AECL Research

Fluid SealingTechnology Unit

SEVERE SERVICE SEALING SOLUTIONS
AECL-11181
Presented at Seminar on Applied Tribology in Engineering
Toronto, Ontario, September 1994
by
R. Metcalfe and R. Wensel
REACTOR DEVELOPMENT
ENGINEERING TECHNOLOGIES DIVISION
Chalk River Laboratories
Chalk River, Ontario KOJ 1J 0
1994 September
AECL- 11181
Severe Service Sealing Solutions
R. Metcalfe and R. Wensel
Fluid Sealing Technology Unit
Atomic Energy of Canada Limited
Chalk River Laboratories, Ontario
Successful sealing usually requires much more than initial leak-
tightness. Friction and wear must also be acceptable, requiring a good
understanding oftribology at the sealing interface. This paper describes
various sealing solutions for severe service conditions. The CAN2A
and CAN8 rotary face seals use tungsten carbide against carbon-graphite
to achieve low leakage and long lifetime in nuclear main coolant pumps.
The smaller CAN6 seal successfully uses tungsten carbide against silicon
carbide in reactor water cleanup pump service. Where friction in
CANDU fuelling machine rams must be essentially zero, a hydrostatic
seal using two silicon carbide faces is the solution. In the NRU reactor
moderator pumps, where pressure is much lower, eccentric seals that
prevent boiling at the seal faces are giving excellent service. All these
rotary face seals rely on supplementary elastomer seals between their
parts. An integrated engineering approach to high performance sealing
with O-rings is described. This is epitomized in critical Space Shuttle
applications, but is increasingly being applied in CANDU plants. It
includes gland design, selection and qualification of material, quality
assurance, detection of defects and the effects of lubrication, surface
finish, squeeze, stretch and volume constraints. In conclusion, for the
severe service applications described, customized solutions have more
than paid for themselves by higher reliability, lower maintenance
requirements and reduced outage time.
Introduction
Although high performance sealing is
required in many places, two industries lead all
others in terms of their demandsnuclear and
aerospace. More than 25 years ago a group of
specialists was formed at AECL's Chalk River
Laboratories to pursue seal research, product
development and field application, resulting in
a vast improvement in the performance, safety,
reliability and maintainability of a wide variety
of fluid seals (Ross-Ross and Metcalfe, 1983).
This was necessary to resolve major
shortcomings of commercial seals, which were
evident in early CANDU nuclear plants. The
resulting technology has become widely used
throughout the sealing industry.
Rotary seals for main coolant pumps
(MCPs) were the first focus of R&D effort
because these pumps are the heart of any
nuclear plant (Earl, 1983). Their size, pressure
CANada Deuterium Uranium
and speed are at the leading edge of face seal
applications, and heavy water is neither the
easiest nor cheapest fluid to seal. Further
rotary seal applications followed for other
nuclear pumps and fuelling machines,
including pump seals for AECL's main
research reactor, NRU. Most recently, a third
generation seal design has been supplied for the
Bruce MCPs. The benefits from these
improvements are ongoing for the life of these
plants.
Elastomers, either supplementary or separate
from rotary seals, have also been vastly
improved in CANDU plants since their early
days. Original work established the
susceptibility of various commercially available
elastomers to radiation, compression set and
extrusion in hot water or humid air-ozone
environments (Wensel, 1977; Wensel and
Cotnam, 1978). Thus, preferred elastomers
could be selected for various static applications.
Later work concentrated on dynamic
performance, where information on stiffness,
damping, resilience, friction and fretting
damage for various elastomers and conditions
could then be applied to design and
maintenance of rotary seals (Wensel, 1985). In
an outgrowth of this technology, AECL made a
major contribution towards improving O-ring
seals for Space Shuttle boosters in the
aftermath of the Challenger accident (Wensel et
al., 1988). Following the successful re-design
of the case segment "field joints, AECL
became prime contractor to Lockheed-Aerojet
and NASA for seal design in the new
generation of "advanced" boosters, employing
numerous O-ring-sealed joints. This included
extensive testing of various elastomer
formulations to optimize the recipe for these
applications and to ensure consistent quality.
Similar methodology is used for selection,
specification and inspection of elastomer seals
for severe nuclear service (Wensel, 1993).
Face Seal Fundamentals
Rotary end face seals for severe service are
"pressure-balanced" to enhance lubrication.
Balancing is accomplished by stepping the so-
called "flexibly-mounted" seal ring to reduce
the hydraulic closing force. In Figure 1, the
flexibly-mounted "rotor" is sealed to the shaft
with an O-ring and runs against a solidly-
supported "stator." Its hydraulic balance ratio
is the ratio of the rear area of the rotor to the
seal face area, normally about 70%. Springs
pushing against the rear of the flexibly-
mounted ring are normally used to hold seal
faces closed whenever pressure alone is not
enough. I f the spring force is large this may
substantially increase the effective balance
ratio, especially at low operating pressures
where hydraulic forces are relatively low
(Mayer, 1977).
Face lubrication is provided by the leaking
fluid. The film between the faces is no more
than a few micrometers, and viscous shearing
of liquid can often generate sufficient heat for
boiling. This can cause dry rubbing and
wearout, unless the faces remain separated by a
film of gas. The boiling regime is particularly
difficult to design for because liquid seals are
generally axisymmetric and relatively narrow,
while gas seals have hydrodynamic features
and wide faces to provide enough "lift" to
avoid rubbing contact (Yasuna and Hughes,
1990).
Large seals for high pressure water have
been the focus in the nuclear industry, where
the hydrostatic effects of seal face deflection
govern their performance (Watson, 1963;
1965). Thus, liquid leaking between the seal
faces causes a separating force dependent on
the pressure gradient. When the faces are flat
and parallel, the pressure gradient is almost
linear (Curve 1, Figure 2) and the separating
force, represented by the area under the curve,
is independent of the face separation. Under
these conditions, hydrostatic lubrication with a
stable, "balanced" separation is not possible.
When the faces deflect to form a diverging gap
in the direction of leakage, the pressure
gradient falls below Curve 1of Figure 2 and
lubrication is reduced further, causing hard
rubbing contact near the outside of the seal
faces. However, when the seal faces deflect to
form a converging gap in the direction of
leakage, the pressure profile is represented by
Curve 2 of Figure 2 and becomes stable.
Smaller separation would move Curve 2
towards a greater separating force, and vice
versa. Therefore, full film lubrication is
achieved at a stable separation dependent on the
amount of convergence and the balance ratio,
whether the seal is rotating or not.
Many further aspects of liquid and gas seal
performance are explained in the literature,
covering tilting (Metcalfe and Brown, 1984),
eccentricity (Metcalfe et al., 1978),
misalignment (Kittmer and Metcalfe, 1980),
whirl (Etsion and Constantinescu, 1984), and
the response to various environmental
conditions and transients (Lebeck, 1991).
Success is usually a matter of providing
sufficient lubrication without too much leakage
penalty. Full film lubrication is often
impossible, and performance is then governed
by boundary lubrication. For severe service,
one or more limits must be pushed beyond its
generally accepted norm (Nau, 1980). There
are many different ways that rotary face seals
can and do fail (Metcalfe, 1976).
Bruce NGS Main Coolant Pump Seals
Earlier CANDU plants were designed with
smaller MCPs and at least one spare. With
Bruce-A, the bold precedent of no spare MCPs
was set. However, the pump manufacturer's
(Byron J ackson Canada) "SU" seal design
from Pickering, which was upscaled for Bruce,
proved to be too unreliable (Figure 3, at right).
This was when the earlier years of seal R&D
at Chalk River began to pay big dividends.
The seal's deflection characteristics were
analyzed by the newly available finite element
technique and found to be highly sensitive to
pressure and temperature changes (Metcalfe,
1976; Metcalfe et al., 1982). With this
ammunition, several innovative modifications
were quickly designed and tested. The aim
was to maintain a converging gap between the
sealing faces over as much of the operating
cycle as possible by re-shaping the parts and
their means of support, by matching the
stiffnesses and thermal expansions of the
materials, and by improving the flow of heat
away from the sealing interface. This became
the "CAN1" seal (Figure 3, at left), whose
parts were fully compatible with the existing
seal cartridges for Bruce-A. It saw commercial
service from 1976 until a few years ago, when
it was finally phased out in favour of the CAN2
design (Inset, Figure 3), which has a slightly
higher pressure rating (Metcalfe, 1979).
CAN2 seals began serving 600 MW
CANDU needs at Point Lepreau and
Wolsong-1 after extensive rig testing, followed
by in-pump testing at the pump manufacturer's
Toronto plant. Bruce-B and later, Darlington,
used this design from the start. Its reliability
has been excellentless than 1/2 day/unit/year
of forced outage at Bruce, and virtually none
elsewhere (Metcalfe and Burchett, 1987).
However, recent changes in service conditions
have measurably reduced seal lifetime at
Bruce-A (intermittent low pressure running,
which was not an original requirement for the
CAN2 seals). Therefore, further upgrading to
the CAN8 seal has begun, as described later.
CAN2A Seal for BWRs
By the early eighties, performance of the
CAN2 seals at Bruce was well established and
noticeably better than the pump manufacturers'
seals being used in US Boiling Water Reactors
(BWRs). Nine Mile Point Unit One (NMP-1)
was having particular difficulties because of
seals that were causing an average of one
forced outage per fuel cycle, with consequential
bearing failures. They contracted AECL to
develop the CAN2A seal (Figure 4)a CAN2
derivative that was optimized to withstand
extensive running at reactor head pressure
(<0.3 MPa) and to close off leakage if
cooling to the seal cavity was reduced. This
latter feature was firstly to protect the close-
clearance carbon bearing, and secondly to
withstand loss of cooling during emergency
"station blackout" conditions (Metcalfe, 1991).
