- Email: [email protected]
Towards absolute minimum ignition energies for dust clouds?
COMBUSTION AND FLAME 24, 53-64 (1975)
53
Towards Absolute Minimum Ignition Energies For Dust Clouds? ROLF K. ECKHOFF Chr. Michelsen Institute, Department of Applied Physics, N-5000 Bergen, Norway
A review of the literature related to spark ignition of dust clouds revealed that contrary to what is the case of gases, the spark energy required for igniting a given dust cloud is markedly dependent, even by orders of magnitude, upon the characteristics of the spark discharge, in particular the discharge time. By using a specially designed spark generator producing sparks of optimal discharge characteristics, it was found that airborne clouds of various dusts could be ignited by considerably smaller spark energies than hitherto published minimum ignition energies for similar dust clouds. Thus, with these optimal sparks the minimum ignition energy for lycopodium was found to be about 6 mJ, for wheat grain dust containing a considerable fraction of coarse material, about 12 m J, for sulphur less than 0.3 mJ, and for a fine quality flake aluminium somewhere between 0.3 and 0.9 mJ. By keeping in mind that this typ2eof optimal sparks is associated with capacitive discharge circuits in which more than 90% of ½CV is lost in the external circuit resistance, and therefore only less than 10% delivered in the spark, an approximate estimate of the electrostatic hazard may be obtained by multiplying by a factor of ten the "absolute" minimum ignition energies found.
I. Introduction The concept of minimum ignition energy, which has been discussed by Lewis and yon Elbe [I ], implies that if thermal energy is instantaneously delivered within a small volume o f an inflammable gas mixture, the minimum energy required to cause ignition is constant for a definite gas mixture and thermodynamic state. As will be shown in the following, the concept of minimum ignition energy for gas mixtures is resting on convincing experimental evidence obtained from ignition o f such mixtures by means of electric sparks. As will also be shown in the following, similar evidence appears to be lacking for the ignition of explosible dust clouds by means o f electric sparks. Nevertheless, the concept of minimum ignition energy has been adopted for the characterization of such systems, and considerable amounts o f data have been published [2, 3, 4, 5, 6].
H. Evidence Related to Spark Ignition of Gases In order to demonstrate the type o f evidence available, investigations on the ignition o f 8.5%
methane in air has been chosen as a typical example. The minimum stored capacitor energy required for the ignition of this mixture, at normal temperature and pressure, was determined, independently, by Lewis and yon Elbe [ 1 ] and Litchfield [7], by means of capacitive spark techniques. In both investigations the minimum ignition energy was found to be 0.28 mJ. Since the order o f spark discharge times in both investigations was probably 1 #sec or less, and since it has been shown by Lintin and Wooding [8] and Litchfield [9], that the time between the initiation o f the discharge and the inflammation in an 8.5% methane/air mixture at atmospheric pressure probably is o f the order o f 0.1 msec, the requirement of approximately instantaneous energy delivery was satisfactorily fulfilled. At SMRE [10], ignition experiments with the same gas mixture were carried out by means o f the standard inductive break flash apparatus. Initially a rather large minimum ignition energy was found, namely 2.3 mJ, which is about eight times the value obtained by capacitive sparks.
Copyright © 1975 by the Combustion Institute Published by American Elsevier Publishing Company, Inc.
54 Later, however, Gordon et al. [11] reported that they had been able to reduce this figure considerably by increasing the separation speed of the electrodes in the break flash technique, and minimum ignition energies comparable with those obtained by capacitive sparks were recorded. This reduction in minimum ignition energy was explained in terms of reduction of the heat loss to the electrodes. Similar experiments were carried out by Berz [ 12, 13], who, instead of increasing the separation speed of the electrodes, reduced the electrode size. Finally, by breaking wires of 0.025 mm diam Berz was able to ignite the 8.5% methane/air mixture with inductive sparks having an energy, in terms of ~.Li 2 , where L is the inductance and i, the current, of about 0.30 m J, which is practically identical to the value of 0.28 mJ obtained with capacitive sparks. Again, this result can be explained in terms of heat loss to electrodes, since, at a given electrode separation speed, and for a given electrode material, the heat flow into the electrodes must obviously decrease with decreasing electrode dimensions. It can, therefore, be concluded that the minimum spark energy needed for the initiation of explosions in an 8.5% methane/air mixture, at normal temperature and pressure, appears to be largely independent of the characteristics of the spark, which in turn supports the assumption that the minimum ignition energy for an explosible gas mixture is, in fact, a sensible concept. The evidence gained from igniting methane/air mixtures is slightly opposed by the findings of Priede [14], who studied the electric spark ignition of hydrogen/air mixtures. Priede was able to demonstrate that if a series resistance of the order of 5 X 104 I2 was introduced in the capacitive discharge circuit, the minimum spark energy for ignition was reduced to about half the value found without the resistance in the circuit. This effect may be explained in terms of the resistance producing a prolongation of the spark discharge time, which in turn means a reduction of the energy losses from the spark due to thermal radiation, heat sink effect of electrodes, shock waves and expansion work prior to ignition.
ROLF K. ECKHOFF Although the evidence put forward by Priede to some extent contradicts the concept of minimum ignition energy, this concept, when applied to gases, is nevertheless, an acceptable approximation. The question to be considered in the following is then whether this is the case even with dust clouds. III. Earlier Investigations on Electric Spark Ignition of Dust Clouds A. The Work of Boyle and Liewellyn
The first set of data indicating that the application of the concept of minimum ignition energy to dust clouds requires considerable care was probably produced by Boyle and Llewellyn [ 15]. These workers were able to demonstrate that the minimum capacitor energy ½ C V 2, C being the capacitance and V the initial capacitor voltage, capable of igniting dust clouds of various powders, decreased quite considerably when a series resistance was introduced in the discharge circuit. Some of the results obtained by Boyle and Llewellyn for transient dust clouds, produced by dispersing a quantity of powder by means of a blast of air, are shown in Fig. 1. As can be seen, the minimum %CV 2 for ignition decreased by a factor of about ten both with granular aluminium and magnesium, when a series resistance of 104 to l0 s was added to the discharge circuit. Similar trends were also found by these workers for dust clouds of ferromanganese, zinc, silicon, and sulphur. Boyle and Llewellyn expressed their conclusions in terms of capacitor energy. It is, however, to be expected that an appreciable series resistance in the spark discharge circuit will, during discharge, absorb a significant fraction of the capacitor energy, so that the energy delivered in the spark gap will be significantly less than the theoretical value ½ C V 2 . In fact, this fraction has been determined experimentally by various workers. Some results for spark discharge in air at normal room temperature are shown in Fig. 2. The relationships established by Priede [14] and Moore et al. [17] for capacitors of 1000 pF and V-zCV2 values of 12.8 and 18.0 m J, respectively, are particularly interesting, since the most incendiary sparks produced by Boyle and Llewellyn were obtained
MINIMUM IGNITION ENERGIES FOR DUST CLOUDS ~q
cuit, reduced the minimum net spark energy for ignition to only one per cent, or perhaps even less, of the energy required with a small series resistance. The question now arises whether a plausible explanation of the rather dramatic influence of the size of the series resistance can be found. An observation made by Boyle and Llewellyn during their experiments with sulphur may throw some light on the problem. In their paper these workers wrote as follows:
I
~. 0.6i
z o p-
•
0t|
_o e.- 0o
,o
--. 0.: Z z
~r0.0 101
10z
i113
111~
t11s
t06
SERIES RESISTANCE IN OHMS
The spark produced by direct discharge of a condenser is a short, sharp, noisy emission of energy. If the capacity is small and the voltage high the spark appears to push the cloud away from it; ff a spark of the same energy is derived from a condenser of greater capacity, i.e.,at a low voltage, this effect is greatly reduced.
Fig. 1. Results from ignition o f dust clouds by capacitive sparks, using an additional series resistance in the discharge circuit. Data from Boyle and Liewellyn [ l 5 ].
with similar capacitances and ½ C F 2 values. Although there is some discrepancy between the data of Priede and those of Moore et al., it is clear that, with a series resistance in the range of 10 to 107 I2, the net spark energies were only of the order of 5% to 10% of the initial capacitor energy ½CF 2 . This, in turn, means that in the experiments of Boyle and Llewellyn, an inclusion of a series resistance of 104 to l0 s 12 in the discharge cir-
-n-
o0
55
B. Investigations with Lycopodium Being a naturally occurring dust with constant properties such as particle size, shape, and chemical composition, lycopodium is very suitable for dust explosion research, since a sensible comparison of data obtained with this material in different laboratories can be performed. Thus Line et al. [19] in their investigations used steady-state wall free and wall confined 1
0~,
J, 4.2 pF, 1141
i
u) 80 z
so
~
L
"
1,1
c,~
I
~
,Oo
i~ dL
12.5
o:
1114
\/
115mJ, 9pF, 1141
200
lo0
1111
lo 2
8__
lo 3
ioz.
tos
m5
lo?
