Cowles Dissolver Fire Involving
IR Flare Mix
Did ESD Cause Military
IR Flare Mixtures to Ignite?
Paper by Dr. Dillehay and R.C.
Leander on Early 1990's Incident Investigation
Dr. D.R. Dillehay and R.C. Leander
Marshall, Texas 75670
In November, 1993, a fire occurred
in a Cowles Dissolver while mixing infrared decoy flare composition.
The major constituents of the mix were magnesium, polytetra-fluoroethylene
(PTFE), and a fluoroelastomer binder dissolved in acetone. Hexane
is used to precipitate the binder and wash the mix. During the wash
cycle, ignition occurred and consumed the mix. There were no injuries
but significant heat damage was evident in the mixer bay. Deluge,
blow-out walls and other safety features functioned as intended.
A Board of Investigation composed of
both Government and Contractor personnel was convened to determine
the most probable cause. The Board concluded that the most probable
cause for the fire was either ESD or friction. Subsequent efforts
by Contractor personnel to quantify the electrostatic hazards of
the system were not able to duplicate an ESD scenario. Field tests
to evaluate the friction scenario were not completed.
This paper will summarize the event
and the subsequent testing to isolate the cause. The final conclusion
was that, while ESD seems unlikely, it cannot be ruled out, nor
can friction conclusively be proven to be the cause.
Thiokol Corporation is the operating
contractor of Longhorn Army Ammunition Plant in Karnack, Texas.
Thiokol is a long time producer of pyrotechnic items and rocket
motors at this location. One such family of items is infrared decoy
flares. Several mixing methods are used to manufacture the flare
compositions, including the use of a Cowles Dissolver in the "shock
The major ingredients in the flare
composition are magnesium powder, polytetrafluoro-ethylene (PTFE),
and a fluoroelastomer binder. The fluoroelastomer binder is dissolved
in acetone before the mix is made. The PTFE and magnesium powder
are suspended in the binder solution by the action of the high shear
Cowles Dissolver blade. Hexane is rapidly added to the stirred suspension
and the binder precipitates on the surface of the suspended particles.
The solids are allowed to settle and the liquid is siphoned from
the mix bowl. Hexane is used to wash the solids and the wash is
siphoned from the bowl two times. The solids are then removed and
dried for transfer to an extruder for processing.
The location where the mixing of the
composition is performed is Building 54-H, Bays R-109 and R-110.
The mixing of the flare composition is a remote operation performed
by a crew leader and two operators located in a control room. These
operators control the mixing process with the aid of video monitors
and electronic remote control devices. Flare composition is normally
mixed concurrently in Bays R-109 and R-110. The mixing operation
is protected by a combination of passive and active features. The
two mixer bays have substantial dividing walls on three sides and
a pressure relief wall on the front of the bays. There were approximately
450 square feet of surface area for pressure relief. There are fire
doors located on both ends of the mixing bays plus a third fire
door in the hallway leading to the control room. The building was
equipped with a sprinkler system and each bay had a rapid response
deluge system utilizing UV detectors and high speed valves.
Figure 1. Flow Diagram for Cowles Dissolver Process
The mixing cycle begins with the weigh
up of the fuel and oxidizer in separate facilities. The solids are
transported to Building 54-H as required. The binder/acetone mixture
is procured as a solution, ready to mix.
The mixing day begins with the crew
performing pre-operational checks of the building and equipment.
The fuel (magnesium) hoppers located in the mixing bays are loaded
manually by the operators. The mixing bowls are positioned in Bay
108 and loaded with pre-metered amounts of the acetone/binder solution.
The pre-weighed PTFE is added to the bowls and the bowls are transferred
to the mixing bays. The bowls are manually positioned under the
Cowles Dissolver and the cart frame is locked into position against
fixed stops. The operators return to the control room and begin
the remote mixing process.
The process is controlled and timed
by computer interface. The operator is required to push indicator
buttons at the end of each processing step to confirm completion
of each operation before the next operation can begin. The mix procedure
is detailed in the following steps:
A three (3) minute mix to
combine the PTFE and binder mixture.
