CFD Simulation for Gas Explosions

7/26/2019 CFD Simulation for Gas Explosions

1/15

AutoReaGas ACFD TOOLFOR GASEXPLOSION HAZARD ANALYSIS

A.C.van den Berg , H .G.The and W.P.M .Mercx

TN O Pr ins Maur i t s Labora tory

P . O . B o x 4 5

2280 AA Ri jswi jk, The Netherlands

Y. Mouilleau

I N E P J S

P.O.Box 2

60550 Veraeui l -en-Halat te , France

C.J. Hajrhurst

Century Dynamics Ltd.

12 Ci ty Business Centre, Hors ham

West Sussex RH13 5BA, UK

1 . I N T R O D U C T I O N

Gas explosions const i tute a major hazard for offshore gas and oi l producing instal lat ions. A

gas explosion is the consequence of an accidental release of a flammable gas, the mixing with

ai r and a subsequent igni t ion. Under appropriate boundary condi t ions the resul t ing f lame

propagat ion process may develop explosive combust ion and damaging blast loadings. In

spaces containing a lot of equipment, this is a part icular problem and a small quanti ty of fuel

may be sufficient to give rise to the development of high explosion overpressures. If such

overpressures are not ant icipated in the design they may have fatal consequences for both

crew and rig.

The hazard of gas explosions offshore was demonst rated by the incident wi th the Piper

Alpha rig in 1988 (Pet rie^). A smal l -scale gas explosion caused the fai lure of vi tal cont rol

and communicat ion funct ions on board. In consequence of this , the incident escalated to

unforesee n ci rcumstances leading to the total loss of the r ig and the death of 167 people.

However, gas explosion effects can be cont rol led by a proper design of the instal lat ion.

Modern offshore instal lat ions consist of a number of separate modules of l imi ted size.

Present understanding of the phenomena indicates that the module shape, the posi t ioning of

the equipment inside the module and the posit ioning and the size of vents largely affect the

5.8.1

Author manuscript, published in "International Conference and Exhibition Offshore Structural Design. Hazards safety andengineering, Londres : United Kingdom (1994)"
http://hal.archives-ouvertes.fr/


2/15

development of an in ternal gas explosion.

So far , s imple venting guidel ines (e .g . Cubbage and Simmonds

9

; Brad ley and Mi tcheson

4

'

5

)

are widely used to assess the consequences of possib le gas explosions on board of offshore

insta l la t ions. Venting guidel ines are empir ical corre la t ions based on experimental data . Most

of the experimental data have been obta ined in small-scale tes ts using near cubical , empty

vessels . These guidel ines , however , do not a l low applicat ion beyond the experimental

condit ions they were derived from. Applicat ion to larger volumes of more complex

geometr ies which conta in many objects may lead to substant ia l underest imation of effects .

For the design of adequate gas explosion control provis ions in the offshore , mo re

sophis t ica ted methods are essent ia l .

AutoReaGas is a sof tware package capable of userfr iendly , in teract ive , 3-D numerical

s imulat ion of any aspect of gas explosion phenomena. AutoReaGas conta ins both a gas

explo sion sim ulator and a blast simulator, eac h tailored to specific proble m features.

After a general descr ip t ion of the phenomena and how they are modelled , in th is paper the

software is demons tra ted in a pract ical offshore case s tudy.

2 . P H E N O M E N A

2.1 Gas Exp los ion

In a gas explosion a f lammable gas mixture is consumed by a combust ion process which

pro pag ates th roug h the mixture in the form of a f lame front . The f lame front is the in terface

be tween co ld reac tan ts and ho t combus t ion p roduc ts . Because combus t ion p roduc ts a re o f

high temperature , the cold f lammable medium expands s trongly on combust ion. The

expansion induces a flow field whose structure is fully determined by the nature of its rigid

boundaries . In th is f low f ie ld the combust ion process is carr ied a long. The ra te of

combust ion is s t rongly affected by the f low structure (veloci ty gradients and turbulence)

met. Flow velocity gradients stretch the flame front, enlarge its interface and increase the

effect ive combust ion ra te . Low intensi ty turbulence wrinkles the f lame front with a s imilar

effect on the combust ion ra te . Higher combust ion ra tes in tensify the expansion. Higher f low

veloci t ies go hand in hand with more in tense turbulence levels . Higher turbulence levels

speed up the comb ust ion, e tc . e tc In o ther wo rds: unde r the approp ria te ( turbulence

generat ive) boundary condit ions, a posi t ive feedback mechanism is t r iggered by which a gas

explosion develops exponentia l ly both in speed and overpressure .

5 8


3/15


4/15

Turbulent combust ion i s model led by an expression which relates the combust ion

rate to turbulence. Several opt ions are avai lable varying from theoret ical relat ions

such as t he Eddy Break Up model (Spa ld ing^) and the Eddy Di ss ipa t ion model

(M ag n u ssen an d H j e r t ag e r

1 2

and Hjertager et al .

