Probabilistiske brannrisikoanalyser til bruk i design 2013/Brannsikkerhet/7... · Probabilistiske...
Transcript of Probabilistiske brannrisikoanalyser til bruk i design 2013/Brannsikkerhet/7... · Probabilistiske...
Probabilistiske brannrisikoanalyser til bruk i design
Robust brannsikkerhet for offshoreinnretninger
Temadag Petroleumstilsynet 24. oktober 2013
Asmund Huser og Jens J. Garstad DNV GL
© Det Norske Veritas AS. All rights reserved.
Content
Overview
- NORSOK vs FABIG heat flux
- Approaches in DNV and general
Why probabilistic method
DNVs advanced probabilistic method
- Design Accidental Load (DAL)
- Response analysis
- PFP optimzation
Simplified probabilistic analysis
- Fire DAL
Example
Conclusions
2
© Det Norske Veritas AS. All rights reserved.
Standards and guidelines
Norsok S-001:
- Gives heat fluxes
- Opens for probabilistic DAL to find
fire sizes
Guideline for the protection of
pressurized systems exposed to
fire
- Use scenario with Max duration
(worst case)
- Duration set by cut off leak rates
and actual inventory
FABIG Technical Note 11
- Gives heat flux for gas jet, two
phase jet and pool
- Gives probabilistic procedure
3
© Det Norske Veritas AS. All rights reserved.
Comparisons FABIG and NORSOK
Leak rate (kg/s): 0.1 1 10 30
FABIG Gas jet fires 180 250 300 350
Two phase jet fires 200 300 350 400
Pool fires 125 125 250 250
NORSOK Jet fire 250 250 350 350
Pool fire 150 150 150 150
4
© Det Norske Veritas AS. All rights reserved.
Total heat flux (kW/m2) comparison FABIG vs. NORSOK
0
50
100
150
200
250
300
350
400
450
Gas jet firesFABIG
Two phase jetfires FABIG
Jet fireNORSOK
Pool fire FABIG Pool fireNORSOK
0.1 1 10 30 kg/s
5
© Det Norske Veritas AS. All rights reserved.
Fire Design Accidental Load challenges
Definition: Fire DAL is a flux (kW/m2) and duration
- Flux is given in guidelines or from CFD models
- Duration needs to be decided based on actual process
- different strategies:
Approaches to set DAL
- Scenario based:
- Credible scenario picked from QRA
- Worst case scenario picked from QRA
- Probabilistic approach:
- Different x-axis strategies
- Different levels of detail
Different application areas needs different approaches
- For general DAL spec
- PFP optimization on structure
- PFP optimization on piping and equipment
Evolution in approach and knowledge
6
© Det Norske Veritas AS. All rights reserved. Slide 7
Highlights of DNVs advanced procedure
Includes dynamic, 3D effects of all possible fires.
- Each fire described by one number: Max dose received
Fire analysis with same level of detail as NORSOK
explosion analysis
Probabilistic fire analysis is used in optimization of:
- Design Accidental Load (DAL) for fires – also duration
- Flare; ESD, blowdown
- PFP on structure and pipes
- Fire/explosion dividers, and arrangement
© Det Norske Veritas AS. All rights reserved.
Why Probabilistic and CFD?
Optimization of mitigation
Consistent quantitative method
- All areas protected to the same risk level
Fire and explosion risk presented on the same format
- Opposing mitigating measures
Research show large fires can give very high heat
fluxes
- Captures only with CFD
© Det Norske Veritas AS. All rights reserved.
Why CFD and FE fire model?
Research gives (Ragnar Wighus 2008):
- Heat flux most dependent on size of flames
- Heat flux also dependent on amount of equipment
and turbulence generating objects
- Oil (and wood) can give same heat flux as gas
- More PFP gives more heat – vicious cycle
Generalized fire loads can be too conservative wrt
fire protection – not explosion
© Det Norske Veritas AS. All rights reserved.
”Confined fires” measurements, Phase 2, 1998
© Det Norske Veritas AS. All rights reserved.
Test Barrels w/wo PFP
With PFP, 50 kW/m2 higher radiation
With permission from Ragnar Wighus, Sintef NBL
© Det Norske Veritas AS. All rights reserved. Slide 12
Combined analysis: Optimal mitigation
Schematic illustration of optimal passive fire protection. In this example, fire and
explosion escalation frequencies are crossing each other.