The operator dumps the magnesium
powder into the mix from a remote hopper.
Mix the slurry for approximately
10 minutes to assure thorough dispersion of all of the solids
in the binder solution.
The addition of approximately
32 gallons of hexane to the mix. This step "shock precipitates"
the binder out of the acetone while the Cowles blade keeps
all of the particles in suspension.
Shut the Cowles blade off
to allow the coated solids to settle in the mix bowl. After
the mix has settled, a decant tube is lowered into the bowl
and the hexane/acetone mixture is siphoned from the bowl with
Start the Cowles blade turning
again and add additional hexane to wash the mix.
Stop the Cowles blade and
again remove the hexane/acetone mixture.
Start the Cowles blade and
wash the mix a second time with hexane.
Stop the Cowles blade and
raise the mixing head to allow removal of the bowl to Bay
R-111 for dumping and mix recovery.
Another operation occurs when "cross
mixing" is used to blend mixes for improved homogeneity. In
cross mixing, half of a previous mix (approximately 62 lb) is added
to the current mix just prior to the second hexane wash. This gives
a total mix of about 187 lb.
On November 2, 1993, the mixing day
began at 0600. The first mixes of the day (both bays) were made
without incident and set aside for "cross mixing" with
the following mixes. The second pair of mixes was started in both
bays and proceeded normally up to the point of hexane addition.
The hexane failed to pump for the initial "shock" of the
binder. Maintenance personnel were called and the problem was corrected
by bleeding the hexane pump in the dispensing system. The mixing
process was continued through the first hexane wash. After the first
wash, the Cowles Dissolver mixing head was raised and approximately
half of one of the first mixes was manually added to the mix bowl.
The mixing head was lowered and mixing was ready to resume. About
this time, Ordnance Operations Safety Director entered the building
to conduct a safety audit. The Production Foreman and the mix crew
were also in the control room.
At approximately 1030 hours, during
the hexane addition for the second wash, a fire occurred in Bay
R-110. All personnel exited the building safely and there were no
injuries. Personnel working in the other end of the building also
exited safely. The Safety Manager was the last to leave the building
after observing that the control room did not experience any overpressures,
smoke or flame. Other than the video evidence, those in the control
room only noted that it sounded like a big "whoosh".
At this point, the Safety Manager assumed
control of the scene and verified the personnel count and that there
were no injuries. After a short period of waiting, the control room
was re-entered and it was observed that the television monitor in
Bay R-109 was still functioning. A small fire was observed on top
of the mixer bowl. It appeared that this mix had not burned. Since
the initial concern was that the second mix still had the potential
to burn, it was decided to back off from the building and evaluate
a course of action. It could not be determined what was feeding
the small fire in the bay. It appeared to fluctuate around the top
of the mixer bowl. The deluge systems were flowing water in both
bays. After about two hours, it was decided that the best course
was to attempt to extinguish this small fire. This was accomplished
by the Fire Department personnel spraying a small amount of foam
from a protected position outside the bay through the blown out
An initial survey of the fire scene
indicated that all safety measures functioned as designed. The sprinkler
and deluge systems activated, pressure relief walls released, fire
doors prevented the spread of the fire and the equipment shut down
as designed. At this time, the area was secured awaiting the formation
of a joint Government/Contractor Board of Investigation.
In any incident investigation, it is
common practice to develop a fault tree to try to determine the
most probable cause of the incident. Infrared decoy flare compositions
are very hazardous pyrotechnic compositions that are treated with
a great deal of care in manufacture and processing. Indeed, there
is a significant history of damage and death from processing these
In the second wash cycle with hexane,
there was an ignition in the bay that resulted in loss of the mix
and some damage to the mixing bays. Deluge activation and other
safety features worked as planned and there were no personnel injuries
due to the incident. After clearing the building of live material,
an investigating team composed of AMCCOM safety personnel, local
Army safety personnel and contractor personnel was convened. Data
were gathered and collated for analysis. A fault tree was established
for the incident as shown in Figure 2. The results of the fault
tree analysis are outlined in the following paragraphs.