The two "original" CAN2A seal cartridges
were installed in two of the five MCPs (termed
Reactor Recirculation Pumps in BWRs) at
NMP-1 in 1986. One was removed for
assessment after one fuel cycle and found to
have an extrapolated lifetime of more than ten
years. CAN2A seals were then installed in the
remaining MCPsall have performed perfectly
ever since. At the refuelling outage in 1993,
the second of the "original" (1986 vintage)
CAN2A seal cartridges was removed for
assessment. It was in excellent condition and
looked capable of 15 to 20 years of service
(Figure 5).
NMP-1 's sister plant, Oyster Creek, has
also installed CAN2A seals, beginning in
1990. Instead of replacing seals every fuelling
outage (and nevertheless suffering forced
outages) these plants have now extended seal
replacement plans to five fuel cyclesan
outstanding improvement. The savings to date
according to NMP-1: More than $35 million!
Because of station blackout concerns in the
US, CAN2A seals in a full-scale pump mock-
up were extensively tested for response to loss
of cooling while the pump slows to a stop, but
remains hot and pressurized (Figure 6). The
seal faces did not "pop-open" or give high
leakage under any circumstances, unless
already detectably deteriorated (Rhodes, 1992;
Green and Inch, 1992).
CAN8 Seal Development
Background. The CAN8 seal was
developed in 1990 for the Grand Gulf BWR as
a scaled-up and improved CAN2A. It is
shown in a seal cartridge for the Bruce plant in
Figure 7. It was installed in both Grand Gulf
pumps in 1991 and ran smoothly through the
fuel cycle. Both were then assessed while the
pumps were being disassembled for shaft
inspections. Their extrapolated lifetime was
estimated as four fuel cycles. They were re
built with further minor improvements, and
continue to perform reliably.
Features of the CAN8 seal that made it
suitable for Grand Gulf also make it suitable
for CANDUs. Parts are fully compatible from
Grand Gulf to Bruce, Darlington, Lepreau and
Wolsong-1. They can be adapted to fit
Sulzer-Bingham pumps at Embalse,
Cemavoda, Wolsong-2, -3, -4 or future
CANDUs.
The basic requirement in CANDUs is for
reliable lifetime exceeding five years. The
CAN2 seal does not quite achieve this. Its
three main drawbacks are addressed by the
CAN8 sealdamage to the stator holder
through spinning of the stator, higher wear rate
when subjected to varying conditions with low
pressures, and occasional displacement of its
U-cup during rapid transients.
The particular features that overcome these
and similar drawbacks of the pump
manufacturers' seals are described in detail as
follows (see Figure 8).
Reduced sensitivity to transients. A
major improvement achieved with the CAN8
design is its reduced sensitivity to transients,
shown schematically in Figure 8. This is
primarily responsible for its stability and long
lifetime under normal running, and its
capability for long-term back-up operation with
full pressure drop across a single stage. When
pressure or temperature in the seal gland is
changed, its seal faces remain relatively flat. If
they twisted and deformed, leakage would
change and, more importantly, the faces would
rub and wear more rapidly. The result is seen
in the minimal size of the hysteresis loop
(Figure 8) compared with previous designs.
On the stationary side of the seal assembly,
the much stiffer holder of the CAN2 seal was a
major improvement over the more flimsy
stainless steel back-up ring of the SU seal
(Metcalfe, 1985). It also allowed the stator to
be much smaller than in the SU seal, with
correspondingly less of the unpredictable
deformation (swelling, warpage) that occurs
with carbon-graphite. The CAN8 seal uses an
extremely stiff tungsten carbide stator support
in place of stainless steel. This isolates it from
the seal cartridge flanges, which warp when
service conditions change and severely affect
the pump manufacturers' seals (Martinson,
1980).
On the rotating side, the contact area
between the rotor and rotor support of the
CAN8 seal is roughly a factor of five less than
for SU or CAN2 seals. Thus, the CAN8 rotor
is supported mostly by water pressure. Being
made of very stiff tungsten carbide, it is
therefore relatively unaffected by twisting of
the rotor support, which is much less stiff.
Reduced wear at low pressure. The
CAN8 seal is much better for running at low
pressures than the CAN2 or pump
manufacturers' seals because of improved
cooling of the rotating parts. Its seal faces do
not heat-check.
When seals run at low pressure, less water
leaks between the faces and they axe more
inclined to rub against each other. More heat
can be generated, which can cause all the water
to vaporize, particularly in the upper stage
which vents to atmosphere.
In the CAN8 seal, the faces rub less because
the faces remain flatter. Any heat is rapidly
conducted to the back of the tungsten carbide
rotor because of its higher conductivity. A
secondary cooling flow behind the rotor
removes the heat. This flow is centrifugally
pumped through holes in the rotor support.
No spinning of the stator. To prevent
the CAN8 carbon-graphite stator from spinning
in its holder, as happens frequently with CAN2
seals, it is held gently around its outer
circumference by a composite elastomer-
stainless steel key ring with sixteen soft lugs.
This has been shown to be capable of torques
many times higher than would occur in service,
without causing damage to the seal. The
shapes of the elastomer lugs and the mating
slots in the stator have been carefully designed
to avoid stress concentrations, since carbon-
graphite is relatively weak and brittle.
Refurbishment. Only the stator, O-rings
and U-cups (Figure 7) must be replaced every
time when servicing the CAN8 seal. The rotor
may normally be used twice by reversing it. In
contrast with the CAN2 seal, the CAN8
tungsten carbide stator support is re-usable
without re-lapping, and will not scratch or
warp. Re-lapping of the seal flange or rotor
support faces is unnecessary for the CAN8
seal, unless to remove burrs caused by careless
handling. This seal is much less dependent on
their flatness than are other seals.
U-cup retention. The CAN8 seal is not
prone to its U-cups being displaced ("rolled"
out of their original seated position and unable
subsequently to function). This happens with
the SU seal through reverse pressurization,
(e.g., if the reactor system pressure drops
rapidly, the lower stage seal in the cartridge
may briefly be pressurized in reverse, which
displaces the U-cup). This happened more
than once with CAN2 seals during
commissioning of the four Bruce-B units.
Although BWRs were thought to be immune,
this also happened once at Oyster Creek
through a "not-to-be-repeated event."
Verification tests have demonstrated that the
CAN8 seal is not susceptible to this.
Loss of cooling. Temperature and
radiation-resistant, functionally qualified,
elastomer parts are supplied with the CAN8
seal. The ethylene propylene (EPDM)
compound that is used is uniquely formulated
to AECL's specification. It was chosen
through extensive testing to maximize
performance in hot water and to minimize long
term deterioration. The O-rings and U-cups are
quality-assured and batch-tested.
The material is qualified by demonstrating
that (in the form of a 3.5 mm cross-section
O-ring) it is capable of resisting extrusion
through a 0.38 mm gap for 18 hours at 290C
and 7.5 MPa. CAN8 elastomers, therefore,
are qualified for the worst case of complete loss
of cooling to a stationary pump, as could
happen during a "loss of cooling" or station
blackout scenario.
CAN6 SealReactor Water Cleanup
B ackground. In contrast to MCP
applications, seals for Reactor Water Cleanup
Pumps (RWCP) are required by their BWR
users to have very low leakage as well as long,
reliable lifetime. Size is only 65 mm diameter,
compared with 100 to 240 mm for MCPs.
Speed is 3600 rpm, there is no injection
cooling, and a single seal takes the full pressure
drop. MCPs are less than half this speed and
use at least two seals in tandem.
The CAN6 seal (Figure 9) was developed
under contract to five US utilities with
particular needs for improved lifetime and ease
of maintenance. In some cases these pumps
were the single greatest source of radiation
dose at the plant.
As an improvement over the conventional
seal it replaced, the CAN6 seal is a cartridge
design that is installed as a unit, rather than
piece by piece. This not only reduces
maintenance time, but ensures the cartridge is
correctly assembled in clean conditions on the
bench.
Cooling of the RWCP seal gland is primarily
by recirculation through an external cooler,
forced by a pumping ring built into the seal.
There is also a cooling water jacket, but this
alone does not keep the gland below 100C.
The seal faces are tungsten carbide against
silicon carbidetwo hard faces for added
abrasion resistance, since unlike CANDU
MCPs, there is no injection of clean water into
the seal gland. For integrity, the faces are
shrink-fitted inside and out with high strength
stainless steel. Unwanted deformations due to
pressure or temperature are prevented by the
symmetrical, reversible shape of the seal rings
and the way they are supported hydraulically.
The only deformations are due to thermal
gradients, designed to be self-relieving through
"twisting open" of the seal faces in response to
frictional heating between them.
Instead of the springs rotating, in this case
they are stationary in order to isolate the seal
from distortion or misalignment of the gland
ring, which is not changed when the CAN6
seal is installed. This puts the sliding O-ring
(0-4 in Figure 9) in the coolest possible
location to accommodate the inevitable tilt,
axial movement and wear. Details of O-ring
design for this application are given in a later
section.
CAN6 seals are currently operating in ten
pumps at five plants (see Figure 10). Of these,
all except the pumps at Grand Gulf are "hot-
leg" installations, where pumpage is at full
reactor temperature of almost 300C. At Grand
Gulf, the RWCPs are mounted in the "cool leg"
of the system, making recirculation cooling
superfluous.
The first two CAN6 seal installations were at
Hope Creek in mid-1990. Figure 10
summarizes the service record here and
elsewhere, showing just one seal failure (after
two years) due to normal operation.
