26

FIRE HAZARDS IN CHEMICAL PLANT FROM

FRICTION SPARKS INVOLVING THE
THERMITE REACTION
By N . G I B S O N , P h . D . , D.I.C., B . S c , A.Inst.P.,* F . C. L L O Y D , L.Inst.P.,* and G. R. P E R R Y *

SYNOPSIS
Impacts involving aluminium and rusty mild steel can initiate a thermite reaction. It is shown that the glancing
impact of stainless steel, mild steel, brass, copper-beryllium, bronze, aluminium, copper, and zinc on to
aluminium smears on rusty mild steel can initiate a thermite reaction of sufficient thermal energy to ignite
flammable gas-air and solvent-air atmospheres and dust clouds typical of those found in the chemical industry.
The conditions of impact under which the different metals are most likely to produce an incendive thermite
reaction are described.
The data indicate that although the No. 1 Wheeler Test and Godbert-Greenwald Furnace Ignition Tempera-
tures of a dust cloud may not indicate its sensitivity to ignition by this form of friction " spark ", Class I dusts
are much more likely to be ignited than Class II dusts.

Introduction Rusty mild steel target and aluminium smear

T h e potential fire hazard from friction " sparks " has been The rusty targets were prepared by allowing the mild steel
recognised for many years in the mining and petroleum indus- blocks to rust slowly under atmospheric conditions. T h e
tries (see bibliography). Experiments with m e t h a n e - a i r aluminium smear was produced by rubbing a piece of alumin-
mixtures have shown that a particularly dangerous form of ium rod (aluminium content > 9 9 . 5 % ) across the rusty steel
friction " s p a r k " can be produced by impacts involving surface under hand pressure until the smear contained the
aluminium a n d rusty mild steel because of the initiation of a m a x i m u m a m o u n t of aluminium that would ahdere to the
thermite r e a c t i o n . 3 5 ' 3 6 , 4 1 rusty surface. R u b b i n g the aluminium across the rusty sur-
face produced a n u m b e r of loose aluminium particles on top
A thermite reaction can be produced by the simple action
of the smear. It was found that a more extensive thermite
of striking a n aluminium smear on a piece of rusty mild steel
reaction was obtained when these loose particles were allowed
with a h a m m e r . Burning aluminium particles produced in
to remain in the impact area. T h e incendivity experiments
this manner are shown in Fig. 1. Sulphur dust scattered over
were normally carried out with a n d without the loose alumin-
the impact area can be readily ignited by the thermite reaction
ium in the impact area.
produced by impact (Fig. 2).
In order to assess more fully the potential hazard of this
type of friction spark, experiments have been carried out to Impact conditions
determine: A glancing blow that did not stop the forward movement of
(a) the conditions under which a thermite reaction can the h a m m e r was found to be the most successful in initiating
be produced during glancing impacts between a metal a n d thermite reactions, and this type of impact was used through-
an aluminium smear on rusty mild steel, and out the investigations. A typical velocity profile for the
h a m m e r head during impact is shown in Fig. 4—the h a m m e r
(b) the incendivity of the friction " spark " in the pre-
sence of flammable gas-air mixtures, solvent v a p o u r - a i r head was in contact with the target surface for a distance of
mixtures, a n d chemical dust clouds. a b o u t three centimetres.

Experimental Conditions Production of the Thermite Reaction with Different Striker

Apparatus Metals
T h e a p p a r a t u s used in this study is shown in Fig. 3; it con-
sists essentially of a spring-loaded hardened steel h a m m e r In order to assess whether or not the production of a
(weight 1.4 kg) which, when released, accelerates rapidly and thermite reaction was dependent on the material of the
strikes a nearly tangential glancing blow on the flat, horizon- striker, metal caps m a d e from stainless steel, mild steel, brass,
tal, upper surface of a rusty mild steel block. Only the leading copper-beryllium, bronze, aluminium, copper, and zinc were
edge of the h a m m e r head makes contact with the test sample. fitted in turn on to the striking face of the h a m m e r .
T h e test sample is clamped in a vice bolted to the upper surface
of the lower plate. A rigid construction minimises any Characteristics of metal strikers
absorption of energy at the m o m e n t of impact by the relative A n analysis of the major components in each metal, the
movements of the various parts of the equipment. D i a m o n d Pyramid Hardness (D.P.H.) data a n d melting points
T h e striking metal can be varied by fixing caps m a d e from of the metals are given in Appendix I.
different metals on to the striking face of the h a m m e r .
Experimental procedure and results
* Imperial Chemical Industries Ltd., Dyestuffs Division, The aluminium smears on the rusty mild steel target were
Hexagon House, Blackley, Manchester 9. struck glancing blows with each metal cap on the h a m m e r

