You are on page 1of 10

Fire Safety Journal, 3 (1980181) 67 - 76 67

Elsevier Sequoia S.A., Lausanne --Printed in the Netherlands

Fires and Flashover in R o o m s -- A Simplified T h e o r y *

P. H. THOMAS
Building Research Establishment, Fire Research Station, Borehamwood, Herts. WD6 2BL (Gt. Britain)
(Received May 26, 1980; in revised form August 18, 1980}

SUMMARY could be regarded as one form of 'flashover',


and this has been shown by T hom as et al. [3,
A simplified form of energy balance for a 4] to be a general result so long as the burning
quasi steady compartment fire is presented. It rate is a strong enough function of tempera-
allows for mass transfer in a partially vitiated ture of the h o t gases in t he room .
but uniform atmosphere and for a thermal Quintiere et al. [1] drew at t ent i on to pos-
feedback approximately representing the sible mixing between the u p p e r and lower
radiation from hot gases and walls. layers which brings vitiated air t o the low level
According to the magnitudes o f two dimen- fire, and remarked on its possible significance.
sionless parameters, one, characteristic o f the Here, a simple model of a r o o m fire is de-
ratio o f heat generation to wall loss, and the scribed which includes in simplified form t he
other, characteristic o f the non-linear feedback, effects of t hat vitiation t oget her with thermal
one obtains eight types o f relation between feedback (both convective and radiative).
burning rate and ventilation rate. Several o f The model identifies several regimes, in
these regimes exhibit a thermal instability that particular one which exhibits t w o pairs of
has already been taken to represent one kind criticalities. In order t o d e m o n s t r a t e and dis-
o f 'flashover'. It arises from an increase in fire cuss this regime and its properties we shall
area or a reduction in the ventilation. One first describe the t h e o r y o f the uni form tem-
regime, however, exhibits a second instability perature fire in detail.
that arises from an increase in ventilation. The
first is thus consequent, primarily, on reducing
cooling, the second on increasing the oxygen THE MODEL
concentration, which, in turn, leads to a fall.
Both these falls in oxygen concentration lead We consider an idealised c o m p a r t m e n t
to ventilation-con trolled burning. having no thermal capacity in its walls nor in
Newer materials include some which may the condensed and gaseous phases. The tem-
be more prone to these instabilities than are perature and the gas composition are con-
traditional materials. sidered as uniform -- i.e., we regard the fire as
a well stirred reactor. Infinite kinetics are
assumed and the air flow and the reaction
INTRODUCTION products are assumed i n d e p e n d e n t o f tem-
perature. The energy equation is then written
Quintiere, McCaffery and den Braven [1] as:
have made c o m p u t a t i o n s of t he temperatures , • ma
[hwAw + c ( m * + m ~ ) ] 0 =m~AHc ms<--
in rooms containing fires of c o n s t a n t area but r
subject to heating f r om the h o t smoke, h o t (1)
walls, etc. The model e m p l o y e d was Quintiere's
• ma
[2] two layer mo d e l in which the lower layer _ ma AH c m~ > - -
remained cold and at a m bi ent oxyge n concen- r r
tration. A thermal instability occurred which (2)

we neglect radiation loss through the window


*@)Crown copyright 1980. and linearise the dependence of m~ on B. Here
68

~°~" 2

10:

\,
\
\ .......

IA

y = ma{l+l/r) 10 -1 1 133 10
o<

Fig. 1. Stoichiometric solutions (eqn. (8)). Fig. 2.0~,~ Regimes (r = 2). The numbers in the spaces
refer to the number of intersections of rnf(ma) on the
stoichiometric line.

f o r simplicity we neglect t h e 'blockage e f f e c t ' f o r the mass transfer, an a s s u m p t i o n o n which we


c o m m e n t below, and write

+ Cg (Og -- Of
m f, - pgDgAf lNox/a r InAf (3)
5 Qf _

mf
hw is strictly n o t i n d e p e n d e n t o f 0, b u t in a s t e a d y state or in a fire a f t e r a long t i m e [4] t h e
main resistance t o h e a t f l o w is in t h e walls and a c o n s t a n t c o e f f i c i e n t m a y t h e n be t a k e n as a first
approximation.
T h e c o n c e n t r a t i o n o f o x y g e n , relative t o the e x t e r n a l c o n c e n t r a t i o n , is readily s h o w n t o be
Nox m*~ + rm*f if m : ,
- - - >mr
gox/a ma + m~ r