Grand Gulf, the "cool-leg" installation, has
the best recordboth pumps have been
trouble-free for almost three years since
installation of CAN6 seals. Other plants have
all had some "consequential seal failures" due
to inadequate cooling, crud or bearing failures.
It is revealing to study these in more detail,
since there are many lessons to be learned
about avoiding similar problems in the future.
Cooling lessons learned. Cooling
problems have been most prevalent at River
Bend, where a cluster of seal failures occurred
in the fall of 1993 because heat transfer
surfaces had degraded through fouling.
Cooling was marginal previous to this, and
because there is no recirculation cooling
whenever the pump is stopped, standby
conditions caused steam-cutting erosion of the
faces of several of these seals. The start-up
procedures (venting, heat-up, timing of bearing
oil changes, etc.) were revised, and clean
coolers were installed. Subsequently, both
CAN6 seals at River Bend ran normally for
eight months, although their gland ring
temperatures were still around 120C. The
seals were replaced for preventive maintenance,
both to install larger external coolers and to
make piping changes to inject clean, cool water
into the seal cavity from an external source.
This cools the seal cavity regardless of whether
the pump is running or not, and if it fails,
recirculation cooling now takes over. There
have been no further problems.
The lesson for CAN6 seals is that to achieve
their potential they must be kept from boiling
across their faces at all times. This requires
that the running temperature in the seal cavity
be below 100C (i.e., gland ring temperature
below about 120C). Because seal gland
temperature will rise to essentially that of the
gland ring when the pump is stopped, hot
standby time must be minimized.
Other lessons learned from the field.
Bearing failures have impacted on the CAN6
seals at River Bend and Perry. Heat from a
worn bearing boosts the temperature of the seal
by conduction along the pump shaft. The
increased gland ring temperature can be
mistakenly interpreted as a seal problem, even
before the higher temperature has damaged the
seal. Any significant, relatively sudden and
systematic increase of temperature should
therefore be suspected as a bearing problem
unless there is also noticeable seal leakage.
Misalignment led to one aborted CAN6 seal
installation at River Bend and rubbing-galling
damage between the sleeve and spring holder
(Figure 9). This was surprising, since the
minimum clearance is elsewhere. However, it
was discovered that this gland ring and some
others have very loose spigoted fits into the
pump cover. Eccentricity is now checked
carefully after each gland ring has been bolted-
up. I f the gland ring or bore of the seal gland
is more than 0.1 mm eccentric to the pump
shaft, centering is called for by loosening and
re-tightening the bolts. Otherwise, there may
be difficulty installing the CAN6 seal cartridge,
and its internal parts may rub and gall.
Crud (dirty water) appears in the CAN6
service record (Figure 10) as the reason for
seals in both pumps at Hope Creek "failing" in
late 1990. In fact they were preventively
replaced following chemical cleaning of the
system. Normal crud levels in the reactor
water cleanup system do not seem to have
caused any problems in CAN6 seal
performance, even though none but River Bend
have installed an injection system.
At Hope Creek, custom handling jigs are
used to facilitate maintenance. At Grand Gulf,
Hope Creek and Limerick, a training pump
helps installation to be practised, unhindered by
space, protective clothing or other constraints.
These have helped avoid installation snags.
CAN6 service conclusions. The only
failure mode in almost thirty pump-years of
service of the CAN6 design has been
inadequate cooling to the seal cavity, where
boiling temperatures have led to steam-cutting
erosion of the seal faces, particularly during
standby conditions. Otherwise, there have
been no failures in normal service and no
wom-out or broken parts.
It has been crucial to follow-up closely, and
to ensure through good communication and
field service that each user learns not to repeat
another's problems.
The CAN6 design has proven itself robust
and durable, with service lifetime of two or
more years under normal conditions.
Fuelling Machine Ram Seals
In order to refuel CANDUs on power,
fuelling machines attach to each end of each
hot, pressurized fuel channel in turn, remove a
seal disc, then proceed to push fuel in one end
and out the other. The two concentric rams
doing this are driven through four ball screws,
all within the ram housing pressure boundary.
Their four drive shafts are each sealed by an
end face seal where they pass through this
pressure boundary.
For the fuelling machines in Pickering and
the 600 MW CANDUs, the severe requirement
is for seal torque to be no more than
0.6 N.mlow enough to achieve accurate
positioning of the rams. Seal gland conditions
are:
- 25 to 55C
- 0, 3 or 10 MPa (gauge)
-24, 36 or 246 rpm, reversible
Reliability is essential, as fuelling proceeds
round-the-clock, and the machines are needed
almost 40% of the time and "visit" almost a
thousand fuel channels per year (Ross-Ross et
al., 1975).
Many seal types were tested during early
development. The most acceptable was a six-
pocket-with-orifice hydrostatic design, similar
to those used in the first CANDU MCPs at
NPD (Nuclear Power Demonstration), then at
Pickering (Billington and Fitzsimmons, 1966;
Pothier and Rod, 1980). Its face materials
were bronze vs. carbide-coated stainless steel.
Leakage was about 2 L/min at full pressure,
and its main failure mode was dirt or erosion
affecting the orifice flow and de-stabilizing the
seal, thus causing face rubbing, friction and
wear. It was also difficult and expensive to
manufacture, with quality assurance of the
orifices, face coating, and the bellows for axial
movement being particular problems
Performance was tolerable throughout the
seventies and early eighties because each plant
kept a spare machine available, but the days of
generous spares came to an end in the mid
eightieshigher reliability was demanded.
A different type of hydrostatic seal
(Figure 11: CAN13 seal) was quickly
developed using analysis techniques and
materials not available to the early CANDU
designers. The bellows was substituted by an
O-ring of material with demonstrated capability
for the service. Silicon carbide was chosen for
the new seal faces based on extensive
hydrostatic seal testing in the seventies, when it
was found to be both resistant to erosion and
tolerant of moderate rubbing contact. The need
and means to hold it securely in compression
(by shrink-fitting) to prevent breakage had also
been understood.
Requirements for low torque and low speed
necessitated a hydrostatic seal; the extreme
range of pressure required a seal face that
would neither erode while on pressurized
standby, nor wear when rubbing at pressures
as low as atmospheric.
The flat-conical seal face configuration was
developed so that the flat region of the face
would rub lightly and not wear out-of-flat
while running at atmospheric pressure and
246 rpm. The conical region was optimized to
meet the torque requirement yet, for three
reasons, to have minimum leakage at the higher
pressures:
(1) to minimize the erosive flow,
(2) to minimize the amount of dirt
consequentially filtered out between the
seal faces,
(3) to reduce the 2 L/min leakage per seal
suffered with the previous seal design.
CAN 13 seals were first installed in
Pickering in 1988 after extensive testing in
laboratory rigs and fuelling machines. Full
commitment to change to this design was made
in 1989 and completed by 1992. There have
since been no failures of the 64 seals running in
these eight reactors, despite dirty operating
conditions in Pickering-B that have caused
many pump seals to fail. The 600 MW
CANDUs have similarly changed to the new
seals.
Savings due to installation of CAN 13 seals
at Pickering are estimated to be at least M$1 per
year, with about a third coming from
maintenance savings and two thirds from extra
kW-h of production. Extrapolating over the
expected lifetimes of the existing plants now
using this seal, the savings are of the order of
M$ 20. The economic viability of future
plants has also been much enhancedall this
coming from an original 1985 investment of
$53,230 for design, development, rig-testing
and supply of the first two seals!
NRU Eccentric Seal
The NRU research reactor at Chalk River
Laboratories has been the workhorse of
CANDU system development since it was built
in the fifties. The moderator (at atmospheric
pressure in its vessel) carries away the reactor
heat, pumped by seven Main Heavy-Water
Pumps. These operate at 1700 rpm and no
more than 0.35 MPa, with frequent stop-
starts. Requirements for the seals are less than
0.2 mL/min leakage, high integrity and a
reliable lifetime of seven years.
The seal used for these pumps until 1992
was difficult to install and often exceeded the
desired leakage. Spare parts were becoming
unobtainable, and its lack of a back-up seal
compromised its integrity, as seen when
565 kg of tritiated heavy water, worth
~$250 000, spilled into the Ottawa River in
1988 following a pump bearing failure.
Although the activity of this spill (4xl014Bq
total) had been diluted to less than 2% of the
maximum in Canadian Drinking Water
Guidelines by the time it reached the next
community downstream, public concerns gave
further impetus to system improvements,
including better seals (Champ et al., 1989).
Because boiling at the seal faces is a severe
problem at low pressure, AECL's patented
eccentric concept was applied (Figure 12)
(Metcalfe, 1993). This idea traces its origins to
elliptical seals developed by AECL for use as
back-up seals at NPD, Douglas Point,
KANUPP, RAPP and Pickering (Vilim and
Zeller, 1963). Later, the same sweeping action
across the seal face was adapted into an
eccentric seal by Hayward Tyler, who supplied
thus-equipped MCPs for the Caorso BWR
(Italy). Success came after contract assistance
by AECL to control deformations, which had
been a similar problem for elliptical seals at
high pressures.
Recent testing at Chalk River has
demonstrated that eccentric seals are suitable
for CANDU MCP service, with deformation
no longer a problem. They are also simple to
manufacture.
The eccentric seeds for NRU are designed
into a single-unit cartridge to facilitate
installation (Figure 13). They are arranged to
expose the outboard seal to full pressure
initially, with the inboard seal providing
complete back-up in the event of outboard seal
failure. This outboard "main seal," works as
the primary seal while the inboard "backup
seal remains flooded with water at full
pressure on both sides (Metcalfe et al. 1994).
Material and geometric details of both seals
in the cartridge are given in Table 1. Each
stationary seal ring (stator) is axisymmetric and
eccentric. It is shaped to give a geometric
balance ratio of 0.75. Adding the spring load,
the effective balance ratio exceeds unity for
sealing pressures below 0.26 MPa.