I.Chem.E. S Y M P O S I U M S E R I E S N o . 25 (1968: Instn chem. Engrs, London)

 GIBSON, LLOYD, and PERRY. FIRE HAZARDS IN CHEMICAL PLANT FROM FRICTION SPARKS 27

and a visual assessment made as to whether or not the impact (a) The presence of loose aluminium on the smear: this
produced a thermite reaction. appeared to aid the initiation of a thermite reaction.
The visual evidence could be placed into one of three
categories: (b) The number of impacts to which the striking area on
(a) N o visible evidence of a spark of any kind—symbol- the hammer head had been subject: the zinc and aluminium
ised by N. hammer heads rarely, if ever, gave a thermite reaction on
the first impact. However, if the same area of the hammer
(6) A red spark similar in appearance to the spark pro- was used for a number of blows without being cleaned
duced when two metals impact in the absence of aluminium. between each blow a thermite reaction could be produced.
When this occurred it was considered doubtful whether a During repeated impact the softer metals become impreg-
thermite reaction had been initiated by the impact and even nated with aluminium and rust and this appears to aid the
if it had, it had not propagated through the aluminium dust initiation of the thermite reaction. The effect of these two
cloud—symbolised by S. factors can be seen from the results given in Table I, for
(c) A white flash indicating a propagating thermite reac- five successive impacts with each experimental situation.
tion—symbolised by R.
The harder metals (stainless steel, mild steel, and brass) not
All the metals produced thermite reactions but with the only initiated reactions under all conditions of impact but
softer metals with lower melting points (i.e. aluminium, also on the first impact. Because of their hardness there was
bronze, copper, and zinc) a thermite reaction did not occur at little, if any, impregnation of the impact area by aluminium
every impact. Photographs of two thermite reactions pro- and rust and this was not necessary for a thermite reaction to
duced by each metal striker are shown in Figs 5 and 6. In be produced. Two reactions were obtained from five impacts
one photograph of each pair background lighting has been with the copper striker having a clean surface and no loose
used to show the position of the hammer. The size of the aluminium. Five successive reactions were obtained with an
thermite reaction shown in the photographs should not be impregnated surface. The same increase in reaction frequency
taken as characterising the relative magnitudes of the thermite was obtained from a clean surface when loose aluminium was
reactions that are possible using the different metals because present in the target area. The aluminium and bronze
the size of the thermite reaction for each metal varied some- strikers did not give reactions under either condition with
what from impact to impact. clean surfaces, but both readily initiated reactions when
Two factors were found to be important in determining aluminium and rust had become embedded in the impacting
whether or not a reaction was produced. faces. Five successive reactions could be obtained with the

I.Chem.E. SYMPOSIUM SERIES No. 25 (1968: lnstn chem. Engrs, London)

28 GIBSON, LLOYD, and PERRY. FIRE HAZARDS IN CHEMICAL PLANT FROM FRICTION SPARKS

TABLE I.--Effect of Cleaning Hammer Heads between Impacts

Impact Condition

hammer Clean Metal Impregnated Metal Clean Metal Impregnated Metal

— loose Al — loose Al + loose Al + loose Al
Bronze NNNNN RRRRR NNNNN RRRRR
Cu-Be SRRRR RRRRR RRRRR RRRRR
Stainless steel RRRRR RRRRR RRRRR RRRRR
Mild steel RRRRR RRRRR RRRRR RRRRR
Copper RNRSN RRRRR RRRRS RRRRR
Brass RSRRR RRRRR RRRRS RRRRR
Aluminium NNNNN RRRRR NNNNN RRRRR
Zinc NNNNN NNNNN NNNNN RRRRR

