You are on page 1of 30

PHAST & SAFETI

Development Roadmap

Mike Johnson, DNV GL – Head of Product Management and Strategy


August 23, 2016

1 DNV GL © 2016 27 August 2016 SAFER, SMARTER, GREENER


Safeti and Phast 8.0 Scope

2 DNV GL © 2016 27 August 2016


7.2

 Performance work had to be completed as part of NL and 8.0 release, so we are tackling the
problem sooner rather than later.
 Working to expose risk database to connect to offerings like BI as well as working with customers
to enhance their custom reporting.

3 DNV GL © 2016 27 August 2016


8.0

 Scheduled for early Q2 2017.


 Key improvement is the AWD which is a fundamental shift in how we perform non- steady state
dispersion analyses.
 Full MC enhancements will be revisited following v8.0 – AWD is needed to finalize MC.

4 DNV GL © 2016 27 August 2016


8.0

 Remaining scope for v8.0:


– Radiation shielding (verified and validated)
– Improved risk integration for the contouring (Irisk)
– Improved ATEX expansion
– INEX dispersion for instantaneous pressurised releases
– Jet fire enhanced (rotation when hits the ground)
– Time varying jet fire and fireball (initially for the long pipeline)
– Warehouse fire

5 DNV GL © 2016 27 August 2016


New Leak Model for Pipes Between Vessels

T (total)
A Pump at
upstream end
A B B
Valve Breach – Valve downstream
orifice end

Automated model choice: picks the right combination of models based on user
inputs
No new models as such

6 DNV GL © 2016 27 August 2016


Crater Model for Pipeline Releases
Pipe crater exit plane (initial dilution, momentum loss)

Wcrater
Lcrater
SOIL

CRATER
Hcrater
PIPE
Hrelease

 Cases: rupture or punctures (at top, middle or bottom)


 Crater dimensions: length Lcrater, width Wcrater (=Lcrater for punctures), depth Hcrater
 Release depth Hrelease (soil depth to pipe centre for ruptures, soil depth to hole centre for
punctures)
 Presume vertical upwards jet (presume plume not falling back on itself)

7 DNV GL © 2016 27 August 2016


Crater model and dispersion linking

 Model
– Crater geometry (width, depth, etc.) set based on correlations of post-expansion diameter,
fracture length and soil type
– A ‘path length’ is calculated (reflecting the distance travelled between rupture and soil surface
 UDM then uses as its input modified discharge results based on empirical correlations:
– Initial mass fraction is reduced to between 45% and 100% of its post-expansion value
– Initial velocity is reduced to between 15% and 60% of its post-expansion value

8 DNV GL © 2016 27 August 2016


Atmospheric expansion (ATEX) extended validation

Discharge ATEX Dispersion

vessel pipe

flow

Vessel / pipe Atmospheric


orifice pressure

9 DNV GL © 2016 27 August 2016


ATEX Current Models

 Isentropic
– Predicts lower velocities
– Used as the basis of the work used to derive the droplet size correlation for flashing releases
 Conservation of momentum.
– Predicts high velocities, subject to the cut-off applied in Phast (default – 500 m/s)
– Preferred model following literature reviews
 Near-field dispersion
– Turbulence can drive air entrainment, applying a cut-off velocity acts to constrain this
– Kinetic energy changes currently neglected: applies an isenthalpic model instead of
conservation of energy

10 DNV GL © 2016 27 August 2016


Conclusions and possible scope of work for ATEX

 Conservation of momentum gives best results against experiments


 Always apply conservation of momentum (possibly without the velocity cap)
 Requires re-deriving droplet size correlation from experiments based on conservation of
momentum
 Add kinetic energy term to near-field dispersion thermodynamic model

11 DNV GL © 2016 27 August 2016


Dispersion Improvements

 The UDM has been extensively modified:


– Inclusion of a replacement Instantaneous Expansion model (INEX) for ruptures of pressurised
vessels
– Modelling of time-varying releases (including rainout cases) no longer based on continuous
segments
– Improved linking between dispersing clouds and pools
– Inclusion of along-wind diffusion effects due to:
– Passive dispersion (mainly far-field effect)
– Gravity spreading of heavy clouds (mainly near-field effect)

12 DNV GL © 2016 27 August 2016


INEX Model (Instantaneous Expansion) – Old Model

 Old INEX Model had significant limitations. The most important were:
– Droplets always moved upwards with a fixed angle (i.e. no rainout during the INEX phase;
rainout under-estimated)
– Droplets started at the downwind edge of the cloud (rainout could occur outside the bund)

Fixed droplet trajectory during


INEX phase

Droplets start at
edge of cloud Vapour cloud centreline
horizontal

13 DNV GL © 2016 27 August 2016


INEX Model (Instantaneous Expansion) – New Model

 Liquid distributed throughout the cloud


 Droplets expand radially in all directions
 Rainout occurs as the expanding cloud intersects the ground
 Cloud centreline height varies as per standard UDM modelling

Droplets expand in all directions


– rainout as droplets hit the ground
Elevated release

Start of rainout
Cloud is grounded and
Touching down – truncated sphere INEX rainout ends

14 DNV GL © 2016 27 August 2016


AWD: Introduction

 Real clouds can spread in the along-wind direction


– Atmospheric turbulence and mixing
– Gravity-driven spreading of ‘heavy’ clouds
 In Phast / Safeti, approaches to handle this are limited:
– Quasi-Instantaneous transition (QI)
– Finite Duration Correction (FDC)
– Neither work for time-varying or rainout cases; FDC only provides maximum concentration vs
distance (can’t be used for risk)
 Consequences ...
– Stable conditions can produce clouds too short and too wide
– Far-field concentrations can be very significantly over-estimated

15 DNV GL © 2016 27 August 2016


Along Wind Diffusion: What is it and why do we need it?

 What is it?
– AWD is the modelling of dispersing clouds spreading in the along-wind direction due to
atmospheric mixing and turbulence.
 Why do we need it?
– The UDM predicts only steady-state concentration profiles
– For short duration or multi-segment cases we use a simplification: take a ‘slice’ out of the
steady state results for all segments:
– The result:
– Clouds spreads crosswind, but not at all in the release direction
– This keeps concentrations artificially high
– Problem exacerbated in the far field and for short duration releases

16 DNV GL © 2016 27 August 2016


Understanding AWD

 Dispersion involve two mechanisms: Advection (moving/bulk movement) and Spreading (Diffusion)
 Advection typically occurs alongwind (i.e. in the x or release direction)
 Diffusion in the alongwind (x), crosswind (y) and vertical directions (z) (Gaussian)
 Diffusion in the x direction is typically << advection; hence, often ignored
– Most dispersion models account for σy and σz , but not σx
– However, AWD can be significant. especially where advection forces are weak

17 DNV GL © 2016 27 August 2016


Consequences of no AWD

 Stable conditions can produce clouds too short and too wide
– If AWD accounted for – more dilute clouds, increased downwind length
 Far-field concentrations can be very significantly over-estimated
– Especially important for toxic releases where dispersion to low concentrations is the norm
 Lack of AWD is significant for the following releases types
– Short duration releases
– Time varying scenarios
– Releases with rainout
 In Phast / Safeti, approaches to handle this are limited:
– Quasi-Instantaneous transition (QI) and Finite Duration
Correction (FDC)
– Neither work for time-varying or rainout cases; FDC only provides maximum concentration vs.
distance (can’t be used for risk)

