than those 
outside so that the fluid on either side of the seal is nearly stagnant. Finally, it is assumed that heat 
transfer effects are negligible. That is, that the flow is adiabatic. The information necessary to 
calculate the flow through the seal includes the pressure drop across the seal, the seal clearance, and 
length, the fluid viscosity, ratio of specific heats, and real gas constant and the coefficient of discharge 
for the entrance nozzle. For the nozzle region, the equation for the flow rate may be written as 


rti = ndcC D P 0 


RT„ 


Mt 


(‘ * ^ M >) 


1/2 


( 1 ) 


This equation assumes that the nozzle flow is unchoked, however; it may be shown that, when the 
nozzle is joined to a constant area duct with friction, choking can occur only in the duct. In the duct 
region the relation between the entrance and exit mach numbers (regions 1 and 2 respectively in 


4 



Figure 2) is 


1/2 


1 

'll' 

+ Y+1 ln 

M?(2 * [r-l]M| 

_ 2.656 

trdcjiL 

Y 

Mj Ml 

2y 

M^(2 . [y-ljM?) 

c 

ifa 


( 2 ) 


where a friction factor corresponding to laminar boundary layer flow on a flat plat has been used due 
to the entry region type flow in the seal annulus. Also, since the flow is assumed adiabatic and the 
gases are assumed perfect, the stagnation temperature remains constant throughout the seal and, 
provided the flow is not choked, it is found that 



1 

Y 7 ! 


-1 + 


1 + 


2(y-l)M 


V2 


T+l P„ 


Kf I 


1/2 


( 3 ) 


These are three equations and three unknowns (Mj, M 2 , and ift) which may be solved 
simultaneously, however; due to their complexity their solution must be obtained iteratively. The 
procedure is as follows. A guess is made at rii and equation 2 is solved for Mj on the assumption 
that M 2 =1.0 (e.g. choked flow). Equation 3 is then solved for the pressure ratio which corresponds 
to choking. If the actual pressure ratio Pg/Po is less than or equal to the choking value then the 
actual pressure ratio is irrelevant and the value of M| obtained above may be used in equation 1 to 
find the new rti. If the actual pressure ratio is greater than the choking value then with P e /P 0 andrii 
known, 2 may be substituted into 3 and this equation may be solved for Mj using a technique such 
as the False Point method. With Mj known, equation 1 may then be used to find a new guess at rh. 
This process is repeated until the new and old mass flow values agree to within a specified tolerance. 
The above procedure is illustrated clearly using the flowchart shown in Figure 3. Comparisons of this 
model with data obtained experimentally for several different gases [8] are shown in Figures 4, 5, and 
6. The format of Figure 6 is different from the other two because in the experiment several different 
clearances were used as well as different pressure ratios. Information concerning the graphs can be 
found in Table 1. Also shown in Figure 6 are the predicted flow values obtained using another 
program called FLOWCAL described in [8]. As is clearly evident, the present model performed 
significantly better. The ordinate axis label SCFM refers to Standard Cubic Feet per Minute and is 
simply the mass flow per minute divided by the density of the gas at standard temperature and 
pressure. 

Labyrinth Seal 


5 


The Labyrinth seal, shown in Figure 7, was modeled according to an analysis performed by Martin 
[9]. Here it is assumed that each seal within the Labyrinth may be treated as an orifice (e.g. the 
length is negligible) and that the fluid stagnates completely on either side of the seal tooth. 
Furthermore, it is assumed that the pressure drop across any one tooth is not large. This is a valid 
assumption for the HPOTP labyrinth since it has seventeen teeth. The equation for the mass flow 
across the entire seal may be written as 


1 - 


( p 

r e 


rfi = irdcC 


D 




N - In 




1/2 


( 4 ) 


This is an established and accepted model and therefore no experimental data is presented here for 
verification. It is noted however, that modifications exist in the literature [10] which account for 
residual kinetic energy (non stagnant flow between teeth) and geometrical variations such as teeth 
which fit into grooves on the shaft instead of the smooth shaft type shown in Figure 7. These 
modifications were used in the present investigation. For reasons to be explained later it was 
necessary to solve this equation for the pressure ratio with the flow rate given. This was done by 
moving the right side of equation 4 to the left and solving for the roots of the resulting equation 
using the False Point method. 