zinc hammer only when it had an impregnated striking surface and the percentage of reactions causing ignition was calcu-
and loose aluminium was present in the target impact area. lated. If an impact did not initiate a thermite reaction then
It cannot be concluded from these results that a thermite it was not included in the results.
reaction could never be initiated under those conditions for
which five successive non reactions were obtained. Neverthe- Coal gas-air atmospheres
less the results do show the relative frequency with which The glancing impact apparatus was enclosed in a metal
impacts involving the different metals can be expected to explosion cubicle filled with coal gas-air mixtures of known
initiate a reaction under the different impact conditions. The concentration in the range 4-20% (v/v).
results also indicate that, with soft metals, the chance of Thermite reactions produced by all the metals caused igni-
obtaining a reaction increases with the number of impacts tion of certain concentrations of the coal gas-air atmosphere
between the hammer surface and the aluminium smear. and a similar relationship between percentage ignitions and
The other conclusion of importance is that the use of coal gas concentration was obtained with each metal. The
copper-beryllium in so-called " non-sparking " tools does not relationship shown in Fig. 7 obtained with the copper-
significantly decrease the possibility of a thermite reaction beryllium striker is typical. Ignition was quickly obtained in
when the impact is on to aluminium-coated rusty mild steel. every case once the gas concentration exceeded a certain
minimum value. The minimum ignitable concentrations for
the different metals were:
The Incendivity of the Thermite Reactions
hardened mild steel, 5-0%
The incendivity of the thermite reactions produced by stainless steel, 5-0%
impacts involving the various metals has been examined using brass, 5-0%
flammable coal gas-air, methane-air, acetone-air, toluene- copper, 5-0%
air, and methanol-air atmospheres, and dust clouds. aluminium, 6-0%
Impacts were produced in each flammable atmosphere, a copper-beryllium, 6-0%
note made of the number of reactions that produced ignition, bronze, 6-0%
zinc, 6-0%.
All the minimum concentrations on the threshold of igni-
tion are at or just greater than the lower limit of flammability
of coal gas-air atmospheres (5%). The thermite reactions
produced in impacts involving the four metals with the lowest
D.P.H. hardness ratings—namely, aluminium, copper-
beryllium, bronze, and zinc—required slightly greater concen-
trations of coal gas before ignition was obtained and could
therefore be considered less incendive than those from the
harder metals. The change in coal gas concentration from
5 % to 6 % to give ignition is however too small to be of any
practical significance. It must be concluded that when the
concentration of coal gas is above the lower limit of flamma-
bility the thermite reaction produced by the impact of any
of the metals tested could ignite the flammable atmosphere.

Methane-air atmospheres
The equipment and test method for methane-air atmos-
pheres were identical with those used in the coal gas experi-
ments. Methane-air mixtures in the concentration range
4 % - 1 5 % (v/v) were examined.
Rae has shown 37 that strikers made from steel, brass, and
aluminium can ignite methane-air atmospheres. Tests were
only carried out therefore on the metals not used by Rae
(i.e. bronze and copper-beryllium) and on zinc, the metal
with which he could not obtain ignitions. The copper striker
was not available during the experiments with methane.
The experimental results were similar in form to those
Fig. 3.—Frictional impact apparatus obtained with coal gas. The three metals again produced

LChem.E. SYMPOSIUM SERIES No. 25 (1968: Instn chem. Engrs, London)

 GIBSON, LLOYD, and PERRY. FIRE HAZARDS IN CHEMICAL PLANT FROM FRICTION SPARKS 29

Start of impact
TEnd of impact

Fig. 4.—Velocity profile of hammer

head during impact

10 20 30 40
TIME (ms)

thermite reactions that could ignite the flammable atmosphere that can burn with explosive violence when exposed to a
provided the methane concentration was above a certain source of ignition (e.g. a hot surface, an electric spark) in a
minimum value (zinc, 6-2%; copper-beryllium, 5-3%; confined space.
bronze, 6-1 %). The lower flammability limit of methane-air The sensitivity of a dust cloud to ignition by a heat source
is 5-3%: thus, as with coal gas, it can be concluded that for can be measured by the No. 1 Wheeler Test and the Godbert-
all practical purposes the thermite reaction can ignite Greenwald Furnace Test. A dust cloud that ignites in both
methane-air mixtures in the flammable concentration range. the No. 1 Wheeler and the Godbert-Greenwald Furnace Test
Thermite reactions and ignitions were only obtained with the is classified as Class I, a dust cloud that does not ignite in
zinc striker after it had become impregnated with aluminium the No. 1 Wheeler Test but ignites in the Furnace Test is
and rust by contact with the aluminium smear. classified as Class II. A dust cloud that does not ignite in
either test is classified as Class III and will not ignite in normal
Solvent vapour-air atmospheres manufacturing processes.
Flammable acetone-air, toluene-air, and methanol-air In order to determine whether the thermite reaction pro-
atmospheres were used to test the possibility of ignition with duced during impact is a potential source of ignition, thermite
solvents. The concentrations of acetone and toluene were reactions have been produced in dust clouds from 95 products,
4-7 % and 2-3 % respectively. The concentration of methanol selected to be representative samples of Class I and II dusts.
vapour tended to vary due to condensation in the explosion In all the tests the stainless steel striker was used to produce the
chamber. A reaction giving non-ignition was not included in thermite reaction.
the data if an electric spark released in the impact area imme- A dust cloud of the powder under test was produced by
diately after impact did not cause ignition. placing a small cone-shaped heap of powder on the mild steel
The incendivity was determined with and without loose target at the edge of the impact area (Fig. 8). As the hammer
aluminium on the mild steel target. The experimental results moves in an arc it strikes the aluminium smear a glancing
from five successive thermite reactions are summarised in blow, initiates the thermite reaction, and projects burning
Table II. aluminium particles forward. These pass through the dust
These results indicate that if the impact involving any of cloud produced as the hammer continues its forward move-
these metals except zinc produces a thermite reaction then ment and strikes the heap of powder.
ignition of a surrounding flammable vapour-air atmosphere Up to five attempts were made to ignite each powder; if
will almost certainly follow. With a zinc striker it is more ignition occurred before the fifth reaction the tests were
difficult to produce a thermite reaction but when this occurs stopped and the number of the reaction that caused ignition
there is a reasonable probability that ignition will follow was noted. If ignition did not occur after five reactions " no
although it is possible to produce a small thermite reaction ignition " was recorded. The visual evidence after impact
that does not propagate throughout all the combustible depends not only on whether or not ignition occurs, but also
aluminium and which does not generate sufficient heat to on the form in which the burning is propagated through the
ignite these solvent vapour-air atmospheres. dust cloud. It was possible to divide the results into three
categories:
Clouds of chemical dust (1). Those in which, after impact, flame propagated clear
A wide number of powder products manufactured in the of the region of the thermite reaction, in some cases up to
chemical industry, whilst not explosives, can form dust clouds distances of two to three feet. This evidence indicated that