#
SERIES RESISTANCE IN OHMS
Fig. 2. Results from determination ofl~rcentage of total capacitor energy dissipated in the spark for a range of capacitances, spark energies and additional series resistance in the discharge circuit.
56
ROLF K. ECKHOFF
and 2 in diam columns of lycopodium suspended in the oxiding gas. Some of the results obtained by these workers for 1 in columns in air are shown in Fig. 3. 1.0
111,,, I z 0
0.6 ,
~- O.4: z -- 0.2
~ ,°.~
NO ADOITIONAL RESISTANCE
111111
I
,t?!. ! I1!!1
~,' W~LL CONF NEO I I
I
l.}llllll (o)
III1[11 III ./I
IIII l lliilill
u~ 0.6
{b) LL
2o
10-3
,11 10"2
Ill 10-1
1
10
CAPACITOR ENERGY ~ J
Fig. 3. Frequency of ignition of a 1 in diam. stationary column of 80 mg per litre of lycopodium in air, as a function of capacitor energy. Effect of wall confinement and additional series resistance in the discharge circuit. Data from Line et al. [19]. As can be seen both for the wall free and the wall confined case the stored capacitor energies required for producing a given probability of ignition, decreased roughly by a factor of ten if a series resistance of l0 s I2 was included in the discharge circuit. It is interesting to note that both the order of magnitude of the decrease, as well as the order of magnitude of the series resistance giving this maximum decrease, agree with the corresponding figures found by Boyle and Llewellyn [15] for completely different powders and an entirely different experimental technique. Line et al. produced direct evidence, by means of high speed photography, supporting the suggestion that a short spark can in fact distrub the dust cloud significantly. Hence, in the case of a capacitor energy of 1.5 J and only a very small series resistance in the discharge circuit, the high speed photographs revealed that a dust free zone of about 4 em a was created around the spark channel within the first two msec after the initiation of the spark discharge. This means that the
dust particles and the spark were effectively separated, at this stage, so that ignition could not occur. It was also demonstrated by high speed photography that with a l0 s ~2 series resistance, or a considerable series inductance, the disturbance effect was practically absent. Some results produced by Eckhoff [20] may further elucidate the problem in hand. This worker studied the ignition of transient clouds of lycopodium in air, produced by dispersing a quantity of the dust in a one liter bomb, by a short blast of air. The ignition sources were capacitive sparks of discharge times of the order of a few psec. Some of Eckhoff's results are given in Figs. 4(a) and 4(b). Each frequency of ignition is based on at least one hundred single trials. By comparing the data in figs. 4(a) and 4(b) with the data for the wall free case in fig. 3(a), a remarkable coincidence is observed. Both Line et al. and Eckhoff found that the probability of ignition remained in the range 0.3 to 0.7 over the considerable range of ½CF 2 values from 0.3 to 8 J. A possible explanation for this effect might be that the increase in the incendivity to be expected from the increase in the spark energy is nearly
1.0 0°8
H/I | IIIIIIII
0.6 z
Io)
OJ ,
i-
z i
o.2
o.o I I IIIIlll I I[lllll[
°i
I I IILII
14.
10-3
Ib)
10"2
tO-I
1
tO
CAPACITOR ENERGYt J
Fig. 4. Frequency of ignition of transient clouds of lycopodium in air in a one litre bomb, as a function of capacitor energy. Series resistance in discharge circuit 0.01 ,.(2. Data from Eckhoff [20].
MINIMUM IGNITION ENERGIES FOR DUST CLOUDS compensated by the increased ability of the spark to disturb the dust cloud. In fact, as indicated by the decrease of probability of ignition with increasing spark energy shown in Fig. 4(b), it might even be possible that the increased incendivity is more than compensated by the increased disturbance. Eckhoff [20] has considered the problem of combining the two opposing mechanisms in a unified picture. IV. Optimum Spark Discharge Time for Ignition By considering the work of Line et al. [19] and Boyle and Llewellyn [15], in combination with Fig. 2, one would anticipate that if a spark discharge circuit was designed to supply well defined sparks with discharge times of the order of the sparks produced by these workers with a series resistance of l0 s I2, very low minimum ignition energies would be found. In fact, considering Fig. 1 and Fig. 3(b), and taking into account that only 10% or less of ~ C V 2 was delivered in the spark, one would expect that ignition would be possible with net spark energies of the order of 1 mJ and perhaps even less. The current in a markedly overdamped R - C - L series discharge circuit, after the initial rapid rise to its maximum value, is given by the approximate equation: Eo
i .... R
e
-t/RC
,
(I)
where E o is the initial capacitor voltage, R the circuit resistance, C the capacitance, and t is the time. By defining the discharge time as the time required for the current to decrease to 1% of the initial value (t=0), it is found that the discharge time equals t=4.6 • R - C..
57
series resistance of 5 × 104 [2. The lowest ½CV 2 for ignition in the whole series of experiments was found with the 0.00055 #F capacitance, and in this case the discharge time, as defined in Eq. (2), was about 0.13 msec and the net spark energy probably of the order of 1 mJ. For a magnesium dust it was found t h a t R = 15 × 10 a [2 and C = 0.01/IF produced the lowest ½CV 2 for ignition, and in this case the discharge time was about 0.7 msec. and the net spark energy probably of the order of 2-3 mJ. For a granular aluminium, the optimum occurred a t R = 75 X 10 a I2 and 0.01 /aF, i.e., with a discharge time of 3.5 msec. and a net spark energy probably of the order of 5 mJ. It is thus seen that the optimum discharge time for ignition varies within the approximate range from a tenth of a msec to a few msec. A systematic increase in the optimum discharge time with increasing spark energy is indicated.
V. The CMI Spark Generator The various features of the generator will be discussed in detail elsewhere [21 ]. In the present context, it is sufficient to focus on the basic layout of the discharge circuit and on the main characteristics of the sparks produced by it. The discharge circuit is illustrated in Fig. 5, in which C is the discharge capacitor having an initial voltage of V. By having a range of capacitors from 40/~F down to 0.005/IF and variable voltages V from 1000V and downwards, a continuous range of ½CV 2 values covering about four decades is available.
V
(2)
By inserting the optimum values found by Line et al., namely C = 0.0008/~F and r = l0 s I2, one finds a discharge time of 0.37 msec. Boyle and Llewellyn give some interesting results for sulphur dusts, showing that for a range of capacitances from 0.01 to 0.00055/aF, the lowest ½CV 2 for ignition at each capacitance was obtained with a
v
w
Fig. 5. Diagram of the CMI circuit for generation of optimal synchronized capacitive sparks for ignition of dust clouds.
58
ROLF K. ECKHOFF
Initiation of the spark discharge at the desired moment, which is essential if synchronization of spark with formation of a transient dust cloud is required, is accomplished by means of a trigger circuit in which a capacitor C T, a switch S, and the primary coil of a transformer T constitute the essential elements. By closing the switch, a high voltage pulse of approximately 15 kV peak value, is induced in the secondary coil of the transformer, causing breakdown of the spark gap G, and thereby, discharge of the main capacitor C. In order to produce sparks similar to those from overdamped circuits (high series resistance) a diode D was applied across the discharge capacitor. Figure 6 shows two typical pairs of oscilloscope traces of spark current and voltage across spark gap for sparks produced by the CMI generator. Figure 6(a) refers to a 3.3 J spark, whereas the spark energy was only 2.4 mJ in the case shown in Fig. 6(b). It is seen that the discharge times vary
considerably, being about 1.7 msec. for the 3.3 J spark and only about 60/~sec. for the 2.4 mJ spark. The net spark energies produced by the generator were determined by recording pairs of oscilloscope traces of the type shown in Fig. 6. In Fig. 7, the resulting calibration curve, relating the net spark energy to the corresponding ½CI/2, is given. As ½CF 2 decreases, the energy supplied by the triggering circuit becomes significant, and in the range of very low energies, the net spark energy is larger than the ½CI/2 of the main discharge capacitors. In fact, the lowest net spark energy used in the tests, 0.3 mJ, was obtained by means of the triggering circuit alone. All the energy values quoted in the following account of ignition experiments are the net spark energies obtained from the average calibration curve in Fig. 7. It is seen that in the lower range, there is a noticeable scatter in the measured energies, which indicates the degree of accuracy
120
0.6
100
0.5 f ' ~ k l l
80
/---,
60
/
04 0.3
SPARK CURRENT
SPARK CUI~RENT
~0.2
~-
2Q
oo S(]
70 6O
0.5
I
1.0
I
< o.1
1.5
o.o o
2.0 4oo
I
250
o> , I (
200
3O
150
20
IO0 50
05
60
80
I
VOLTAGE ACROSS SPARK GAP
300
I
,~ so
0.0
~
60
I
350
VOLTAGE ACROSS SPARK GAP
20
1.0 TIME. MILUSEC
(a)
1.5
2-0
0
f 2O
t.0
60
80
TIME. MICROSEC
(.b)
Fig. 6. Two examples of simultaneous oscilloscopetraces of spark current and voltage across spark gap for discharges produced by the CMI generator. (a): Spark energy 3.3 J; (b): Spark energy 2.4 mJ.