Slinger Seal 

The Slinger seal was extremely difficult to model and it is noted at the outset that this aspect of the 
seal system model is probably the least accurate. At best it is claimed that reasonable qualitative 
results may be expected. An analysis by Voss [5] provided the starting point for the present model. 
For reference a schematic drawing of the slinger seal is shown in Figure 8. As shown by the arrows, 
the fluid flow path originates at the high pressure point just upstream the turbine side bearing set. 
As it travels radially outward along the smooth side of the slinger there is an increase in the pressure 
owing to the centrifugal force of rotation. On the other side of the slinger there is a set of vanes. 
Because of the vanes there is much less fluid slip than on the smooth side. Consequently, the 
centrifugal effects are multiplied considerably. If both sides of the slinger were filled with liquid it 
can be seen that for a given radial position the pressure on the vaned side would be less than on the 
smooth side. Furthermore, this difference increases as one moves closer to the shaft. This is the 
operating principle of the slinger, however; there is another important feature of this particular 
system. The vanes also "stir" the fluid like the whisks of a beater. Consequently, work is done on 
the fluid thereby raising its internal energy. When enough energy has been added, and the proper 
pressure conditions exist (e.g. at some radial point) the liquid oxygen will turn to vapor. When this 


6 



happens the density drops substantially, thereby greatly reducing the centrifugal effects. The vanes 
however, can continue to heat the gas. On the smooth side of the seal it is assumed that no work 
is done and that the flow is isentropic. The change in pressure due to the centrifugal force is 
expressed as 

P r,4r ' P r * k^p^rir (5) 

Here it is assumed that the change in radius is small so that, over the interval, the density may be 
assumed constant. The slip factor kj used in the present study was 0.5. To find conditions at the 
outer radius of the slinger, it was divided into 20 intervals. At each new radial position the pressure 
was calculated using equation 5 and the value of density and pressure from the previous position. 
A new density and temperature were then found using the new pressure and constant entropy. This 
was accomplished by utilizing a series of software routines known collectively as GASPLUS available 
on the VAX cluster at NASA [11]. These routines are designed to calculate the properties of many 
gases and liquids over a wide range of conditions. The process was repeated until the outside of the 
slinger was reached. On the vaned side equation 5 could again be used, however,; this time the 
direction of travel is inward and the value of the slip factor ky was much higher (e.g. less slip). In 
the present investigation the value was chosen as 0.98 if the fluid was liquid and 0.93 if it was vapor 
or saturated vapor. An energy balance on a "thin ring" control volume yields 

•Wr * 1>, * GC,££!(l -k.fr 3 Ar (6) 

This equation is based on the assumption that the work done per mass on the fluid flowing through 
the ring is equal to the work rate of the torque. A simple analysis of the torque induced on the 
slinger vanes is outlined in [12]. The procedure was thus to again divide the vaned side of the slinger 
into 35 intervals beginning at the outer edge. For each new radial position equations 5 and 6 were 
used to calculate a new pressure and enthalpy based on the density at the last position. Using the 
GASPLUS routines the new temperature and density were then found. This process was repeated 
until the shaft radius was reached. Of course, at some point the liquid would change to a vapor. 
When this happened the new ky for gas was then used to continue the calculations. For this 
calculation to proceed, the flow rate through the seal, and the state of the LOX on the slinger’s 
smooth side are needed. The result is the conditions at the base of the vaned side just upstream of 
the Labyrinth seal. Since the Slinger and the Labyrinth seal are in series (Figure 1) it is necessary 
that the flow through them be the same. This fact was used to solve for the flow through the 
Slinger/Labyrinth combination. A guess was made at the flow rate, ih and equations 5 and 6 were 
used to find the properties upstream of the Labyrinth seal. With the same flow rate and the known 
downstream pressure, equation 4 was also used to find the pressure upstream of the Labyrinth. If 
this pressure matched that calculated with equations 5 and 6 then the flow was solved. If not, a new 
guess (based on the magnitude of the pressure difference) was made and the cycle was repeated until 


