UVCE (Roberts1982)
UVCE (Roberts1982)
UVCE (Roberts1982)
ABSTRACT
Roberts, A.F. and Pritchard, D.K., 1982. Blast effect from unconfined vapour cloud
explosions. Journal of Occupational Accidents, 3: 231-247.
INTRODUCTION
served damage and that the rise-times and positive phase durations are
longer in the blast wave from a UVCE than in the blast wave from high
explosives, at similar overpressures. These differences affect the damage
caused by the blast wave. However, to date no convincing experimental
simulation of a UVCE has been made nor has any adequate explanation of
flame acceleration effects been given.
The major sources of data on the overpressure-time histories of UVCEs
are actual accidents. The damage caused to pieces of equipment or struc-
tural elements can be used diagnostically for determining the magnitude and
possibly the duration of the overpressure at given locations and it is im-
portant that maximum use should be made of this diagnostic information.
A problem with the diagnostic approach is relating damage to overpres-
sure or to impulse (the pressure/time integral). There is a considerable body
of information on the effects of blast from high explosives and nuclear
bombs that may be used to estimate overpressures. However, due to the dif-
ferent characteristics of the UVCE blast wave such data may lead to errors
in estimates of overpressure in this context, e.g. a blast wave of long dura-
tion but low peak pressure could cause more damage to certain objects, de-
pending on their frequency of response, than a blast wave of higher peak
value but shorter duration. There is therefore a need for more detailed
analyses in this area.
In this paper an analysis is given of the explosion damage to a set of
17 lamp posts located on the Flixborough site. The significance of these
lamp posts is that they were objects of standardised dimensions and form
of construction distributed fairly uniformly around the site. As rigidly
mounted vertical cylinders, they are also representative of many chemical
plant structures, e.g. distillation columns, flare stacks. The observed defor-
mation (Roberts, 1974) was mostly of a simple plastic type at a hinge
formed by the blast forces and the simple construction of the posts enabled
an analysis to be developed linking the applied forces to the observed defor-
mation. Since the deformation is progressive, an actual estimate of blast
loading is obtained, rather than a greater than estimate as with many
forms of damage (e.g. windows). The results of this analysis are related to
other information on the development of the explosion.
Gugan (1979) has also put forward an analysis of damage to some of the
same lamp posts but because of entirely different assumptions about blast
wave form and mechanisms causing deformation the two accounts are not
directly comparable.
The lamp posts were of a tubular steel construction with a larger diam-
eter base tapering to a smaller diameter upper portion.
Significant dimensions were:
233
SECTIONS 25A
I
(a)
I /
(b)
a mass per unit length of m and it is subjected to a steady force per unit
length (u)) perpendicular to its major axis for a period of duration td. For
an explosion, it is assumed that the force w results from aerodynamic drag
from the blast wind, which is reasonable where the difference in static pres-
sure across the object is of a lower order of magnitude than the stagnation
pressure. Clearly if the external force does not cause the elastic limit of the
object to be exceeded there will be no permanent deformation; if however
the elastic limit is exceeded there will be a permanent deformation, from
the magnitude of which the external force can be calculated if the ap-
propriate information is available.
When the elastic limit is exceeded it is conventionally assumed that a
plastic hinge is formed in the structure which exerts a constant resistance to
motion independent of the degree of rotation about the hinge. This
moment will be denoted MO. For a hollow cylinder, as in the case under
consideration
MO = + (R23 - R13)
where u0 is the yield stress of the material and Rz and RI are the outer
and inner radii of the cylinder, respectively.
If the length of the structure is L then the moment exerted by the
external load at the plastic hinge is wL2/2. Hence the equation of motion
for 0 < t < td may be written
mL3 wL2
e= -- MO (2)
3 2
3
(j= __
(3)
mL3
and
e=_-!.- -wL2
mL3 ( 4
-- M0 t2
2 )
(4)
3 WL2 td
(j= - -Mot
~ (5)
mL3 2
and
Motion of the displaced object ceases when 6 = 0, i.e. when t = wL2 t&2&&
Substituting this value of t into eqn. (6) gives the following equation for
e, the permanent angular deformation:
3w wL2
e^= - hi2 ( - -1
4t?lL 2s 1
z
-=
(8)
0d
; = 26&d (9)
For the above example of f3d = 18 and assuming a value of td = 200 ms
(on the basis of the analysis to be presented below)
Since the lamp posts considered have a height of 9 m, the maximum tip
velocity on the basis of these figures would have been 28 m s-l - a value
si~ific~t in comparison with the estimated wind speeds of 200 m s- but
not one to seriously limit the assumption of w being constant for
0 < t < td, i.e. a constant aerodynamic loading.