59
MINIMUM IGNITION ENERGIES FOR DUST CLOUDS associated with the energy values quoted. Figure 8 shows a plot of the discharge times as a function
of the net spark energies for sparks produced by the CMI generator.
I0
/ /J
,/
>.
W z
cI 1/2 V2
AC,UAL SPARK
r,,
ENERGIES
I,--
lO-3g
/ /
'//// 10-4
10-4
10 -3
10 - 2
10-!
I
10
100
CAPACITOR ENERGY. J
Fig. 7. Measured spark energies as a function of initial capacitor energies for discharges produced by the CMI generator.
10"z
CJ UJ u) Z 0
f
-3
0
.2.
X
U~
I
,,.M
f
10.4
/ 10-s
I0 -~
10-3
10":
10 -t
1
SPARK ENERGY,JOULES
Fig. 8. Duration of spark discharge as a function of measured spark energy for sparks produced by the CM1 generator.
10
60
ROLF K. ECKHOFF
VI. Ignition of Dust Clouds with Sparks from the CMI Generator A. Experimental Procedure A transient dust cloud was generated by dispersing a given quantity of powder, by means of a given blast of air in a standard Hartmann apparatus [22, 23], modified by inserting a conical ring between the dispersion cup and the vertical perspex explosion tube in order to guide the dust cloud towards the spark gap [24]. For each type of powder an optimum combination for ignition of amount of powder, dispersing air pressure (fixed 500 cm 3 reservoir), and time lag between the release of the dispersing air and the spark discharge, was determined in a series of initial tests. B. Experimental Results. During the period of time in which the spark generator has been in operation, a wide range of dusts have been tested. In the present paper four characteristic sets of data will be presented. Figure 9 shows the results obtained with lycopodium having its natural moisture content of about 3.5%. As can be seen, ignition occurred even with a spark energy as low as 7 mJ. This
i.0
i.o i
~_ 0.e, ....... z
O.8
~ 0.4
11 OI
~ 0.2 u_ 0.0 10-3
extracted from Fig. 3, these workers were able to produce ignition, with a series resistance of l0 s I2 in the discharge circuit, by using ½ C V z values down to the range of 20 to 50 mJ, depending on whether a wall confined or a wall free dust column was used. By assuming, as indicated by Fig. 2, that less than 10% of ½ C V 2 was delivered in the spark, it follows that the minimum ignition energies found by Line et al. were of the order of a few mJ. This conclusion is in reasonable agreement with the results produced in the present investigation and given in Fig. 9. Attention is now drawn to Fig. 10, which gives the results from tests with a dust from Australian wheat collected from the top section of a bucket elevator in a grain silo plant during operation. Prior to the tests the dust was dried, but otherwise it was tested in its original state. The particle size distribution was quite wide, 50%, by weight, being coarser than 55/am. As can be seen from Figure 10, it was possible to ignite this dust with spark energies as low as 20 m J, whereas the minimum ignition energies quoted in the literature [2] for wheat flour and dusts from untreated wheat and wheat straw are in the range of 50 to 60 mJ. Although the percentage difference between the two values is less than that for lyco-
IHt titH q .
If 2
i~ I SPARK ENERGY=J
.
.
.
~ 0./., 0.2
.
i
10
10-3
If ; I
J
1(:
./1
I
i'i iili IIIH wi 10-i
I
I0
SPARK ENERGY, J
Fig. 9. Frequency of ignition as a function of spark energy for transient clouds of lycopodium in air, using sparks from the CMI generator as ignition sources. Ten trials per plotted point.
Fig. 10. Frequency of ignition as a function of spark energy for transient clouds of Australian wheat grain dust in air, using sparks from the CMI generator as ignition sources. Ten trials per plotted point.
is considerably less, by a factor of six, than the minimum ignition energy of 40 mJ quoted in the literature [2] for this particular dust. It is, however, of considerable interest to compare this rather low value with the results of Line et al. [19] discussed earlier in this paper. As can be
podium, it, nevertheless, represents a factor of three, in spite of the fact that the dust tested in the present work contained a considerable fraction of coarse material. The next series of results to be discussed is shown in Fig. 11. In this case the powder tested
61
MINIMUM IGNITION ENERGIES FOR DUST CLOUDS z
1,0
UJ
~" 0.2
14.
o.o I0 -t,
-7
AA
IIIll i
I IllJ
fiIllll l t
illll 1111 10-3
tfil
10-2 SPARK
A A
illl fll !11!1s!L lIJl Ii II I I I;ii till
[ I:Jl J,I
10-1
1
10
ENERGY t J
Fig. 11. Frequency of ignition as a function of spark energy for transient clouds of sulphur in air, using sparks from the CMI generator as ignition sources. Ten trials per plotted point.
was finely ground sulphur. As can be seen, ignition occurred even with spark energies as low as 0.3 mJ. This value is, in fact, nearly two orders of magnitude less than the value of 15 mJ quoted in the literature [6] as the minimum ignition energy for clouds of a very fine ("4/am") sulphur powder. This is quite a dramatic result, since it implies that the energy required for ignition in this case probably is even less than the minimum ignition energy for 8.5% methane in air. It is, in this context, of substantial interest to consider the results obtained by Boyle and Llewellyn [15] for the minimum ½ C V ~ for the ignition of sulphur. These workers found that by using the optimum combihation of capacitance and series resistance for ignition, namely 550 pF and 5 X 104 I2, the minimum ~ C V 2 for ignition was as low as 9 mJ. By combining this result with the evidence provided by Fig. 2, it can be concluded that the minimum net spark energy for ignition was, in this case, probably less than 1 mJ. The result produced by the CMI generator is therefore in agreement with the findings of Boyle and Llewellyn. Finally, attention is focused on some results shown in Fig. 12 from tests with a very f'me aluminium powder. Scanning electron microscopy revealed that the powder was composed of flaky particles with typical diameter 10/am, and thickness of the order 0.1/am. As can be observed in Fig. 12, clouds of this powder were readily ignited with spark energies of the order o f l mJ. The lowest published value of the minimum ignition energy for very free aluminium with
flaky particles hitherto traced, is that produced by US Bureau of Mines [4], namely 10 mJ. This value is of the same order as the 13 mJ obtained by Raftery [23] for a "6/am" aluminium powder by means of a spark generator very similar to that used by US Bureau of Mines. The figures quoted for the maximum rate of pressure rise for these two powders were > 1400 kp/cm 2 sec [4], and 1330 kp/cm 2 see [23], respectively. This is somewhat lower than the maximum value obtained with the aluminium powder in the present investigation, namely 2500 kp/cm 2 sec. Although'it appears that the aluminium powder tested in the present investigation is perhaps somewhat more explosible than the two others, it is felt that all three powders are basically of the same nature, and that the considerable discrepancy between published minimum ignition energies and the results shown in Fig. 12 is mainly due to the different spark generators used. VII. Discussion A. The Object of Minimum Ignition Energy Measurements
The question of why the "minimum ignition energy" of a dust cloud ought to be measured appears at least to have the following three answers: (a) The object is to assess the relative ease with which a dust cloud is ignited, compared with other dust clouds. For this purpose, it is not necessary to specify the ignition source in absolute terms. A relative measurement of the energy required for ignition is sufficient. It appears that this philosophy forms the basis of
62
ROLF K. ECKHOFF z 1.0
_0
OJ ;D
I W.I
~ 0.2 U-
0.0 10-4
10-3
10-2
I
10-1
ill1 I
I
I
I I 10
SPARK ENERGY~J
Fig. 12. Frequency of ignition as a function of spark energy for transient clouds of fine aluminium flakes in air, using sparks from the CMi generator as ignition sources. Ten trials per plotted point. the large amount of "minimum ignition energy" data published by, e.g., US Bureau of Mines [2, 3, 4, 5, 6]. (b) The object is to assess the hazard o f ignition of a given dust cloud by electrostatic spark discharges from electrically conducting surfaces in an industrial plant. In this case, it is essential to use the type of sparks which can occur in practice, i.e., the type of sparks generated by direct discharge of high voltage capacitors. The sig~aificant quantity will be the stored capacitor energy ½CV 2 required for ignition, rather than the net spark energy, since the essential question is whether the electrostatic energies likely to build up in a given plant are capable of producing sparks which can ignite the dust cloud in question. In this type of experiments it is essential, however, that appropriate attention is paid to the characteristics of the discharge circuit, in particular to the magnitude of the series resistance [15, 19,
281. (c) The object is to assess the "absolute" minimum spark energy required for ignition of a given dust cloud. In this case it is primarily essential to know the net energy delivered to the spark gap, and secondly, to deliver the energy at a rate which provides optimum conditions for ignition. This has been the philosophy adopted in the present investigation, and the results obtained demonstrate that dust clouds may be ignited by spark energies of the same order as the minimum ignition energy for 8.5% methane in air. This conclusion then leads to the question of whether it could in fact be possible to develop a theory for the ignition of dust clouds along the
same lines as those adopted in existing theories for gas ignition. Besides being of considerable fundamental interest, this type of data may, in addition, provide a measure of the electrostatic hazard in practice, since the sparks produced by the CMI generator are very similar to the capacitive sparks obtained with series resistances of the optimal order of 104 to l0 s ~2. Hence, since, in such discharges, only about 10% of less of ½CV 2 is delivered in the spark, an estimate of the minimum ½CV 2 for ignition may be obtained simply by multiplying by a factor of ten the "absolute" minimum ignition energies determined with the CMI spark generator. B. Why are "Short" Sparks Less Incendiary than "Long" Ones?