TABLE II.—Results of Impacts in Solvent vapour-air Atmospheres

Acetone Toluene Methanol

With Without With Without With Without

Metal loose Al loose Al loose Al loose Al loose Al loose Al
Stainless steel inn IIIII IIIII IIIII IIIII IIIII
Mild steel IIIII IIIII IIIII IIIII IIIII IIIRI
Brass inn IIIII mil IIIII IIIII RIIII
Copper IIIII IIIII IIIII IIIII IIIII IIRII
Cu-Be IIIII IIIII IIIII IIIII IIIII IIIII
Bronze IIIII mil IIIII IIIII IIIII IIIII
Aluminium IIIII IIIII IIIII IIIII IIIII IIIII
Zinc IIIII RIIRI RIIII IIRRI IRIII IIRII
Symbols: I—ignition of flammable atmosphere.
R—visible thermite reaction but no ignition of flammable atmosphere.

I.Chem.E. SYMPOSIUM SERIES No. 25 (1968: Instn chem. Engrs, London)

30 GIBSON, LLOYD, and PERRY. FIRE HAZARDS IN CHEMICAL PLANT FROM FRICTION SPARKS
 GIBSON, LLOYD, and PERRY. FIRE HAZARDS IN CHEMICAL PLANT FROM FRICTION SPARKS 31