© Det Norske Veritas AS. All rights reserved.
Background for development of probabilistic fire analysis
Standards and codes gives deterministic, constant fireloads
- Ex. Jet fires > 2 kg/s, use 350 kW/m2
- ”Average maxvalue” is often conservative
Development of NORSOK Z013-G procedure explosion 1997
DNV started to develop this probabilistic fire method in 2001 with Statoil
- Investigate important effects
- Develop a procedure
- Test it on a real platform
FABIG Newsletter 44, 2005 describes the first procedure
FABIG TN 11 2009 Appendix A describes the procedure
ExpressFire program developed 2006
PFPro developed 2007
Apply procedure in many projects
© Det Norske Veritas AS. All rights reserved.
Some QRA and Safety design methods used in DNV
Guidelines/
standards
Simplified
methodsMore detailed
Load AnalysisProcedures, Loads,
tabulated
Experience
based/ empiric/
integral models
CFD, FE,
MonteCarlo,
advanced event
tree
Explosion max pressures DNV-OS-A101 THOR.xls FLACS
Probabilistic explosion
analysis
NORSOK Z013,
DNV-OS-A101ExpressLite Express
Fire load and sizeFABIG TN 11, Scp
GuidelinePhast KFX
Probabilistic fire
analysisFABIG TN 11
Selected
"Credible
Scenario"
ExpressFire
Structural response
FABIG TN 11,
OGUK FX Guidel.,
API RP 2FB,
EUROCODE
PFPro FAHTS/USFOS
All Risk analysis, safety case NORSOK Z013 SOQRATESRisk Framework/
Neptune
Explosion
Fire
PhastRisk
Offshore
Excel
model
© Det Norske Veritas AS. All rights reserved.
FRA procedure
Initial Fire Risk
Assessment
CFD Fire
Modeling
Fire Risk
Assessment
Leak
Frequencies
Ignition
Probabilities
Inventories
Fire Location
Highest Ranked
Cases
Heat Flux
Fire
Frequencies
Risk
Acceptance
Criteria
Fire Frequencies
Risk Rank Cases
Heat Flux for
cases
Fire Exceedance
Curves
DAL Fire and
Load
Input Assessment Output
Sectio
n 1
DAL Fire and
Load
Structure
temperature and
strain response
Simulation
Temperature and
strain response in
structure
Modify
PFP
Modify
Flare,
ESD and
BD
Minimum PFP
Sectio
n 2
Risk
evaluation
Global
Collapse? Strain or
temperature
Criteria
End
Start
Initial Fire Risk
Assessment
CFD Fire
Modeling
Fire Risk
Assessment
Leak
Frequencies
Ignition
Probabilities
Inventories
Fire Location
Highest Ranked
Cases
Heat Flux
Fire
Frequencies
Risk
Acceptance
Criteria
Fire Frequencies
Risk Rank Cases
Heat Flux for
cases
Fire Exceedance
Curves
DAL Fire and
Load
Input Assessment Output
Sectio
n 1
DAL Fire and
Load
Structure
temperature and
strain response
Simulation
Temperature and
strain response in
structure
Modify
PFP
Modify
Flare,
ESD and
BD
Minimum PFP
Sectio
n 2
Risk
evaluation
Global
Collapse? Strain or
temperature
Criteria
End
Start
© Det Norske Veritas AS. All rights reserved.
Section 1 DAL fire
Initial Fire Risk
Assessment
CFD Fire
Modeling
Fire Risk
Assessment
Leak
Frequencies
Ignition
Probabilities
Inventories
Fire Location
Highest Ranked
Cases
Heat Flux
Fire
Frequencies
Risk
Acceptance
Criteria
Fire Frequencies
Risk Rank Cases
Heat Flux for
cases
Fire Exceedance
Curves
DAL Fire and
Load
Input Assessment Output
Sectio
n 1
DAL Fire and
Load
Structure
temperature and
strain response
Simulation
Temperature and
strain response in
structure
Modify
PFP
Modify
Flare,
ESD and
BD
Minimum PFP
Sectio
n 2
Risk
evaluation
Global
Collapse? Strain or
temperature
Criteria
End
Start
Initial Fire Risk
Assessment
CFD Fire
Modeling
Fire Risk
Assessment
Leak
Frequencies
Ignition
Probabilities
Inventories
Fire Location
Highest Ranked
Cases
Heat Flux
Fire
Frequencies
Risk
Acceptance
Criteria
Fire Frequencies
Risk Rank Cases
Heat Flux for
cases
Fire Exceedance
Curves
DAL Fire and
Load
Input Assessment Output
Sectio
n 1
DAL Fire and
Load
Structure
temperature and
strain response
Simulation
Temperature and
strain response in
structure
Modify
PFP
Modify
Flare,
ESD and
BD
Minimum PFP
Sectio
n 2
Risk
evaluation
Global
Collapse? Strain or
temperature
Criteria
End
Start
© Det Norske Veritas AS. All rights reserved.