Attention is again focused on the problem of why short capacitive sparks appear to be much less able to ignite dust clouds than sparks with much longer discharge times. It is felt that two factors ought to be considered, namely, the loss of energy from the spark prior to the onset of ignition, and the question of the dust particles being removed from the spark gap by pressure waves from the spark itself. Typical induction times for ignition of dust clouds by sparks are not known, but assuming that they are of the order of say 0.1 msec to 1.0 msec, the energy dissipated from the spark within this order of time represents a loss. It appears that energy can be dissipated rapidly from the spark channel either as electromagnetic radiation, including thermal, or in the supersonic compression wave moving away from the spark
MINIMUM IGNITION ENERGIES FOR DUST CLOUDS channel during the first few/asec after the spark discharge. Valuable information concerning the radiation loss is given by Krauss and Krempl [25], who measured the temperature decay in a 1.2 J capacitive spark delivered to a spark gap within about 1/~sec. These workers were able to show that at a pressure of one atm, the temperature of the spark channel decreased from the initial value of 5 X 104 °K to 104 °K within the first 3 #see after the discharge, independently of the gas composition. By considering the spark channel as a blackbody of constant heat capacity and initial temperature 5 X 104 °K, simple calculations yield [20] a predicted temperature decay during the first 3/asec, which agrees very well indeed with that measured by Krauss and Krempl. This might indicate that in the case of very short sparks, the major part of the original energy is lost in electromagnetic radiation during the very first/asec after the discharge. Tiffs, however, implies the assumption that the actual radiation is not in itself capable of igniting the dust. When it comes to the question of the compressive pressure wave, Litchfield [9] has estimated the possible energy loss by such waves to less than 30%, perhaps only 5%, of the total spark energy. Similar figures were obtained by Martin [26] and Flowers [27]. It thus appears that the primary significance of the pressure wave in the present context is not a matter of energy loss. It is felt that the direct evidence given by Boyle and Llewellyn [15], and Line et al. [19] of the spark discharge itself being able to push the dust particlesaway from the spark region, is perhaps the most important piece of information available for elucidating the problem in hand. Clearly, if the fuel is effectively separated from the ignition source, ignition is impossible. At present, however, sufficient evidence does not seem to be available for explaining in detail the mechanism of this separation process, although the supersonic compression wave emitted from the spark appears to represent the most sensible element to look at for a start. It appears that essential information can be collected by investigating experimentally the various phenomena taking place at the interface,
63
between the spark and the dust cloud, during the first critical period after the spark discharge. It is the intention of the author to extend the present work in this direction. The author is greatly indebted to Mr. B. Alvestad, who developed the spark generator, to Mr. K. Fuhre, who has carried out the ignition experiments, and to Mr. R. Franck-Petersen and Mr. S. Svendsen, for their invaluable help with the calibration o f the sparks. The author further wishes to express his gratitude to The Royal Norwegian Council for Scien tific and Industrial Research for economic support, making this investigation possible. Further thanks are due to Dyno Industrier A/S, and to a group o f Norwegian grain and feed importers and millers for their invaluable support. Finally the author wishes to thank Dr. techn. 3". A. Andersen, Director of CMI, Dept. o f Applied Physics, for his stimulating interest in the worl~ References 1. Lewis, B., and yon Elbe, G., Combustion, Flames and Explosions of Gases, Academic Press, 1961, 2nd ed. 2. Jacobson, M., Cooper, A. R., and Ball, F. J., Explosibility of Agricultural Dusts, US Bureau of Mines, 1961, Rep. Inv. 5753. 3. Jacobson, M., Nagy, J., and Cooper, A. R., Explosibility of Dusts Used in the Plastic Industry, US Bureau of Mines, 1962, Rep. Inv. 5971. 4. Jacobson, M., Cooper A. R., and Nagy, J.,Explosibility of Metal Powders, US Bureau of Mines, 1964, Rep. Inv. 6516. 5. NaSy,J., Dorsett, H. G., and Cooper, A. R., Explosibility of Carbonaceous Dusts, US Bureau of Mines, 1965, Rep. Inv. 6597. 6. Dorsett¢tL G., and Nagy, J,, Dust Explosibility of Chemicals, Drugs, Dyes and Pesticides, US Bureau of Mines, 1968, Rep. Inv. 7132. 7. Litchfield, E. L., US Bureau of Mines, 1960, Rep. Inv. 5671. 8. Lintin, D. R., and Wooding, E. R., Brit. J. Appl. Phys.
10, 159 (1959). 9. Litehfield, E. L.,Combt. Flame, 5, 235 (1961). 10. Gordon, R. L., West, L. C., and Widginton, D. W.,
Inst. Electr. Eng., Paper No. 3914 M, April, 1962. 11. Gordon, R. L., Lord, H., and Widginton, D. W., Nature 191, 1085 (1961). 12. Berz, L, Combt. Flame 3, 131 (1959). 13. Berz, l.,IEEConf. Rep. Series, 1962, No. 3, p. 5. 14. Priede, T.,Initiation of Explosion in Gases, Ph.D. thesis, Universityof London, 1958. 15. Boyle, A. R.,andLlewellyn, F. J., Trans. Chem. Ind. 69, 173 (1950).
64 16. Riddlestone, H. G., ERA Techn. Rep. 1954, G/T 293. 17. Moore, P. W. J., Sumner, J. F., and Wyatt, R. M. H., ERDE Rep., 1956, 4/R/56. 18. Liddiard, T. P., Report from Explosion Research Dept., US Naval Ordnance Laboratory, White Oak, Maryland, USA. 19. Line, L. E., Gilmer, T. E., and Rhodes, H. A., J. Phys. Chem. 63, 290 (1959). 20. Eckhoff, R. K., The Energy Required for the Initiation o f Explosions in Dust Clouds by Electric Sparks, M. Phil. Thesis, University of London, 1970. 21. Alvestad, B., To be published. 22. Dorsett, H. G., Jacobson, M., Nagy, J., and Williams, R. P., Laboratory Equipment and Test Procedures for Evaluating Explosibility o f Dusts, US Bureau of Mines, 1960, Rep. Inv. 5424.
ROLF K. ECKHOFF 23. Raftery, M. M., Explosibility Tests for Industrial Dusts, Fire Research Technical Paper No. 21, Her Majesty's Stationery Office, London, 1968. 24. Es$1eston, L. A., and Pyor, A. J., Fire Technology 3, 77 (1967). 25. Krauss, L., and Krempi, H., Z. angew. Physik 16, 243 (1963). 26. Martin, E. A., Final Report, ORD Project No. TB3001 (317), University of Michigan, July, 1956. 27. Flowers, J.W.,Phys. Rev. 64, 225 (1943). 28. Palmer, K. N., Dust Explosions and Fires Chapman & Hall, London, 1973, p. 59-60, where reference is made to work by Smelkov, Festisov and Verevkin in USSR.