7 


the pressures matched. Figures 9 and 10 show the Slinger/Labyrinth mass flow and exit temperature 
respectively as functions of exit pressure for two different Labyrinth seal clearances. The LOX 
conditions at the base of the slinger (inlet) as well as the stinger rpm were held constant for these 
calculations. The inlet conditions are presented for reference in Table 2 following the figures. Note 
that this is in contrast to the other seals discussed where the exit pressure was held constant and the 
inlet pressure was varied. The inlet information was obtained from the SSME Power Balance 
Model. Figure 9 shows an expected trend in that flow rate increases as the exit pressure is dropped 
and eventually chokes at very low exit pressures. One interesting aspect however, is fact that the 
calculations show that it is possible for the exit pressure to exceed the inlet pressure and still maintain 
positive through flow. In these cases the centrifugal effects on the smooth side remain the same and 
increase the pressure at the tip of the slinger above the inlet pressure at the base. In the low flow 
cases however, the work done per unit mass on the gas is large on the vaned side and the interface 
occurs near the slinger tip. With the gasification of the Oxygen, the centrifugal effect on the vaned 
side is substantially reduced and the pressure at the base of the slinger on the vaned side can remain 
higher than that at the inlet. It is noted here that the "waviness" of the data is a result of the rather 
large finite intervals into which the slinger was divided in order to calculate properties up and down 
the sides (e.g. the deltas in equations 5 and 6). Smaller divisions lead to better results but far more 
computing time. Figure 10 is presented because, unlike the annular seals described earlier in which 
the temperature changes vary little across them, this seal combination shows a fairly strong 
dependence of exit temperature on the flow rate. This, of course, makes intuitive sense for as the 
flow rate decreases more work is done to heat the fluid. 

Drain Lines 


The seal drain lines, of which there are three (see Figure 1), and the helium purge feed line were 
all modeled using a simple resistance type equation based upon an incompressible flow analysis. 
Some allowance was made for compressibility by using an average density for the flow in the drains 
based upon inlet and exit pressures and the inlet temperature (assumed constant throughout the 
drain). Although the flows in the seals were not, in fact incompressible, it was assumed that in the 
drain lines, where velocities and mach numbers are relatively low, the assumption was nearly valid. 
According to this analysis, the flow rate may be expressed in the form 


ifc 


P 2 - P 2 
o e 

2RT 0 8t 


(?) 


The resistances (St’s) for the various drain lines were obtained either through discussions with MSFC 
personnel [13] or by matching experimental flow rates with known pressure differences. 


8 



The System 


The final step in the modelling process was simply to put all of the various smaller models described 
above into one complete representation of the seal system. This was accomplished using iterations 
on the equations of continuity. Figure 11 is a redrawing of Figure 1, however; here the system has 
been broken into three sections. Also shown in the figure are the points at which the fluid properties 
and flow rates are calculated. The procedure begins by guessing at the pressure at point 3. With this 
pressure know, all of the flow quantities in region III may be found. Next a guess is made at the 
pressure at point 6. With this known, the flow through the oxygen drain (7) and the 
Labyrinth/SIinger may be found. The Pressure at point 4 may then be adjusted until the net flow into 
point 6 is zero. Hence, all flow quantites in region I become known. With the pressure at point 4 
known, the net flow into point 3 may then be found. If this is zero, the iteration is complete. If not, 
a new guess must be made for the pressure at point 3 (e.g. using the false point method) and the 
procedure is repeated. This process is done until the net flows into points 1, 4, and 6 are all zero. 


Results 

Although the flow rate was calculated at each of the prescribed points, it will not be presented in any 
of the results. Flow rate measurements are not currently practical within the small confines of the 
seal system passages and thus, it is not expected that they would be used in a health 
monitoring/failure detection capacity. It is noted however, that in the case of a Primary Turbine Seal 
failure of the magnitude (i.e. clearance) discussed here, the flow rate across the seal is large and 
amount to approximately 6% of the total flow through the HPOTP turbine. Figures 12 and 13 show 
the pressure and temperature values respectively at each of the calculation points for four different 
engine power levels, using the nominal seal clearances listed in Table 3. The calculation points as 
well as the points where conditions are assumed known are labeled in Figure 11. Also shown in 
Figures 12 and 13 are the points (marked with asterisks) where actual transducers are located (this 
is also shown in Figure 1). As mentioned earlier, the information for the known conditions was 
obtained from the Power Balance Model. The Primary Turbine Seal inlet conditions were taken to 
be those at the exit of the HPOTP turbine with 400° R subtracted from the temperature. This 
accounts for the mixing of the gas with some coolant flow from other parts of the pump [13]. Slinger 
inlet conditions were taken to be those just upstream of the number 4 bearing set. These were 
estimated as follows. The pressure was assumed that at the HPOTP LOX pump inlet, with an 
accounting for area change due to the location of the passage and a flat 20 psi added to the pressure 
to account for the flow over the bearings. Temperature was taken to be that at the inlet to the 
booster pump. Finally, the exit pressure for all of the drain lines was assumed to be 1 atmosphere 
or 14.7 psia. It is seen from these figures that the conditions stay relatively constant over most of the 
engine power levels. The one exception is the 65% level where a sharp drop in pressure is occurs 
at points 1 and 8 as well as a significant temperature drop at nearly all points. The pressure and 


