Professional Documents
Culture Documents
FABIG TM79 DDTandDetonationPredictionswithCFD ORHansen
FABIG TM79 DDTandDetonationPredictionswithCFD ORHansen
E: olav.hansen@lr.org
T: +47 91 17 17 87
Working together
for a safer world
FLACS validation against full scale tests1997-98, HSE Phase 3A test 4, 19 and 22
©Lloyd’s Register Consulting
17 barg
2.0 barg
TEST
Pressure at end of test rig Pressure 12m, 24m and 48m outside
FLACS
1.0 barg
0.6 barg
2.4 barg
Deflagrations
Blast energy
Unit 3 Unit 4
DNV GL video
Detonation
Detonations propagate by ignition of unburnt gas ahead of flame due to shock waves
16-20 barg pressure & 1600-2000 m/s (detonation limits similar to flammability limits - LFL:UFL)
From OECD hydrogen safety report and other sources
~1 cm ~3 cm ~ 5-10 cm ~ 30 cm
At
Flame speed > 600 m/s DDT BLASTENERGY Unit 1 Unit 2
Detonation
DNV GL video
Buncefield
There are likely more …
Sunrise
Puerto Rico???
2012: Method to simulate DDT and detonation with FLACS (Hansen & Johnson 2013)
©Lloyd’s Register Consulting
mm % vol/m m/m3 m2/m3 m/m3 adj m2/m3 adj Diameter model m/m3 m2/m3 Pr 3 x 3 x 4m
170 0,00 0,023 0,00 0,00 0,00 0,00
120 0,41 0,011 0,36 0,14 0,12 0,05 100 1 0,314 V 36
85 1,04 0,006 1,83 0,49 0,61 0,16
60 0,40 0,003 1,42 0,27 0,47 0,09 50 2 0,314 H 72
42 0,21 0,001 1,52 0,20 0,51 0,07
30 0,24 0,001 3,40 0,32 1,13 0,11 25 4 0,314 H+V 144
22 0,18 0,000 4,74 0,33 1,58 0,11
18 0,19 0,000 7,47 0,42 2,49 0,14 10 20 0,628 H+V 720
12 0,23 0,000 20,35 0,77 6,78 0,26
8 0,23 0,000 45,78 1,15 15,26 0,38 5 52 0,816 1872
3 0,23 0,000 325,55 3,07 108,52 1,02
7,15 2,382 Original estimate 2,386
3 x high 3 x high 2,65 Final estimate
©Lloyd’s Register Consulting Method used November 2012 to develop tree-model used by Hansen & Johnson (2013)
Video frames
From Hansen & Johnson (2013) Mike Johnson FABIG presentation March 2014
©Lloyd’s Register Consulting
Automatic scripts for Safety Gap Issue + Blast Smearing also developed
Deflagration
Visual look
Detonation
Offshore platforms
Illustration
Already optimized using CFD (?) Not caused by DDT
Higher loads expected with DDT
However, already in “trouble” at DDT
Could make automatic deluge more attractive! DDT can represent a significant potential for complete loss
©Lloyd’s Register Consulting
Study approach:
1 Ventilation & dispersion study
=> frequency for ignitied clouds estimated
Leak frequencies:
• Typical distribution
• Somewhat low (0.5/year)
Ignition model:
• Time dependent OGP (UKOOA)
• 55% of ignition frequency from 10s to 150s
• Not sensitive to gas cloud size, only initial release rate
Cloud development
• Natural gas rises, propane/condensate fall when losing momentum
• Comparable maximum cloud sizes, longer duration for propane
• Significant clouds at hull deck for large propane releases
Natural gas
96 kg/s
Propane 96 kg/s
Observations:
• 2-3 times larger Q9 cloud during low winds
• Flammable cloud (LFL:UFL) ~ twice the size of Q9
• NG & propane have similar maximum cloud sizes
• 10-4/year ignited cloud comparable for NG & propane
NG release frequency 50% higher
Longer exposure for propane clouds
Main reasons:
Conservative choice of free horizontal jet into domain
Choice of transient ignition model (normally OLF used)
©Lloyd’s Register Consulting
~45 FLACS explosion calculations performed Cloud size Base case Base case
Q9 (m3) NG Propane
• 13-15 different cloud sizes, 1-3 ignition positions 400
• propane and natural gas 700
• conservative ignition location 1000
1400
• some simulations repeated as DDT 2000
Too few scenarios for quantitative DDT evaluation 3000
4500 Near DDT
6000
Observations: 8000 Near DDT
10000 DDT DDT
• DDT seems likely for all propane clouds Q9 > 10,000m3
12000 DDT
• DDT seems likely or near for some natural gas clouds 14000 DDT
17000 DDT
20000 N/A DDT
24000 N/A DDT
The figure shows identified DDT potential.