18 DNV GL © 2016 27 August 2016


AWD: Modelling a short duration release – Phast vs ‘real’ clouds

Phast / Safeti
Actual
Concentration

Sharp upwind and downwind edges


are eroded as material diffuses along-
wind

Time 1
Concentration over-estimated to an
increasing degree further downwind

Time 2
x(t1-d) x(t1) x(t2-d) x(t2) Distance
19 DNV GL © 2016 27 August 2016
Example run: AWD vs Phast 7 (5 minute phosgene release)

Phast 7
results
Phast AWD results

20 DNV GL © 2016 27 August 2016


Benefits of AWD in Phast & Safeti

 In general reduced concentrations in the medium and far field


 Reduced toxic effects where N values > 1
 More realistic near-field and far-field dispersion
– Elimination of the ‘too short, too wide’ cloud problem
 Better modelling of time-varying releases
 Releases represented as continuous clouds rather than a collection of individual segments…

Phast 7
Phast AWD

21 DNV GL © 2016 27 August 2016


AWD: Verification and Validation

 Verification
– Against FDC results (i.e. maximum concentration vs distance)
 Validation
– Kitfox (short duration CO2 area source)
– Higher predictions;
– Better results for F stability
– LNG / LPG spills (Burro, Coyote, Maplin Sands)
– Overall very similar results
– Expected – Mainly near-field measurements, though physics of pool-cloud linking within the
UDM has been improved

22 DNV GL © 2016 27 August 2016


Gravity spreading

 Clouds can also spread in the along-wind direction due to the ‘slumping’ of heavy clouds
– Analogous to atmospheric mixing, but a different mechanism
– Applies to the near-field rather than far-field
– Only crosswind spreading modelled in Phast 7

Dense cloud released

Spreading upwind and downwind Density differences reduced – along-wind


due to density difference with air spreading becomes dominated by
atmospheric mixing

23 DNV GL © 2016 27 August 2016


Gravity spreading correction: method

 Problem caused by excessive crosswind spreading relative to downwind travel speed (hence a
problem primarily for low windspeed releases)

 Application of a post-dispersion correction ensures the extend of crosswind spreading equals


downwind spreading
Before cloud-shape-correction

 Constraint: incremental cloud area is conserved

Weff(x+Δx)
Crosswind y

Weff cor(x)+0.5ΔS
Weff(x)
add gravity-shape

Weff cor(x)
 Effect is to extend the cloud downwind and
correction such as to
conserve cloud area
make it narrower.
Downwind x
Δxcor = Δx+ΔS

Δx

24 DNV GL © 2016 27 August 2016


Gravity spreading example: continuous chlorine release

 After 100 and 200 seconds, with and without gravity spreading correction

25 DNV GL © 2016 27 August 2016


3.6.2. Time-varying fireballs and jet fires for long pipelines

27 DNV GL © 2016 27 August 2016


Model based on Martinsen and Marx plus standard jet model

 Phases
– Ground-level fireball [before lift-off time tLo = tflame/3]
– increasing fireball radius
– constant SEP (close to HSE model)
– Rising fireball [after lift-off time until fire duration tflame]
– constant fireball radius (close to TNO model)
– rising with constant velocity until maximum flame centre height Hflame=3rflame
– SEP linearly decreasing to 0
– Correlation for fire duration (close to TNO model)
– Followed by standard jet fire model but allowing for changes with time

28 DNV GL © 2016 27 August 2016


Validation of fireball radiation model - Roberts HSE experiments

29 DNV GL © 2016 27 August 2016


Validation of fireball radiation model - Roberts HSE experiments
200
Experiment (Line A)
180 Experiment (Line B)
Experiment (Line C)
160 Experiment (Line D)
Experiment (Line E)
140 TVFM_M&M

120
Radiation dose (kJ/m2)

100

80

60

40

20

0
0 50 100 150 200 250 300
Distance (m)

30 DNV GL © 2016 27 August 2016


Questions or comments?

DNV GL Software Support


software.support@dnvgl.com
1-800-800-7764

www.dnvgl.com

31 DNV GL © 2016 27 August 2016

You might also like