= 0 if ma
< : m f, .
r
T h e t e r m in b r a c k e t s is t h e c o n v e n t i o n a l mass t r a n s f e r n u m b e r c o r r e c t e d f o r a n e t r a d i a t i o n flux
In o n t o t h e fuel surface.
In t e r m s o f dimensionless variables these e q u a t i o n s are r e w r i t t e n f o r m*f < m*/r as
(1 + ma + m f ) ¢ = m~ (4)
and

mf . . . . . +~ + -- (5)
r \ ma + ~~f ] r
69

where, for simplicity, we have neglected the difference between fuel surface temperature and
ambient temperature by comparison with other temperature differences and have taken an ap-
proximation form for radiation in terms of the temperature rise to a power p where p is of order
4. We shall n o t restrict the initial discussion to mf < ma/r but any solutions found outside this
condition are, of course, invalid physically.
For mf > ma/r the equations are taken as
(1 + ma + mf)¢ = mJr (6)

mf = ~b + ~ Cp.
-- (7)
r
Likewise, any solutions f o u n d outside the condition mf > m Jr are invalid.
All these equations can be discussed analytically. Babrauskas and Wickstrom [5] have dis-
cussed a version of the above w i t h o u t vitiation and w i t h o u t wall loss other than a temperature
d e p e n d e n t radiation loss through the window. They, like Thomas [6], only included the last
term in eqns. (5) and (7). For a constant m~ the second, ventilation-controlled pair give only one
value of ¢ and, hence, of mf. The mf(ma) characteristic is tangential to a line through the original
having a slope a.
On the stoichiometric line mf = mJr we m u s t have from eqns. (6) and (7) (or from eqns. (4)
and (5))
+ a ~ ¢p-1 = 1
r 1 - - ¢ ( I + r)
(~/~ _ ( I
v 1)P-i[1--a +ma(1 + 1 ) ] . (8)
rP(l + r) p-1 ~ ma(1 + 1/r)

has two stationary values (see Fig. 1)


/3L., = --r (1 + r)P-11[p--2~+-x/p2+ 4 ~ ( p - - 1 ) ] [ p - - 2 ( p - - 1 ) a + X / p 2 + 4 a ( p - - 1 ) l P - 1 I. (9)
2 2(p -- I)(I -- ~)
The + sign gives/3L. The higher root/~, does not exist for ~ < 1. Considering ~ as dependent on ¢
with the requirement that 0 < ¢(1 + r) < 1, eqn. (8) shows that there can be only one inter-
section on the stoichiometric if a > p 2 1 4 ( p - - 1). If ~ lies between 1 and p 2 [ 4 ( p - - 1) there can be
one, or for/] between values/3, and/]L 3 intersections, and for ~ < I (for which flu does not
exist) no intersection or, if ~ lies above ~L, 2 intersections (see Fig. 2). This is necessarily the
same for the fuel-controlled regime: both regimes meet on the stoichiometric line. However, for
the fuel-controlled regime the general treatment of the equations is more complicated. Equations
(4) and (5) can be arranged to give

(Y+Yu)(Y+YL)y = ~ ( ~ +~¢p-1) (10i)

where

2Yu,L = 1 + -- -+ 1- --4~. (10ii)


r

For ventilation-controlled fires, from eqns. (6) and (7)

y = [(r+~)¢+ ~-~¢P]/(1--r¢). (10iii)

Equations (10i) and (10ii) are shown diagrammatically in Fig. 3 and depend on whether the
m i n i m u m of the left hand side of eqn. (10i) is greater or less than the m i n i m u m of the right hand
side, a condition that defines a saddle point
70

Little vitiation

Radiation High
fne;dibga• Ike ; aed~jal~'a°~

Hpgh vitiation "~


Temperature q~ Temperature ¢~

Fig. 3. Various regimes. (left) ~ < ~s, (right) ~ > ~3s.