9 


temperature values for a primary turbine seal failure and a secondary turbine seal failure are shown 
in Figures 14 and 15. The broken line at point 1 for the primary turbine seal failure indicates that 
the pressure exceeds the maximum value in the figure (it is 728.6 psia). Failures for all the seals were 
manifested by changing the clearance. The values used for failed seals are shown in Table 3. As 
expected, a primary turbine seal failure increases the flow through the primary seal drain and thus 
raises the pressure at points 1 and 2 considerably. With this pressure elevated, the flow through the 
secondary turbine seal is also increased thereby increasing the flow through the secondary drain and 
subsequently the pressure at point 3. The increased temperature at point 3 is the result of increased 
hot gas flow through the drain line. The secondary turbine seal failure results also display expected 
trends. Here, the pressure at point 1 drops due to the increased flow across the secondaiy turbine 
seal and hence reduced flow through the primary seal drain. Similarly, the increased secondary 
turbine seal flow increases flow through the secondary seal drain and therefore increases pressure and 
temperature at point 3. It is found that if both primary and secondary turbine seals failed, the redline 
pressure value of 100 psi is exceeded at point 3 (secondary seal drain cavity). Figures 16 and 17 show 
the results of purge seal failures. For either purge seal failure a large drop in the pressures at points 
4 and 5 are seen. This is explained by the fact that the resistance to flow is greatly reduced as the 
seal clearance is increased. Thus it takes less pressure difference across the seal to drive more flow. 
Again it is found that the pressure calculated at point 5 falls below the redline value for a failure of 
either seal. The remainder of the deviations from the nominal pressures and temperatures 
throughout the seal system for the purge seal failures may be explained using analogous arguments 
to those used for the turbine seal failures. Finally, the results of a'slinger/labyrinth seal failure are 
shown in figures 18 and 19. Since a failure of the slinger seal itself is extremely unlikely it was 
assumed here that failure was brought about by a change in the clearance of the labyrinth seal. It 
is seen in the figures that a pressure rise occurs at point 6 due to the increased flow through the 
drain line but that a significant pressure drop occurs at point 8. This drop is due to the fact that 
increased flow through the slinger decreases the work done per mass of fluid, which in turn moves 
the interface toward the shaft and hence allows the centrifugal effects to provide a greater pressure 
drop. Since less work is done on the fluid it is also discharged at a lower temperature which is 
reflected at point 6, 7, and 8 of Figure 19. Although the results presented above are in qualitative 
agreement with expectations, the lack of available experimental data makes them difficult to evaluate 
objectively. 

It is interesting to note that if the actual magnitude of the pressure and temperature changes 
associated with a given failure are ignored and instead only their direction (greater than or less than) 
away from the nominal values are considered, each failure seems to have a characteristic pattern. 
This result is illustrated in Tables 4 and 5. In these tables, the respective pressure and temperature 
deviations of each point from the nominal value (at RPL), are shown as either upward or downward 
facing arrows for positive or negative deviations, or as a dash indicating no change. These tables 
suggest that a potential failure detection and isolation scheme for this seal system could rely on more 
qualitative measurements and still be successful. 


