Professional Documents
Culture Documents
Camille Azzi
Gexcon
E: camille.azzi@gexcon.com
T: +47 94825662
Robert Brewerton
Natabelle Technology
E: rwb@natabelle.co.uk
T: +44 1732465465
Camille Azzi
June 2014
© GexCon AS
2
www.gexcon.com
© GexCon AS
3
www.gexcon.com
BLASTWALL
© GexCon AS
4
www.gexcon.com
• Dispersion
• Described in more details (topic of presentation)
• Explosion
• 7 cloud size categories distributed over the whole area with
centre and edge ignitons
• Risk calculations
• Time dependent ignition model
© GexCon AS
5
www.gexcon.com
© GexCon AS
6
www.gexcon.com
FORWARD
WIND
© GexCon AS
7
www.gexcon.com
© GexCon AS
8
www.gexcon.com
© GexCon AS
9
www.gexcon.com
© GexCon AS
10
www.gexcon.com
© GexCon AS
11
www.gexcon.com
© GexCon AS
12
www.gexcon.com
BLASTWALL (1E‐04/yr)
0.180
0.160
0.140
Overpressure [barg]
0.120
0.100
0.080
0.060
Various Configurations
0.040
Central Rows
0.020
Central Rows & Colums
0.000
0 2 4 6 8 10 12 14 16 18 20 22 24
Number of Leak Locations
© GexCon AS
13
www.gexcon.com
Explosion Risk:
1 location / All directions
BLASTWALL (1E‐04/yr)
0.180
0.160
0.140
Overpressure [barg]
0.120
0.100
0.080
0.060
Various Configurations
0.040
Central Rows
0.020
Central Rows & Colums
0.000
0 2 4 6 8 10 12 14 16 18 20 22 24
BLASTWALL (1E‐04/yr)
0.180
0.160
0.140
Overpressure [barg]
0.120
0.100
0.080
0.060
Various Configurations
0.040
Central Rows
0.020
Central Rows & Colums
0.000
0 2 4 6 8 10 12 14 16 18 20 22 24
Explosion Risk:
Various locations / directions
BLASTWALL (1E‐04/yr)
0.180
0.160
0.140
Overpressure [barg]
0.120
0.100
0.080
0.060
0.040
Each analysis: ~180
0.020
sim
0.000 (720 with FC)
© GexCon AS
17
www.gexcon.com
Acknowledgements
© GexCon AS
18
www.gexcon.com
25-06-14 1
Outline
• The use of anticipated congestion and application of
safety factors to derived explosion loads
• Defining DAL loads
• What to do about high drag pressures
• NLFEA examples and DAL input requirements
• The probabilistic / deterministic dilemma
• Conclusions
25-06-14 2
Acknowledgements
• Gexcon (advances in load definition in FLACS)
• Arup advanced Technology Group (LSDYNA NLFEA)
• Booth Industries Ltd (some design examples)
25-06-14 3
25-06-14 4
Structural
Equip 1 + layout 1
Implement
large pipes
anticipated
congestion
CAD topsides
Layout 1
Equip 2 + 1st Explosion
pressures Response
medium pipes
analysis analysis 1
Implement in
CAD model Structural
layout 2
Implement
anticipated
congestion
CAD topsides
Layout 2
Equip 3 + 2nd Explosion Response
small pipes pressures analysis 2
analysis
Implement in Structural
CAD model details final
Final expl.
CAD topsides
pressures
Layout 3
analysis
Start
Safety Case Construction
25-06-14 5
25-06-14 6
25-06-14 7
Comment: 60 km pipe
means about 5000 lines,
15000 piping Iso’s and
35000 pipe supports to
engineer
25-06-14 8
25-06-14 9
For significance of
the dotted yellow
lines see text
Dotted lines indicate loadings with 1.5 safety factor and show that, in this case,
applying it is like using 3 x 10-5 pressure in place of 10-4 value
25-06-14 10
Pmax
Impulse value
pressure
td = shaded area
Using 3 x 10-5 exceedence pushes up impulse by 50%,
25-06-14 12
In this case applying 1.5 safety factor is like using 5 x 10-5 pressure in place of
10-4 value
25-06-14 13
25-06-14 14
25-06-14 16
25-06-14 17
25-06-14 20
25-06-14 21
Comments
• Level 1 model is simple and essential to develop the
design
• Level 1 model potentially lends itself to probabilistic
response analysis
• Level 2 model is the definitive structural analysis and
definitely has to be deterministic with scenarios selected
from the Level 1 analysis
• Level 3 analysis is essential to check local failure modes /
code check.
25-06-14 22
25-06-14 23
25-06-14 24
25-06-14 25
506
503 504
501 502
505
25-06-14 26
Point 501
0.20
Point 502
Point 503
0.15
Point 504
P ressu re [b ar]
Point 505
0.10
Point 506
0.05
0.00
-0.05
-0.10
0.75 0.80 0.85 0.90 0.95 1.00 1.05
Time [s]
25-06-14 27
25-06-14 28
25-06-14 29
25-06-14 30
Piping Design:
Implement Monitor Point Arrays into FLACS for Drag
25-06-14 31
25-06-14 32
Conclusions
1. Always apply a safety factor to explosion overpressures and only get rid
of it at as built stage.
2. NLFEA methods require carefully selected DAL input.
3. If you want to run NLFEA probabilistically you need very coarse models.
4. Coarse models will not properly code check the response and
demonstrate survival: you need refined models and deterministic load
cases.
5. Piping and equipment analysis can be difficult: there are potentially
many thousands of lines, nozzles and supports to check, hence keep
pressures down so that the task is manageable and late design
changes are reduced.
6. Be very careful about your SCE designation: it affects both safety and
workload.
7. Beware of thin walled pipe in high blast areas.
25-06-14 33
Thank you
Robert Brewerton
Natabelle Technology Ltd
www.natabelle.com
25-06-14 35
Q: You talked about applying a safety factor of 1.5 to the explosion loads. I understand
that a safety factor can either be applied by increasing the load or by using a lower
frequency criterion. You opted for the former; what is the reason for this?
A: In projects, you actually have to use a lower frequency criterion and I tried to show in
my presentation a way of determining which reduced frequency can be used. There
are also other ways; one can take a more conservative approach on the gas cloud
size for instance. The problem with such approach is that this can lead to arguments
with the regulator as it may look like you are shifting the goal posts over the course
of the project.
Q: Why does Gexcon use the Q9 approach to quantify the level of volume of flammable
gas cloud?
A: This question has been ongoing for some time and I am probably not be the best
person to answer. I would however say that it has been used for many years and has
been tested and compared to inhomogeneous clouds. I understand that some
companies use other approaches.
A: This is because ignition sources will be detected and isolated with time, and
personnel will also be evacuated.
Q: You showed an escape tunnel in your presentation. For a long tunnel, could you
please explain how you take into account the changes in explosion load distribution
as the explosion propagates along the tunnel?
A: This is carried out by having a panel for each section of the tunnel, so that each span
of the tunnel is analysed according to the load on that particular panel. Regarding
time considerations, if you have not done the analysis comprehensively using non-
linear FEA, we generally find that the speed of the explosion wave along the tunnel is
not sufficient to generate a “snake like” response.