1 [(r 2V~.+r))(p_~_l)]P
/~s - +1+ (11)
p--1 P ~-
i.e.,

(p -- 1)p-I
This is s h o w n in Fig. 2 f o r r = 2.
A t t h e saddle p o i n t , s,

(m~ + mf)s = Ys = x (12)


and
p
- - ( 1 +x)
r (p--l)
m~s= l--~+x 2+2x (13)

where
x = x/~(1 + l/r).
From eqns. (12) and (13), the saddle point lies on or above the stoichiometricline only
when

~ > (1 + x ) 1
( x)
-- . (14i)
p 1
This c o n d i t i o n is i n v a r i a b l y satisfied if ~ > p 2 / 4 ( p - - 1), i.e., 1.33 a p p r o x . I f ~ < 1 t h e c o n d i t i o n
e a n n o t be satisfied i f x < 2, i.e., if r < g1. H e n c e , f o r ~ < 1 a n d n o r m a l fuels n o t c o n t a i n i n g t h e i r
o w n o x y g e n , t h e saddle m a y lie on, or b e l o w , t h e s t o i c h i o m e t r i c line a n d so in t h e *valid' fuel-
c o n t r o l l e d region. I t d o e s so if ~ is less t h a n a value r e a d i l y e v a l u a t e d f r o m eqn. (14i) as

2(p - - 1)(1 + rp) + (p - - 2)2(1 + r) + (p - - 2)x/(1 + r ) [ p 2 ( 5 r + 1) - - 8 p r + 4r] l.


(14ii)
a < r 2(1 + rp) 2 1
F r o m Fig. 3 o n e sees h o w o n e o b t a i n s s o l u t i o n s t h a t e x c l u d e a c e r t a i n range o f y or a c e r t a i n
r a n g e o f ¢. A p a r t f r o m this, a n y value o f y gives t w o v a l u e s o f ¢ a n d vice versa. In Fig. 4(a) a n d
(b) t h e s e fuel r e g i m e s i t u a t i o n s are s h o w n d i a g r a m m a t i c a l l y in t h e m~(ma) plane, t h o u g h f o r
m~/ma > 1/r t h e results h a v e n o p h y s i c a l m e a n i n g a n d m u s t be r e p l a c e d b y t h e v e n t i l a t i o n ~ o n -
t r o l l e d regime.
71

(l+rna+rnf)q)= Heatloss
" ~ IStn°'Cmh;°=me:r}rc /
m~ Heatgeneration
Invald ~ion /'

m,l // q~
Fig. 5. Stability of ventilation-controlled fire.

\ va,,ro0,oo
~ ° n C : t ~ ~ m f =rna/r~

rna (1÷ma÷mf}q~= ~

(a) mf

~ U nSaetbls t a b l . ~ _ e

Invalidregi /
Fig. 6. Stability of fuel-controlled fires.

a(ma--r¢) a
mf = (1 -- 4) + -- ~.P (15)
r \ m a +~ r
/ We note that in fuel controlled fires ~ < m~,
// Validregion so, necessarily, ma > r#. The comparison in
Fig. 6 of heat loss and generation shows that if
there are three equilibria for a given ma (irres-
Illa pective of whether or not m~ < m J r ) , the
middle one only is unstable.
(b) If there is only one equilibrium, it is stable.
In Babrauskas and Wickstrom's [5] model
Fig. 4. Fuel-controlled fires. (a) ~ < ~s, (b) ~ > ~s- the upper invalid regime is suppressed, and
the lowest solution excluded. They, like
Thomas [6], found only the unstable fuel-
controlled regime and a stable ventilation-
controlled one. The upper fuel-controlled
STABILITY equilibrium is associated with high tempera-
tures. It normally does n o t exist in a valid
Since m a is taken as independent of # and regime, but if it does, it can be shown t h a t it
mf increases with 4, all ventilation¢ontroUed is associated with low values of r, i.e., fuels
equilibria are stable. This is seen from the requiring little added oxygen.
heat balance (Fig. 5). A small shift from equi-
librium restores equilibrium. For fuel-
controlled fires we substitute for mf in terms
of ¢ on the right hand side of eqn. (5) and DISCUSSION OF m f ( m a ) EQUILIBRIA
obtain the heat generation explicitly in terms Consider fuel-controlled equilibrium with
of ¢ m f < ma/r.
72

/
Invalid branch of solution
/// /~

mf
........."i'" mf

o:

ma
ma

(a)
(a)

lnvalid branch of solution

Ventilation
i
~// ~/~
controlled , / / /~uel controlled
mf
rnf
~'" /j /

ma
ma

(b)
(b)

Fig. 7. Typical relationships between mf and rna, Fig. 8. Typical relationships between mf and rn a
< ~s and not between ~u and ~L. (a) O~> 1, (b) (~ < 1. (~s > ~ > ~ L ) ' ( a ) ( ~ > 1 , ( b ) 0 < a < 1.