Initial Fire Risk Assessment
Coarse screening to find high risk cases
Risk is represented by frequency and mass of HC in segment
Use high risk cases as representative in CFD model
© Det Norske Veritas AS. All rights reserved.
CFD fire modeling
Use parameters from cases with highest risk (composition, etc.)
Obtain effects of varying leak rates
Model impinging and non impinging jets, and pools
- Obtain effects of varying jet direction
- Obtain effects of varying leak location
Gives two main results:
- Gives max heat loads for all fire case as a function of time: Input to fire risk analysis
- Real 3D dynamic heat load fields: Input to Structure response analysis
© Det Norske Veritas AS. All rights reserved.
CFD model captures the fire dynamics
Ventilation and air/fuel flow
Combustion
Congestion
Flame thickness
Scale effects
Radiation
Convection
Deluge
© Det Norske Veritas AS. All rights reserved.
Dynamic temperature plots
N
Z
30 kg/s
10 kg/s
3 kg/s
1. kg/s
0.3 kg/s
© Det Norske Veritas AS. All rights reserved.
Heat flux histories
0
50
100
150
200
250
300
350
0 1 2 3 4 5
Dimensionless time
Q (
kW
/m2)
1
2
3
4
30 kg/s 10 kg/s 3 kg/s 1 kg/s 0.3 kg/s
© Det Norske Veritas AS. All rights reserved.
Heat flux histories
0
100
200
300
400
500
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Tbar (-)
Q (
kW
/m2)
Qmax_L1
Qmax_L2
Qmax_L3
Qmax_L4
Qmax_L5
© Det Norske Veritas AS. All rights reserved.
Fire risk analysis principles
Exceedance curve is based on one representative number describing
severity for each case
Rank all fires according to this one number
This number is on x-axis on exceedance plot
Y- axis is the accumulated frequency
Use separate curves for oil and gas
Where total curve crosses 10-4 per year is the DAL fire
The target also decides the x-axis number used
For structure, the total received heat during scenario is used
Can be described by a dose
When the DAL dose is found, it must be converted back to a flux and
duration.
23
1.0E-05
1.0E-04
1.0E-03
10 100 1000
Heat dose received by structure (MJ/m2)
Acc
um
ula
ted
fir
e im
pa
ct f
req
uen
cy (
per
yea
r)
© Det Norske Veritas AS. All rights reserved.
Fire risk analysis
Sorted
Case number and name Phase Size Heat dose (kWs/m
2)
Collapse frequency
(per year)
Cumulative frequency
(per year)
Case 1, LP stages gas Large 20 1.04E-04 3.59E-04
Case 2, HP stage gas Large 51 1.07E-04 2.56E-04
Case 1, LP stages gas Medium 108 5.71E-05 1.48E-04
Case 2, HP stage gas Medium 127 6.44E-05 9.12E-05
Case 1, LP stages gas Small 142 1.56E-05 2.68E-05
Case 2, HP stage gas Small 177 1.12E-05 1.12E-05
Find dose for each case. Integrate in each CFD cell and use max:
- Dose: 𝑫𝑽 = 𝑸 𝒕 𝒅𝒕 (J/m2), where Q(t) (kW/m2) is heat flux
Find fire impact frequency for each case
- Impact freq. = leak frequency*ignition probability*impact probability
Sort cases with increasing heat dose
- from case study:
© Det Norske Veritas AS. All rights reserved.
Plot fire impact frequency vs. dose
1.0E-05
1.0E-04
1.0E-03
10 100 1000
Heat dose received by structure (MJ/m2)
Acc
um
ula
ted
fir
e im
pact
fre
qu
ency
(p
er y
ear)
DAL Dose = Q t
t = Dose/250 kW/m2
DAL:
Q = 250 kW/m2
t = 8.3 min
© Det Norske Veritas AS. All rights reserved.