10 



Another noteworthy aspect of the results was the reliability of the seal system as a whole. Recall that 
the main purpose of this system was to keep the Oxygen flow through the pump separate from the 
hot, Hydrogen rich gas flow through the turbine. From the results presented it appears that this 
separation is maintained despite the failure of any one of the seals. In fact, when further test were 
conducted it was found that potential mixing of the two gases exists only when both of the purge seals 
fail or the Helium supply pressure is reduced to values of under 100 psia. These are highly unlikely 
occurances. Thus, from a health monitoring/failure detection standpoint it would seem that 
monitoring these seal failures during flight is unnecessary. Since a failure is not catastrophic to the 
engine and, in fact, does not even seem to imply minor damage, it is enough to monitor the condition 
of the seals in a post flight mode using stored data. 


Conclusions 

A model of the HPOTP Shaft Seal System for the SSME was successfully developed. The model 
predicts the fluid properties and flow rates throughout this system for a number of conditions 
simulating failed seals. The results agree well with qualitative expectations and redline values but 
cannot be verified with actual data due to the lack thereof. The results indicate that each failure 
mode results in a unique distribution of properties throughout the seal system and can therfore be 
individually identified given the proper instrumentation. Furthermore, the detection process can be 
built on the principle of qualitative reasoning without the use of exact fluid property values. A 
simplified implementation of the model which does not include the slinger/labyrinth seal combination 
has been developed and will be useful for inclusion in a real time diagnostic system. 


11 


References 


1. System Controls Technology Inc., "Failure Modes Definition for Reusable Rocket Engine 
Diagnostic system," NASA Contract No. NAS3-25813, June, 1990. 

2. Perry, J. G., "Reusable Rocket Engine Turbopump Health Monitoring System," NASA 
Contract No. NAS3-25279, March, 1989. 

3. Martin Marietta Inc., "Failure Mode and Effects Analysis and Critical Items List," NASA 
Contract No. NAS8-30300, April, 1987. 

4. Reshotko, E. and Rosenthal, R. L., "Fluid Dynamic Considerations in the Design of Slinger 
Seals," Journal of the American Society of Lubrication Engineers, Vol. 24, July, 1968, pp. 
303-314. 

5. Voss, J. S., "Fluid Dynamic Analysis of the Space Shuttle Main Engine High Pressure 
Oxidizer Turbopump Slinger Seal," NASA University Grant No. NGT-01-002-099, August, 
1980. 

6. Thew, M. T., and Saunders, M. G., "The Hydrodynamic Disk Seal," Proc. 3rd Int. Conf. on 
Fluid Sealing, British Hydromechanics Research assoc., Paper H'-5, April, 1967. 

7. Due, H. F., "An Emperical Method for Calculating Radial Pressure Distribution on Rotating 
Disks," Journal of Engineering for Power, April, 1966, pp.188-197. 

8. Proctor, M. P., "Leakage Predictions for Rayleigh-Step Helium Purge Seals," NASA 
Technical Memorandum No. 101352, December, 1988. 

9. Martin, H. M., "Labyrinth Packings," Engineering, January, 1908, pp. 35-36. 

10. Vermes, G., "A Fluid Mechanics Approach to the Labyrinth Seal Leakage Problem," ASME 
Journal of Engineering for Power, April, 1961, pp. 161-169. 

11. Fowler, J. R., "GASPLUS User’s Manual," NASA Lewis Research Center, Cleveland, Ohio, 
August, 1988. 

12. Thew, M. T., "Further Experiments on the Hydrodynamic Disk Seal," Presented at Fourth 
Inti. Conf. on Fluid Sealing, Paper No. 39, April, 1969. 

13. Wilmer, G., Marshall Space Flight Center, Huntsville, Alabama, 1990. 


12 



Downstream Pressure, P e psia 
Upstream Temperature, T 0 °R 


Seal length, L in. 


Shaft diameter, d in. 


Radial clearance, c in. 


Discharge coefficient, C D 


Viscosity, p Reyns 


Ratio of specific heats, y 
Real gas constant, R ft-lbf/lbm-°R 


Figure 4 
Hydrogen 


Figure 5 
Air 


14.7 


14.7 


555.0 


530.0 


0.28 


2.0 


1.97 


6.0 


2.26 x 10' 3 


0.029 


0.6 


1.0 


1.21 x 10' 9 


2.76 x 10' 9 


1.39 


1.40 


772.5 


53.46 


Figure 6 
Helium 


530.0 

0.077 

1.968 


1.0 

2.89 x 10‘ 9 
1.66 
386.0 


Table 1 Annular Seal Conditions 


Inlet pressure, psia 

354.8 

Inlet temperature, °R 

191.1 

Inlet density, lbm/ft 3 

66.41 

Shaft Speed, rpm 

27210 

Number of Labyrinth teeth 

17 


Table 2 Labyrinth Seal Inlet Conditions 


13 














































Nominal Clearance 
(in.) 