Firstly we consider the/3(~) plane, shown calities, and weak feedback effects, e.g., w o o d
in Fig. 2 as ~ vs. ,* for r = 2. It is possible to cribs as the fuel.
show algebraically that for ~ < 1, ~s has a If ~s > ~ > ~L the m ~ ( m a ) characteristics
minimum value with respect to 'r' and that are as previously described, with the addition
these minima are never less than ~L. Seven of a detached zone produced by the second
regimes are identifiable plus an eighth on the and middle (unstable) equilibrium lying below
limit ~ -* 0. According to the value of r,/3~ the stoichiometric line (see Fig. 8). The point P
m a y pass above or below the cusp where ~L is an invalid non-stoichiometric c o m m o n solu-
and ~u meet. tion to both fuel- and ventilation-controlled
We will first consider the 'weak' feedback regimes: the equality of the two expressions
fires/~ </~s. For these, there is a region of ¢ for m~ in eqns. (5) and (7) (with ¢ defined
b e t w e e n the valid and the invalid region differently in the two equations) can be
where there are no solutions, and if ~ lies out- factorised by m a - - rm~. If 'r' is sufficiently
side the region between/~u and ~L, m ~ ( m a ) is small there is an upper stable fuel-controlled
as sketched in Fig. 7. regime, see Fig. 4(a) and Fig. 9.
There is no criticality for ~ > 1, this only For ~ > 1 and ~ > ~s, criticalities may
appears when ~ > /3~ (see Fig. 4). The fuel- appear as in Fig. 10(a) and (b).
controlled equilibrium shown is either the As before, a low 'r' modifies the m t ( m ~ )
only one or the lowest of three, it is therefore zone to include a stable fuel controlled region
stable. Figure 7(a) is the conventional relation (not shown).
for m a n y fires having t w o modes of behaviour The last main region, ~ < 1 and/~ >>/3~ is
(ventilation- and fuel-controlled), no criti- discussed in the next Section. Apart from this
73

\
\
/
.~Stable regime exists only
~/~j/; if 7' is small . . . . gh
7 / / /i
~// , / f l ~ Unstable
mf

ma

Fig. 9. Effect of small 'r' on upper part of fuel-


controlled region. ma

(a)
t h e o n l y o t h e r result we n o t e h e r e is t h a t f o r
= 0 and a~ finite (= ~) t h e r e is n o solution \
at all unless ~ e x c e e d s \

r (p(1 + r)lP-1
, //

w h e n t h e s o l u t i o n consists o n l y o f t h e / J /.~/-Unstable
d e t a c h e d z o n e . It is this z e r o ~ c o n d i t i o n t h a t //;/
T h o m a s [61 (with p ; 2) a n d Babrauskas and
W i c k s t r o m [ 5] c o n s i d e r e d earlier.

ma

DISCUSSION OF FLASHOVER
(b)
T h o m a s et al. [3, 4] have i d e n t i f i e d the
unstable f u e l - c o n t r o l l e d regime as o n e pos- Fig. 10. Typical relationships between mf and ma,
sible s o u r c e o f flashover. Figure 11 shows >~sand~> 1.(a)~u>/3>~sand1.33>~> 1,
h o w t h e r e m a y , in fact, be t w o t h e r m a l insta- (b) ~ > ~u-
bilities. If ~ > 1 t h e l o w e r p a r t o f t h e mt(ma)
curve with its criticality m a y be w h o l l y w i t h i n
t h e v e n t i l a t i o n - c o n t r o l l e d region and so is
e x c l u d e d f r o m physical c o n s i d e r a t i o n .
If t h e saddle lies b e l o w the s t o i c h i o m e t r i c / "//i ~ iI

line (see e q n . (14)) t h e raising of/3 f r o m just


b e l o w t o just a b o v e ~s i n t r o d u c e s a criticality
at A and t h e s i t u a t i o n changes as in Fig. 11. mf
/(/;(
W h e t h e r t h e j u m p at the criticality is large
e n o u g h t o be d e s c r i b e d c o r r e c t l y as ignition
w o u l d involve us in a discussion o f t r a n s i e n t
p h e n o m e n a -- pressure changes, etc, it is how-
ever a d i s c o n t i n u i t y b e t w e e n s t e a d y states.
If t h e saddle lies a b o v e t h e s t o i c h i o m e t r i c
/ ma