Fourthly, e,sn. (7) apparently shows that one needs to know both td and
w to predict 8. However;within the limits of accuracy feasible in this type
of work, if wL2~2M~ > 5 (say), eqn. (7) may be written
--& 0
(wtd)2
237
All the relevant ~mensions and properties of the F~~borough lamp posts
are known or can be estimated for substitution into eqns. (l)-(10). w, the
external loading on the posts derived from aerodynamic drag from the blast
wind, is given by
W = titPd&C (11)
Pd =$ pu2 (12)
where p = air density and U = air velocity. Since p is not known at each
location and time (being a function of temperature and pressure) a value of
1.3 kg mV3 will be assumed throughout, as a means of estimating approx-
imate local blast velocities.
In analysing the behaviour of the posts it was assumed that w and m
were constant with height (i.e. the effect of the broader base on w and M
was neglected - it is relatively small in numerical terms) but different
values of MO were used for bending at A and B (58.8 kN m and 15.9 kH m,
respectively).
The equations then predict that permanent deformation takes place at A
if Pd > 11.5 kPa (U =.133 m s-l) and at B if Pd > 5.7 kPa (U = 94 m s-l).
The data given in Table 1 show that in three cases(L2, LlO and Lll) bend-
ing was observed at both A and B, which is not a feature that the analytical
model predicts. One possible explanation is that the blast wind originally
caused bending at A and then decreased in velocity, so that bending ceased
at A but began at B. In calculating the total impulse exerted by the blast
wind, the independent contributions of bending at A and B have been
added together for the above 3 posts.
The formulae derived for bending at A and B (from eqn, (7)) - with g
in degrees, Pd in kPa and td in seconds - are:
Bending at A
TABLE 1
; = 3.33 I2 (13)
Bending at B
; = 8.53 I2 (14)
where I = Pdtd, the impulseAor pressure/time integral. Equations (13) and
(14) show a dependence of 0 on I only, hence they would also apply to
wave forms other than the one assumed in the derivation of the equation.
Equations (13) and (14) apply to a reasonable degree of accuracy where
6 > 2, i.e. in all circumstances where reasonably accurate measurements
can be taken in the field.
Values of I calculated from the observations of damage listed in the
Annex are tabulated in Table 1 in units of kPa s. IA denotes an impulse
calculated for bending at A, IB the equivalent for: bending at B, and
IA + IB the sum of the two. The distance of each post.from the explosion
centre, (see Fig. 1) as denoted by Sadee et al. (1976) is listed, together
with the value of Pd calculated from (IA + 1~) + td, assuming td = 200 ms.
The final column gives the value of U equivalent to Pd as described above.
No estimates could be obtained for L4 and L5 because of the complex de-
formation of these posts.
In Table 1, the inequality sign for greater than is used for cases where
8, = 90 or 19~= 130 because in these cases the tip of the lamp post
struck the ground and an additional decelerating force was applied to the
lamp post which is not included in eqn. (2).
239
Fig. 3. Comparison of dynamic pressure impulse for TNT explosion and estimates for
Flixborough explosion.
240
was about 100 kPa, for which the particle velocity in the shock wave is
180 m s- . It would, therefore, be physically inadmissible on this basis for
the blast winds at the 100 m distance to exceed 180 m s-l. From this argu-
ment it follows that for lamp post L2, to keep the estimated particle
velocity below this value, td > 280 ms. For lamp post L3 td > 190 ms. On
the other hand a value of Pd > 5.7 kPa would be required to produce the
slight bending observed in L17, in which case a value of td < 160 ms is in-
dicated. A value of td = 200 ms has therefore been used to estimate the
values of Pd and U given in Table 1. It is not necessarily the same as the
duration of the period of energy release in the explosion since the locus of
energy release was moving and creating gas movement in different regions at
different times. The above values of td and the estimates of Pd and U in
Table 1 are all affected by the assumption of p = 1.3 kg mm3, introduced
following eqn. (12), and are therefore not particularly accurate. Neither the
pressure/time history nor the temperature/time history at a particular loca-
tion can be estimated with any accuracy but it is considered that the estima-
tion of pd and U is a useful exercise, even if it can only be done approx-
imately, which justifies the assumption of a representative value of p,
L14. Less displacement than L15 and L13 which are further away.
Robably protected by the pipe bridge.
L6. No observable displacement although closest to centre of explosion.
Heavily impacted with soot on face away from office block. This suggests
that post was in a stagnation region with flow direction perpendicular to
the main wall of the office block rather than parallel to it (this is consistent
with the observed displacement of the office block).
L4 and L5. Twisted, apparently acted on by more than one set of forces.
Consistent with a flow parallel to N wall of project offices setting up a
region of intense shear with rapid &anges in flow direction.
If the general description of the explosion development given in this sec-
tion is valid, then Flixborough was certainly not an unconfined vapour
cloud explosion - partially confined would be a better description in
view of the solid floors of Sections 7 and 27, the walls in parts of Section 7,
the nearby office block, control room, pipe bridge, warehouse, etc. The pat-
tern of blast development would have been considerably affected by the pre-
sence of all these obstacles.