Received June 6, 1974
53
Towards Absolute Minimum Ignition Energies For Dust Clouds? ROLF K. ECKHOFF Chr. Michelsen Institute, Department of Applied Physics, N-5000 Bergen, Norway
A review of the literature related to spark ignition of dust clouds revealed that contrary to what is the case of gases, the spark energy required for igniting a given dust cloud is markedly dependent, even by orders of magnitude, upon the characteristics of the spark discharge, in particular the discharge time. By using a specially designed spark generator producing sparks of optimal discharge characteristics, it was found that airborne clouds of various dusts could be ignited by considerably smaller spark energies than hitherto published minimum ignition energies for similar dust clouds. Thus, with these optimal sparks the minimum ignition energy for lycopodium was found to be about 6 mJ, for wheat grain dust containing a considerable fraction of coarse material, about 12 m J, for sulphur less than 0.3 mJ, and for a fine quality flake aluminium somewhere between 0.3 and 0.9 mJ. By keeping in mind that this typ2eof optimal sparks is associated with capacitive discharge circuits in which more than 90% of ½CV is lost in the external circuit resistance, and therefore only less than 10% delivered in the spark, an approximate estimate of the electrostatic hazard may be obtained by multiplying by a factor of ten the "absolute" minimum ignition energies found.
I. Introduction The concept of minimum ignition energy, which has been discussed by Lewis and yon Elbe [I ], implies that if thermal energy is instantaneously delivered within a small volume o f an inflammable gas mixture, the minimum energy required to cause ignition is constant for a definite gas mixture and thermodynamic state. As will be shown in the following, the concept of minimum ignition energy for gas mixtures is resting on convincing experimental evidence obtained from ignition o f such mixtures by means of electric sparks. As will also be shown in the following, similar evidence appears to be lacking for the ignition of explosible dust clouds by means o f electric sparks. Nevertheless, the concept of minimum ignition energy has been adopted for the characterization of such systems, and considerable amounts o f data have been published [2, 3, 4, 5, 6].
H. Evidence Related to Spark Ignition of Gases In order to demonstrate the type o f evidence available, investigations on the ignition o f 8.5%
methane in air has been chosen as a typical example. The minimum stored capacitor energy required for the ignition of this mixture, at normal temperature and pressure, was determined, independently, by Lewis and yon Elbe [ 1 ] and Litchfield [7], by means of capacitive spark techniques. In both investigations the minimum ignition energy was found to be 0.28 mJ. Since the order o f spark discharge times in both investigations was probably 1 #sec or less, and since it has been shown by Lintin and Wooding [8] and Litchfield [9], that the time between the initiation o f the discharge and the inflammation in an 8.5% methane/air mixture at atmospheric pressure probably is o f the order o f 0.1 msec, the requirement of approximately instantaneous energy delivery was satisfactorily fulfilled. At SMRE [10], ignition experiments with the same gas mixture were carried out by means o f the standard inductive break flash apparatus. Initially a rather large minimum ignition energy was found, namely 2.3 mJ, which is about eight times the value obtained by capacitive sparks.
Copyright © 1975 by the Combustion Institute Published by American Elsevier Publishing Company, Inc.
54 Later, however, Gordon et al. [11] reported that they had been able to reduce this figure considerably by increasing the separation speed of the electrodes in the break flash technique, and minimum ignition energies comparable with those obtained by capacitive sparks were recorded. This reduction in minimum ignition energy was explained in terms of reduction of the heat loss to the electrodes. Similar experiments were carried out by Berz [ 12, 13], who, instead of increasing the separation speed of the electrodes, reduced the electrode size. Finally, by breaking wires of 0.025 mm diam Berz was able to ignite the 8.5% methane/air mixture with inductive sparks having an energy, in terms of ~.Li 2 , where L is the inductance and i, the current, of about 0.30 m J, which is practically identical to the value of 0.28 mJ obtained with capacitive sparks. Again, this result can be explained in terms of heat loss to electrodes, since, at a given electrode separation speed, and for a given electrode material, the heat flow into the electrodes must obviously decrease with decreasing electrode dimensions. It can, therefore, be concluded that the minimum spark energy needed for the initiation of explosions in an 8.5% methane/air mixture, at normal temperature and pressure, appears to be largely independent of the characteristics of the spark, which in turn supports the assumption that the minimum ignition energy for an explosible gas mixture is, in fact, a sensible concept. The evidence gained from igniting methane/air mixtures is slightly opposed by the findings of Priede [14], who studied the electric spark ignition of hydrogen/air mixtures. Priede was able to demonstrate that if a series resistance of the order of 5 X 104 I2 was introduced in the capacitive discharge circuit, the minimum spark energy for ignition was reduced to about half the value found without the resistance in the circuit. This effect may be explained in terms of the resistance producing a prolongation of the spark discharge time, which in turn means a reduction of the energy losses from the spark due to thermal radiation, heat sink effect of electrodes, shock waves and expansion work prior to ignition.
ROLF K. ECKHOFF Although the evidence put forward by Priede to some extent contradicts the concept of minimum ignition energy, this concept, when applied to gases, is nevertheless, an acceptable approximation. The question to be considered in the following is then whether this is the case even with dust clouds. III. Earlier Investigations on Electric Spark Ignition of Dust Clouds A. The Work of Boyle and Liewellyn
The first set of data indicating that the application of the concept of minimum ignition energy to dust clouds requires considerable care was probably produced by Boyle and Llewellyn [ 15]. These workers were able to demonstrate that the minimum capacitor energy ½ C V 2, C being the capacitance and V the initial capacitor voltage, capable of igniting dust clouds of various powders, decreased quite considerably when a series resistance was introduced in the discharge circuit. Some of the results obtained by Boyle and Llewellyn for transient dust clouds, produced by dispersing a quantity of powder by means of a blast of air, are shown in Fig. 1. As can be seen, the minimum %CV 2 for ignition decreased by a factor of about ten both with granular aluminium and magnesium, when a series resistance of 104 to l0 s was added to the discharge circuit. Similar trends were also found by these workers for dust clouds of ferromanganese, zinc, silicon, and sulphur. Boyle and Llewellyn expressed their conclusions in terms of capacitor energy. It is, however, to be expected that an appreciable series resistance in the spark discharge circuit will, during discharge, absorb a significant fraction of the capacitor energy, so that the energy delivered in the spark gap will be significantly less than the theoretical value ½ C V 2 . In fact, this fraction has been determined experimentally by various workers. Some results for spark discharge in air at normal room temperature are shown in Fig. 2. The relationships established by Priede [14] and Moore et al. [17] for capacitors of 1000 pF and V-zCV2 values of 12.8 and 18.0 m J, respectively, are particularly interesting, since the most incendiary sparks produced by Boyle and Llewellyn were obtained
MINIMUM IGNITION ENERGIES FOR DUST CLOUDS ~q
cuit, reduced the minimum net spark energy for ignition to only one per cent, or perhaps even less, of the energy required with a small series resistance. The question now arises whether a plausible explanation of the rather dramatic influence of the size of the series resistance can be found. An observation made by Boyle and Llewellyn during their experiments with sulphur may throw some light on the problem. In their paper these workers wrote as follows:
I
~. 0.6i
z o p-
•
0t|
_o e.- 0o
,o
--. 0.: Z z
~r0.0 101
10z
i113
111~
t11s
t06
SERIES RESISTANCE IN OHMS
The spark produced by direct discharge of a condenser is a short, sharp, noisy emission of energy. If the capacity is small and the voltage high the spark appears to push the cloud away from it; ff a spark of the same energy is derived from a condenser of greater capacity, i.e.,at a low voltage, this effect is greatly reduced.
Fig. 1. Results from ignition o f dust clouds by capacitive sparks, using an additional series resistance in the discharge circuit. Data from Boyle and Liewellyn [ l 5 ].
with similar capacitances and ½ C F 2 values. Although there is some discrepancy between the data of Priede and those of Moore et al., it is clear that, with a series resistance in the range of 10 to 107 I2, the net spark energies were only of the order of 5% to 10% of the initial capacitor energy ½CF 2 . This, in turn, means that in the experiments of Boyle and Llewellyn, an inclusion of a series resistance of 104 to l0 s 12 in the discharge cir-
-n-
o0
55
B. Investigations with Lycopodium Being a naturally occurring dust with constant properties such as particle size, shape, and chemical composition, lycopodium is very suitable for dust explosion research, since a sensible comparison of data obtained with this material in different laboratories can be performed. Thus Line et al. [19] in their investigations used steady-state wall free and wall confined 1
0~,
J, 4.2 pF, 1141
i
u) 80 z
so
~
L
"
1,1
c,~
I
~
,Oo
i~ dL
12.5
o:
1114
\/
115mJ, 9pF, 1141
200
lo0
1111
lo 2
8__
lo 3
ioz.
tos
m5
lo?