Without eccentricity, such seals in NRU
would operate under boundary lubrication
conditions with danger of boiling between the
faces. With eccentricity, the faces are
hydrodynamically lubricated and cooled under
all operating conditions. Extra cooling caused
by the eccentricity was calculated to be about
five times the cooling from all other effects
(i.e., the temperature of the seal faces above
ambient is five times less than for an otherwise
similar concentric seal).
The seal cartridge was tested as shown in
Figure 13. A short-term test evaluated its
general suitability, and the leakage and friction
characteristics of each seal stage in the cartridge
in response to pressure and temperature
changes, then a long-term test verified the
reliability of the final design.
The short-term tests included pressures from
0.04 to 0.4 MPa; rotational speeds of 500,
1000 and 1700 rpm; temperatures between 24
and 70C; and spring load between 85 and 135
N. Leakage was less than 0.1 mL/min for all
cases, and power requirement variations were
less than 0.2 kW over and above the drive
power for fully lubricated conditions (~0.5 kW
for two seals and bearings). The sealing faces
after tests were seen to be in excellent shape,
with barely measurable wear.
The long-term test that followed was a
2300-hour test, including stop-starts and
intermittent axial movements. Mean leak rate
was less than 0.05 mL/min, and power
remained low, as before. Post-test examination
showed the faces of both seals to be in good
condition. Carbon stator thicknesses were
measured for overall wear0.025 mm for the
main seal and 0.013 mm for the back-up,
easily meeting the requirement for seven years
of extrapolated seal lifetime. Carbide rotors
were barely worn.
The first eccentric seal cartridge was
installed in the NRU pumps in November 1992
and continues to operate successfully. Others
are being installed in the remaining pumps.
Table 1: Material and Geometric Details of NRU Seal for Main Heavy-Water Pumps
DETAILS OF NRU SEAL
Seal Face Material Sealing Face Details
Part Type Density
(kg/m3)
Thermal
Conductivity
(W/m.K)
Specific
Heat
(J /kg.K)
Outer
dia.
(mm)
Inner
dia.
(mm)
Balance
dia.
(mm)
Stator
Eccentricity
(mm)
Spring
Force
(N)
Rotor Tungsten
Carbide
14900 90 180
102 93 95. 3 2. 4 89
Stator Carbon
Graphite
1400 20 750
O-Ring Seal Fundamentals
There is much more to sealing than simply
squeezing an O-ring according to handbook
guidelines, as was dramatically pointed out to
the world when the Challenger exploded in
J anuary 1986 (Metcalfe et al., 1989).
The design of O-ring sealed joints is always
a compromise, but significant improvements
are usually feasible if, from the outset, the seal
design is integrated with the associated parts
not added as an afterthought, transcribed or
scaled from another application.
For optimal performance, the priorities and
trade-offs required for design decisions must
be well understood. Fundamental
understanding of the sealing mechanism is
crucial (Metcalfe and Wensel, 1994).
Sealing is achieved when the leakage paths
remaining between an O-ring and the
counterface against which it is pressed are
small enough to allow insignificant leakage.
There must therefore be enough force to cause
the relatively soft O-ring material to conform to
the topography of the counterface (i.e., to in
fill the machining marks, scratches and other
roughness). Initially, this depends on the
amount by which the O-ring is squeezed from
its original shape, which becomes flattened
across each sealing "footprint," creating contact
force (Kuran et al., 1994).
O-rings are able to transmit hydrostatic
pressure independently of their deformation,
since the hydrostatic component of stress on an
O-ring creates no deformation (elastomers are
essentially incompressible). Typical deformed
shapes are usually the result of pressure
variations of about 0.7 to 1.4 MPa (100 to
200 psi). In higher pressure applications,
therefore, O-rings act almost as fluid-filled
bags. This emphasizes their difference from
conventional gaskets (e.g., spiral-wound),
which rely on large clamping forces and require
heavy flanges.
WTien hydraulic pressure is applied to the
upstream side of an O-ring, this may either
increase or reduce the contact force (contact
pressure), depending on whether the
arrangement promotes "pressure-assist" or not.
This is a crucial consideration for soft seals in
general. Too much pressure-assist can cause
friction problems, for example during sliding
of piston-type seals. On the other hand,
negative pressure-assist inevitably leads to
leakage at higher pressures and is generally
undesirable unless the sealed pressure is
significantly lower than the initial contact
pressure.
The wedge-shaped seal in Figure 14
illustrates pressure-assist behaviour. This will
not seal well if installed as shown in
Figure 14b, whereas when installed the
opposite way, as in Figure 14a, the greater
squeeze at the upstream end of the sealing
footprint acts to reduce penetration of hydraulic
pressure and thus increase the contact force and
sealing effectiveness.
Space Shuttle Booster O-Rings
Elastomeric O-rings are suitable for most
quasi-static joints at temperatures below about
150C. They continue to be the most suitable
option for Space Shuttle boosters (SRMs,
Solid Rocket Motors), provided the hot gas
inside is kept away from the joints!
The contrast between the field joints (those
assembled at the launch site) of the "Re
designed" SRM (RSRM) and those of the next
generation ("Advanced") ASRM, now in
development, illustrates how much
improvement a new, integrated approach can
give compared with even the most thorough
"afterthought" (Metcalfe and Wensel, 1994):
(1) The RSRM joint (Figure 15) required
structural changethe addition of a
"capture feature" to the upper half of the
jointto reduce the pre-Challenger gap
opening problem during ignition
transients. Many new case sections had
to be manufactured to implement this.
(2) O-ring squeeze was increased for the
RSRM by using a larger cross-section in
the same depth, but wider, groove. This
made the assembly procedures more
problematic because of greater friction
and the chance of cutting the O-rings,
which then fuelled a thorough revamping
of the post-assembly leak-checking
procedures and their efficacy.
(3) Various elastomers, manufacturing
processes and lubricants were
investigated for the RSRM in order to
boost resilience and subjugate defects
compared with the original SRM.
However, the manufacturing process of
splicing ground cord stock into finished
O-rings presented such a restriction that
only the defect detection and acceptance
criteria were significantly changed.
Heaters therefore became an RSRM
requirement for cold launch conditions.
By comparison, there are fewer field joints
in each ASRM, and they have been designed
(Figure 15) with the following sealing
considerations in mind:
(1)The ASRM case structure is designed
for essentially zero gap opening at all the
O-rings.
(2) High squeeze is used without creating
assembly problems because of the face
seal arrangement.
(3) The elastomer formulation is
considerably improved and the O-rings
are net-moulded at their full size (not
spliced from smaller pieces).
Another ASRM joint where vast
improvement over the "handbook" design was
possible is the igniter-to-case joint, shown in
Figure 16. This is a static application. Since
the only significant movement is during
assembly, it is desirable to create high contact
forces in the sealing footprints. It is also
desirable that these be as wide as possible.
High squeeze and large O-ring cross-section
are therefore two important design goals.
As in the field joints, the basic shape of this
joint is essentially fixed by the structural
requirement for zero gap opening. Positioning
the seals in a conventional bore seal
arrangement (Figure 16a) raises concerns for
damage to a critical sealing surface or an O-ring
during assembly. The sealing bore of the
uppermost member could easily be scratched
due to metal-to-metal contact as it slides across
the upper lip of the lower member. This is
unlikely to be noticed during assembly due to
the large size and weight of the components.
Such scratches in the cross-footprint direction
could cause leakage. O-ring damage could
also occur during assembly because of the
sliding distance, and particularly when passing
a leak check port. High squeeze exacerbates
this.
Considering the effects of machining
tolerances, including out-of-roundness and
eccentricity of these two large-diameter
components, the squeeze range for the
conventional bore seals will be about 15%.
This means that if for assembly reasons 25%
squeeze is not to be exceeded, then at the
opposite extreme of tolerances the squeeze will
only be 10%. This range is too wide.
In considering potential accident conditions,
a conventional design would incorporate two
identical seals having critical sealing surfaces
on the same member, directly in line with one
another. These are prime candidates for
common-mode failure. For example, should a
leak path be established by erosion of the first
seal by a jet of high temperature gas, erosion of
the second seal could be quick to follow.
Recognizing these problems of the bore seal
design, and returning to the original goals of
high squeeze and large footprint, it is clear that
this conventional arrangement is marginal at
best. Neither can the cross-sectional size of the
O-rings simply be increased to improve the
situation. Available axial space is already used
up by:
(1) lead-in chamfers for assembly,
(2) groove width (to provide sufficient axial
space to prevent overfill of the groove by
the squeezed O-rings),
(3) the lip on the uppermost edge of the
uppermost groove (to ensure that it will
not be damaged by bending or breaking
if bumped).
By moving the O-rings into the comers
(Figure 16b), much more axial space becomes
available for the sealsthe lead-in chamfers
become part of the seal cavities, the lip above
the uppermost seal no longer exists and the
axial lengths of the grooves become less
(relative to the O-ring sectional diameter).
Much larger seals can thus be accommodated in
the given spacein fact, in addition to the 30%
increase in seal size incorporated in the
alternative design shown, the axial length and
weight of the joint is reduced.
The comer seal arrangement allows much
greater squeeze to be applied to the O-rings,
since this comes from axial compression rather
than radial interference. It is also much better
controlled, being governed by short axial
dimensions and tolerances rather than large
diameters.
The potential for common-mode failure has
been reduced by the alternative design, which
avoids an "in-line" arrangement. Damage to
the lower seal due to "jetting" of hot gases past
a leak in the first seal is much less likely
because of the tortuous path. The potential for
assembly damage due to sliding, or to the leak
check port, is eliminated.
