Vous êtes sur la page 1sur 31

Blast and quasi-static pressure in

partially confined geometries:


Empirical, analytical and numerical modeling

E. LAPBIE, A. OSMONT, L. YOUINOU, R. SOULI, A. GENOT : CEA/DAM/Gramat


S. BARROT : LCPP
Corresponding author : emmanuel.lapebie@cea.fr

IPSF - APRIL 2015


18 MAI 2015

CEA | 10 AVRIL 2012 | PAGE 1

HEADLINES

CEA/Gramat overview
Main activities

P.04

Blast-related studies

P.05

Blast wave and QSP modelling


Empirical models

P.08

Numerical models

P.16

Analytical approaches

P.20

Conclusions
Choosing a model

P.27

Way forward

P.29

18 MAY 2015

CEA | APRIL 2015 | PAGE 2

CEA/Gramat overview

CEA | 10 AVRIL 2012


| PAGE 3
18 MAI 2015

CEA/Gramat overview
Main activities: weapon effects
CEA/G history began just after WW II with the study of nuclear weapons effects
through the use of large-scale experimental facilities:
Blast and thermal effects.
Ground shock propagation / interactions with buried structures.
E/M Impulse and X-Ray generation.

The activities extended to the effects of conventional weapons:


HE physics (initiation, detonation propagation, direct effects, ).
Lethality of warheads and vulnerability of systems (aircrafts, infrastructures, ).

More recently, our scope broadened to global safety:


Physics of transient CBR Source Terms.
Improvised explosives.
Effects of IEDs on light / unprotected structures.
Domino effects in chemical industries.

Our expertise is based on both experiments and modelling.


18 MAI 2015

CEA | APRIL 2015 | PAGE 4

CEA/Gramat overview
Blast-related studies

1/2

Free-field HoB setup

Free-field:
Blast effect of OTS and improvised high explosives.
Characterization of blast (and fragments) from ammunitions.
Semi-confined setup

Urban:
Various projects on the consequences of bombings.
Effects of VB-IEDs on buildings.
Prediction of collateral effects from air strikes.

Confined geometries:

Bunker setup

Detonations in a vented bunker.


HE in commercial aircrafts.
Weapons effects in multi-room facilities, including EBXs.
Effects of IEDs in tunnels / metro stations (with LCPP).

Urban setup

18 MAI 2015

CEA | APRIL 2015 | PAGE 5

CEA/Gramat overview
Blast-related studies

2/2

Free-field:
Overpressure time-history = f(R) or f(Z).
Second shock (modelling of afterburning ).
Empirical TNT equivalence for pressure and impulse.

Free-field results

Urban:
Reflection (normal, Mach ).
Canyon propagation.
Diffraction.
Interaction.

Bunker setup : QSP build-up

Confined geometries:
Superposition of:
- Blast waves (complex due to multiple reflections).
- Quasi-static Pressure (QSP) build-up.
Effects of:
- Venting (doors, windows)
- Collapse of partition walls / main walls.
18 MAI 2015

Bunker setup : venting


CEA | APRIL 2015 | PAGE 6

Blast wave and QSP modelling

CEA | 10 AVRIL 2012


| PAGE 7
18 MAI 2015

Blast wave and QSP modelling


Empirical models Free field

(1/4)

Numerous empirical models


Modified Friedlander [from Baker]
Friedlander shape only.
All parameters (P, I, t ) or some of them only.
Abaci / explicit formulations / Excel worksheet (BEC).
Assume TNT equivalence to compare HEs.

Historical abaci

Issues
Scaling (Hopkinson, Sachs, ).
Reproducibility of experiments.
Nature of the HE & initiation train (booster ?).
Confidence in experimental setup / measures.

From Cooper 1994 see also MSIAC L-132


Blast
18 MAI
2015wave parameters

Hopkinson scaling
(the simplest approach)

CEA | APRIL 2015 | PAGE 8

Blast wave and QSP modelling


Empirical models Free field

(2/4)

Comparison of empirical models Overpressure


From spherical or hemispherical bursts (ground reflection coefficient < 2).
Huge discrepancies for small to intermediate Z values but reproducibility ?
Even worse for other parameters (I+, t+, ) when the models are available.