i 1 to form a combustible mixture capable of propagating fire or

explosion. Some caution is necessary therefore in using the
100 data in Table III as an indication of the distribution of sensi-
tivities within a group of typical products, For products
categorised as I / P a thermite reaction in the presence of a
combustible mixture of the dust in air could clearly cause a
80 - propagating fire or explosion. With products categorised as
I /NP it may well be that the absence of flame propagation
CO was due to the poor dust dispersion produced in the test
o equipment and not due to an intrinsic inability of the product
to propagate flame. On the other hand the product may not
60 - - in fact be capable of extensive propagation of fire, and a
o thermite reaction or any other local ignition source could only
cause a localised fire. For the same reason categorisation of
a product as N 5 should not be taken as conclusive evidence
that ignition could not occur under all conditions of dust
40 - dispersion.
o Despite this imprecision, the results do indicate that the
UJ
thermite reaction resulting from the impact of the metal onto
an aluminium-coated rusty mild steel surface can ignite a
20 significant proportion of the dust clouds present in powder
manufacturing units.
The results also showed that the majority of dusts that were
ignited were from Class I powders; out of 42 Class II powders
tested only one ignited. Plots of the sensitivity ratings of the
n —-OQ- .. 1 1 samples to ignition by the thermite reaction (i.e. Iit I 2 , I3,14,
0 10 20
I 5 , and N 5 ) against the Godbert-Greenwald and No. 1
PERCENTAGE GAS CONCENTRATION Wheeler ignition temperatures failed however to show any
Fig. 7.—Percentage ignition/coal gas-air concentration. Copper-beryllium precise relationship between these variables (see Figs 10 and
hammer, mixture of coal gas and air 11). It is known that the relative sensitivity of dust clouds
from different powders depends very much on the nature of
not only could the product be ignited by the thermite reac- the source of ignition; this lack of correlation between igni-
tion but also that the combustion would propagate through tions obtained with the thermite reaction and the No. 1
a dust cloud of the material. These results were recorded Wheeler and Godbert-Greenwald Furnace Test is not there-
as I/P with a suffix on I denoting the reaction at which fore unexpected. It does mean, however, that the ignition
ignition occurred. temperatures determined on these latter tests cannot be used
(2). Those in which flame, distinct from the thermite to assess the sensitivity of a powder to ignition by a thermite
white " flash " could be seen but in which the propagation reaction.
was less than two to three inches long. These results were
recorded as I/NP with an appropriate suffix on 1. Source of ignition
The experiments with dust clouds show that although the
(3). The third type was that in which no evidence of major part of the thermite reaction occurs in the impact area
flame or propagation was seen—recorded as a non-ignition the burning particles of aluminium projected clear of this area
and symbolised by N 5 . In this category were included a can act as sources of ignition. In these experiments a number
few products that were charred by the burning aluminium of discrete burning particles were projected over distances
but where there was no evidence that the powder had sus- greater than 12 in. from the impact area. The weight distribu-
tained the combustion. tion of particles projected clear of the impact area in this
The data for the 95 dusts tested are summarised in Table III. apparatus is shown in Fig. 12. Although the majority of the
This table shows that 46 of the 95 dusts tested were ignited by particles are in the weight range 1-8 //g, individual particles
the thermite reaction and 27 of these propagated the burning weighing up to 16 //g were found.
outside the impact area. Fig. 9 shows the ignition of a powder
categorised as I/P. This powder is typical of many processed Impact area
in the chemical industry—No. 1 Wheeler Test temperature
975°C, Godbert-Greenwald Furnace Temperature 550°C. Path of hammer Powder
The classification of a test result as propagating or non-
propagating requires some comment. The initiation of burn-
ing and its subsequent propagation depends not only upon
the sensitivity of the product but also on the form of the dust
cloud into which the burning aluminium is projected. The
present method of producing a dust cloud by the hammer
movement is far from ideal. Under certain plant conditions Rusty mild steel Aluminium
(e.g. in the filter bag of a grinding unit or during pneumatic target smear
transfer) powders that were not well dispersed by the hammer Fig. 8 . -Schematic diagram showing position of impact area and powder
movement could be entrained in the air in sufficient quantity under test.

TABLE III.—Test Results on Dust Clouds

Test Result Ix/P I2/P I3/P U/P I5/P Ix/NP I2/NP I3/NP U/NP Is/NP N5
No. of Samples 25 1 1 0 0 16 2 1. 0 0 49

I.Chem.E. SYMPOSIUM SERIES No. 25 (1968: Instn chem. Engrs, London)

 GIBSON, LLOYD, and PERRY. FIRE HAZARDS IN CHEMICAL PLANT FROM FRICTION SPARKS

Fig. 9.—Ignition of powder sample

1 1 1 1
- I - O 0 o
o
<
r-:
UJ I , -
L
j O o -

- O 0 -

I,' 4 -
=
°° Class I dusts
c
xx Class 13 dusts
_
1

I. -
Fig. 10.—Comparison of sensitivity
to ignition by thermite reaction and
<si N .
-_ ..„.:i.|i:.;.g.g >i\ . X _ Godbert-Greenwald furnace
tion temperature
igni-
i • . i

200 400 600 800 1000

GODBERT-GREENWALD FURNACE IGNITION TEMPERATURE ("CI

| oo Class I dusts

Fig. I I.—Comparison of sensitivity

to ignition by thermite reaction and
No. I Wheeler test ignition tempera-
o o o o
ture. Class I dusts

200 400 600 800 1000 1200

No. I WHEELER TEST IGNITION TEMPERATURE CO

LChem.E. SYMPOSIUM SERIES No. 25 (1968: Instn chem. Engrs, London)