From the perspective of transient response to
sudden pressurization at SRM ignition, the
comer seals have a number of additional
advantages over the rectangular groove bore
seal arrangement. The first is "diagonal
sealing." The "inlet" to each seal cavity is
diagonally opposite the "outlet." On sudden
pressurization then, there is a larger pressure
force urging the O-ring to move towards, and
thus seal, the outlet. With a bore seal
arrangement, such "pressure-assistance" can be
delayed, particularly when the O-ring is pre-set
to the upstream edge of its groove (as is the
case in this application because a high pressure
leak check is conducted prior to flight). A
further advantage is that once pressure is
established, the contact force in the footprint
region is magnified due to the acute angle, over
that for rectangular grooves. High contact
force means better in-filling of asperities and
flaws and better "bridging" of contaminants.
Other benefits of the alternative design are:
(1) It eliminates clearance at the lower of the
two seals, thus minimizing any concerns
over possible extrusion.
(2) Critical sealing surfaces are more
accessible, both for initial machining and
for refurbishment (ASRM components
are intended for multiple use).
(3) Internal stresses in the O-ring are
lessened; triangular confinement avoids
the damaging stresses of high squeeze in
an unconstrained rectangular groove.
Looked at as a whole, many benefits accrue
from the comer seal design. However, parts
must be carefully toleranced to avoid damage
by exceeding 100% fill. As stated previously,
elastomers are essentially incompressible;
however, they may swell, and their thermal
expansion is -2x10"^/C (i.e., an order of
magnitude greater than most metals).
Pump Seal O-Rings
Similar improvements are obtainable in
many other sealing applications through open-
minded exploration of the geometrical options,
and adherence to basic seal design principles
(Metcalfe et al., 1992). A good example is
given by the CAN6 seal for reactor water
cleanup pumps (Figure 9), whose service
history was described earlier.
These required less radical treatment than the
Space Shuttle O-ring examples, but each O-ring
is tailored to its specific function. They were
designed for high reliability, not for particularly
high leak-tightness. In the CAN6 seal, slight
additional leakage from the O-rings is not
significant compared with water leaking
between the rotating seal faces.
Design parameters for the seven O-rings that
supplement the rotary face seal are shown in
Figure 9. Note that a wide range of squeezes
is used. For good sealing, the higher the
squeeze the better. The trade-off, however, is
friction, which is a prime consideration for the
axially moving O-ring 0-4.
By contrast, for static face-type O-rings such
as O-l, where no sliding occurs either during
assembly or in service, high squeeze can be
used. The upper limit then depends on factors
such as tear strength at the service temperature,
since the O-ring will split apart if squeezed too
highly in a wide, rectangular groove (Stevens,
1989).
O-ring 0-1 has negative stretch
(compression), which creates elastic
interference at its outside diameter to hold it in
place; otherwise, assembly would be very
difficult. The principle of using stretch to
locate an O-ring is enhanced for 0-6 and -7 by
using a 10 degree slope (semi-dovetail). Only
minimal squeeze is needed for these, since they
provide only back-up sealing.
O-ring 0-5 is stretched 10% and has zero
squeeze. This is because the CAN6 seal
cartridge is inserted as a unit. There must be
minimal friction as 0-5 slides into place, and it
must be stretched tightly or it could fall out of
position. Sealing relies on the strong pressure-
assist behaviour induced by the "diagonal
sealing" arrangement, whereby any leakage
must pass from one comer to the diagonally
opposite comer of the space that 0-5 occupies.
Optimum Elastomer Selection
Elastomer Characteristics. Elastomer
seal "compounds" are intermittently produced
in relatively small batches, compared to most
industrial products. They are made by adding a
variety of ingredients to a base elastomer stock.
The choice of ingredients and their proportions
govern the physical, chemical, and functional
properties of the finished product. Since there
are many base elastomers and many additives,
the variety of compounds having different
properties is often overwhelming. Industry
wide standardization by formula does not exist,
nor is it feasible.
Being organic, elastomers are subject to
deterioration with time, temperature, fluid
contact, and other environmental influences.
To select the best elastomer compound for a
particular set of service requirements,
compound-specific testing is required, with
careful choice of performance criteria and test
methodology (Wensel, 1993).
Readily measured physical properties, like
hardness and tensile strength, have traditionally
been favoured in the selection process, rather
than functional performance like (for seals)
leak-tightness and extrusion resistance in actual
service conditions. There is often little
correlation.
Hence, functionally related testing
procedures must be used if high performance is
needed. They must be based on the particular
conditions under which the O-ring will or could
be required to serve. These "functional tests"
must cover all relevant effects of the actual
service conditions, including accelerated
lifetime testing (Wensel, 1985). Each test
should reproduce a type of failure encountered
or expected in actual service.
Although functional tests must provide the
basis of the methodology for selection of
compounds for severe service, readily
measured properties can then help to
"fingerprint" the specific compound, once
selected, to ensure that subsequent batches are
the same.
Space Shuttle O-ring selection. As an
example of an extremely comprehensive
process to derive the best fluorocarbon
compound for all O-rings of the ASRM,
thirteen measurable, function-related properties
and their corresponding test methods were
devised that encompassed all the service
requirements for these seals (see rows and
columns of Figure 17). Next, weighting
factors were estimated for each "box" of this
matrix, based on the expected influences of
each property on each sealing requirement.
Trade-offs were appraised and quantified
during this process.
The thirteen properties of sixteen different
candidate compounds were then measured.
Each formulation and manufacturing process
was strictly controlled. The quantifiable
results, analyzed to ensure statistical
significance, were then used with the pre-
established weighting factors to select the
optimum compound for ASRM service. They
also demonstrated that most of the sixteen
compounds could be considered "qualified" for
the intended service, but with reduced margins.
Next, this optimum compound was
evaluated for variations, especially batch-to-
batch, to establish rejection limits for quality
assurance. Three functional properties
(compression-recovery, hot jet resistance, high
temperature extrusion resistance) and eight
physical/chemical properties (specific gravity,
tensile strength, elongation, modulus,
hardness, compression set, rheometry,
volumetric swell) were found sufficient for this
purpose. Together, these properties and their
respective rejection limits served to
"fingerprint" the compound.
Nuclear O-ring selection. For nuclear
applications, ethylene propylene (EPDM) is
often the preferred basic elastomer class due to
its resistance to hot water and aging. In its
formulation, there are trade-offs such as
extrusion resistance gained at the expense of
thermal aging. Selecting the optimum
commercially available compound requires both
testing and application-specific prioritizing of
performance requirements (Wensel 1994).
Figure 18 illustrates the wide range of
performance that occurs. In this case, three
EPDMs are compared on the basis of thermal
aging resistance when used as static seals in hot
water. The range of failure times, particularly
at the higher temperature condition, is striking:
37 weeks for the best and only 2 weeks for the
worst. Similarly, extrusion resistance tests in
hot water for loss of cooling in a typical nuclear
plant (Figure 19) give a wide range of
performance between EPDMs. One compound
resists extrusion at gaps four to five times those
of another, although in common industrial
practice such materials are often regarded as
equivalent. There is clearly a strong need to be
compound-specific.
Specific testing to qualify elastomer seals in
MCPs for station blackout conditions has been
performed for several US PWRs (Pressurized
Water Reactors). The problem is that because
no external power is available for an extended
period to drive pumps for cooling, the MCP
seals can reach main coolant temperatures of
>250C (Rhodes et al., 1987). Seal leakage is
the safety concern, since this could uncover the
reactor core. There are two possibilities, one is
leakage directly past the elastomer seals, the
other, through opening of the end face seals as
a consequence of "binding" of the elastomer
seals when the shaft moves axially.
The qualification process defined potential
worst case events, then tested to determine
whether the elastomer (i) maintained a seal, and
(ii) did not "bind" excessively to the shaft.
This included accelerated aging of the elastomer
seals.
Figure 20 is an example of a test matrix to
assess integrity of the elastomer seals in a
particular MCP seal under station blackout
conditions. Analysis (thermal-hydraulic,
structural, etc.) provided predicted "best-
estimate" pressure, temperature and gap
conditions for each of the many O-ring seals in
this assembly. From this, four possible
"worst-case plus margin" combinations of
pressure, temperature, gap and pre-history
(thermal and radiation aging) were identified.
Statistically significant numbers of tests were
run to determine failure pressures at each of the
gap-temperature conditions and thus to
demonstrate that the seals would continue to
function during and after a station blackout.
To assess the possibility of "binding" of any
axially moving elastomer seals, axial stroking
tests were conducted at predicted worst-case
conditions, and forces were compared to the
balance of hydraulic forces that essentially
maintain an equilibrium small separation
between the rotary face seal rings (Rhodes and
Metcalfe, 1987).
Figure 21 shows schematics of some of the
seals that have been tested, and Figure 22
shows the basic arrangement of the apparatus
(a pair of seals back-to-back in a test cell
connected to a pressurized water source with
heat provided by a laboratory oven and
stroking by an electromechanical device
connected through a load cell outside the oven).
From these tests it was found that axial friction
forces are often two to three times greater at
station blackout temperatures than at normal
operating temperatures. This can be a relatively
large de-stabilizing force, causing the seal rings
to be dragged apart and leak excessively. Each
seal design must therefore be assessed
individually from this standpoint.
