Sie sind auf Seite 1von 25

CFD Based Approach for VCE

Risk Assessment

2009 MKOPSC International Symposium

Anna Qiao, Steven Zhang, Asmund Huser


DNV Energy
Objective
 Determine the maximum design load at a specified
frequency
 Provide a detailed risk picture for large offshore platform
 Provide recommendations on design issues based on CFD
analysis results

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 2
CFD Based Approach Adopted
Start

CFD references database Geometry modelling FLACS Probabilistic


distribution of
leak sizes,
wind rose, etc.

CFD explosion CFD ventilation and


dispersion

Explosion load Ventilation and cloud size


Response surfaces distribution

Probabilistic
assessment
EXPRESS

Acceptec criteria,
QRA, design

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 3
Major Steps
 Identification of critical leak scenarios
 CFD model (FLACS)
- Geometry model – grid calculation
- Ventilation and dispersion analysis
- Explosion analysis

 Probabilistic analysis (EXPRESS)


 Sensitivity studies
 Design recommendations

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 4
Ventilation

 Purpose
- Select the leak location,
leak rate, and wind
speed to find the largest
flammable cloud
- Determine the time to
start leak in dispersion
model

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 5
Dispersion Analysis

 Purpose
– to find the
maximum
flammable
cloud size

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 6
Dispersion Analysis
 Set up a series of scenarios
- Wind speed, wind direction, leak rate, leak direction and leak location are
varied systematically
- Select condition giving largest cloud size
90
CFD Results Buoyant Gas
80 Buoyant Gas Trend
CFD Results Dense Gas
70
Dense Gas Trend
Filling degree Q9 (%)

60

50

40

30

20

10

0
0 1 2 3 4 5 6 7 8 9 10 11
Leak rate (kg/s)

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 7
Approach Limiting CFD Runs
 Problem: CFD analyses are major calculations and time consuming
 Solution: DNV adopted a “Frozen Cloud” approach
 Assumption: a linear relation between gas concentration, c, and leak rate
and the wind speed is assumed for each leak scenario
m
c ( x, y , z ) 
u
 The effect on the flammable gas cloud size from running slightly smaller
and larger leak rates can be assessed without CFD simulation of
additional cases

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 8
Frozen Cloud Results
0.7

0.6

0.5

0.4
Vf/V

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6
R=Qg/Qa

 Case study shows that for this area, the highest filling, 60% of the process area, is
expected to occur when R=0.15 (Ratio of volume flow of combustible gas to air)

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 9
Complete Cloud Response Surface

Huser et al, OTC 12951, 2001

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 10
Explosion Analysis
 Once cloud dispersion is established, the explosion
calculation can also be executed by CFD modeling
 Consider the effect of
- Geometry: size, confinement, obstruction, minimum distance between
decks/plates, corner constant
- Gas mixture: composition, location, quantity

 Critical locations for explosion load determination


- Control rooms, vessels, columns, decks

 Critical loads is a combination of:


- The maximum pressure pulse measured on monitors is used as
reference pressure
- Maximum drag

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 11
Explosion Analysis from FLACS

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 12
Explosion Database of CFD Cases and Experiments

 19 explosion models
- One process platform (plant) for each model
- 15 to 50 explosion simulations performed in FLACS for each model
- Maximum explosion overpressure used to develop the correlations

 15 experiments
- Various module sizes
- Different congestion levels
- Different explosion venting areas

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 13
Simulation Overpressure vs. Cloud Size Trend

5/3
P ~V f

Huser et al, Journal of Loss Prevention in the Process Industries, 22 (2009), 324
© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 14
Smart Process

Database
New Installation

Conditions New Ventilation New Dispersion New Explosion

Exceedence Curve
Frequency

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 15
Explosion Risk Analysis, EXPRESS
 Purpose – Frequency of exceeding explosion load vs the
load
 Selected critical loads are used in probabilistic analysis
 DNV Probabilistic tool EXPRESS is applied to find pressure
exceedence curves
- EXPRESS establishes a response surface using a limited number of
CFD runs
- The response surface rapidly generates the exceedance results
- Pressure on other targets are found as a linear relationship between
the reference pressure and the other target

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 16
CFD Based Approach Adopted
Start

CFD references database Geometry modelling FLACS Probabilistic


distribution of
leak sizes,
wind rose, etc.

CFD explosion CFD ventilation and


dispersion

Explosion load Ventilation and cloud size


Response surfaces distribution

Probabilistic
assessment
EXPRESS

Acceptec criteria,
QRA, design

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 17
Frequency/Probability
 Leak frequency (for each leak category)
 Probability of ignition – Time Dependent Internal Ignition Model (TDIIM)
- Pump
- Compressor
- Generator
- Electrical equipment
- Personnel
- Hot work
- Other

 Explosion frequency

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 18
Time Dependent Internal Ignition Model (TDIIM)
 Intermittent ignition sources
- The total cloud size

 Continuous ignition sources


- The increase in cloud size

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 19
Monte Carlo Simulation
 Select specific outcomes of all stochastic variables from their statistical
distributions
 Calculate ignitable time dependent cloud size and ignition probability
 Select ignition time from the ignition time probability distribution
 Calculate explosive cloud size at time of ignition
 Calculate explosion loading from explosion load response surface

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 20
Pressure Exceedence Curves
1.E-02
Accumulated frequency (1/year) .

ESD close after 60 s

ESD close after 30 s


1.E-03

1.E-04

1.E-05

1.E-06
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Explosion pressure load (barg)

Reduction in explosion load due to rapid initiation of ESD

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 21
Prevention/Mitigation
 Leak frequency reduction
 Localized enclosure
 Ventilation
 Deluge
 Gas detection, shutdown and blowdown
 Reduced number of ignition sources
 Blast wall

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 22
Why Method is Appropriate
 Experimental explosion results used for the development of
the correlation
 CFD results directly applied to find cloud dispersion/size and
resulting maximum explosion pressure
 Special techniques allow fewer CFD runs
- Frozen cloud approach
- Response surface approach

 Combined CFD and probabilistic analysis gives the detailed


risk picture

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 23
Typical Risk Based Design Activities for Offshore
 Explosion – FLACS & EXPRESS
 Fire – KFX, PFPro & EXPRESS
 WCI – Wind Chill Index
 Ventilation - ACH
 Gas detector location - GDOZ

© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 24
© Det Norske Veritas AS. All rights reserved 03 December 2009 Slide 25

Das könnte Ihnen auch gefallen