The view expressed above is partially consistent with that of Sadee et al.
(1976). Their explosion centre is fairly near the region in which it is esti-
mated that the explosion began to develop and the region of major blast ef-
fects lies within their estimate of the cloud boundary (see Fig. 1). Near its
limit, crop scorching probably resulted from a short exposure to a premixed
flame; the relative positions of the cloud boundary and the crop scorch-
ing boundary in Fig. 1 are consistent with their suggested causes. The sug-
gested pattern of explosion development is not consistent with the TNT
models indication of an elevated source (45 rt 24 m). The point of ignition
is to some extent irrelevant since many eye witness accounts describe the
existence of flames prior to the blast. A reasonable hypothesis is that flames
spreading through the cloud without significant blast effects reached the
portion of the flammable cloud in the vicinity of Sections 7 and 27 and
were there subjected to flame acceleration effects that gave rise to the blast-
producing high flame speeds necessary to create the explosion. Candidate
flame acceleration mechanisms would be partial confinement by the floors
of these sections and turbulence produced by the passage of flame past the
many items of plant in these sections (see Fig. 4).
Thus, the general pattern of explosion development indicated by the
lamp post displacement and other factors is not consistent with an elevated
point source of energy release but with a source of appreciable volume with
the local rate of energy release differing appreciably from one point to an-
other. As the source of energy release moved through the cloud, its ap-
parent origin as indicated by damage in the near field was affected by the
presence of major obstacles creating anomalous displacement directions.
However, in the medium to far field the significance of these effects de-
creased and the effects of the explosion approached those of a point source.
243
Fig. 4. Suggested location of rapid flame acceleration and major source of blast.
FUTURE WORK
0 / I , , 'A
TNne
01Seconds
:t ~ '~
Fig. 5. Typical overpressure-time profiles for dust and gas explosions in test gallery,
Fig. 6. Dynamic pressure-time profile for 9.8% methane/air explosion, measured in the
test section (37 m from closed end) with the reed anemometer.
Another obvious limitation with tests of this type is the size of test
specimen relative to the gallery diameter. To avoid boundary layer effects
and non-uniform loading, the specimens must be well clear of the gallery
walls. Specimens have therefore been limited to a height of 0.75 m. The
strength of the specimens that can be tested is also limited. As an indica-
tion of this, it was found that for the maximum pressures generated in the
gallery a 0.75 m by 9.5 mm thick mild steel cantilever would just suffer
permanent deformation. The range of size of test specimens could be ex-
tended by using the Buxton full-scale surface explosion gallery (366 m long
with a U-shaped cross-section of 2.7 m by 2.4 m high) as an alternative to
the 1.2 m diameter gallery described above.
Preliminary experiments with full instrumentation have been carried out
in the 1.2 m diameter gallery with mild steel solid rectangular (3.2 and
6.4 mm thick) and tubular (12.7 mm o.d. by 0.94 mm thick) cantilevers.
The results from these experiments are still being analysed and will be re-
ported at a later date. Further work with cantilevers is planned, which will
include looking at the influence of a change of cross-section on the
deformation.
For the rigid-plastic analysis of the lamp posts, described in Section 3, a
square overpressure-time history was assumed. Work is in progress on ex-
tending the analysis to other wave forms, including ones with a negative
phase. Analytical solutions have been obtained for a triangular overpressure-
time history. Results from the above experimental programme will be used
to test and refine these solutions.
246
CONCLUSIONS
ANNEX
Figure 1 shows the location of the seventeen lamp posts surveyed and
the estimated lines of displacement, where measured, These lamp posts are
designated Ll-L17 and further observations made on them are given below.
The angle between the lower portion of the lamp post and the vertical is de-
signated el, the angle between the lower portion and the upper portion of
the post is designated e2, while the height above ground level at which the
latter displacement starts is designated H (see Fig. 2(b)).
Ll. Moved since the explosion, but the original line of displacement was
estimated from an impact mark where the top of the lamp post struck the
surface of the car park. The top 2.4 m of the column had been rendered
concave on the surface facing the source of the blast.
81 = 90, e* = o, H = 0.
L2. Moved since the explosion, but the original line of displacement was
estimated from an impact mark where the top of the lamp post struck a
kerb stone on the opposite side of the road.
e1 = 25, e2 = 90, H = 2.25 m.
L3. Moved since the explosion but no impact mark located.
8, = O, ez = 140, H = 2.2 m.
L4. Deformed in a more complex way. The post appears to have been
twisted and bent in more than one place. No estimate for el, Bz and H.
L5. Apparently subjected to more than one displacement effect. The
lower part of the post was displaced in the direction shown in Fig. 1 and
the upper part in a somewhat different direction.
247
REFERENCES