Go back to the accident listing
You do not have a membership or you are not logged in
Become a member
General accident information
|| * * * *
Explosion of a mulitshell
Ammonia separator caused
0 / >1
|Occurrences or events
||Blast-wave/Shock-wave, Blow-away, Chemical reaction, Chemical reaction,
Chemical reaction, Corrosion, Crack, Explosion,
Fire, Penetrate/Puncture, Release, Stench-emission
Full accident information
Click here to view a document containing the full accident information
Date : 1988 0418
After 10/years operation an ammonia separator in a 1.000/metric tons per
day plant ruptured. The process is based on steam reforming of natural gas,
followed by CO-shift, CO2-scrubbing, methanation, compression and ammonia
synthesis coupled with a compressor refrigeration unit. Materials of
construction are according to normal practice, e.g. no aluminium components
are used anywhere in the plant.
The separator is of the multi shell design with a 22/mm thick core layer
and 13 wrapping layers of 7.1/mm thickness each, operated at 29800/kPa and
-5/C. The wrapping layers failed in a ductile manner and approximately 90%,
heavily deformed, weighing 23600/kg were thrown 150/m away. The blast wave
was directed to the opposite side. The rest of the separator, i.e. core
layer and both hemispherical heads, failed predominantly in a brittle
manner leading to fragmentation.
In total 81 fragments could be identified weighing altogether 42000/kg
whereas the total vessel weight was 48000/kg. Some fragments reached a
distance of up to 500/m. Many uncounted small pieces (shrapnels) have
punched walls of the reformer section 50/m away and the feed gas heater
including the coil. The feed gas heater burnt out and suffered also from
the blast wave as did the reformer. Walls buckled, steel profiles were bent
and the refractory was affected to an extent that complete replacement was
required. The immediate neighbourhood of the separator showed damages from
nil to minor up to complete indicating that flying fragments are mainly
responsible for the damages in this area, whereas the reformer area
suffered heavily from the blast. The fire originating from the blow out
added not too much to the total damage. There were no fatalities and only
In case mercury is accepted as a factor of influence, then the events can
be explained in a way rather consistent with all observations and findings.
Mercury collected at the bottom of the separator changing the
electrochemical behaviour of the system steel/liquid ammonia to the effect
that stress corrosion becomes possible similar to the SCC in ammonia
storage tanks (mercury instead of oxygen?).
The fact that mercury has penetrated into the steel has to be taken as
such. Perhaps somebody is able to calculate from the given figures the time
for which it must have been present.
The leak was observed 15-20/sec. before the rupture, ammonia smell was
sensed by an operator in a distance of 75/m. When he started to move to the
synthesis the blast released from the exploding vessel knocked him down.
Therefore it may well be that the leaking occurred earlier than observed.
However, the developing crack must have reached a critical size, but not to
explain the fragmentation. This time a critical crack size is necessary
only to explain pressure built-up between the different wrapping layers,
especially behind the most outer one. Estimating calculations show that the
pressure can reach at this point the bursting pressure of the thin wrapping
layer, because the size of the weep holes limits the possible pressure
reduction over the length of the weep holes. As in case of the first route
the wrapping layers will then rupture successively under overload
As mentioned earlier it is not unlikely that mercury may form explosive
Hg-N or Hg-N-O compounds with high detonating violence.
Indeed, it was reported that during cleaning of a valve in the ammonia
storage area some small mercury bubbles rolled out and the attempt to
scrape off some scale led to small explosions.
Experts told me that 1 kg of such compound is equivalent to approximately
2/kg/TNT or 9.2/MJ. With the feed 60-72/kg Hg are introduced per year. If
we assume 10/kg to have been present at the time of rupture at the bottom
of the separator we have a charge of 473/MJ for 11/m3 gas volume
Whatever assumptions are made with respect to blast or missile energy it
can be postulated, that the range where fragments were found is smaller
than it should be based only on the stored energy. This leads to the
speculation that within seconds a steep depressurization may have occurred
before final rupture. It has been estimated that a crack 2/cm wide and 2/m
long would depressurize the system within approximately 6/seconds,
sufficiently short to happen within the mentioned 15-20/sec. Assuming for
example that the pressure dropped to below 2000/kPa then the remaining
stored energy is reduced to approximately 50/MJ, not sufficient to explain
the fragment distribution.
Would there be still any need for the vessel to explode?
There is no answer to that question. However, the steep pressure fall and
the percussions would be sufficient to release the detonation of an
explosive Hg-N-compound which is said to be violent and would destroy the
vessel wall. The contribution to fragment distribution on the other hand is
considered to be small. To assume that the final rupture of the vessel was
accompanied by such an internal explosion is consistent with:
. the complete fragmentation of both heads,
. formation of many shrapnels of bullet size,
. the suspicion from fracture path studies that unstable brittle crack
growth may also have been launched from the bottom head,
. the fact that a specific nozzle was not found.
Lessons learned (conclusion)
The conclusion to give in the form of messages and they consider as usual
all possible effects which may have lead to failure.
Weep holes are designed for the release of diffusing hydrogen and to
indicate small leaks. Large leaks (crack size larger than weep hole size)
can lead to pressure built up between the wrapping layers and subsequent
Repair of a core layer of a multi shell may be performed successfully, if
the core layer needs no heat treatment (no experience available). In case
heat treatment is required for the material, a reliable repair cannot be
a. with a heat treatment the composite structure of the multi shell is
b. with no heat treatment the required material properties cannot be
A required multi shell vessel should be considered as a safety risk and be
replaced as soon as possible.
It is suspected that mercury in liquid ammonia changes electrochemical
behaviour of the system steel-liquid ammonia and makes the steel prone to
SCC similar to ammonia storage tanks.
It is further suspected that mercury and liquid ammonia form explosive
compounds of the type Hg-N or Hg-N-O which detonate under a stroke, a rapid
pressure fall or a concussion and decompose slowly under normal conditions.
If mercury is detected in the feed it should be removed to the lowest