Comparison of some empirical overpressure models at intermediate (left) and long range (right).
18 MAI 2015

CEA | APRIL 2015 | PAGE 9

Blast wave and QSP modelling


Empirical models Free field

(3/4)

Adding consequence models [TNO Green book]


Probabilistic models for casualties / injuries + levels of damage to buildings.
Models depend on DP and I+ (impulse scales with the lengthscale).
Zmin threshold for 100% casualties and 100% building collapse for a given GE mass
no need for precise computations below this limit (but still not sufficient to
compensate from variations among models).

1,3
ton

100%
Death
100%
Collapse

CEA | APRIL 2015 | PAGE 10

18 MAI 2015
Comparison of some empirical overpressure models

Blast wave and QSP modelling


Empirical models Free field

(4/4)

Scaling considerations : mass or energy ?

R
E P0 1/ 3

Sachs scaling

Cube-root scaling based on HE mass is the simplest scaling, and is prominently used for HEs.
Energy scaling is a true (non-dimensional) scaling and accounts for variations of P0 (many
experiments have been performed with P0 standard atmospheric pressure)
Replacing mass scaling by energy scaling allows for comparison with other explosions (BLEVEs,
VCEs, pressurized vessels bursts) provided that the proper energy is considered many
references in the literature on industry accidents.

Applications to high explosives


For HEs, the energy transferred in the blast wave is only a fraction of the total energy released at
short times (radiation, etc.).
Comparing HEs using a measured of total energy assumes that the fraction of energy released in the
blast is the same for different formulations

Geometry considerations : 1D plane / cylindrical / spherical propagations.


SPHERICAL

CYLINDRICAL

PLANE

Geometry parameter "n"

Characteristic dimension "R0"

sphere radius

cylinder radius

slab half-width

"Mass" formula

rho.4/3.Pi.R0^3

rho.Pi.R0^2

rho.2.R0

"Mass" unit

kg

kg/m

kg/m^2

Scaled distance formula

R/M^(1/3)

R/M^(1/2)

R/M

Scaled distance unit

m/kg^(1/3)

m^(3/2)/kg^(1/2)

m^3/kg

Generic formula

Explosion types, from Strehlow and Baker, 1976


18 MAI 2015

m^(3/n)/kg^(1/n)

Geometry considerations
CEA | APRIL 2015 | PAGE 11

Blast wave and QSP modelling


Empirical models QSP

(1/4)

Interests in QSP modelling for the consequences on buildings


Detonations in closed volumes result in the superposition of a transient signal (multiple
reflections and interactions of blast waves) and a quasi-static one (QSP build-up).
The characteristics QSP duration leads to a large QSP impulses.
Pressure discharge surfaces (venting) may help limiting structural damage.

Typical pressure signal obtained


in the 2 m3 CEA-G bunker

Influence of a discharge surface in the damage resulting from an internal detonation


(CEA-Gramat experiments)

CEA | APRIL 2015 | PAGE 12

Blast wave and QSP modelling


Empirical models QSP

(2/4)

Interests in QSP modelling for the consequences on people


The studies of human vulnerability to blast usually focus on external detonation,
sometimes accounting for a nearby reflecting surface.
The analysis of terror bombings shows that primary blast injuries (PBIs) are much
more frequent during bombings in partially confined geometries (buses, trains, ).
Pressure discharge surfaces (venting) does not help that much

Katz & al. Primary blast injury after a bomb explosion in a civilian bus
The estimated QSP and duration are 3.8 to 5.2 bar and 2 to 3 ms.
Eardrum injury: 76%, Lung injury: 38%, Abdomen injury: 14%.
Much more than for open-space detonations [reflected pressures also play a role].

Frequency of blast lung injury in various studies

Bombing forensics
CEA | APRIL 2015 | PAGE 13

Empirical models QSP

Initial state

Blast wave and QSP modelling


(3/4)

The QSP depends on M/V AND atmosphere (afterburning).


final state

Single room, no venting.