Fault Tree Analysis for 54H Incident
Human Initiated Error (HIE)
The personnel in the building all had
significant training and experience in the operations. The building
foreman is intimately familiar with the operational aspects of all
phases of mixing and handling of the raw materials and compositions.
Procedures have been reviewed and no evidence of deviations have
been noted. The Safety Manager for Thiokol Ordnance Operations and
the Thiokol Corporation Safety Director were reviewing safety procedures
in the Control Room at the time of the incident. Neither noted any
unusual occurrence and the mixer crew followed SOP's through-out
the mixing and during the incident. No evidence has been found that
indicates any operator error.
Operator awareness of the hazards of
these materials makes them very alert for contamination or unusual
Electrostatic Discharge (ESD)
A previous fire in 54H was attributed
to ESD initiation during the addition of hexane during the mix cycle.
In this previous case, the video tape showed the spark initiation
at the surface of the liquid and even followed the spark down the
mixer vortex followed by eruption of the entire mix from the bowl.
This event led to the discovery that the hexane generated a very
large static charge, especially as it flowed from the feed pipe.
It was felt that the position of the pipe discharge contributed
to the spark discharge. From this incident, the addition of hexane
was changed to reduce the chance of the configuration contributing
to the spark discharge.
Later, an additive made by DuPont (Stadis
450 Conductivity Improver) was identified that increased the static
dissipation of hexane. This material is now routinely added to the
hexane as it is pumped from tank trucks into the hexane storage
tanks. This additive does not prevent the buildup of static charge
in hexane but does promote dissipation of the charge by increasing
the conductivity of the hexane. The time constant for dissipation
is a function of concentration, temperature, grounding path and
amount of charge.
In the current incident, the flare
composition had been mixed and rinsed once with hexane. The mix
was in the process of recharging with hexane to complete the second
wash. Of the 20 gallons of hexane required, it is estimated that
18 gallons had been added to the mixer when the ignition occurred.
The video tape of the incident shows the hexane addition was delayed
for approximately 2.5 minutes waiting for the Cowles mechanism to
seat so the control panel light would enable the hexane feed switch.
The hexane is observed to flow slightly to the left of the shaft.
The path appears to be approximately 6 inches from the end of the
nozzle to the rotating shaft of the mixer blade. The hexane appeared
to spread slightly as opposed to the flow observed from the previous
In the process, the Cowles dissolver
blade is turning at 800 to 1000 rpm during the addition of hexane.
The particles of fuel and oxidizer are being swirled in the mix
during the addition. The hexane may pickup a charge as it is discharged
from the pipe and this charge will be carried into the bowl.
There are several grounded discharge
paths for the charge to follow for dissipation, however, the solvent
is only "static dissipative", not conductive. During the mix cycle,
the swirling of the particles in the hexane may cause a buildup
of charge in the liquid. This charge will continually try to bleed
off, but may vary in intensity as a function of speed of mix blade,
absolute humidity of the air, ground integrity, etc. If the charge
exceeds the breakdown potential for air (25,000 volts/cm), a discharge
may occur. The vortex created by the mixer blade/shaft provides
a ready path for the discharge to occur to the grounded shaft. The
figure below shows this graphically.
In this case, the "semi-conductive"
nature of the hexane may allow buildup of a considerable charge
in the bulk of the liquid since only a small portion of the hexane
actually contacts a grounded path for dissipation. Under the right
circumstances, a discharge could occur between the liquid and the
grounded shaft. The ignition energy for a flammable concentration
of hexane and air is 0.29 millijoules (0.00029 joules). If the concentration
of hexane is between 1.25% and 6.90% by volume in air (Lange's Handbook
of Chemistry, Ninth Edition), ignition of the hexane vapor will
occur from a very small spark. The flash point of hexane is -7 °
F. As shown previously, the sweeping of the hexane flame down into
the bulk of the slurry will cause ignition of the entire mixture.
The pressure from the burning mixture throws most of the mix into
the air where complete mixing with oxygen in the air results in
a huge fireball with pressure to force the fire into other areas.