1 0

) up to experimental correlat ions

between turbulence and combust ion (Bray^). Because the appl ied cel l s ize i s often

too large to ful ly resolve a turbulent combust ion zone, the combust ion rate i s

correc t ed .

The init ial stage of combustion upon ignit ion is modelled by a process of laminar

flame p ropa gat ion w hose speed i s cont rol led on the basis of experimental data.

Objects too smal l to be represented by sol id boundaries in the computat ional mesh,

are model led by a subgrid representat ion. The presence of a subgrid object i s

modelled by the specification of appropriate flow condit ions: i .e. : a fluid dynamic

drag and a source of turbulence.

Num erical solut ion of the set of equat ions is accomp l ished by means of the pow er

l aw scheme appl ied wi th in a f in ite vo lume approach (P a t an ka r^ ) .

3.2 Blast

As long as objects wi th large cross-flow dimensions are considered, the interact ion wi th gas

explosion blast i s predominant ly governed by the pressure wave character of the blast . The

drag component can be neglected. The pressure wave character of blast f low fields can be

accurately represented by the assumption of inviscid flow. Often, blast flow fields are

characterized by the presence of gas dynamic discont inui t ies such as shocks. Model l ing of

blast -object interact ion requi res careful descript ion of such phenomena. Therefore, the blast

simulator in AutoReaGas models blast -object interact ion as fol lows:

The gas dynamics is modelled as inviscid compressible flow of a perfect gaseous fluid

which can be formulated in the conservat ion equat ions for mass, momentum and

energy for inviscid flow, i .e. the Euler-equations.

Descript ion of shock phenomena requires a sophist icated numerical technique

tai lored to proper representat ion of steep gradients. To this end, the blast s imulator

u t il izes F lux-Correc t ed T ranspor t (FCT) (Bo r i s and Bo ok

2

and Boris-^). FCT makes

an optimized use of numerical diffusion so that steep gradients present in shocks are

retained. Numerical diffusion is added only where i t is required for numerical

stability.

5.8.4


5/15

4 . A N A L Y S I S O F A G A S E X P L O S I O N O N A N O F F S H O R E P L A T F O R M

4.1 P rob lem

Modern offshore insta l la t ions are character ized by a modular s tructure . The various aspects

of the oil and gas production process take place in different areas separated by fire/blast

res is tant walls . The in tent ion is to keep the consequences of a possib le incident with in

bounds - the modu le .

In case of a gas explosion the in ternal overpressure can be control led by venting any

expanding gases . Therefore , modern modules are constructed so that they are as open as

possible . Outer walls of ten consis t of l ight-weight windcladding or windscreens which are

at tached to the main s tructure in such a way that they may easi ly fa i l and are b lown off a t a

low in ternal overpressure .

A vented gas explosion gives r ise to an external explosion. As soon as the combust ion

process in the module is in i t ia ted , the f lammable mixture inside the module s tar ts vent ing in

th e form o f a turbulent flammable jet. T his jet exp lodes w hen it is ignited at the time th e

combust ing gas mixture vents . The resul t ing blast may do damage to , for instance, nearby

equipment and s tructures .

A vented gas explosion is the subject in the present analysis carr ied out with A utoR eaG as.

Figure 1 shows a h ighly s implif ied , made up representa t ion of an offshore production

platform. The pla tform consis ts of a main deck and a cel lar deck. The main deck consis ts of

several mo dules . On e of these modules is a lmost complete ly bui lt in . Th e only possib i li ty for

venting for th is module is the space on deck betw een the m odules and the l iv ing quarters . A t

th is s ide the module is lef t complete ly open as a vent . The consequences of a gas explosion

in th is modu le are analyzed by applying the AutoR eaG as software .

This exercise addresses the fo l lowing quest ions. What is the overpressure developed by a

gas explosion in the module? What is the b last loading of a 3 m diameter , 8 m long vessel

and a 0.3 m diameter tube present on deck in front of the vent opening and what are blast

over press ures a t the wall of the l iv ing quarters?

4.2 Analysis

A com putat iona l domain is specif ied . T he domain, consis t ing of 40* 20* 20 cells of

1

m^ size,

5.8.5


6/15

covers the module as well as the space between module and l iv ing quarters . Within th is

dom ain, th e softw are allow s the specification of the physical layou t of any system o f rigid

bou ndarie s e .g . boxes, beam s, vessels and tubes by means of a CAD -like in terface .

Figure 2 represents an AutoReaGas configurat ion of the domain showing only the larger

pieces of equipment. The module (left) is filled with a number of horizontal and vertical

vessels , in terconnected with a lo t of p ip ing and appenda ges. A 3 m diam eter , 8 m long vessel

(A) as well as a 0 .3 m diameter tube are defined in the space on deck between module and

liv ing quarters ( r ight) .