line, which it always does f o r ~ > 1.33, it is


in t h e region w h e r e f u e l - c o n t r o l l e d b e h a v i o u r Fig. 11. Behaviour near ~s./3 ~ ~s, ~ < 1 and saddle
does n o t o b t a i n . below stoichiometric line. - - / 3 < ~s, -- -- --/3 > ~s.
It m a y readily be s h o w n t h a t t h e m e a n con-
c e n t r a t i o n o f o x y g e n increases so long a s
mf/ma > dm~ldma, w h e r e t h e m~(ma) is tan- increase in o x y g e n c o n c e n t r a t i o n , a greater
gential t o t h e origin is a s t a t i o n a r y value. h e a t release, a n d a h i g h e r t e m p e r a t u r e . Even-
H e n c e , b e l o w A a rise in m~ first leads t o an tually, t h e o x y g e n c o n c e n t r a t i o n falls and
74

criticality A is reached and the fire becomes


ventilation controlled. Further increase in
ventilation leads to a drop in burning rate at
the upper stoichiometric condition.
The saddle point m a y be restated in terms
of the original variables. For this we take mf
/
In = keoO 4 where o is the S t e f a n - B o l t z m a n n
constant and, hence,

kero(AHc) 3
{is - he c J~
ma

27
(r
1+--+2 (16)
Fig. 12. The double criticality.
256
where
heAHcAf heat transfer processes. However, the model
Ol--
QehwAw as described has identified several modes of
fire behaviour, which provide a basis for con-
and where we have written ducting such analyses and interpreting them.
Figure 2 suggests t h a t for high fl, the transient
cpgD
- hf growth though increasing ~ is a more complex
5 process than for low/~. The regime with two
pairs of criticalities is especially interesting.
a convection mass transfer coefficient to the We consider two mf(ma) curves of the same
fuel bed. /~ but, by virtue of a change in burning area
If we write two slightly different values of ~. We assume
ngo that/3 is greater than both the two values of
OA D -- /3s. If ma is in the region shaded in Fig. 12 the
(1 + r)c
increase in Ae leads to a 'jump' to the ventila-
for the adiabatic temperature rise we have, tion controlled regime.
for r = 5, In both cases the temperature rises and the
oxygen concentration eventually falls to zero.
kaO~e However, a possibly important distinction
- 8.07
he may be noted. 'Ignition' (B) is essentially as-
sociated with a reduction in ventilation and so
and 0.22 for a = 1 and for ~ = 5, respectively. in cooling, raising the temperature and hence
In the former case a ratio of radiation transfer the mf, as a result of greater radiation feed-
to local convection transfer of nearly 10 is back.
required for criticality, and in the latter case a 'Ignition' (A) is associated with an increase
relatively small feedback contribution from in ventilation, which suggests t h a t it acts by
the radiation is sufficient. initially increasing the oxygen concentration
A necessary condition for fires being in the - so raising the concentration-dependent term
-

region o f criticalities or growing through them and the heat release through the greater local
is t h a t fl exceeds fls (eqn. (11)). The movement effects on mass transfer. Eventually this in-
from wood to fuels with smaller effective Qf crease in mf leads to falling oxygen concen-
(and possibly larger AHc) and improved insu- tration.
lation {lower hw), raises a and lowers/3s, and A quasi-steady model will not make much
so allows the region of criticalities to occur more of a distinction between (A) and (B),
with smaller fires. but clearly the time constant associated with
To make definitive conclusions one needs the two processes m a y be different if radia-
to c o n d u c t numerical analysis with a more tion feedback is primarily from the walls and
detailed model which represents better the n o t the h o t smoke. In so far as ignition (A) is
growing fire in a c o m p a r t m e n t and t h e various associated more directly with gas phase reac-
75