In this case, the blow-out panels in the mixer bays released the
pressure to the outside and the event was basically confined to
the mixer bay. A small amount of smoke was seen above the fire door
to the hallway. The appearance of the bays after the incident, including
the apparent lack of damage to the mixer bowl, supports this scenario.
This scenario was provided to the Thiokol Science and Engineering
Division for concurrence.
The video tape appears to show ignition
at the surface near the center of the bowl. If ignition had occurred
under the blade, the fire would have had to flow up around the outer
edge of the bowl, since the blade would shield the shaft.
It was observed that there is a layer
of charred material on the mixer shaft. This layer is in the region
where the hexane pours onto the shaft. It was measured at 22 inches
above the bottom of the shaft. The mixer bowl measures approximately
34 inches from the center to the top. The bottom of the shaft is
estimated at 2 inches above the bottom of the bowl which means that
the top of the layer is estimated at 10 inches below the top of
the bowl. It was observed, also, that there is char on the mixer
bowl walls that could be related to this binder coating the bowl
walls. The area directly under the mixer blade appears to be relatively
free of char. When the acetone/binder solution is shocked by the
hexane, some of the binder will precipitate on the shaft. Likewise,
some of the binder will precipitate on the walls of the mixer bowl.
These layers may serve to insulate the shaft and the bowl and reduce
the grounding potential of these surfaces. This would aggravate
the build-up of electrical charge by reducing the grounding paths
to drain the charge. By the time the hexane is added to the mixer
bowl, the ground path could be less effective. The area of the mixer
shaft above the mix would be uncoated, however, and could serve
as a ground point for the electrostatic charge to discharge through.
As the last of the hexane enters the bowl, the top of the solvent
layer is observed to be about the right level to allow a discharge
from the surface of the solvent to the uncoated mixer shaft. This
scenario fits the observed ignition on the video tape and the overall
pattern for an ESD event to ignite the hexane vapors. Figure 3.
shows the position of the pipes, mixer blade and bowl with indicators
marked for binder deposits.
A computer simulation was run on the
ignition of the hexane vapors in the top of the mixer bowl. The
following assumptions were made:
The volume above the mix was 2 ft in
diameter by 1 ft in depth for a volume of 5400 in.3 or 88000 cc.
Assuming 4% by volume for the hexane (the middle of the explosive
limits) gives 3700 cc of hexane vapors. With a molecular weight
of 86 gm/mol, this translates into about 12 grams or 0.03 lb of
hexane. Using the Vapor Cloud Explosion Damage Assessment Model
(VEXDAM+) from Engineering Analysis, Inc, the following scenario
was set up. Using 0.03 lb of hexane at a category 7 ignition (where
category 10 is a detonation with shock wave), a calculation of the
pressure generated by ignition of this vapor cloud showed 15 psi
of pressure within the mixer bowl. Changing to a category 8 event
shows a calculated pressure in the mixer bowl of 29 psi. The category
6 event predicts only 1.7 psi in the mixer bowl. Based on the observations
from the video tapes, it is felt that the event was probably between
the category 7 and category 8 events. This means that the pressure
generated from the hexane ignition alone would have exerted about
9900 ± 3500 lbs of force down on the surface of the mix.
Figure 3. Details of the Cowles
From the tape, the ignition occurring
on the front side of the bowl would have allowed the shaft to shield
the back side of the bowl from the pressure wave and generated an
unbalanced force on the mix. This unbalanced force would have forced
the mix down and out the back of the bowl and into the back side
of the bay. Examination of the heavy ash deposits in the back of
Bay R-110 suggests that this did occur. In this scenario, the hexane
ignition alone would have been sufficient to blow the panels off
the outside wall and the mix would have ignited as the hexane vapors
were driven off of the mix as it splattered throughout the back
of the bay. The main fireball from the flare composition would have
been reduced somewhat by the action of the deluge water absorbing
some of the heat. The video tape frames after the ignition also
support continued burning within the bay for a significant time
after the main fireball from the mixer bowl.
These conclusions do not alter the
recommended actions to improve the handling and safety of the mix.