#
SERIES RESISTANCE IN OHMS
Fig. 2. Results from determination ofl~rcentage of total capacitor energy dissipated in the spark for a range of capacitances, spark energies and additional series resistance in the discharge circuit.
56
ROLF K. ECKHOFF
and 2 in diam columns of lycopodium suspended in the oxiding gas. Some of the results obtained by these workers for 1 in columns in air are shown in Fig. 3. 1.0
111,,, I z 0
0.6 ,
~- O.4: z -- 0.2
~ ,°.~
NO ADOITIONAL RESISTANCE
111111
I
,t?!. ! I1!!1
~,' W~LL CONF NEO I I
I
l.}llllll (o)
III1[11 III ./I
IIII l lliilill
u~ 0.6
{b) LL
2o
10-3
,11 10"2
Ill 10-1
1
10
CAPACITOR ENERGY ~ J
Fig. 3. Frequency of ignition of a 1 in diam. stationary column of 80 mg per litre of lycopodium in air, as a function of capacitor energy. Effect of wall confinement and additional series resistance in the discharge circuit. Data from Line et al. [19]. As can be seen both for the wall free and the wall confined case the stored capacitor energies required for producing a given probability of ignition, decreased roughly by a factor of ten if a series resistance of l0 s I2 was included in the discharge circuit. It is interesting to note that both the order of magnitude of the decrease, as well as the order of magnitude of the series resistance giving this maximum decrease, agree with the corresponding figures found by Boyle and Llewellyn [15] for completely different powders and an entirely different experimental technique. Line et al. produced direct evidence, by means of high speed photography, supporting the suggestion that a short spark can in fact distrub the dust cloud significantly. Hence, in the case of a capacitor energy of 1.5 J and only a very small series resistance in the discharge circuit, the high speed photographs revealed that a dust free zone of about 4 em a was created around the spark channel within the first two msec after the initiation of the spark discharge. This means that the
dust particles and the spark were effectively separated, at this stage, so that ignition could not occur. It was also demonstrated by high speed photography that with a l0 s ~2 series resistance, or a considerable series inductance, the disturbance effect was practically absent. Some results produced by Eckhoff [20] may further elucidate the problem in hand. This worker studied the ignition of transient clouds of lycopodium in air, produced by dispersing a quantity of the dust in a one liter bomb, by a short blast of air. The ignition sources were capacitive sparks of discharge times of the order of a few psec. Some of Eckhoff's results are given in Figs. 4(a) and 4(b). Each frequency of ignition is based on at least one hundred single trials. By comparing the data in figs. 4(a) and 4(b) with the data for the wall free case in fig. 3(a), a remarkable coincidence is observed. Both Line et al. and Eckhoff found that the probability of ignition remained in the range 0.3 to 0.7 over the considerable range of ½CF 2 values from 0.3 to 8 J. A possible explanation for this effect might be that the increase in the incendivity to be expected from the increase in the spark energy is nearly
1.0 0°8
H/I | IIIIIIII
0.6 z
Io)
OJ ,
i-
z i
o.2
o.o I I IIIIlll I I[lllll[
°i
I I IILII
14.
10-3
Ib)
10"2
tO-I
1
tO
CAPACITOR ENERGYt J
Fig. 4. Frequency of ignition of transient clouds of lycopodium in air in a one litre bomb, as a function of capacitor energy. Series resistance in discharge circuit 0.01 ,.(2. Data from Eckhoff [20].
MINIMUM IGNITION ENERGIES FOR DUST CLOUDS compensated by the increased ability of the spark to disturb the dust cloud. In fact, as indicated by the decrease of probability of ignition with increasing spark energy shown in Fig. 4(b), it might even be possible that the increased incendivity is more than compensated by the increased disturbance. Eckhoff [20] has considered the problem of combining the two opposing mechanisms in a unified picture. IV. Optimum Spark Discharge Time for Ignition By considering the work of Line et al. [19] and Boyle and Llewellyn [15], in combination with Fig. 2, one would anticipate that if a spark discharge circuit was designed to supply well defined sparks with discharge times of the order of the sparks produced by these workers with a series resistance of l0 s I2, very low minimum ignition energies would be found. In fact, considering Fig. 1 and Fig. 3(b), and taking into account that only 10% or less of ~ C V 2 was delivered in the spark, one would expect that ignition would be possible with net spark energies of the order of 1 mJ and perhaps even less. The current in a markedly overdamped R - C - L series discharge circuit, after the initial rapid rise to its maximum value, is given by the approximate equation: Eo
i .... R
e
-t/RC
,
(I)
where E o is the initial capacitor voltage, R the circuit resistance, C the capacitance, and t is the time. By defining the discharge time as the time required for the current to decrease to 1% of the initial value (t=0), it is found that the discharge time equals t=4.6 • R - C..
57
series resistance of 5 × 104 [2. The lowest ½CV 2 for ignition in the whole series of experiments was found with the 0.00055 #F capacitance, and in this case the discharge time, as defined in Eq. (2), was about 0.13 msec and the net spark energy probably of the order of 1 mJ. For a magnesium dust it was found t h a t R = 15 × 10 a [2 and C = 0.01/IF produced the lowest ½CV 2 for ignition, and in this case the discharge time was about 0.7 msec. and the net spark energy probably of the order of 2-3 mJ. For a granular aluminium, the optimum occurred a t R = 75 X 10 a I2 and 0.01 /aF, i.e., with a discharge time of 3.5 msec. and a net spark energy probably of the order of 5 mJ. It is thus seen that the optimum discharge time for ignition varies within the approximate range from a tenth of a msec to a few msec. A systematic increase in the optimum discharge time with increasing spark energy is indicated.
V. The CMI Spark Generator The various features of the generator will be discussed in detail elsewhere [21 ]. In the present context, it is sufficient to focus on the basic layout of the discharge circuit and on the main characteristics of the sparks produced by it. The discharge circuit is illustrated in Fig. 5, in which C is the discharge capacitor having an initial voltage of V. By having a range of capacitors from 40/~F down to 0.005/IF and variable voltages V from 1000V and downwards, a continuous range of ½CV 2 values covering about four decades is available.
V
(2)
By inserting the optimum values found by Line et al., namely C = 0.0008/~F and r = l0 s I2, one finds a discharge time of 0.37 msec. Boyle and Llewellyn give some interesting results for sulphur dusts, showing that for a range of capacitances from 0.01 to 0.00055/aF, the lowest ½CV 2 for ignition at each capacitance was obtained with a
v
w
Fig. 5. Diagram of the CMI circuit for generation of optimal synchronized capacitive sparks for ignition of dust clouds.
58
ROLF K. ECKHOFF
Initiation of the spark discharge at the desired moment, which is essential if synchronization of spark with formation of a transient dust cloud is required, is accomplished by means of a trigger circuit in which a capacitor C T, a switch S, and the primary coil of a transformer T constitute the essential elements. By closing the switch, a high voltage pulse of approximately 15 kV peak value, is induced in the secondary coil of the transformer, causing breakdown of the spark gap G, and thereby, discharge of the main capacitor C. In order to produce sparks similar to those from overdamped circuits (high series resistance) a diode D was applied across the discharge capacitor. Figure 6 shows two typical pairs of oscilloscope traces of spark current and voltage across spark gap for sparks produced by the CMI generator. Figure 6(a) refers to a 3.3 J spark, whereas the spark energy was only 2.4 mJ in the case shown in Fig. 6(b). It is seen that the discharge times vary
considerably, being about 1.7 msec. for the 3.3 J spark and only about 60/~sec. for the 2.4 mJ spark. The net spark energies produced by the generator were determined by recording pairs of oscilloscope traces of the type shown in Fig. 6. In Fig. 7, the resulting calibration curve, relating the net spark energy to the corresponding ½CI/2, is given. As ½CF 2 decreases, the energy supplied by the triggering circuit becomes significant, and in the range of very low energies, the net spark energy is larger than the ½CI/2 of the main discharge capacitors. In fact, the lowest net spark energy used in the tests, 0.3 mJ, was obtained by means of the triggering circuit alone. All the energy values quoted in the following account of ignition experiments are the net spark energies obtained from the average calibration curve in Fig. 7. It is seen that in the lower range, there is a noticeable scatter in the measured energies, which indicates the degree of accuracy
120
0.6
100
0.5 f ' ~ k l l
80
/---,
60
/
04 0.3
SPARK CURRENT
SPARK CUI~RENT
~0.2
~-
2Q
oo S(]
70 6O
0.5
I
1.0
I
< o.1
1.5
o.o o
2.0 4oo
I
250
o> , I (
200
3O
150
20
IO0 50
05
60
80
I
VOLTAGE ACROSS SPARK GAP
300
I
,~ so
0.0
~
60
I
350
VOLTAGE ACROSS SPARK GAP
20
1.0 TIME. MILUSEC
(a)
1.5
2-0
0
f 2O
t.0
60
80
TIME. MICROSEC
(.b)
Fig. 6. Two examples of simultaneous oscilloscopetraces of spark current and voltage across spark gap for discharges produced by the CMI generator. (a): Spark energy 3.3 J; (b): Spark energy 2.4 mJ.