The specified configuration of objects in the domain is automatically converted by the

software in to the proper input for the explosion s imulator . Large objects are represented by

rig id bo unda ries while the presence of small objects is modelled by the subgrid formulat ion.

The software allows the specification of any distribution of fuel in the domain which can be

ignited in any desired locat ion. However , to approach worst case condit ions in th is problem,

the module is assumed to be f i l led with a s to ichiometr ic propane-air mixture and igni ted in

the cen t re o f the back wa l l .

Th e Au toR eaG as software a l lows f ix l ly in teract ive s imulat ion, show ing the d is tr ibut ions of

any specif ied process parameter on the screen, any wanted number of t ime s teps again .

Figu re 3 sho ws a com pila t ion of such a series of p ic tures . The pic tures show the tem peratu re

field in bo th a horizontal and vertical cross-sectio n at a num ber of consecu tive poin ts of

t ime. The temperature is v isual ized by means of a suggest ive colour gradat ion. The t iming of

the p ic tures indicates how the f lame propagation process develops. After a s low laminar

s tar t , i t speeds up under the inf luence of the equipment in the module . The combust ion

process vents in the form of a mushroom-like shaped flame front, which is fully in line with

experimental observat ions (Catl in^ nd Bimson e t a l . l ) .

During the s imulat ion, process parameters can be monitored throughout the domain. F igure

4 represents the overpressure t races recorded inside the module in the gauges 1 ( igni t ion

point) and 2 (vent open ing) . The traces show the character is tic behaviour of a gas explosion:

a re la t ively long in i t ia l phase of s low development and low overpressure progressing in to a

more violent development character ized by a sudden pressure pulse . A maximum internal

overpressure of approximately 70 kPa is observed a t the back wall of the module .

The Figure 5 represents f ive overpressure t races (3 - 7) recorded a t the l iv ing quarterns wall

in front of vent opening (Figure 2). All traces are more or less similar showing a double-peak

5.8.6


7/15

shape with a maximum of approximately 40 kPa.

The blast loading of the 3 m diameter vessel on deck is monitored by recording the pressure

differential between the front and back area of the vessel. This pressure differential is

rep rese nted in Fig ure 6. M ultiplication of this pres sure differential w ith the c ross-s ection al

area o f the vessel immediate ly resul ts in the horizonta l force induced by the vent f low.

The blast loading of the 0 .3 m diameter tube is recorded by monitor ing both the densi ty and

gas veloci ty components in three d ifferent gauges in the row of cel ls in which the tube is

specified as a subgrid object. The force on the tube per meter length is calculated from these

parameters assuming a drag coeff ic ient equal to 1 . The drag force on the tube per meter

length as a funct ion of t ime is represented in Figure 7 . The dou ble-peak shape, ampli tude

and durat ion of the tube load are in l ine with experimental observat ions (Catl in) .

The double-peak shape of the various loading traces seems to be character is t ic and can be

explained considering the process parameters in more deta i l . S tagnat ion pressures are

proportional to both the density and the square of the flow velocity. Initially, stagnation

pressures are the result of relatively low-velocity/high-density flow, i.e. the blast from the

internal and external explosion. At the instant the gauges are reached by low-densi ty

combust ion products , s tagnat ion pressures drop. S tagnat ion pressures r ise again in the

growing burned gas vent veloci t ies . Gas vent veloci t ies tend to increase s trongly a t the

instant combust ion products s tar t vent ing.

5 C O N C L U S I O N

AutoReaGas is a CFD-tool for analysis of gas explosion problems. AutoReaGas consis ts of

a gas exp losion simu lator and a blast simulator, pla ced in a use r friendly environ me nt.

Several possib i l it ies of the software w ere dem onstra ted in a practical case s tudy. P roblem s

can be defined in a userfr iendly CAD-like environment . Computat ional resul ts indicate that

the software is capable of realistic simulation of (vented) gas explosions. The exercise in this

paper demonstra ted the possib i l i ty of deta i led computat ion of the b last loading of objects

specif ied in the com putat ional domain.

5 8 7


8/15

6. REFERENCES

1. Bimson S.J. et al. (1993)

An experim ental study of the physics of gaseous deflagration in a very large vented enclosure,

14the Int.Coll.on the Dy namics of Explosions and Reactive Systems,

Coimb ra, Portugal, Aug. 1-6, 1993

2. Boris J.P. and Book D.L., (1976),

Solution of continuity equations by the m ethod of Flux-Corrected Transport

Meth ods in Co mpu tational Physics, Vol.16, Academic Press,New York, 1976

3.

Boris J.P., (1976)

Flux-Corrected Transp ort m odules for solving enerailized continuity equations,

NRL M emorandum report 3237, Naval Research Laboratory, Washington,D.C.