It appears unlikely that absolutely
positive elimination of static charge buildup in a "semi-conductive"
solvent like hexane can be achieved. Addition of the DuPont additive
should be continued to minimize the time required for dissipation
of charge from the hexane.
Discussions with vendors and some laboratory
tests showed that blanketing of the mixer bowl with an inert gas
is not practical and the concept was abandoned. A series of tests
were arranged using sophisticated ESD detection equipment to try
to duplicate the conditions during mixing and determine the probability
of an ESD event. A technical paper describing these tests will be
presented at a later date. The test measurements did not indicate
that significant electrostatic charge was being generated in the
acetone, hexane or combinations thereof. While this these tests
were unable to detect buildup of a charge sufficient to cause a
discharge, they did not completely rule out the possibility that
some test variables were not duplicated from the incident. Accordingly,
additional tests are needed to better quantify the conditions required
for a static discharge to occur.
Examination of the mixer bowl from
Bay R-110 showed no evidence of metal-to-metal contact. No significant
foreign material was found in the mixer bowl that was not likely
to have fallen in after the fire.
Previous experience has shown that
accidental contact of the mixer blade with the bowl wall did not
produce an ignition. The wet flare composition is not very friction
or impact sensitive (based on laboratory tests.) Ignition of the
hexane by metal-to-metal contact is not possible. The video tapes
of the mixing operation do not support metal-to-metal contact.
A small scratch observed in the mixer
bowl from Bay R-110 was caused after the incident. This is evident
by the fact that the scratch is not deep in the metal and goes across
two areas of soot. If the scratch had existed before the incident,
the scratch would have been covered by the soot.
At one point in the past, it was found
that some of the oxidizer added to a mixer bowl in 54H had become
trapped under the Cowles blade and had apparently remained under
the blade during the mix cycle. The material had fused into a solid
mass. The way the oxidizer is added in the 54H complex requires
the mix blade to be above the mixer bowl during the dump. As the
blade is lowered, rotation is started when the blade contacts the
liquid surface. This is intended to move the material from under
the blade before it reaches the full down position.
In at least one documented case, oxidizer
did not clear from under the blade and the curved bottom of the
mixer bowl apparently kept the material in place during the run.
In this case, there was a fire in the opposite bay and the mixer
bowl again sat for several days. Under normal mix procedures, the
operators raise the mixer blade and adjust the rpms to "lift" the
mix and disperse it. This compacted material has never been seen
at the end of a normal (complete) mix cycle. It is likely that this
occurs to some degree in every mix at some point in the run.
In the blending operation being run
in 54H at the time of the incident, wet pre-mixed flare composition
is added to the mixer bowl while the mixer blade is in the raised
position. The mixer blade starts rotation as it comes down to the
hexane surface. It appears likely that some material becomes trapped
under the blade as it reaches the full down position. If the material
is packed tightly and cannot flow from under the blade, there would
be differential rotation between either the bottom of the bowl and
the composition on the blade, or between the blade and the composition
on the bottom of the bowl. If solvent could not flow in the area,
heat could build up. The flare composition is not friction sensitive
and it is questionable that sufficient heat could be developed in
a localized spot to ignite the flare composition. If ignition occurred
at the bottom of the mixer blade, however, there should not have
been any indication of a surface glow before the full mix erupted
from the bowl.
In reviewing the video of the incident,
it was noted that the Cowles mechanism did not settle readily into
the mix. It appeared to "hang up" for approximately 2.5 minutes
beyond the normal seating time. The operators reported that this
was unusual and it explains why they waited so long to begin hexane
addition. The Bay R-109 mechanism seated in the normal time frame.
The Bay R-110 mechanism took longer. Until the mechanism is seated,
the hexane addition cannot be started. There is a limit switch to
preclude addition before the mechanism is seated.
While the Cowles mechanism was seating
in Bay R-110, the hexane pipe was observed to undergo severe vibration.
The vibration seemed to lessen and intensify alternately over a
significant time period. After an extended time, the vibration apparently
stopped and the mechanism was seated.