 GIBSON, LLOYD, and PERRY. FIRE HAZARDS IN CHEMICAL PLANT FROM FRICTION SPARKS 33

I I I It is not necessary for the impact to be in the flammable

atmosphere for ignition to occur. Individual particles of mass
7 - 9 greater than 7 ug are capable of igniting flammable solvent
and gas atmospheres.
3. Incendive thermite reactions could be produced by all
the metals examined but the proportion of impacts that caused
an incendive reaction was less for the metals of low hardness
6 - and low melting point. Soft metals {e.g. aluminium and zinc)
are more likely to initiate a thermite reaction if the striking
surface is impregnated with rust and aluminium from pre-
vious impacts.
4. The use of " non sparking " tools made from copper-
beryllium does not significantly decrease the possibility of a
dangerous friction " spark" when the impact is on to
en aluminium-coated rusty mild steel.
ID
_J
O 5. In the case of powder products, present evidence indi-
cates that the No. 1 Wheel Test and Godbert-Greenwald
< Furnace ignition temperatures do not provide a measure of
Q- the sensitivity of the product to ignition by the thermite
LL reaction.
O
01 6. For a thermite reaction produced by impact to initiate
3 a fire in a plant the following conditions must be satisfied:
m
(1). Not only must aluminium and rust be present in the
area but they must also come into contact to produce a
smear of aluminium on a rusty metal surface.
(2). The aluminium smear must be struck a blow by a
second object—if the striker is aluminium it is possible for
the smear and the thermite reaction to result from the same
blow or from two blows in succession.
(3). The subsequent thermite reaction must occur in or
propagate into a flammable atmosphere.
If any one of these conditions is not fulfilled a fire cannot
occur. The probability of fulfilling all three conditions will
vary from plant to plant and will depend on the nature of the
plant, its mode of operation and the type of product being
0 2 U 6 8 10 12 14 16 18 20 manufactured. A separate assessment of the hazard is there-
WEIGHT (ug)
fore required for each plant /product combination in order to
ensure safe operation and yet not impose unnecessary restric-
Fig. 12.—We/'g/it distribution of burning particles projected clear by tions on the use of aluminium.
impact area

Bowden and Lewis7 have shown that the minimum weight

of an aluminium particle capable of igniting 5 %-9 % methane- References and Bibliography
air atmosphere is 1 ug and tests by ourselves have shown that Anon. "Sparks from Hand Tools", 1956. (New York:
less than 7 ug of aluminium is required to ignite the stoichio- American Petroleum Institute),
metric vapour-air mixtures of acetone, toluene, and methanol.
This is of practical importance in that impacts outside a Anon. " Ampco Spark-resistant Tools. Spark Resistance.
flammable atmosphere, e.g. on the outer surface of a reaction Results of Comparative Tests of Material'", 1959. (Mil-
vessel, can still initiate a fire if burning particles of aluminium waukee: Ampco).
are projected into the flammable atmosphere e.g. enter a Anon. " Report of the Fire Research Board with the Report of
reaction vessel via the charge hole. the Director of Fire Research", 1962, p. 65. (London:
H.M.S.O.).
Anon. Report NSS 651-200/1958. (New York: N.Y. Naval
Shipyard).
Conclusions Allsop, G. A. Trans. Instn min. Engrs, 1938/9, 96, 310.
1. The glancing impact of stainless steel, mild steel, brass, Anfenger, M. B., and Johnson, C. W. Proc. Am. Petrol. Inst.,
copper-beryllium, bronze, aluminium, copper, and zinc on to 1941, 22, 54.
aluminium smears on rusty mild steel can initiate a thermite Bailey, J. C. Trans. Instn min. Engrs, 1958/9, 118, 223,
reaction and cause the ignition of flammable gas and solvent Bernstein, H. and Young, G. C. Report NGF/T/1/57.
atmospheres and dust clouds formed from certain powders (Washington: U.S. Department of Commerce, Office of
(particularly Class I products) manufactured in the chemical Technical Services).
industry. Bowden, F. P. and Lewis, R. D. Engineering, Lond., 1958,
2. The sources of ignition are: 186, 241.
Bowden, F. P. and Ridler, K. E. W. Proc. R. Soc, 1936,
(a) a dense cloud of burning aluminium immediately A154, 640.
surrounding the point of impact, and Bowden, F. P., Stone, M. A., and Tudor, G. K. Proc. R. Soc,
(6) individual particles of burning aluminium that may 1946/47, A188, 329.
be projected some distance from the point of impact. Brenner, P. and Eversheim, P. Gliickhauf 1958, 94, 180.
LChem.E. SYMPOSIUM SERIES No. 25 (1968: Instn chem. Engrs, London)
34 GIBSON, L L O Y D , and PERRY. FIRE H A Z A R D S I N C H E M I C A L P L A N T F R O M FRICTION SPARKS