NATO AASTP-04 model

PQST
QSP variables

2 x 6 parameters for TNT (PQST in kPa, M/V in kg/m3))

PQST = exp(A+B.[ln(M/V)]+C.[ln(M/V)]2+D.[ln(M/V)]3+E.[ln(M/V)]4+F*[ln(M/V)]5)
Other HEs accounted for through TNT equivalence = f(M/V).
1,8

AASTP-04
TNT coefficients

1,6

1,4

1,2

0,8

0,6

0,4

0,001

0,01

0,1

10

ANFO
Comp A3
Comp B
Comp C4
Cyclotol 70/30
DESTEX
H-6
HBX-1
HBX-3
Poudre M1
MINOL-II
Octol 75/25
Pentolite 50/50
Picratol
Tritonal

0,2
100

Equivalence coefficients = f(M/V)


Comparison of the AASP-04 model with TNT experiments

CEA | APRIL 2015 | PAGE 14

Blast wave and QSP modelling


Empirical models QSP

(4/4)

Alternative approach : empirical fit to the results of a thermochemistry solver


US code CHEETAH (the widely available version 2 is sufficient) or other codes.
Considering HE mass + room atmosphere as an equivalent explosive.
Parametric constant volume explosion runs = f(M/V).
Physics-based fit for any (non-aluminized) explosive much easier than AASTP-04.

Large discrepancies at small M/V (solver problem).

Pqst polynomial fit.

2 m3 CEA-Gramat Bunker
Comparison of CHEETAH fits with TNT experiments

CEA | APRIL 2015 | PAGE 15

Blast wave and QSP modelling


Numerical models

(1/4)

Numerical models
From first-principles reliability.
Not limited to simple geometries (3D).
Added physics if required (multiphase ).
Possible integration of simple consequence models.
Handles reflections, diffractions, interactions, etc.

Current issues
Non ideal HEs.
Mesh size and peak pressure.
- AMR, shock capturing, a.s.o.
Weak or strong fluid / structure coupling.
Computational burden (no fast answer !).

Illustrations are from the HI2LO hydrocode

HI2LO results : blast propagation


in urban geometries from GIS data.
TOP : flat terrain; BOTTOM : with DEM.

Developed by the RS2N company for CEA-Gramat.


Dedicated to the transient dispersion of pollutants (high Mach flows).
Checked against an almost exhaustive set of urban experiments from the literature.

CEA | APRIL 2015 | PAGE 16


18 MAI 2015

Blast wave and QSP modelling


Numerical models

(2/4)

Example: 1 ton HE in Paris.

Examples of consequence models


(two large detonations):
TOP : % of broken windows.
BOTTOM: % of eardrum ruptures.

18 MAY 2015

HI2LO ; Pressure contours on the ground


(12 M cells, for the last step, after 2 remaps).

CEA | APRIL 2015 | PAGE 17

Blast wave and QSP modelling


Numerical models

(3/4)

Example: Comparison with INSA-CVL experiments (propane-oxygen detonations).


Configuration 1 : Centered charge, with roof opening, with obstacle.

Configuration 2 : Centered charge, without roof opening, with obstacle.

Configuration 3 : Centered charge, without roof opening, without obstacle.

Sochet & Sauvan

Configuration 4 : Charge in the upper-left corner, without roof opening, with obstacle.

CEA | APRIL 2015 | PAGE 18


18 MAY 2015

Blast wave and QSP modelling


Numerical models

(4/4)

Specificity of indoor detonations.


Afterburning MUST be modelled.
From models of intermediate complexity (no gas/gas interpenetration, mixed is burnt
approach with infinite rate chemistry) to full models including chemical kinetics
(computationally expensive and not always robust) Still a difficult problem !

Pressure evolution [A. Milne, FGE]


18 MAY 2015

Temperature maps
[A. Milne, FGE]

CEA | APRIL 2015 | PAGE 19

Blast wave and QSP modelling


Analytical approaches

(1/6)

CEA-Gramat interests in Fast-Running Models of blast consequences

DEMOCRI E

Enhancement of our Vulnerability and Lethality Assessment Models as well as specific


tools developed for the Air Force.
Frequent interactions with end-users, asking for fast answers (but as accurate as
possible) for more and more complex problems.
CEA leads the DEMOCRITE project, funded by the French National Research Agency,
for the Paris Fire-fighters Brigade. Among many other things, the project aims at
developing of a fast-running algorithm for blast consequences in urban environments.