Pick DAL fire scenario
1.E-06
1.E-05
1.E-04
1.E-03
Large Large Medium Medium Small Small
gas gas gas gas gas gas
Case 1, LP
stages
Case 2, HP
stage
Case 1, LP
stages
Case 2, HP
stage
Case 1, LP
stages
Case 2, HP
stage
Sorted after heat dose received by structure (MJ/m2)
Acc
um
ula
ted
im
pa
ct f
req
uen
cy (
per
yea
r)
© Det Norske Veritas AS. All rights reserved.
Leak profile: Medium is DAL fire
0.1
1
10
100
0 5 10 15 20
Time (min)
Lea
k r
ate
(k
g/s
)
Very large
Large
Medium
Small
© Det Norske Veritas AS. All rights reserved.
Heat flux from DAL fire (medium)
5 kg/s 1.25 kg/s 0.5 kg/s
© Det Norske Veritas AS. All rights reserved.
Max heat flux medium fire (DAL fire)
0
100
200
300
400
500
0 0.5 1 1.5 2 2.5
Tbar (-)
Q (kW
/m2
)
0 6 10 min
© Det Norske Veritas AS. All rights reserved.
Section 2: Structure response analysis
© Det Norske Veritas AS. All rights reserved.
Section 2: Structure response analysis
Initial Fire Risk
Assessment
CFD Fire
Modeling
Fire Risk
Assessment
Leak
Frequencies
Ignition
Probabilities
Inventories
Fire Location
Highest Ranked
Cases
Heat Flux
Fire
Frequencies
Risk
Acceptance
Criteria
Fire Frequencies
Risk Rank Cases
Heat Flux for
cases
Fire Exceedance
Curves
DAL Fire and
Load
Input Assessment Output
Sectio
n 1
DAL Fire and
Load
Structure
temperature and
strain response
Simulation
Temperature and
strain response in
structure
Modify
PFP
Modify
mitigation
measures
Minimum PFP
Sectio
n 2
Risk
evaluation
Global
Collapse? Strain or
temperature
Criteria
End
Start
Initial Fire Risk
Assessment
CFD Fire
Modeling
Fire Risk
Assessment
Leak
Frequencies
Ignition
Probabilities
Inventories
Fire Location
Highest Ranked
Cases
Heat Flux
Fire
Frequencies
Risk
Acceptance
Criteria
Fire Frequencies
Risk Rank Cases
Heat Flux for
cases
Fire Exceedance
Curves
DAL Fire and
Load
Input Assessment Output
Sectio
n 1
DAL Fire and
Load
Structure
temperature and
strain response
Simulation
Temperature and
strain response in
structure
Modify
PFP
Modify
mitigation
measures
Minimum PFP
Sectio
n 2
Risk
evaluation
Global
Collapse? Strain or
temperature
Criteria
End
Start
© Det Norske Veritas AS. All rights reserved.
Two acceptance criteria:
Temperature < 500 C in steel (see Chapter 6 in TN11)
- Can do local heat transfer analyses of single beams
Strain criteria: Global collapse not accepted
- Stress < Yield stress,
- Plastic utilization < 1.
- Use FE model
- Collapse of single beams is accepted if loads are re distributed
© Det Norske Veritas AS. All rights reserved. 33
Structural Response analysis
Context and input
• Calculate Structural response for structure, equipment and
pipe support given the dimensioning fire scenario
© Det Norske Veritas AS. All rights reserved. 34
Structural Response analysis
Context and input
Interface fire analysis - structure analysis
• Calculate Structural response for structure, equipment and
pipe support given the dimensioning fire scenario
• Identify the fire case at 10-4 per year. (duration and KFX case
with 3D transient heat flux development)
• Dimensioning fire deveolopment represented by a series of KFX
simulation snap shots with corresponding duration
© Det Norske Veritas AS. All rights reserved. 35
Structural Response analysis
Context and input
Interface fire analysis - structure analysis
Short summary
• Calculate Structural response for structure, equipment and
pipe support given the dimensioning fire scenario
• Identify the fire case resulting in the dimensioning fire dose
(duration and transient heat flux development)
• Dimensioning fire case represented by a series of KFX
simulation snap shots with corresponding duration
i. Temperature development in steel calculated using FAHTS
ii. Structural response calculated using USFOS
iii. Fire location moved around to reflect many possilbe leak
locations
iv. Optimize design – ESD and PFP
© Det Norske Veritas AS. All rights reserved.