Specifications, Inspection and Defects
To ensure the correctly performing
compound is received for critical applications,
specifications must not unwittingly open the
door to other compounds in the same class of
elastomer. Otherwise, service may be
unacceptablelow safety margins, unreliability
and frequent replacement. I f alternative
compounds are needed as back-up, then each
must be separately qualified. As a minimum,
purchase specifications should require that each
elastomer seal be traceable to the particular
"batch" of ingredients mixed and processed
together to form the unvulcanized stock from
which the seal was made, and that a
manufacturer's certificate of conformance be
provided confirming the seal's size, compound
identification, batch number, date of cure,
hardness, specific gravity and tensile strength
relative to the manufacturer's expected values.
In AECL-produced MCP seal assemblies for
example, all elastomer seals are batch qualified
by functional testing. In addition, the most
critical elastomer seals in the assembly are
individually serialized, and their moulds have
cavities for simultaneous production of
O-rings, which can then be functionally tested.
This, and positive identification of the parts,
serves to verify the moulding process in
addition to verifying the batch of unvulcanized
elastomer. Receiving inspectors at a plant, in
this case, need only confirm the
documentation.
Defectsoverall nuclear approach.
An oft-neglected area of quality assurance that
impacts heavily on seal integrity and reliability
is inspection methodology and acceptance
criteria for defects. This includes defects in the
seal gland as well as in the elastomer seal, since
machining marks and other irregularities in the
gland surfaces can disrupt sealing just as much
as can isolated scratches, pits, abrasions, and
cuts in the seal itself.
Although defect acceptance criteria for seals
and seal glands are best established by
application-specific functional testing, some
general guidelines apply. Defects like
scratches, machining marks ("chatter marks" in
a lathe-turned part) or cuts that extend across
the "footprint" of the seal are a prime concern.
Whether or not these are acceptable depends
largely on their steepness, i.e., their depth to
width ratio. This applies equally to defects in
the surfaces of the seal gland. The deeper and
narrower the scratch or groove, the more
difficult for the elastomer to deform into it.
Limits can be assigned based on test data and
analysis (Kuran et al., 1994).
For surface defects, examination by unaided
eye is a satisfactory inspection method. This
requires stretching of each region, since many
defects such as cuts and tears are difficult to see
in unstrained parts. Size can be compared to
reference standards that simulate acceptance
limits, and (if necessary) measured using
optical and mechanical aids (e.g., calibrated
magnifier, depth measuring microscope, stylus
profilometer).
For detection of internal defects such as
inclusions, voids and regions of inhomogeneity
in elastomer seals, a non-destructive tool called
an Elastodynamic Tester has been found most
useful. This measures reaction force resulting
from an imposed deformation. Figure 23
shows an automated system for elastodynamic
testing of O-ringsa 1994 R&D 100 Award
winner (Anon, 1994). The reaction force on
two pinch rollers is measured as the O-ring is
driven and squeezed between them. Internal
defects are signalled by spikes in reaction
force. Any generally high or low force, or
variation around the O-ring, signifies abnormal
properties when compared with a known
baseline.
Space Shuttle O-ring defects. The
approach followed in developing rejection
criteria and defect inspection methods for Space
Shuttle O-rings was first to develop functional
sealing tests that subjected the O-ring to
simulated conditions exceeding "worst-case,"
then to use these tests to evaluate intentional
defects of all plausible and potentially
dangerous types (Wensel et al., 1988). For
each type of defect, the critical size for
unacceptable leakage was generally found by
bracketing a failure at one size with a pass at a
smaller size.
Supplementary tests with transparent
fixtures allowed defect behaviour to be
observed under the simulated service
conditions, both statically and during ignition
transient simulations. O-rings were viewed
and photographed to measure the shape and
width of the footprint (region of contact) under
a range of squeeze, pressure and gap opening
conditions, thus establishing the size of each
type of defect at which bridging of the footprint
(i.e., leakage) occurs.
"Defect detection requirements" were
established by applying a safety factor (i.e., a
reduction of at least two in the largest defect
that consistently passed the sealing tests and
was also seen to be safe in the footprint
observations). I f the O-rings could
consistently and economically be manufactured
to a more stringent requirement, and detection
method sensitivity was sufficiently reliable at
these smaller defect sizes, then a higher safety
factor than two was applied.
The detection capabilities of inspection
techniques were assessed relative to the
detection requirements. Rejection limits
specific to each adopted detection technique
were determined for each defect type. For
visible, directly measurable surface defects
(abrasions, grooves, voids, cuts), rejection
limits were defined simply in terms of defect
dimensions. For sub-surface defects (voids,
inclusions, regions of inhomogeneity, sub
standard material where depth and orientation
within the cross section were important
factors), rejection limits were defined in terms
of elastodynamic test results. This measured
the effect of defects on the static and dynamic
contact (sealing) force response, and was
therefore most closely related to sealing ability.
In summary, inspection requirements and
defect limits for severe service should be
customized to functional, application-specific
testing. This assures safe service without
imposing overly stringent standards and
wastefully high rejection rates.
Elastomer Seal Storage and Usage
Elastomers are subject to deterioration with
time, temperature, and other environmental
influences. Ideal storage conditions are cool,
dark, and free from contaminants (such as
ozone, solvent vapors, etc.). Elastomer seals
should be stored in a relaxed state, free from
strain (i.e., not folded, twisted, or hanging on
a rack). Under such conditions, most
elastomer compounds will remain serviceable
for many years. Measurement of critical
sealing properties (e.g., compression set,
extrusion resistance, hardness) of O-rings of
certain ethylene propylene and nitrile
compounds stored under such conditions by
AECL has shown them to be essentially
unchanged after more than twenty years. Not
all elastomer compounds are this stable,
however, particularly not all nitrile compounds.
Immediately prior to installation, after
verifying that an elastomer seal taken from
storage is of the required size and material, it
should be thoroughly examined for defects that
might previously have gone unnoticed,
particularly for crazing or cracking that might
have occurred during storage. All contacting
surfaces of the seal gland should be examined
for edge break at the comers of grooves, and
smoothness of lead-in chamfers and
countersurfaces over which the elastomer must
slide during assembly.
Rolling or twisting of an O-ring in its groove
with bore-type assemblies can be avoided by
ensuring smooth surfaces and correct
lubrication. Only a thin film is needed, and
once applied, care is necessary to avoid
introducing contaminants, particularly hairs and
fibres, which readily create leak paths.
Excessive or inappropriate lubricant, however,
can cause failures (e.g., by causing swell or by
restricting small flow passages or
contaminating the rotary face seals in MCPs).
Conclusion
The Canadian nuclear industry had special
needs to develop greater fundamental
understanding, improved products and better
practice in the field of fluid sealing. Various
solutions for severe service have been
described, covering both rotary face seals and
elastomer seals.
For face seals, controlled lubrication is the
key to long seal lifeachievable in several
distinct ways, but generally by considering all
possible operating conditions and ensuring
through good design that a favourable
alignment and lubricant film face is maintained
between their faces.
For elastomer seals, a thorough
understanding of principles is needed, coupled
with an open-minded, custom-design
approach, and close attention to delivering
consistent, well-qualified material.
The economic benefits of better sealing
given by just the examples described here are in
file hundreds of millions of dollars to Canada
a not-inconsiderable fraction of the total $1.3
billion annual savings estimated as feasible
through improved tribological practices
industry-wide (Anon, 1987). In addition,
there has been immeasurable spill-over, with
many industries now using and benefitting
from nuclear-developed sealing technology
and nuclear plants benefitting from the greater
knowledge gained through Space Shuttle
O-ring improvement programs.
Acknowledgement
The contributions of many colleagues
towards the sealing solutions presented here is
gratefully acknowledged.
References
Anon, 1987, "Enhancing Reliability and Efficiency
Through the Reduction of Wear and Friction,National
Research Council Canada, Ottawa.
Anon, 1994, "Elastodynamic Inspection System,"
R&D Magazine, (a Cahners Publication), September.
Billington, I.J., and Fitzsimmons, T.E., 1966, "Final
Report on Shaft Seal Development," Dilworth, Secord,
Meagher & Associates Limited, Toronto, Ontario,
Report No. DSM-203-358 Revised, (AECL Report
AECL-2549).
Champ, D.R., Brown, R.M., Cooper, E.L., and
Cornett, R.J., 1989, "Emergency Response to a Spill
of Tritiated Heavy Water - The Interface Between
Emergency Response, Routine Monitoring and
Research," Proceedings of an International Symposium
on Recovery Operations in the Event of a Nuclear
Accident or Radiological Emergency organized by the
International Atomic Energy Agency, Vienna,
November (AECL Report AECL-10250).
Earl, A.H., 1983, "Reliability of CANDU Heat
Transport Pumps," Proceedings of International
Symposium on Reliability of Reactor Pressure
Components, Stuttgait, Federal Republic of Germany,
March 21-25.
Etsion, I., and Constantinescu, I., 1984,
"Experimental Observation of the Dynamic Behavior of
Noncontacting Coned-Face Mechanical Seals," ASLE
Transactions, Vol. 27, 3, pp. 263-270, July, (AECL
Report AECL-8604).
Greene, T.B., and Inch, G.B., 1992, "Evaluation of
Shaft-Seal Leakage Under Station-Blackout Conditions
for the Reactor-Recirculation Pumps at Nine Mile Point
Unit One, Proceedings of the Fifth International
Workshop on Main Coolant Pumps, Electric Power
Research Institute, Palo Alto, California.
Kittmer, C.A., and Metcalfe, R., 1980, "An Inside
View of Rotary Seal Dynamics," Proceedings of the
Symposium on Engineering Applications of
Mechanics, Ottawa, June, (AECL Report AECL-6947).
Kuran, S., Grade, B.J., and Metcalfe, R., 1994,
"Low Pressure Sealing Integrity of O-Rings Based on
Initial Squeeze and Counterface Finish." Paper accepted
by STLE for publication in Tribology transactions.