Strategy adopted since 2014


Database of results from the literature + development of new sets of experiments:
- Urban geometries and tunnels (up to a few kg of solid HE): CEA-Gramat
- Small-scale urban geometries at the Institute Von Karman (RP-80 detonators).
- Multi-room facilities (2015): parametric experiments at INSA-CVL (propane-O2).
Validation of the HI2LO code and parametric simulations on various geometries.
Analysis of existing works and development of our own approach for blast waves in
urban geometries, coupled to GIS tools.
Enhancement of an unstructured solver for multi-room geometries.
CEA | APRIL 2015 | PAGE 20
18 MAY 2015

Blast wave and QSP modelling


Analytical approaches

(2/6)

Existing (more or less) simplified approaches :


Theoretical : Geometrical Shock Dynamics (Whitham, 1957 !) and Shock Ray Theory.
Empirical: Neural network learning of simulated scenarios (Remnikov & Rose).
Empirical approaches based on volume equivalence: EVA (Equivalent Volume
Approach, Borgers, 2008) ECF (Energy Concentration Factor, Silvestrini 2009).
Analytical: Mirror-images for indoor explosions (from acoustic propagation).
Analytical: Shock waves addition rules from LAMB model (Hikida & Needham).
Numerically-based empirical approach: Coarse Grain Method (Flood): propagation of
Friedlander shape parameters on coarse grids (preliminary fits on numerical results).
Mixed approach: Ray-tracing & shortest path + specific rules for canyon detection
(Frank & al.).

Whitham book [1974]


18 MAY 2015

Mirror-image principle
[from Pope, 2010]

CEA | APRIL 2015 | PAGE 21

Blast wave and QSP modelling


Analytical approaches

(3/6)

A closer look at EVA / ECF Methods


A spherical P(Z3D) fit is chosen.
At a given distance X in tunnels, canyons, etc.:
- The total volume swept by the blast wave is V(X).
- Req is the equivalent sphere radius: 4/3.p.Req3 = V(X).
- Zeq is computed from Req and the HE mass M.
- The pressure at distance X is given by P(X)=P(Zeq).

Playing with EVA / ECF : blast in tunnels

Assessment of ECF against data sets


[from Silvestrini & al, 2009]

ISTSS 2008

Comparison of EVA results


with numerical simulations.

CEA | APRIL 2015 | PAGE 22


18 MAY 2015

Blast wave and QSP modelling


Analytical approaches

(4/6)

First results of CEA-G model compared to experiments.


Mixed method with automated canyon detection.
Scale 1/10th , streets are 2m wide.
Run time < 1 second (HI2LO : 106 seconds).
A
2 kg
hemisphere,
T junction.

Tir 1
Zones with
100% damage
at real scale

B
A
1 kg
hemisphere,
L turn.

Tir 4
B
Overpressure and impulse comparisons: Markers: experiments, Lines: CEA model.

CEA | APRIL 2015 | PAGE 23

Blast wave and QSP modelling


Analytical approaches - QSP

(5/6)

Remember the 2x6 coefficients for TNT in the AASTP-04 QSP model ?
CEA-Gramat QSP DAMOCLES model : all range of M/V, with NO PARAMETER.

CEA-Gramat
detonation calorimeter

The HE decomposition energy is used to heat the gases (Cv(T) from NIST data).
Decomposition energy means total combustion energy when oxygen is in excess
the combustion energy is measured in a combustion or a detonation calorimeter.
For large M/V ratios (oxygen-deficient atmospheres), decomposition energy means
(1-a).Edetonation + a.Ecombustion, a computed for total combustion with available O2.
the detonation energy is measured in a detonation calorimeter.
a decomposition scheme is required (empirical or from a thermochemistry code).