4. Bradley D. and Mitcheson A .,(1978a)

The ve nting of gaseous explosions in sherical vessels. I - Theory

Combustion and Flame,Vol.32,(1978),pp.221-236

5. Bradley D. and Mitcheson A.,(1978b)

The ven ting of gaseous explosions in sherical vessels. II - Theory and experiment

Com bustion an d Flam e, Vol.32,(1978),pp.237-255

6. Bray K.N .C., (1990)

Studies of turbulent bu rning velocity

Proc.Roy.Soc.London,Vol.A431,(1990),pp.315-325

7. Ca t l inC.A. (1991)

Scale effects on the external combustion caused by venting of a confined explosion

Combustion and Flame,Vol.83,(1991),pp.399-411

8. Catlin C.A. (1993)

Th e blast loading imparted to a cylinder by venting of a confined explosion

2nd Int.Con f.and Ex hibition Offshore Structural Design agains Extreme Loads

Nov. 3-4, 1993 , London

9. Cubbage P.A. and Simmonds W.A. (1955)

An investigation of explosion reliefs for industrial drying ovens:

I - Top reliefs in box ovens

Trans.Inst.Gas Eng.,Vol. 105,(1955),pp.470

10 .

Hjertager B.H. et al. ,(1992 )

Com puter mod elling of gas explosion propagation in offshore modules

J.LossPrev.Process Ind.,Vol.5,No.3,(1992),pp. 165-174

11 .Laun der B.E. and Spalding D.B. , (1972)

Mathematical models of turbulence

Academic Press, London, 1972

5.8.8


9/15

12 . Magnussen B.F. and Hjertager B.H.,(1976)

On the mathematical modelling of turbulent combustion

with special emphasis on soot formation an d combustion,

16th Symp .(Int)on Combustion,pp.719-729,

The Com bustion Institute, Pittsburgh (PA), 1976

D. P a ta nka r S . V . , ( 1980)

Nu merical h eat transfer and fluid flow

Hemisphere Publishing Corporation, Washington, 1980

14 .Petrie J.R., (19 88)

Piper Alpha technical investigation interim report

Petroleum Engineering Divis ion,

Department of Energy, September, 1988

15.

Spalding D.B., (1977)

Develop ment of the eddy break up m odel of turbulent combustion

16th Sy mp .(Int) on C ombustion,

pp.

1657-1663

The Combustion Institute, Pittsburgh (PA), 1977

5.8.9


10/15

Figure 1 Offshore o i l and gas production pla tform

Figure 2 Au toR eaGas p rocess equ ipm en ts rep resen ta t ion .

5 8 10


11/15

i

g = f e .

s

L L

i S:S >;

m m

Figure 2 Au toR eaGas p rocess equ ipmen ts rep resen ta tion (con t ).

5 8 11


12/15

Hor izo n ta l c ross sec t ion Ver t ica l c ross sec t ion

Time742.9ms after ignition

Time781.5 ms after ignition.

T ime

813.9

ms after ignition

Temperature [K]

300 500 700 900 1100 1300 1500 1700 1900 2100

Figure 3 Com pi la t ion o f Au toR eaGas p rocess mon i to r ing .

5 8 12


13/15

Ho rizon tal cross sect ion Vert ical cros s sect ion

Time

840.8

ms after ignition

Time865.6 ms after ignition

Time958.5ms after ignition

Temperature [K]

300 500

700 900 1100 1300 1500 1700 1900 210 0

Figure 3 Com pi la tion o f Au toR eaGas p rocess mon i to r ing (con t ).

5 8 3


14/15

s

s

o

70

o

5O

4O

3O

2O

1O

O

1O

2O

3O

2 O O 4OO BOO

Time[fn

8 O O

10OO 12OO

Figure 4 Ov erpressure- t im e traces recorde d a t

gauge s 1 and 2 inside the mod ule

BO

7O

SO

SO

40

30

ZQ

1O

O

10

2O

3O

4OO BOO

ms

Im

8 0O

1OOO 120O

200 4oo eoo B O O 1000 1200

Figure 5 Ov erpressure- t im e traces recorde d a t gauge s 3

up to 7 located at the wall of the living quarters.

5 8 14


15/15

i

1

so

4

3

2

1

O

1

2O

F i g u r e

d P P front)-P back)

2 OO B O O

Time [ms]

BOO

10OO

Pressu re dif ferent ial be tween f ront

and

back area

of

vessel

A on

deck .

12OO

2OO BOO 3 0 O

1OOO 12OO

Time[ms]

Figure7 Blas t loadingat three gauges alongap ipe above ves se lA

Date post:	01-Mar-2018
Category:	Documents
Upload:	nazeemlng
View:	234 times
Download:	0 times

CFD Simulation for Gas Explosions

Documents