One suggested explanation for the delay
in seating and the hexane pipe vibration is the entrapment of material
under the Cowles blade. The blade is observed to be spinning from
the time it entered the slurry until the mechanism finally seats.
While not conclusive, it seems fairly unlikely that material could
be packed under the blade in this circumstance. Since the vibration
stopped before the mechanism was seated, trapped material contributing
to a frictional heat build-up seems remote.
Another possibility for the vibration
involves the flow controller for the hydraulics that raise the mechanism.
The Cowles mechanism is raised by applying hydraulic pressure to
a cylinder. The system closes, however, by gravity pushing the hydraulic
back through a flow controller. A check valve in the system allows
the flow to quickly raise the cylinder while the needle valve on
the flow controller slows the lowering of the mechanism. If the
flow controller was having a problem, this could also induce the
vibrations in the system and the hexane pipe is attached at the
top of the mechanism and has a long moment arm that would cause
it to respond to the vibration from the flow controller. The flow
controller should be checked for proper operation after removal
from the system.
Since the video review appears to indicate
a surface glow before the eruption of the mix, it is considered
unlikely that ignition occurred from friction at the bottom of the
mix. Close examination of the mixer blade when the bowl was removed
supports this postulation. Enhancement of the video frames during
the ignition sequence appears to confirm the area near the mixer
blade shaft as the ignition site.
Tests have been proposed to investigate
the frictional heating mechanism in simulation, however funding
and scheduling have not permitted these to be conducted at this
The requirement to raise the blades
for ingredient addition in 54H stems from the use of removable bowls.
The removable bowls were implemented in the installation to permit
removal of the mixer bowls for dumping and recharging while a new
set of bowls were positioned in the mixer bays for mixing. On a
practical basis, it has been found unnecessary and impractical to
run the building in this fashion.
The positioning of the vent system
and other items mounted on the Cowles mixing system was chosen for
convenience in the operations. If the vent system is moved around
to the backside of the mixer bowl, room will be provided near the
front side of the mixer bowl to allow dumping of materials into
the mixer bowl with the mixing blade in the full-down and seated
This configuration would eliminate
the possibility of material being trapped under the mixer blade.
A checklist item should be added to
the SOP to observe the movement of the Cowles mixing system as it
lowers. If excessive time or vibration is noticed, the flow controller
should be checked for proper operation.
Examination of the interior of the
mixer bowls was the primary means of investigating the presence
of any foreign material in the mixer bowl. Both bowls were screened
and checked for the presence of any foreign material. The few pieces
of material found in the Bay R-110 mixer bowl entered after the
fire and would not have caused an ignition. No foreign material
was found in the Bay R-109 mixer bowl.
Since the mix is not very impact sensitive
and the presence of a large volume of liquid solvent would tend
to absorb heat generated by impact of something with the mixer blade,
this mechanism is considered unlikely.
Nothing was observed to drop into the
mixer bowl during the video tape review of the mix. While not conclusive,
it appears unlikely that any large item could have been introduced
into the mixer bowl during material addition. The raw materials
are screened through an 8 mesh screen before addition to the mixer
bowl. The previous mix that was added for cross-blending had been
made under the same screening conditions and SOP controls throughout
the plant should preclude foreign material contamination of a mix
by any significant item.
On entering the Bay R-110, it was noted
that the fuel hopper was out of normal position. The fuel hopper
is usually raised clear of the mixer bowl by a pair of hydraulic
cylinders mounted on the wall. The long cylinder lowers the hopper
to the load position for addition of the fuel prior to starting
a mix. After loading, the cylinder raises the hopper to the dump
position. An upper hydraulic cylinder raises the assembly higher
to allow positioning of the mixer bowl and access to the bowl for
dumping material into the bowl.
If the hydraulic system failed and
allowed the fuel hopper to fall during a mix cycle, the weight of
the hopper could deflect the mixer bowl out of position, causing
the blade to impact the bowl, the bowl wall be brought nearer to
the hexane tube thus increasing the possibility of an electrostatic
discharge, or causing the bowl wall to impact the hexane tube.