Presented at 1994 STLE Annual Meeting, Pittsburgh,
PA, (AECL Report AECL-11154).
Lebeck, A.O., 1991, "Principles and Design of
Mechanical Face Seals," John Wiley & Sons, New
York.
Martinson, A.R., 1980, "Design, Development and
Testing of Large-Diameter, High-Pressure Seals for
Nuclear-Reactor, Primary-Coolant Pumps-A Challenge
to the Pump Manufacturer," Lubrication Engineering,
Vol. 36, 6, pp. 325-340.
Mayer, E., 1977, "Mechanical Seals," 3rd English
Edition, Newnes-Butterworths, London, Boston.
Metcalfe, R., 1976, "End Face Seals in High Pressure
Water-Learning from Those Failures," Lubrication
Engineering, Vol. 32, 12, ASLE, (AECL Report
AECL-5607).
Metcalfe, R., 1976, "The Use of Finite Element
Deflection Analysis in Performance Predictions for End
Face Seals," Proceedings of Third Symposium on
Engineering Applications of Solid Mechanics,
University of Toronto, Toronto, Ontario, June 7-8,
(AECL Report AECL-5563).
Metcalfe, R., Pothier, N.E., and Rod, B.H., 1978,
"Diametral Tilt and Leakage of End Face Seals With
Convergent Sealing Gaps," Proceedings of 8th
International Conference of Fluid Sealing, Cranfield,
England, (AECL Report AECL-6234).
Metcalfe, R., 1979, "Development of a Better Seal
for CANDU," Proceedings of Canadian Nuclear
Association, 19th Annual International Conference,
Toronto, June.
Metcalfe, R., Kittmer, C.A., and Brown, G.W.,
1982, "Effects of Pressure and Temperature Changes on
End-Face Seal Performance," ASLE Transactions, Vol.
25, 3, pp. 361-371, July, (AECL Report AECL-7644).
Metcalfe, R., and Brown, G.W., 1984, "Eccentricity
of Balanced End Face Seals," Proceedings of 10th
International Conference on Fluid Sealing, BHRA,
Cranfield, England, (AECL Report AECL-8289).
Metcalfe, R., 1985, "Sensitivity of a Reactor Coolant
Pump Seal to Changes of Operating Conditions,"
Electric Power Research Institute Report, EPRI NP-
4244, Palo Alto, California, September, (AECL Report
AECL-11184).
Metcalfe, R., and Burchett, P.R., 1987, "Primary
Heat Transport Pump Seal Reliability, Performance and
Development," Proceedings of 14th Inter-Ram
Conference for the Electric Power Industry, Toronto,
May, (AECL Report AECL-9547).
Metcalfe, R., Baset, S.B., Lesco, R., and Selander,
W.N., 1989, "Modelling of Space Shuttle Solid Rocket
O-Rings," Proceedings of 12th International Conference
on Fluid Sealing, BHR Group, Cranfield, England,
(AECL Report AECL-11167).
Metcalfe, R., Rhodes, D.B., and Peterson, A.J.,
1991, "Development of a Nuclear Pump Seal for BWR
Application," Proceedings of the Fourth International
Workshop on Main Coolant Pumps, Electric Power
Research Institute , Palo Alto, California, (AECL
Report AECL-11149).
Metcalfe, R., Baset, S 3 . , and Kuran, S., 1992,
"Contact hydraulics in the Sealing Footprint-Effects on
Deformation, Leakage and Friction of Soft Seals,"
Proceedings of 13th International Conference on Fluid
Sealing, BHR Group, Cranfield, England, (AECL
Report AECL-11166).
Metcalfe, R., 1993, "Eccentric Face Seal With
Asymmetric Closing Force," United States Patent
Number 5,180,172, January.
Metcalfe, R., Graham, T.A., and Wong, W.C., 1994,
"Eccentric Seals For Nuclear Pumps," Proceedings of
14th International Conference on Fluid Sealing, BHR
Group, Cranfield, England, (AECL Report AECL-
11139).
Metcalfe, R., and Wensel, R., 1994, "High
Performance O-Ring Sealed Joints," Lubrication
Engineering, Vol. 50, 2, STLE (AECL Report AECL-
10865).
Nau, B.S., 1980, "Observations and Analysis of
Mechanical Seal Film Characteristics," ASME Journal
of Lubrication Technology, Vol. 102, July.
Pothier, N.E., and Rod, B.H., 1980, "Development
Testing of a Pocket-Type, Hydrostatic Rotary End Face
Seal in High-Pressure Water," ASLE Transactions, Vol.
23, 1, pp. 77-85 January, (AECL Report AECL-6378).
Rhodes, D.B., Hill, R.C., and Wensel, R.G., 1987,
"Reactor Coolant Pump Shaft Seal Stability During
Station Blackout," U.S. Nuclear Regulatory
Commission, NUREG/CR-4821, EGG-2492, May,
(AECL Report AECL-9342).
Rhodes, D.B., and Metcalfe, R., 1987, "Preventing
'Dryout' and Popping-Open' Failures in End Face
Seals," Proceedings of the 4th Acot Research Seminar,
Tribology Research for Canadian Industry, National
Research Council Canada, Ottawa, May.
Rhodes, D.B., 1992, "Performance of the CAN2A
Recirculation Pump Seal Cartridge During a Station
Blackout," Proceedings of the Fifth International
Workshop on Main Coolant Pumps, Electric Power
Research Institute , Palo Alto, California.
Ross-Ross, P.A., Adams, E.J., Dixon, D.F., and
Metcalfe, R., 1976, "Performance and Reliability of
Primary Circuit Components in CANDU Reactors,"
reprinted from Nuclear Energy Maturity, Proceedings in
Nuclear Energy Series, Pergamon Press, Oxford and
New York, (AECL Report AECL-5092).
Ross-Ross, P.A., and Metcalfe, R., 1983,
"Mechanical Development for Reliable Reactor
Components," paper presented at the CSME Council
and Annual General Meeting in Calgary, 1980 April,
(AECL Report AECL-8186).
Stevens, C.A., 1989, "Prediction of Fracture in
Unconstrained Elastomeric O-Ring Seals," Proceedings
of 12th International Conference on Fluid Sealing,
BHRA, Cranfield, England.
Vilim, P., and Zeller, L.A., 1965, "Shaft Seal for A
Pump," Canadian Patent Number 716.850, August.
Watson, R.D., 1963, "Effect of Seal Ring Deflection
on the Characteristics of Face-Type Mechanical Shaft
Seals in High Pressure Water," (AECL Report AECL-
2242).
Watson, R.D., 1965, 'Terformance Testing of
Conical-Face High Pressure Rotary Shaft Seals,"
(AECL Report AECL-2551).
Wensel, R.G., 1977, "Resistance of O-Ring Seal
Compounds to Heat and Ionizing Radiation in Air and
in Water," (AECL Report AECL-5697).
LEAKAGE
^ HOUSING \N
PRESSURE
STATIC
O-RING
STATOR
ROTOR
s AXIALLY
MOVING
O-RING
SPRINGS
AXI AL
FORCES
AND
BALANCE
PRESSURE
PROFILE
O-RING
FRICTION
CLOSING
PRESSURE
SPRING
FORCE
Figure 1. A Rotary End Face Seal
with Flexibly-Mounted Rotor. The axial
forces on the rotor must be in equilibium. This
is possible without any contact between the seal
faces if the hydraulic pressure profile has the
shape shown.
Wensel, R.G., and Cotnam, B.M., 1978, "Evaluation
of O-Rings of Various Elastomer Compounds for
Service in Hot Water," (AECL Report AECL-6214).
Wensel, R.G., 1985, "Factors Affecting Wear
Damage to Elastomer Seals in Main Coolant Pump
Face Seals," Electric Power Research Institute, Report
EPRI NP-4245, Palo Alto, California, September,
(AECL Report AECL-11185).
Wensel, R., Cotnam, B., Gentili, H., and
Constantinescu, I., 1985, "Friction and Axial
Force/Displacement Characteristics of Elastomer Seals
in Water," Lubrication Engineering, Vol. 41, 9.
Wensel, R.G., Metcalfe, R., Pothier, N.E., and
Russell, B.G., 1988, "O-Ring Seal Studies for Space
Shuttle Solid Rocket Booster Joints," Canadian
Aeronautics and Space Journal, Vol. 34, No. 4,1988
December, (AECL Report AECL-9897).
Wensel, R., 1993, "Using Improved Elastomers to
Enhance CANDU Station Reliability," Proceedings of
International Nuclear Congress, INC 93, Toronto,
October, (AECL Report AECL-10912).
Wensel, R.G., Metcalfe, R., and Rhodes, D.B., 1994,
"Overcoming Problems with MCP Elastomer Seals,"
Proceedings of the Seventh International Workshop on
Main Coolant Pumps, Electric Power Research
Institute, Palo Alto, California, (AECL Report
AECL-11182).
Yasuna, J.A., and Hughes, W.F., 1990, "A
Continuous Boiling Model for Face Seals, ASME
Journal ofTribology, Vol. 112, April.
Figure 2. Hydrostatic Pressure
Profile Characteristics between End
Faces of a Rotary Seal.
LOCKING PLATE TERTI ARY OR
BACK-UP SEAL
( CARBON SEGMENTED BUSHING)
GLAND RETURN OR
STAGI NG FLOW (NO. 2)
PRI MARY OR /
NO. 1 STAGE /
SEAL /
SHAFT SLEEVE
Figure 3. Seal Cartridge for Bruce MCPs. The BJ -SU seal parts are shown at right; the
AECL CAN1 parts at left; the AECL CAN2 parts are inset (No. 1stage only).
Figure 4. CAN2A Seal for BWR Service. Two stages are shown, -140 and 150 mm balance
diameters.