59
MINIMUM IGNITION ENERGIES FOR DUST CLOUDS associated with the energy values quoted. Figure 8 shows a plot of the discharge times as a function
of the net spark energies for sparks produced by the CMI generator.
I0
/ /J
,/
>.
W z
cI 1/2 V2
AC,UAL SPARK
r,,
ENERGIES
I,--
lO-3g
/ /
'//// 10-4
10-4
10 -3
10 - 2
10-!
I
10
100
CAPACITOR ENERGY. J
Fig. 7. Measured spark energies as a function of initial capacitor energies for discharges produced by the CMI generator.
10"z
CJ UJ u) Z 0
f
-3
0
.2.
X
U~
I
,,.M
f
10.4
/ 10-s
I0 -~
10-3
10":
10 -t
1
SPARK ENERGY,JOULES
Fig. 8. Duration of spark discharge as a function of measured spark energy for sparks produced by the CM1 generator.
10
60
ROLF K. ECKHOFF
VI. Ignition of Dust Clouds with Sparks from the CMI Generator A. Experimental Procedure A transient dust cloud was generated by dispersing a given quantity of powder, by means of a given blast of air in a standard Hartmann apparatus [22, 23], modified by inserting a conical ring between the dispersion cup and the vertical perspex explosion tube in order to guide the dust cloud towards the spark gap [24]. For each type of powder an optimum combination for ignition of amount of powder, dispersing air pressure (fixed 500 cm 3 reservoir), and time lag between the release of the dispersing air and the spark discharge, was determined in a series of initial tests. B. Experimental Results. During the period of time in which the spark generator has been in operation, a wide range of dusts have been tested. In the present paper four characteristic sets of data will be presented. Figure 9 shows the results obtained with lycopodium having its natural moisture content of about 3.5%. As can be seen, ignition occurred even with a spark energy as low as 7 mJ. This
i.0
i.o i
~_ 0.e, ....... z
O.8
~ 0.4
11 OI
~ 0.2 u_ 0.0 10-3
extracted from Fig. 3, these workers were able to produce ignition, with a series resistance of l0 s I2 in the discharge circuit, by using ½ C V z values down to the range of 20 to 50 mJ, depending on whether a wall confined or a wall free dust column was used. By assuming, as indicated by Fig. 2, that less than 10% of ½ C V 2 was delivered in the spark, it follows that the minimum ignition energies found by Line et al. were of the order of a few mJ. This conclusion is in reasonable agreement with the results produced in the present investigation and given in Fig. 9. Attention is now drawn to Fig. 10, which gives the results from tests with a dust from Australian wheat collected from the top section of a bucket elevator in a grain silo plant during operation. Prior to the tests the dust was dried, but otherwise it was tested in its original state. The particle size distribution was quite wide, 50%, by weight, being coarser than 55/am. As can be seen from Figure 10, it was possible to ignite this dust with spark energies as low as 20 m J, whereas the minimum ignition energies quoted in the literature [2] for wheat flour and dusts from untreated wheat and wheat straw are in the range of 50 to 60 mJ. Although the percentage difference between the two values is less than that for lyco-
IHt titH q .
If 2
i~ I SPARK ENERGY=J
.
.
.
~ 0./., 0.2
.
i
10
10-3
If ; I
J
1(:
./1
I
i'i iili IIIH wi 10-i
I
I0
SPARK ENERGY, J
Fig. 9. Frequency of ignition as a function of spark energy for transient clouds of lycopodium in air, using sparks from the CMI generator as ignition sources. Ten trials per plotted point.
Fig. 10. Frequency of ignition as a function of spark energy for transient clouds of Australian wheat grain dust in air, using sparks from the CMI generator as ignition sources. Ten trials per plotted point.
is considerably less, by a factor of six, than the minimum ignition energy of 40 mJ quoted in the literature [2] for this particular dust. It is, however, of considerable interest to compare this rather low value with the results of Line et al. [19] discussed earlier in this paper. As can be
podium, it, nevertheless, represents a factor of three, in spite of the fact that the dust tested in the present work contained a considerable fraction of coarse material. The next series of results to be discussed is shown in Fig. 11. In this case the powder tested
61
MINIMUM IGNITION ENERGIES FOR DUST CLOUDS z
1,0
UJ
~" 0.2
14.
o.o I0 -t,
-7
AA
IIIll i
I IllJ
fiIllll l t
illll 1111 10-3
tfil
10-2 SPARK
A A
illl fll !11!1s!L lIJl Ii II I I I;ii till
[ I:Jl J,I
10-1
1
10
ENERGY t J
Fig. 11. Frequency of ignition as a function of spark energy for transient clouds of sulphur in air, using sparks from the CMI generator as ignition sources. Ten trials per plotted point.
was finely ground sulphur. As can be seen, ignition occurred even with spark energies as low as 0.3 mJ. This value is, in fact, nearly two orders of magnitude less than the value of 15 mJ quoted in the literature [6] as the minimum ignition energy for clouds of a very fine ("4/am") sulphur powder. This is quite a dramatic result, since it implies that the energy required for ignition in this case probably is even less than the minimum ignition energy for 8.5% methane in air. It is, in this context, of substantial interest to consider the results obtained by Boyle and Llewellyn [15] for the minimum ½ C V ~ for the ignition of sulphur. These workers found that by using the optimum combihation of capacitance and series resistance for ignition, namely 550 pF and 5 X 104 I2, the minimum ~ C V 2 for ignition was as low as 9 mJ. By combining this result with the evidence provided by Fig. 2, it can be concluded that the minimum net spark energy for ignition was, in this case, probably less than 1 mJ. The result produced by the CMI generator is therefore in agreement with the findings of Boyle and Llewellyn. Finally, attention is focused on some results shown in Fig. 12 from tests with a very f'me aluminium powder. Scanning electron microscopy revealed that the powder was composed of flaky particles with typical diameter 10/am, and thickness of the order 0.1/am. As can be observed in Fig. 12, clouds of this powder were readily ignited with spark energies of the order o f l mJ. The lowest published value of the minimum ignition energy for very free aluminium with
flaky particles hitherto traced, is that produced by US Bureau of Mines [4], namely 10 mJ. This value is of the same order as the 13 mJ obtained by Raftery [23] for a "6/am" aluminium powder by means of a spark generator very similar to that used by US Bureau of Mines. The figures quoted for the maximum rate of pressure rise for these two powders were > 1400 kp/cm 2 sec [4], and 1330 kp/cm 2 see [23], respectively. This is somewhat lower than the maximum value obtained with the aluminium powder in the present investigation, namely 2500 kp/cm 2 sec. Although'it appears that the aluminium powder tested in the present investigation is perhaps somewhat more explosible than the two others, it is felt that all three powders are basically of the same nature, and that the considerable discrepancy between published minimum ignition energies and the results shown in Fig. 12 is mainly due to the different spark generators used. VII. Discussion A. The Object of Minimum Ignition Energy Measurements
The question of why the "minimum ignition energy" of a dust cloud ought to be measured appears at least to have the following three answers: (a) The object is to assess the relative ease with which a dust cloud is ignited, compared with other dust clouds. For this purpose, it is not necessary to specify the ignition source in absolute terms. A relative measurement of the energy required for ignition is sufficient. It appears that this philosophy forms the basis of
62
ROLF K. ECKHOFF z 1.0
_0
OJ ;D
I W.I
~ 0.2 U-
0.0 10-4
10-3
10-2
I
10-1
ill1 I
I
I
I I 10
SPARK ENERGY~J
Fig. 12. Frequency of ignition as a function of spark energy for transient clouds of fine aluminium flakes in air, using sparks from the CMi generator as ignition sources. Ten trials per plotted point. the large amount of "minimum ignition energy" data published by, e.g., US Bureau of Mines [2, 3, 4, 5, 6]. (b) The object is to assess the hazard o f ignition of a given dust cloud by electrostatic spark discharges from electrically conducting surfaces in an industrial plant. In this case, it is essential to use the type of sparks which can occur in practice, i.e., the type of sparks generated by direct discharge of high voltage capacitors. The sig~aificant quantity will be the stored capacitor energy ½CV 2 required for ignition, rather than the net spark energy, since the essential question is whether the electrostatic energies likely to build up in a given plant are capable of producing sparks which can ignite the dust cloud in question. In this type of experiments it is essential, however, that appropriate attention is paid to the characteristics of the discharge circuit, in particular to the magnitude of the series resistance [15, 19,
281. (c) The object is to assess the "absolute" minimum spark energy required for ignition of a given dust cloud. In this case it is primarily essential to know the net energy delivered to the spark gap, and secondly, to deliver the energy at a rate which provides optimum conditions for ignition. This has been the philosophy adopted in the present investigation, and the results obtained demonstrate that dust clouds may be ignited by spark energies of the same order as the minimum ignition energy for 8.5% methane in air. This conclusion then leads to the question of whether it could in fact be possible to develop a theory for the ignition of dust clouds along the
same lines as those adopted in existing theories for gas ignition. Besides being of considerable fundamental interest, this type of data may, in addition, provide a measure of the electrostatic hazard in practice, since the sparks produced by the CMI generator are very similar to the capacitive sparks obtained with series resistances of the optimal order of 104 to l0 s ~2. Hence, since, in such discharges, only about 10% of less of ½CV 2 is delivered in the spark, an estimate of the minimum ½CV 2 for ignition may be obtained simply by multiplying by a factor of ten the "absolute" minimum ignition energies determined with the CMI spark generator. B. Why are "Short" Sparks Less Incendiary than "Long" Ones?