Dimensioning fire scenario
Steady state simulation for const. leak
rates: (jet or pool)
- 50 kg/s
- 25 kg/s
- …
Dimensioning fire deveolopment represented by
a series of KFX simulation snap shots with
corresponding duration
36
25 kg/s
10 kg/s
5 kg/s
1.25 kg/s
0
5
10
15
20
25
30
0 500 1000 1500
Lea
k R
ate
(kg
/s)
Time (s)
Transient
Steady
© Det Norske Veritas AS. All rights reserved.
Temperature development in steel
37
Example of temperature development calculated with FAHTS representing the
dimensioning fire with KFX simulation results
© Det Norske Veritas AS. All rights reserved.
Structural response/utilization calculated using USFOS
38
Example of utilization calculation using USFOS to determine possilbe yielding or
plastic deformation of structure
© Det Norske Veritas AS. All rights reserved.
Example of fire locations
39
«moving» the simulated fire around to reflect many possible leak locations
© Det Norske Veritas AS. All rights reserved.
Passive fire protection
40
Established in an iterative process, and with many leak locations of the
dimensioning fire load
© Det Norske Veritas AS. All rights reserved. 41
Optimization in iterative process
• Reduce loads by ESD
isolation and Blowdown
• Passive fire Protection
optimization
Calculate
dimensioning
fire loads
fKFX simulations
Structural
response
Evaluation Update ESD
Update PFP
© Det Norske Veritas AS. All rights reserved.
Simplified FRA assessment
Step 1: Initial fire risk assessment
- List of scenarios with associated fire frequency (leak frequency, ESD/BD failure probabilities
and ignition probabilities)
- List and Locations of fire scenarios (coordinates) and targets
Step 2: Simplified Fire consequence modelling
- Duration based on
- Leak profile calculations with cut-off 0.1 kg/s for jet fires, considering detection, isolation and BD
- Simple pool fire modelling considering size of pool and constant combustion rate per m2 for pool fires
- Apply generic heat flux peak load values from Table 1, NORSOK S-001, Edition 4
- Exposure assessed based on flame lengths calculated either with Phast or simplified
formulaes for jet fire length as function of release rate (depending on congestion and
obstacles)
Step 3: Fire risk assessment
- Coarse assessment by use of Excel calculating heat dose exposing each target from each
scenario case (leak size, ESD/BD failure case)
Step 4: Risk evaluation
- Fire dose exceedance plots and fire DAL heat dose and duration
42
© Det Norske Veritas AS. All rights reserved.
Simplified fire consequence calculations
Leak profile example, after successful isolation and 30 sec delay BD
Flame lengths are estimated f.ex. by use of API RP 521, 1989 method
where Q is the leak rate and Hc is the heat of combustion (J/kg)
43
478.0)(00326.0 cf QxHL
© Det Norske Veritas AS. All rights reserved.
Results
Duration or fire dose exceedance curve
44
1,E-08
1,E-07
1,E-06
1,E-05
1,E-04
1,E-03
0 5 10 15 20 25 30
Ex
ceed
ence
fre
qu
ency
(p
er y
ear)
Duration (minutes)
Separator
© Det Norske Veritas AS. All rights reserved.
Summary – a consistent way of design
Advanced CFD/FE and probabilistic approach gives a more realistic and consistent
fire load than simplified methods
- Includes real fire effects which locally can be worse then averaged
- In general shorter fire durations
- Used for optimization of PFP
Fire exceedence plot is common intermediate result
- Max dose describes size of the fire by one number
- Gives the DAL fire and scenario to be used from CFD
- Gives the escalation probability to QRA
A consistent CFD, FE and probabilistic approach analysis can optimize all
mitigating measures
- Risk for fires and explosions can now be compared on the same detailed foundation
- Easier to make decisions about mitigating measures
- Firewalls and arrangement improved
Discussion
- Develop a NORSOK Z013 Probabilistic fire analysis procedure
© Det Norske Veritas AS. All rights reserved.
46
© Det Norske Veritas AS. All rights reserved.
Safeguarding life, property
and the environment
www.dnv.com
47