Figure 5. Condition of No. 1Stage CAN2A Seal Faces after Three Fuel Cycles Service.
Visually both seal faces looked perfect. They are shown with an optical flat to measure their
lightband flatness. The centre of each seal face is almost original condition; edges are rounded
by ~1fim. The No. 2 stage was very similar.
C LC OUTLET (T10)
CLC INLET (T 11)
CONTROLLED -
BLEED OFF
INTEGRAL
HEAT EXCHANGER -
PUMP COVER
- SEAL LEAKAGE
COLLECTION CAN
RECIRCULATION
FLOW
-RECIRCULATION
FLOW ( T 6)
FILLER
HOT LOOP
INLET
* SEAL CARTRIDGE
Figure 6. Nine Mile Point Reactor Recirculating Pump Mock-Up. Thermal mass and
external piping were included for full simulation of station blackout behaviour.
Figure 7. CAN8 Seal for BWR and CANDU Service. The CAN8 seal parts are shown
Bruce seal cartridge.
ENVELOPE OF
POSSIBLE DEFLECTION
FOR 6.9 MPa (1000psi)
DEFLECTION HYSTERESIS
LOOP FOR PRESSURE
INCREASE THEN DECREASE
SPAN OF POSSIBLE
DEFLECTION FOR
6.9 MPa (1000 psi)
(|iin, where 40 jxin = 1 ^m)
TEMPERATURE SENSITIVITY
Figure 8. Comparison Showing Improved Deflection Characteristics of CAN8 over CAN2
and SU Seals for Bruce No. 1 Stage. The separate plots for each stator face and rotor face show
how they deflect in response to pressure changes. Scales are all equal, and the span of possible
deflection (|iin for 1000 psi) is given by the lightly shaded regions. The direction of arrows
shows deflection hysteresis when pressure is increased to 6.9 MPa (1000 psi) then decreased to
zero (darkly shaded regions). Any environmental temperature variations change deflection
within the lightly shaded region according to the sensitivities shown boxed. Assumed friction
coefficients are: >0.1 for stator to holder; >0.3 for rotor to holder. Deflection cannot go
beyond these regions unless actual friction is higher.
QUENCH
PUMP COVER /
SPACE CONSTRAINT mm
LABEL SIZE SECT.
dia.
INNER
dia.
SEAL
TYPE
INNER
dia.
OUTER
dia.
AXIAL SQU
EEZE
FILL STR
ETCH
LUBR
ICANT
0-1 243 3.5 104.4 Face 85.7 108.0 2.5 29% -3% No
0-2 225 3.5 47.2 Bore 47.6 53.7 4.3 13% 74% 0.7% Yes
0-3 227 3.5 53.6 Bore 53.8 59.9 4.3 14% 74% 0.5% Yes
0-4 230 3.5 63.1 Bore 66. 1 72.4 4.3 10% 68% 4.7% Yes
0-5 230 3.5 63.1 Bore 69.8 76.5 4.6 0% 58% 10.6% Yes
0-6 230 3.5 63.1 Face 66.7 ~ 3.2 6.7% ~ 5.7% No
0-7 230 3.5 63.1 Face 66.7 ~ 3.2 6.7% ~ 5.7% No
Figure 9. CAN6 Rotary Face Seal Assembly Showing Custom O-Ring Gland Parameters.
This seal is for Reactor Water Clean-up Pump Service in BWRs.
Figure 10. CAN6 Service Record, Showing All Replacements and Reasons.
Figure 11. CAN13 Seal for AECL Fuelling Machine Ball Screw Rams. The rotating seal
face is flat-conical to give hydrostatic lubrication when pressurized.
I I
S tator
Eccentric
housing
O-Ring
Springs
Rotating
seal face
Stationary
seal face
SECTION XX
eccentricity
Figure 12. Tilt-Resistant Eccentric Seal Arrangement. The stator is mounted eccentrically,
but because its shape and loading are axisymmetric, it has no tendency to tilt relative to the rotor
which is fixed rigidly to the shaft. The rotating seal face has a velocity component, Vr, normal to
the edge of the stationary seal face.
Figure 13 NRU Main Heavy Water Pump Seal Cartridge in Seal Tester.
SEALING
FOOTPRINT
SEEPAGE
PRESSURE
SEALING
FOOTPRINT
Figure 14. Wedge-Shaped SealEffects of Forward vs. Reverse Pressure.
TOP WALL
(MOVABLE)
" ADVANCED" SOLID ROCKET
MOTOR (ASRM1 FIELD JOINT
-1.2 cm
(0.5 in.)
PRIMARY
RING
2ND
O-RING
EXTRA
O-RING
CUTOUTS
FOR -150
BOLTS
EXTERIOR INTERIOR
' Zero gap opening
' High squeeze
' Potential for assembly
damage greatly reduced
CHALLENGER" SOUP ROCKER
MOTOR (SRM) FIELD JOINT
.TANG
O-RING SEALED JOINTS
IGNITER CHAMBER J OINT
INITIATOR J OINT
///THROUGH-BULKHEAD INITIATOR
/ / / / PRESSURE TRANSDUCER
7 //^jLEAK-CHECK PLUGS
jt / J p ' igniter to case joint
-CASE TO CASE (FIELD J OINTS)
NOZZLE (TO CASE) J OINT 1
LOCATIONS OF NOZZLE
JOINTS 1. 2. 3. 4. 5 & 6
HOLES
FOR177
PINS
CLEVIS
Initial gap still present
Gap opening still present
Axial movement still present
Squeeze limited
Assembly damage concerns
Heaters necessary
Figure 15. Space Shuttle Boosters and their O-Ring Sealed Joints. There are numerous such
joints in each booster. Failure of a "field" joint caused the 1986 Challenger accident. The "re
design" and "advanced" field joints are much more robust.
(a) Conventional Bore Seals
(b) Alternative Corner Seals
GOOD SEALING =HIGH SQUEEZE +LARGE FOOTPRINT
Figure 16. Comparison of Igniter to Case Conventional Bore Seal J oint with Improved
Corner Sealing Arrangement.
MEASURABLE
PROPERTIES
SEALING REQUIREMENTS*
A B c D E F G TOTAL
Compression Recovery 3 2 3 0 1 0 0 9
Age Resistance 3 3 2 0 1 0 0 9
Extrusion Resistance 0 0 3 3 2 0 0 8
Hot Gas J et Resistance 0 0 3 3 0 1 0 7
Compression Stress Resistance 0 2 1 2 1 0 0 6
Surface Toughness 0 0 0 0 3 1 0 4
Sealing Ability 0 0 2 0 0 3 0 5
Permeability 0 0 1 0 0 0 0 1
Abrasive-Erosion 0 0 0 0 1 1 0 2
Modulus 1 0 1 0 1 2 0 5
Elongation 0 0 0 0 1 2 0 3
Hardness 1 0 1 0 1 3 0 6
Swell 1 3 0 0 0 0 3 7
LEGEND*
A - Follow gap opening at 0.5C (33F).
B - Adequate dynamic response after 5-year assembly.
C - Accept pressure without leakage.
D - Resistant to heat and hot gases.
E - Resistant to mechanical damage on assembly.
F - Seal against a single-point machined finish of 1.6 }imRa (63 [iinch
arithmetic average roughness height)
G - Compatible with corrosion-inhibitor.
3 =High Relationship
2 =Moderate Relationship
1=Low Relationship
0 =No Relationship
Figure 17: Example of Matrix Relating Sealing Requirements to Measurable Properties.
EPDM A (177C)
...E3...EPDM B (177C)
EPDM C {177C)
EPDM A (232C)
EPDM B (232C)
EPDM C (232C)
O-Rings sealing 6.9 MPa Hot Water
Sectional Diameter 3.53 mm
Squeezed 20%
0 10 20 30 40 50 60
Ti me (weeks)
Figure 18. Thermal aging comparison of EPDM O-rings. The O-rings were inserted
in a flange-type gland, heated and pressurized, held until end of test or excessive leakage. Every
ten weeks each was temporarily disassembled to measure compression set (initial height-final
height) / (initial height - squeezed height).
P
r
e
s
s
u
r
e

D
i
f
f
e
r
e
n
t
i
a
l

(
M
P
a
)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Ext rusi on Gap (mm)
Figure 19. Extrusion resistance comparison of EPDM O-rings
TEST CONDITIONS: WATER 288C, 18-HOUR HOLD AT NOTED BASE PRESSURE
Figure 20. Extrusion testing of O-rings under four possible "worst-case"
conditions for a particular station blackout event
F
A
I
L
U
R
E

P
R
E
S
S
U
R
E

O
N

R
A
M
P
I
N
G

A
F
T
E
R

1
8
-
H
O
U
R

H
O
L
D

(
M
P
a
)
Hi gh Pr e s s u r e Si de
En er gi zi n g
0 - Ri n g
Channel Seal
Lo w Pr e s s u r e Si de
Hi gh Pr e s s u r e Si de
Lo w Pr e s s u r e Si de
Hi gh Pr e s s u r e Si de
O- Ri n g
^ B a c k u p Ri ng
Lo w Pr e s s u r e Si de
Figure 21. Typical Seals and Test Fixtures.
Figure 22. Apparatus for Stroking of Axially Moving Elastomer Seals. A pair of seals
back-to-back in a test ceil is connected to a pressurized water source with heat provided by
laboratory oven. Stroking is by an electromechanical device outside the oven, with any
"binding" measured by a load cell.
Figure 23. Elastodynamic Inspection System for O-Rings. A Space Shuttle booster O-ring is
being inspected. Anomalies are compared with calibration signals tor reference defects. Overall
properties are compared with known baseline material.