13 33
Burgess, M. J. and Wheeler, R. V. Pap. Saf. Mines Res. Bd, Paterson, S. Phil. Mag., 1939, Ser. 7, 28, 2: Idem ibid., 1940,
46/1928. Ser. 7, 30, 437.
14 34
Burgess, M. J. and Wheeler, R. V. Pap. Saf. Mines Res. Bd, Pechuk, E. I. in " Problems of Mining Electro-mechanics ",
54/1929.
15
160, Third Edition. Vol. 11.
Burgess, M. J. and Wheeler, R. V. Pap. Saf. Mines Res. Bd, 35
62/1930. Rae, D . Res. Pap. Saf. Mines Res. Bd, 129/1956.
36
16
Burgess, M. J. and Wheeler, R. V. Pap. Saf. Mines Res. Bd, Rae, D . Res. Pap. Saf. Mines Res. Bd, 177/1959.
70/1931. 37
Rae, D. Res. Pap. Saf. Mines Res. Bd, 190/1960.
17
Burgess, M. J. and Wheeler, R. V. Pap. Saf. Mines Res. Bd, 38
Rae, D . Combust. Flame, 1961, 5, 341.
81/1933.
39
18 Rae, D . Res. Pap. Saf. Mines Res. Bp. (In preparation).
Coward, H. F. and Guest, P. G. / . Am. chem. Soc, 1927,
40
49, 2479. Rae, D . and Nield, B. J. Res. Pap. Saf. Mines Res. Bd,
1
" Coward, H. F . and Wheeler, R. V. Pap. Saf. Mines Res. Bd, 192/1960.
41
53/1929. Rae, D., Singh, B., and Danson, R. Res. Pap. Saf. Mines
2
Res. Bd. (In preparation).
° Demelenne, E. Annls Mines Belg., 9162, 7/8. 42
21 Riddlestone, H. G. and Bartels, A. / . Inst. Petrol., 1965, 51,
Dittman, P., Voigtsberger, P., and Schulz, F . Arbeitschutz,
1960, 8, 180. 106.
43
22
Evans, W. H., Jeffrey, R., and Titman, H. Res. Pap. Saf. Rosenblek, B. Bergb.-Arch., 1960, 21, 61.
44
Mines Res. Bd, 114/1955. Schreiter, W. Chem. Tech. Beri, 1953, 5, 708.
45
23
Grice, C. S. W. Res. Pap. Saf. Mines Res. Bd, 59/1952. Schultze-Rhonof, H. Gluckhauf, 1956, 92, 777.
46
24
Harrison, P. L. and Yoffe, A. D . Proc. R. Soc, 1961, A261, Schultze-Rhonof, H., and Weichsell, L. Mitt, westf. Berg-
357. werksch-Kasse, 1951, l , N o . 2.
25
Hartmann, I. Bureau of Mines Information Circular 7727/1955. 47
Silver, R. S. Phil. Mag., 1937, 23, 633.
(Washington: Department of the Interior).
48
26
Kingman, F . E. T., Coleman, E. H., and Rogowski, Z. W. Thomas, T. S. E. Colliery Guard., 1941, 163, 202.
49
J. appl. Chem., Lond., 1952, 2, 449. Thomas, W. G. Colliery Engng, 1962, 39, 377.
27 50
Kocherga, N . G. Nauchno-issledouatel'skie raboty Male N.II Thompson, W. and Titman, H. Res. Pap. Saf. Mines Res.
Za, 1959/60, p. 99.
28
Bd, 60/1964.
Konschak, M. and Voigtsberger, P. Arbeitschutz, 1957, 11, 51
Titman, H. Res. Pap. Saf. Mines Res. Bd, 90/1954.
214.
52
29 Titman, H. Trans. Instn Min. Engrs, 1955-56, 115, 535.
Latin, A. Colliery Engng, 1959, 34, 397.
53
30 Titman, H. and Wynn, A. H. A. Res. Pap. Saf. Mines Res.
Margerson, S. N . A., Robinson, H., and Wilkins, H. A. Res.
Pap. Saf. Mines Res. Bd, 75/1953. Bd, 95/1954.
54
31
McCombe, J. Jl R. tech. Coll. Glasg., 1933/6, p. 173. Weichsel, L. Gluckhauf, 1955, 91, 1048.
55
32
Nagy, J. and Kawenski, E. M. Bureau of Mines Report of Wynn, A. H. A. Res. Pap. Saf. Mines Res. Bd, 42/1952.
Investigation 5548. (Washington: Department of the The manuscript of this paper was received on 1 April, 1967.
Interior).