Attention is again focused on the problem of why short capacitive sparks appear to be much less able to ignite dust clouds than sparks with much longer discharge times. It is felt that two factors ought to be considered, namely, the loss of energy from the spark prior to the onset of ignition, and the question of the dust particles being removed from the spark gap by pressure waves from the spark itself. Typical induction times for ignition of dust clouds by sparks are not known, but assuming that they are of the order of say 0.1 msec to 1.0 msec, the energy dissipated from the spark within this order of time represents a loss. It appears that energy can be dissipated rapidly from the spark channel either as electromagnetic radiation, including thermal, or in the supersonic compression wave moving away from the spark
MINIMUM IGNITION ENERGIES FOR DUST CLOUDS channel during the first few/asec after the spark discharge. Valuable information concerning the radiation loss is given by Krauss and Krempl [25], who measured the temperature decay in a 1.2 J capacitive spark delivered to a spark gap within about 1/~sec. These workers were able to show that at a pressure of one atm, the temperature of the spark channel decreased from the initial value of 5 X 104 °K to 104 °K within the first 3 #see after the discharge, independently of the gas composition. By considering the spark channel as a blackbody of constant heat capacity and initial temperature 5 X 104 °K, simple calculations yield [20] a predicted temperature decay during the first 3/asec, which agrees very well indeed with that measured by Krauss and Krempl. This might indicate that in the case of very short sparks, the major part of the original energy is lost in electromagnetic radiation during the very first/asec after the discharge. Tiffs, however, implies the assumption that the actual radiation is not in itself capable of igniting the dust. When it comes to the question of the compressive pressure wave, Litchfield [9] has estimated the possible energy loss by such waves to less than 30%, perhaps only 5%, of the total spark energy. Similar figures were obtained by Martin [26] and Flowers [27]. It thus appears that the primary significance of the pressure wave in the present context is not a matter of energy loss. It is felt that the direct evidence given by Boyle and Llewellyn [15], and Line et al. [19] of the spark discharge itself being able to push the dust particlesaway from the spark region, is perhaps the most important piece of information available for elucidating the problem in hand. Clearly, if the fuel is effectively separated from the ignition source, ignition is impossible. At present, however, sufficient evidence does not seem to be available for explaining in detail the mechanism of this separation process, although the supersonic compression wave emitted from the spark appears to represent the most sensible element to look at for a start. It appears that essential information can be collected by investigating experimentally the various phenomena taking place at the interface,
63
between the spark and the dust cloud, during the first critical period after the spark discharge. It is the intention of the author to extend the present work in this direction. The author is greatly indebted to Mr. B. Alvestad, who developed the spark generator, to Mr. K. Fuhre, who has carried out the ignition experiments, and to Mr. R. Franck-Petersen and Mr. S. Svendsen, for their invaluable help with the calibration o f the sparks. The author further wishes to express his gratitude to The Royal Norwegian Council for Scien tific and Industrial Research for economic support, making this investigation possible. Further thanks are due to Dyno Industrier A/S, and to a group o f Norwegian grain and feed importers and millers for their invaluable support. Finally the author wishes to thank Dr. techn. 3". A. Andersen, Director of CMI, Dept. o f Applied Physics, for his stimulating interest in the worl~ References 1. Lewis, B., and yon Elbe, G., Combustion, Flames and Explosions of Gases, Academic Press, 1961, 2nd ed. 2. Jacobson, M., Cooper, A. R., and Ball, F. J., Explosibility of Agricultural Dusts, US Bureau of Mines, 1961, Rep. Inv. 5753. 3. Jacobson, M., Nagy, J., and Cooper, A. R., Explosibility of Dusts Used in the Plastic Industry, US Bureau of Mines, 1962, Rep. Inv. 5971. 4. Jacobson, M., Cooper A. R., and Nagy, J.,Explosibility of Metal Powders, US Bureau of Mines, 1964, Rep. Inv. 6516. 5. NaSy,J., Dorsett, H. G., and Cooper, A. R., Explosibility of Carbonaceous Dusts, US Bureau of Mines, 1965, Rep. Inv. 6597. 6. Dorsett¢tL G., and Nagy, J,, Dust Explosibility of Chemicals, Drugs, Dyes and Pesticides, US Bureau of Mines, 1968, Rep. Inv. 7132. 7. Litchfield, E. L., US Bureau of Mines, 1960, Rep. Inv. 5671. 8. Lintin, D. R., and Wooding, E. R., Brit. J. Appl. Phys.
10, 159 (1959). 9. Litehfield, E. L.,Combt. Flame, 5, 235 (1961). 10. Gordon, R. L., West, L. C., and Widginton, D. W.,
Inst. Electr. Eng., Paper No. 3914 M, April, 1962. 11. Gordon, R. L., Lord, H., and Widginton, D. W., Nature 191, 1085 (1961). 12. Berz, L, Combt. Flame 3, 131 (1959). 13. Berz, l.,IEEConf. Rep. Series, 1962, No. 3, p. 5. 14. Priede, T.,Initiation of Explosion in Gases, Ph.D. thesis, Universityof London, 1958. 15. Boyle, A. R.,andLlewellyn, F. J., Trans. Chem. Ind. 69, 173 (1950).
64 16. Riddlestone, H. G., ERA Techn. Rep. 1954, G/T 293. 17. Moore, P. W. J., Sumner, J. F., and Wyatt, R. M. H., ERDE Rep., 1956, 4/R/56. 18. Liddiard, T. P., Report from Explosion Research Dept., US Naval Ordnance Laboratory, White Oak, Maryland, USA. 19. Line, L. E., Gilmer, T. E., and Rhodes, H. A., J. Phys. Chem. 63, 290 (1959). 20. Eckhoff, R. K., The Energy Required for the Initiation o f Explosions in Dust Clouds by Electric Sparks, M. Phil. Thesis, University of London, 1970. 21. Alvestad, B., To be published. 22. Dorsett, H. G., Jacobson, M., Nagy, J., and Williams, R. P., Laboratory Equipment and Test Procedures for Evaluating Explosibility o f Dusts, US Bureau of Mines, 1960, Rep. Inv. 5424.
ROLF K. ECKHOFF 23. Raftery, M. M., Explosibility Tests for Industrial Dusts, Fire Research Technical Paper No. 21, Her Majesty's Stationery Office, London, 1968. 24. Es$1eston, L. A., and Pyor, A. J., Fire Technology 3, 77 (1967). 25. Krauss, L., and Krempi, H., Z. angew. Physik 16, 243 (1963). 26. Martin, E. A., Final Report, ORD Project No. TB3001 (317), University of Michigan, July, 1956. 27. Flowers, J.W.,Phys. Rev. 64, 225 (1943). 28. Palmer, K. N., Dust Explosions and Fires Chapman & Hall, London, 1973, p. 59-60, where reference is made to work by Smelkov, Festisov and Verevkin in USSR.
Received June 6, 1974