Failed Clearance 
(in.) 


Primary Turbine Seal 

0.003 

0.015 

Secondary Turbine Seal 

0.003 

0.015 

Hot Gas Side Purge Seal 

0.0015 

0.015 

Oxygen Side Purge 

0.0015 

0.015 

Labyrinth 

0.005 

0.05 


Table 3 Seal Clearances 



1 

2 

3 

4 

§ 

6. 

fl 

8 

Power Level 

i 

1 

- 

- 

- 

I 

- 

1 

Primary Turbine Seal 

t 

t 

■ 

B 

fl 

B 

- 

- 

Secondary Turbine Seal 

1 

- 

t 

- 

- 

- 

- 

- 

Hot Gas Purge Seal 

- 

- 

t 

B 

B 

fl 

- 


Oxygen Purge Seal 

■ 

- 

D 

B 

B 

B 

- 

- 

Slinger/ Labyrinth 

- 

- 

■ 

B 

- 

t 

- 

1 


Table 4 Direction of Pressure Deviations for Various Seal Failures 


14 

































1 2 


3 


5 6 


8 


Power Level 

i 

1 

D 


- 

t 

D 

i 

Primary Turbine Seal 

- 

- 

D 


- 


■ 

m 

Secondary Turbine Seal 

- 

- 

D 


- 


■ 

■ 

Hot Gas Purge Seal 

- 

- 

D 

- 

- 


D 

■ 

Oxygen Purge Seal 

- 

- 

D 


- 


D 

■ 

Slinger/ Labyrinth 

- 

- 

■ 


- 


D 



Table 5 Direction of Temperature Deviations for Various Seal Failures 


©-Turbine Seals 
®=Purge Seals 
(©•Labyrinth Seal 
(S)=Sllnger Seal 



Figure 1. — Seal system schematic. 


15 























Figure 2. — Step seal model. 


16 





Figure 3.— Seal flowchart. 


17 










60 80 100 
d Across Seal (psi) 

ring seal leakage rate. 


8 






Measured Leakage Rate (SCFM) 

Figure 6. — Hefium purge seal leakage rate. 



Figure 7— Labyrinth seal schematic. 


19 



Mass Flow (Ibm/s) 



Pressure Drop Across Seal (psi) 


Rgure 9. — Slinger/labyrinth seal leakage rate. 


20 





Exit Temperature (deg. R) 



Figure 1 1 . — Seal system regions and calculation points. 


21 







300 



200 


100 


0 



1 2 * 3 * 4 5 * 6 7 * 8 


Pressure Point 


Figure 12. — Pressures throughout seal system for various thrust levels. 


22 



Temperature (Deg R) 


1,000 
800 
600 
400 
200 
0 

Figure 13. — Temperatures throughout seal system for various thrust levels. 



1 2 * 3 * 4 5 * 6 7 * 8 

Temperature Point 


23 




1 2 * 3 * 4 5 * 6 7 * 8 

Pressure Point 


Figure 14. — Pressures throughout seal system for primary and secondary 
turbine seal failures at engine RPL. 


24 




Temperature (Deg R) 



1 2 * 3 * 4 5 * 6 7 * 8 

Temperature Point 

Figure 15. — Temperatures throughout seal system for failed primary and 
secondary turbine seal failures at engine RPL. 


25 




1 2 * 3 * 4 5 * 6 7 * 8 

Pressure Point 


Figure 16 —Pressures throughout seal system for failed purge seals at 
engine RPL. 




Temperature (Deg R) 



Temperature Point 

Figure 17. — Temperatures throughout seal system for failed purge seals. 


27 






Temperature (Deg R) 



1 2 * 3 * 4 5 * 6 7 * 8 
Temperature Point 


Figure 19. — Temperatures throughout seal system for failed labyrinth seal 
at engine RPL. 


29 




NASA FORM 1626 OCT 86 


*For sale by the National Technical Information Service, Springfield, Virginia 22161