Too few simulations for quantitative evaluation
and to conclude there is no DDT in green cells.
Hull deck LQ LQ
Process deck
LQ LQ
Separation walls
• Reduce cloud sizes
• Give stronger explosions
• DDT predicted for smaller clouds
• Helps ensuring safe distance to LQ
=> Not convincing for risk reduction
“Unsinkable”, double hull designs Iron ore & oil combination carrier
Conclusion:
Must prevent this from happening on FPSOs/FLNGs
Measures:
Limit significant gas cloud build-up at hull deck
Deluge activation at gas detection may limit DDT risk
Working together
for a safer world
Lloyd’s Register and variants of it are trading names of Lloyd’s Register Group Limited, its subsidiaries and affiliates.
Copyright © Lloyd’s Register Consulting. 2013. A member of the Lloyd’s Register group.
Q: For a FPSO, you advised to design so as to prevent DDT; do you have any
suggestions on how this can be done?
A: The only way I know to prevent DDT is to be very efficient at activating the deluge
system upon gas detection. Another solution is to prevent the gas from filling large
volumes at the hull deck. A DDT high up on the process deck would be significantly
less problematic than one occurring on the hull deck. You should therefore make
sure that gas is prevented from going down to the hull deck. If a gas cloud
detonates, it should be as far as possible from the hull deck as extreme loads are
inside the detonating cloud.
Q: Accidental DDTs are very rare and only seem to occur in large clouds in calm
condition. I have never come across DDTs on offshore structures; can you comment
on this?
A: Perhaps not, but does anyone want to be the first to experience a DDT offshore? After
the Buncefield accident which was the first of its kind, many similar events occurred.
We have seen experimentally that a stoichiometric natural gas can detonate in
modules of 28x12x8m; similar conditions can be achieved for a major gas release in
an offshore structure, so I think that a DDT offshore is a possibility.
Q: I understand that in many cases DDT occurs when the pressure rises ahead of the
flame, for instance when a reflected blast wave hits the flame front. As such, do you
think the DPDX parameter is reliable for predicting DDTs as it looks at the difference
between the pressure at the flame front and ahead of it?
A: We have simulated a lot of cases and the DPDX parameter seems quite reliable. What
you mentioned regarding ignition due to reflection seems to be more related to
laboratory experiments within a channel with pressure waves hitting a corner and
detonating there. In the case of large scale experiments, the occurrence of DDT is
mainly linked to what happens at the flame front.
Q: Are you saying that if the difference between the pressure at the flame front and
ahead of it is more than one bar (for instance), then it is very likely that a DDT
will occur?
A: Not exactly. The parameter that we have developed looks at the next grid cell and a
very high flame speed is required to obtain large pressure differences within one grid
cell. More than two bar in the flame front and no pressure ahead seems to have led
to DDT in some experiments, both from the Spadeadam test site and from Hydrogen
experiments from the literature.
Q: Are you simulating everything with the same solver, or do you use one for
deflagration and another one for detonation?
A: It is the same solver, but we do a dump the results, tweak the parameters and
restart the simulation.
A: Reducing congestion will prevent DDTs as it will take longer to reach a DDT in
reduced congestion.