tions, a transient model with reaction kinetics A vitiated fire can exist in equilibrium,
might show that (A) is more liable to be ex- and if the thermal feedback is strong enough,
plosive than some kinds of (B). (A) might well a special type of instability can exist.
be related to 'flashover' in the original sense For a > 1, a rise in a, i.e., in Af/Q, can take
used by firemen -- an ignition resulting from 'weak' feedback fire across the ~s curve into
admitting air, though we have excluded the the 'strong' feedback zone introducing a
possibility of igniting u n b u r n t gases by our criticality and, hence, a type of 'flashover'.
assumptions of uniform mixing and uniform The lower Q/AHe is, the smaller the fire and
gas phase temperature. These effects require the weaker the coupling, hf, needs to be to
a more complex model for their considera- bring about such a 'flashover'.
tion. The existing model implies t h a t the
'jump' in mf may often be quite small and
perhaps insignificant. High 'jumps' are asso-
NOTATION
ciated with small a and low values of r, being A area
no more than about 25% for a = 1 and r = 5 C specific heat of all gases assumed equal
but over 300% for a = 0.04 and r = 5. D diffusion coefficient
The existence of 6u, 6,. and ~ and their ~Hc calorific value of fuel
basic significance is n o t necessarily altered by h heat transfer coefficient
generalising eqn. (5) by adding a constant k a coefficient of order 1
term and changing, within limits, the coeffi- in net radiation
cients of each term so t h a t the modified equa- m* mass flow rate
tion represents a linearised form of mr(B) or m m* c/h w A w -- dimensionless
an experimentally derived, linearised relation- concentration of oxygen
Nox
ship. concentration of oxygen ambient air
Nox/a
p an index ~ 4
Q heat required to produce unit mass of
CONCLUSION fuel volatiles at fuel surface
r stoichiometric ratio of air/fuel
This model is n o t directly associated with y=m a +rnf
the spread of fire -- the model does n o t in-
clude spread as a transient process. However,
any increase in Af and decrease in wall loss SUBSCRIPTS
can be described in a quasi steady manner by
this model as a first approximation. a air
We have explored the effects of vitiation f fuel or fuel surface
and non-linear thermal feedback and have g gas
seen that there are various regimes defining s saddle
the type of relationship between the fuel L lower of two roots
production rate mf and the air flow m~. There u upper of two roots
is a region, broadly defined by/3 < /~L when w wall (and other heat loss surfaces)
a < i, and by/3 </3s when a > 1.33, which pgcgDgAHcA f hfAf AHc
may be described by 'weak' feedback. For
/3 >> ~s, criticalities, unstable regimes, de- 6hwQfAw hwAw Qf
tached zones of equilibrium exist. For low rln5 c r
values of hf/A~ and Q/AH¢, any prevailing ~¢P - rgpf ~ In
feedback is more likely to be 'strong' than pgDAHc hf A H c
'weak'• The major factors defining the value
[3 - lerea(AHe/c)3/ht whenp ~ 4
of 9, the feedback, are the effective emissivity
and the fuel bed transfer coefficient, which is, 5 boundary layer thickness
in effect, the 'coupling' between the fire and e effective emissivity
the fuel. A high value of hf m a y tend to make 0 temperature rise
feedback 'weak' rather than 'strong' but in a p density
spreading fire, high h~ increases the propensity o S t e f a n - B o l t z m a n n constant
to spread. ¢ cOl~Hc
76

REFERENCES 3 P. H. Thomas, 17th Combustion (Int.) Symp.,


1 J. G. Quintiere, B. J. McCaffrey and K. den Braven, Leeds, U.K., 1979, (discussion of ref. 1).
Experimental and theoretical analysis of quasi -- 4 P. H. Thomas, M. L. Bullen, J. G. Quintiere and
steady small scale enclosure fires, 17th Combustion B. McCaffrey, Combust. Flame, 38 (1980) 159.
(Int.) Syrup., Leeds, U.K., 1979, Combustion 5 V. Babrauskas and U. Wickstrom, Combust. Flame,
Institute, Pittsburg. 37 (1) (1980) 105.
2 J. G. Quintiere, Growth of fire in building com- 6 P. H. Thomas, Some problem aspects of fully
partments, A S T M / N B S Syrup. Fire Standards and developed room fires, ASTM/NBS Symp. Fire
Safety, Spec. Tech. Publ. 614, Am. Soc. Testing Standards and Safety, Spec. Tech. Publ. 614, Am.
Mater., Philadelphia, 1976, pp. 131 - 167. Soc. Testing Mater., Philadelphia, 1976.

You might also like