Syngas Compressor Oil Tank

Explosion in Ammonia Plant

On April 27th, 2017, 02:43 am, a power outage tripped the ammonia plant in Porsgrunn, Norway. 30-
minutes later the syngas compressor oil tank exploded. A thorough investigation identified two
independent root-causes that led to an unexpected and dangerous situation.

Svein Hansen
Yara Norge A.S. (Porsgrunn)

Introduction
Incident Summary

T
he Porsgrunn Ammonia Plant was com-
missioned in 1968 and produces 1450 On the 24th of April 2017, at 02:43 am, a trip oc-
MTPD. The syngas compressor has four curred in the ammonia plant in Porsgrunn due to
stages, a circulator, an electrical motor a major external power supply fluctuation. The
and a steam turbine to increase the pressure from electrical driven process air compressor stopped,
20 to 280 bars (290-4060 psi). The syngas com- a shut-down sequence was initiated, and the syn-
pressor is located in a building with a gross vol- gas compressor tripped.
ume of 24.000 m3 (850.000 ft³). The process air
compressor is in the same building. There are two After stabilizing the ammonia plant’s front-end,
separate oil systems, one for the process air and the shift operators were preparing for a restart of
one for the syngas compressor train. The oil sys- the process air compressor. Suddenly, an explo-
tem for the syngas compressor contains an sion occurred in the main syngas oil tank (T-
amount of 35 m3(1235 ft³), covering lube, seal 401), located at the ground floor of the compres-
and regulator oil. sor hall. The explosion occurred at 03:13, 30
minutes after the trip of the back-end of the am-
Safeguarding the Building and Machines monia plant. An intense fire developed rapidly
inside the building.
Thirteen explosive gas detectors are located in
the inside of the roof, close to the vents in the
Two field operators were present in the compres-
building. The 3rd and 4th stage, circulator driven
sor building at that moment, both were able to es-
by the steam turbine are equipped with a water
cape. The panel operator immediately initiated
mist system, which is activated automatically by
the emergency shutdown of the entire ammonia
flame detectors. These detectors react on fire
plant, contacted the emergency response team
containing hydrocarbons, they thereby target fire
and activated the mitigation device i.e. the foam
originating from oil. A camera monitors the exact
system.
same area and is visible on a screen in the control
room. There is also a foam system installed in the
The Emergency Response Team quickly arrived
building designed to fill the ground floor with fire
on site and after identifying the main fire source
suffocating foam within 5 minutes. This system
and setting up their equipment, they were able to
is activated manually by the operator.

2019 1 AMMONIA TECHNICAL MANUAL

 control the fire. No one from the emergency re-
sponse team and operating shift was physically
injured.

The material damage was concentrated on the

electrical/instrumentation cabling on the west
side of the compressor building. Civil damage to
the building was extensive but impact on the
structural integrity was limited.

Figure 2. Distortion of oil tank T-401

What Had Happened?

The reality was clear: the oil tank ruptured or ex-
ploded (figures 2 and 3) and caught fire. In order
to generate such a rupture of the oil tank several
scenarios were proposed. We ended up exploring
the possibility of an explosion in the vapor phase
of T-401.
Figure 1. Syngas compressor (3rd and 4th stage)
and turbine after explosion and fire. We found that the air separation unit (ASU) also
had a partial trip due to the external power fluc-
The turbine and the 3rd and 4th stage of the syngas tuation at 02:43. This unit supplies the site in-
compressor located near the fire source had been cluding the ammonia plant with nitrogen. In the
protected by the water mist system. In figure 1 ammonia plant this nitrogen is mainly used for
this is clearly visible by the yellow paint being blanketing of several units, among which T-401.
intact on the compressor piping. The process air
compressor - located on the opposite side of the
fire in the building - had damages limited to dirt
and debris. Due to a lot of water and foam in the
building, the oil trap (the ground floor itself)
overflowed to the outside of the building. Minor
environmental pollution was therefore reported
because oil had reached the nearby fjord via the
sewer system.

Figure 3. Top view of ruptured tank T-401

AMMONIA TECHNICAL MANUAL 2 2019

 Figure 4. ASU scheme in the status with an air compressor shutdown and the nitrogen compressor still
running. The flow control valve is forced to open, and air intake occurred via this valve (red).