60000

35000

Detonation + Afterburning of V401


(30 bar O2)

Detonation of V401
(6 bar N2)

30000

50000

y = 1218,2x + 3378,3
R2 = 0,9958

25000

30000

Q(cal)

40000

y = 2424,4x + 6250
R2 = 0,9982

20000

Q cal
Linaire (Q cal)

Q cal

15000

Linaire (Q cal)

20000

10000

5000

m V401 (g)
0
0,00

5,00

10,00

15,00

20,00

25,00

Specific energy of detonation + afterburning


for octoviton. The value obtained by
combustion calorimetry Is 2314 18 cal/g

0
0,00

Q(cal)

10000

m V401 (g)
5,00

10,00

15,00

20,00

Specific energy of detonation


for octoviton.
CEA | APRIL 2015 | PAGE 24

QSP comparisons (CEA model illustrated with combustion only)

25,00

Blast wave and QSP modelling


Analytical approaches - QSP

(6/6)

The icing on the cake: additional phenomena in DAMOCLES


Evaporation or evaporation + combustion in the room are accounted for.
Venting is modelled using a Riemann solver.
- Handles both sonic and subsonic venting regimes in a 0D = f(t) model.

Extension to N rooms (RS2N company for CEA-G, MUZO code).


Unstructured Eulerian code (1 cell / room), runs in seconds.
Ongoing work to extend the model.

Influence of the vent size on QSP evolution (DAMOCLES)

MUZO Multi-room geometry and DP maps


0 ms (a), 8 ms (b), 16 ms (c), 32 ms (d), 64 ms (e), 350 ms (f), 1050 ms (g)
CEA | APRIL 2015 | PAGE 25

Conclusions

CEA | 10 AVRIL 2012


| PAGE 26
18 MAI 2015

Blast wave and QSP modelling


Choosing a model
You have a supercomputer, state-of-the-art hydrocodes and plenty of time ?
Run huge numerical simulations in order to get pretty 3D pictures for your report.
Wonder how you could draw a clear conclusion from billions of data

You are getting short of time ?


Be careful in extrapolating empirical models
Use the best available FREMs.
Then prepare for the next time and improve your tools.

An anthology of Murphys (and others) laws for modellers


Complex problems have simple, easy to understand wrong answers Be wise !
When working toward the solution of a problem, it always helps if you know the
answer Use first order solutions to check more refined results.
Sometimes, where a complex problem can be illuminated by many tools, one can be
forgiven for applying the one he knows best Practice your models to avoid errors.
Given any problem containing n equations, there will be n+1 unknowns Modellers
always have to make assumptions.
There is a solution to every problem; the only difficulty is finding it Good luck !
CEA | APRIL 2015 | PAGE 27

Blast wave and QSP modelling


Choosing a model
Complex problems have simple, easy to understand wrong answers ?
Not always true back to the EVA / ECF approach.
HE charge at the entrance of a tunnel.
A simple EVA analysis gives a speed-up of 50.106 compared to the 3D simulation.
But not easily extended to complex geometries .

EVA approach compared to 3D results


(Excel worksheet, < 1 s runtime)
HPC results with AMR ISTSS 2012
(1024 cores, 13 hours, 110 to 260 M cells)

CEA | APRIL 2015 | PAGE 28

Blast wave and QSP modelling


Way forward
Fast-running models for complex geometries
Specific experiments at CEA-G or in cooperation (LCPP, INSA-CVL, IVK, ).
Further work is required for blast consequences in urban environments.
- Importation of GIS geometries, improvement of sub-models, optimization
In parallel, development of a new version of the multi-room QSP solver.
- It would be interesting to add QSP and the first 2 or 3 dynamic waves
What about indoor / outdoor and outdoor / indoor propagations ?
Human response to complex pressure signals ???

Forensic approach (inversion of the direct model, planned in 2016)

Oklahoma city bombing (1995): building damages

[from Jeske Engineering Inc.]

Oslo bombing (2011) [from Christensen]


CEA | APRIL 2015 | PAGE 29

Blast wave and QSP modelling

Thank you for your attention !

CEA | APRIL 2015 | PAGE 30

| PAGE 31
CEA | 10 AVRIL 2012
18 MAI 2015

Commissariat lnergie atomique et aux nergies alternatives


Centre de Gramat | BP 80200 | 46500 Gramat
T. +33 (0)5 65 10 53 00 | F. +33 (0)5 65 10 54 33
Etablissement public caractre industriel et commercial

| RCS Paris B 775 685 019

DAM
DEA
SDMT

Vous aimerez peut-être aussi