It can be seen from the video tape
that the fuel feed hopper did not fall before the incident. After
examination of the area, it was seen that the hydraulic lines to
the cylinders were burned in two by the fire and the hopper fell
as a result of the fire and this mechanism is excluded.
Any mechanical failure of the mixer
should have produced a visible indication on the video or will be
noted from the examination of the mixer after removal of the mixer
bowl. The external appearance of the mixer and mixer bowl makes
mechanical failure of the mixer system unlikely.
The only unusual occurrence observed
on the video tape of the incident involved the rather vigorous vibration
of the hexane pipe as the mixing blade was lowering and prior to
seating of the Cowles system. The vibration occurred while the system
was nearing the bottom position and stopped before the system seated.
The cause of the vibration and the implications were discussed in
an earlier section. The flow pattern from the nozzle appeared to
be less uniform than that observed on the tape from the previous
Early in the mix cycle, there was a
failure of the hexane pump at the hexane storage tank site. A pipefitter
went to the site to correct the problem. He observed some material
had leaked from the vicinity of one of the pumps. The pump problem
was corrected and the mixing continued.
The flow of hexane was observed to
be less ordered from the end of the pipe. This may have been caused
by the problems with the hexane system, although the flow in Bay
R-109 did not show any change. A more erratic spray of hexane through
the air could contribute a slightly greater charge to the hexane
as it flows into the mixer bowl. From the video tape, the point
of ignition appears to be in the vicinity between the end of the
hexane pipe and the rotating, grounded shaft of the mixer blade.
Contribution of the charged hexane stream cannot be ruled out.
Vapor Removal System
The vapor removal system appeared to
be intact and working at the time of the incident. The sweep of
hexane vapors from the mixer serves mainly to keep the hexane vapor
concentration in the rest of the bay from reaching the lower explosive
limit (LEL = 1.25 volume %) for hexane and air. There is no indication
from the LEL meter in the bay that the hexane limit had been exceeded.
The normal sweep of air across the mixer bowl and into the vent
tube should not cause any additional hazard. The vapor and air concentration
near the surface of the mix is likely to be in the explosive region
during the mix. The system is not designed to preclude operating
between the LEL and upper explosive limit (UEL = 6.90 volume %)
within the mix bowl. The vacuum sweep should be continued to maintain
overall bay conditions below the LEL.
The bowl cart in Bay R-110 was observed
to be out of position after inspection of the bay following the
incident. The rear locking mechanisms were on top of the locking
bar and the cart was observed to be approximately 1 inch back from
the normal seating position. The operation of the mixer is interlocked
to the locking mechanisms, indicating that the cart was in position
at the start of the mix. A review of the video shows that the ignition
at the top of the mixer caused a relative motion of the mixer head
and the bowl cart. The force of the ignition caused the cart to
compress the pneumatic tires on the cart, allowing the cart to drop
below the latches. The rebound of the cart moved the cart back to
cause the latches to be engaged on top of the latching bar.
The movement of the cart did not occur
before the incident and therefore did not contribute to the ignition
of the mix.
The weather was cloudy and cool,
with the relative humidity around 66% and the outside temperature
around 50 ° F. There was no thunderstorm in the vicinity. There
is no chance that a lightning stroke caused any problem within the
The relative humidity and temperature
in the mixer bays is controlled.
The fault tree analysis provides a
method of documenting the consideration of all factors that can
be assigned a value in the investigation. From the fault tree analysis,
additional testing may be indicated or a most probable cause may
be determined. In addition to considering possible causes, a hazards
analysis for each cause allows assignment of a relative probability
for the occurrence. By ranking the causes from the fault tree, the
most probable one or two scenarios is assigned. Corrective actions
are then taken to reduce the probability of ignition even further.
In this incident, there are still two
probable causes that cannot be eliminated. Electrostatic discharge
ignition of the hexane vapors followed by displacement and ignition
of the flare composition cannot be eliminated. Frictional heating
of material trapped below the Cowles Dissolver blade cannot be eliminated.
Steps have been taken to mitigate both mechanisms in future mixing
operations. A new mix procedure that eliminates the need for hexane
is under development.