The partial trip of the ASU evolved as follows: Preliminary Conclusion

• The air compressor supplying the ASU We concluded that we had all the components
stopped, but the nitrogen compressor kept on needed for an explosion in the tank. Oil vapor
running. The flow scheme of the ASU is il- probably containing hydrogen degassed from oil
lustrated in figure 4. to a limited extent, ignited as oxygen enriched
• The nitrogen compressor started to take in ambient air entered the T-401 tank.
ambient air via the flow control valve as the
pressure from the ASU decreased. In addi- The nitrogen compressor did not stop because the
tion, the ambient air was enriched by oxygen pressure switch on the suction side was removed.
from the ASU’s coldbox. The time delay for It proved itself difficult to determine why, or
the analyzer to respond to an oxygen contam- even when the pressure switch was removed. It
ination in the nitrogen was found to be ap- probably happened when we did a PLC upgrade
proximately 10 minutes. in 2010. There was no updated HAZID or PHA
done for the ASU.
The analyzer detect oxygen just before 03:00.
Given the uncertainties around the precise event When It Rains It Pours...
timing and residence time evaluation, this is con- There are several explosive gas detectors in the
sistent with the assumption that air was intro- building where the fire occurred. If two out of
duced in the nitrogen grid at around 02:48. At three detectors exceed a defined limit, a vent sys-
03:08 the nitrogen compressor was manually tem operates to dilute the combustible gases by
stopped, and the nitrogen supply was thereafter adding air. Three fans go to a maximum vent
100% from the main back up supply (liquid ni- mode forcing 170 m3/s of air into the building.
trogen tank). The building has a net volume of 21000 m3
(741,600 ft³).

2019 3 AMMONIA TECHNICAL MANUAL

 Our explosive gas detectors are designed to be
“fail safe”. So, when they were destroyed by the
fire and failed, they indicated high level of com-
bustible gas in the building; the extensive venting
was initiated, and we blew air into the fire! This
unintended fire boost caused some large doors to
open on the ground floor, allowing the foam to
escape to the outside instead of filling the ground
floor with fire suffocating foam.

Repair and Restart

Damages were repaired within seven months. All
Figure 5. Lower/upper flammable limit for oxy-
recommended actions were implemented, and
gen and hydrogen mix
the plant was restarted.
The Suspect
However, as there was a possibility for a hydro-
gen “leak” from the compressor seals to the tank We did a massive amount of studies. Since the
(T-401), we scheduled for a gas sample to be plant was in operation, it gave some limitations
taken from the tank as soon as the plant was up with regards to investigations, but it also gave op-
and running again. Surprisingly, we found a large portunities, for example we could verify the im-
amount of hydrogen in the gas phase of the tank pact of imposed changes immediately.
as shown in table 1. All possible resources were
now allocated to one single task: where is this hy- We ended up with a new theory: there was a leak
drogen coming from? in the 4th stage compressor end lid. There are oil
channels in the end lid and these are routed in an-
Nov 2017 Jun 2018 gles to get desired oil flow as shown in figure 6.
O2 1.5 % 0.9 % The excess holes were plugged after drilling. We
H2 16 % < 0.5 % suspected that one of these plugs was corroded
N2 82 % 98.6 % and started leaking. This led to a connection be-
CH4 0.1 % 0.01 % tween lube oil and process gas. The lube oil re-
NH3 traces - turn is routed straight back to the tank. Our sus-
Table 1. Gas phase composition of T-401 after picion was so strong that we even ordered a new
the start-up in 2017 and in June 2018 end lid.

AMMONIA TECHNICAL MANUAL 4 2019

 Discussion
The incident shows the complexity of identifying
when two errors occur simultaneously creating
“the perfect storm”: The missing pressure switch
trip on the nitrogen compressor and the leaking
end lid. It is very unlikely to identify such a sce-
nario through a HAZOP, where double jeopardy
isn’t usually considered.

Nevertheless, the combination of potential errors

may lead to dangerous situations and expensive
repairs. All unwanted incidents, like a nitrogen
compressor not stopping when it should, must
have two independent technical barriers. In our
case the water mist system limited the damages.

Another fact is that during the 30 minutes be-

tween the plant trip by the power outage and the
fire, the control room operator got 900 alarms
from the DCS. In addition; the nitrogen compres-
sor that did not stop gave no alarm.

Figure 6. End lid oil channels highlighted. Finally, after the repairs in 2017, a photo was
taken from the compressor hall as in figure 8 be-
In June 2018 we had to stop the plant due to op- low.
erational issues. We took the opportunity to dis-
mantle and check the 4th stage end lid. The leak
we found was significant. We re-plugged the
leakage, as the new lid hadn’t arrived yet. The
plant was restarted, and new samples were taken
from the gas phase of T-401 as shown in the last
column of table 1. In figure 7 it is also clear that
since the repair in June 2018 the gas phase is sig-
nificantly less explosive.
Figure 8. Compressor building after repair.

Figure 7. Development of T-401 gas phase com-

position. The June 2018 plant stop, and repair
was a success!

2019 5 AMMONIA TECHNICAL MANUAL

AMMONIA TECHNICAL MANUAL 6 2019