Appendix I
Description of Striker Metals
Hardness
Approximate
Major Melting D.P.H. Data
Metal Constituents Point Remarks
(%) (°C) Load (kg) Hardness
Stainless steel Fe 72 1450 30 224 The hardness of the grades of austenitic chromium nickel steel that are
Cr 17-5 widely used in the chemical industry is 200 D.P.H. maximum in the
Ni 8 fully softened condition in which they are usually supplied.
Mild steel Fe 99-9 1530 10 177 Ordinary low-carbon mild steel which is in the normalised condition
would have a hardness of about 100-120 D.P.H. This material is
probably in a cold worked condition.
Brass Cu 57 900 10 144 In the annealed condition brass would have a hardness in the range
Zn 40 65-75 D.P.H. This sample is in a work hardened condition.
Cu-Be Cu 9S 975 10 140 This sample was machined from the head of a " non-sparking " hammer.
Be 2 The fully heat treated alloy that would be used for chisels, etc., could
have a hardness of 350 D.P.H.
Bronze Cu 87 1032 10 110 This is the order of hardness expected in r. chill cast 10% tin-bronze
Sn 10
commonly used for bearings, etc.
Aluminium Al > 9 9 660 10 106 Typical value for heat treatable alloy in fully heat-treated (WP) condition.
Mg < 0 - l
Copper Cu > 9 9 1083 2-5 54 Typical of commercially pure copper in the annealed (soft) condition.
Zinc Zn > 9 9 420 2-5 40 This is a " mean " value; the hardness varied according to the direction
it was determined relative to the internal structure of the metal.

I.Chem.E. S Y M P O S I U M SERIES No. 25 (1968: Instn chem. Engrs, London)

 GIBSON, LLOYD, and PERRY. FIRE HAZARDS IN CHEMICAL PLANT FROM FRICTON SPARKS 35

DISCUSSION work had been done on this by the Safety in Mines Research
Establishment. His department tended to follow them. As
Mr. F. J. OWEN asked if there was a thermite reaction if there far as aluminium paint on objects was concerned, it depended
was rust on aluminium instead of the other way round, and if on the base of the paint; some were safer than others. At
so had this been investigated? present, in his department, they tended not to use them if they
could avoid it but there was a tradition in the chemical in-
Dr. GIBSON said that it had never been done. Aluminium dustry to use aluminium paint. He was in a research depart-
was such a soft metal that it was very difficult to get rust to ment and there was a credibility gap between them and the
stay on it for any length of time. engineers. He did not think that was peculiar to ICI but it
was fairly common. He did not wish anyone to get the wrong
Dr. H. S. EISNER welcomed the paper because it was clear impression; they were not saying " Don't use the aluminium ",
that the aluminium hazard described by Gibson was still not they were saying: " This could be a danger and it should be
as well known throughout industry as it deserved to be. One used sensibly."
still came across the belief that, on the contrary, aluminium
was a " non-sparking " metal. Indeed, kits of so-called non- Mr. Z. W. ROGOWSKI said that he worked on fire dangers in
sparking tools containing some that were made of aluminium respect of use of aluminium paint many years ago. Sixteen
were still on sale. In the petroleum and gas industries the use commercial paints of various compositions were tested and the
of aluminium containing several per cent of magnesium results indicated that whether the incendive sparks would be
(which increased the hazard) was rapidly gaining ground. produced depended very much on the vehicle incorporated
The safe use of those materials depended on the constant into the paint. No commercial paint produced incendive
awareness of the hazard by all concerned in the design, sparks unless heated. As a result of this work certain re-
construction and use of equipment made of them. commendations were issued by Factory Inspectorate and some
Answering Mr. Owen's question, he thought that if the restrictions were placed on paints based on cellulose nitrate.
conditions of impact were right it would be possible to obtain It should be noted that since then aluminium paints of other
a thermite reaction when powdered rust placed on aluminium compositions had appeared on the market.
was struck by an external striker.
Dr. GIBSON said that the work he was thinking of was the
Mr. P. L. KLAASSEN asked what was Gibson's attitude work of Grice at SMRE (Ref. 23 of the paper). He suggested
towards the use of aluminium paint. Moisture caused rust in his report that the surface did not need to be heated and
behind the paint and could thus create possibly the right that there was sufficient heat from the impact to cause
mixture for ignition. ignition. That was an important point. Very often one could
paint a cold surface—not a radiator or steam pipe—and
Dr. GIBSON replied that aluminium paint had not been Grice had pointed out that the impact itself would give
examined by himself and Messrs Lloyd and Perry. Much sufficient heat and no prior heating was needed.

I.Chem.E. SYMPOSIUM SERIES No. 25 (1968: Instn chem. Engrs, London)