You are on page 1of 4

Simulating Flammability under Process

Conditions
Dan O'Shaughnessey and Bruce Powers
The Dow Chemical Co., 1897C Building, Midland, MI 48667

Establishing safe operating parameters requires the determination of


flammability characteristics under a myriad of process conditions. In order to
meet these needs, a combined approach of literature search, calculational
modeling, and experimental simulation is often utilized. A few specific projects
will be reviewed as example methods for evaluating flammability properties
under challenging en vironments. Flammability diagrams and system
configurationfigures are presented for further review while the interesting
discoveries of each project are covered for interpretation and application to
process conditions.

GENERAL FLAMMABILITY SIMULATION Example Projects Conclusions & Process Applications

Some of the factors to be considered when performing lab- Flammability testing was needed to measure maximum pres-
oratory tests to simulate flammability under process conditions sure (Pm) and maximum rates of pressure rise (dP/dt)m for
include: ignition energy, wall effect, fuel phase, oxidant, reac- relief system design on H2-N2-02mixtures at ambient pressure
tivity, composition region, and products of combustion. Ig- & temperature. Flammability limits and detonation ranges of
nition energy should be sufficiently above the minimum to these mixtures are shown in Figure 1together with experimental
provide reliable flame kernel initiation and is usually a fuse results for Pm and (dP/dt)m.
-
wire of 1 joule energy in the 35 liter system. Quenching wall
effects can be avoided by test vessels of diameter >2" for
For runs with (dP/dt)m> loo0 psilsec, an oscillation fre-
quency of - 1.1 KHz was observed. For runs with (dP/
most hydrocarbons and. larger vessels need to be used when dt)m >200,000 psilsec, three oscillation frequencies were seen
the fuel contains flame inhibiting halogens. The phase of the
fuel becomes less important as the particle size of a solid or
liquid drops to less than 100 microns. Fine mists and dusts,
therefore, have an equivalent concentration range of flam-
mability as vapors when measured in units of fuel mass per
unit air volume. The type of oxidant becomes important when
reviewing thermodynamic calculations of flame temperature
for evaluating flammable limits as the LFL is -600 degrees
lower in chlorine than in air or oxygen. Reactivity of the com-
ponents prior to ignition becomes more critical as the tem-
perature and/or pressure is elevated where reaction rates, slow
oxidation, or nearness to the autoignition temperature can
complicate the mixture stability and hence flammability results.
The composition region should be viewed on a ternary flam-
mability diagram of fuels, inerts, and oxidants similar to Figure
1 or 7 to anticipate the type of testing to be performed, max-
imum Pressures, flame speeds, and products of combustion
before experimentation is attempted.

22 Januarv. 1995
~ -
* I - Process Safety Progress (Vol. 14, No. 1)
Run 3D1 (ALL). 20:40:40.02:H2:N2 MODIFIED 35 LITER SYSTEM
Raw data: 5 rnicroseconds/point
700 ChaprnanJougucl
4, VENT FOR FAN GENERATED TURBULENCE
cnn
”””

500 .-
D~tonatlonspaka:
(on ssaie)
Frequencies
1 0 3 6/46 kHr
I 1
400 --
3 w --
I MAGNEDRIYE

2w ~-
100 --
0 --
-100 -~
-200 ~~ Prn .UllS -1% :1
p.

4 5 6 7 8
TIME (after ignition). Milliseconds

FIGURE 2(a). Hydrogen detonations.


VACUUM CHARCOAL 35 LITER DIRECT FEED STATIONS
PUMP BED VESSEL FEED

FIGURE 3. Turbulent flammability.


PV3SS”W Run 3Dlfl KHz filter). 20:40:40.02:HZ:N2
psia 20 microsecondslpoinl,Smoothed by 40

200 --
Many of the runs in the detonation range did not exhibit
OSClllatlOn
Frequency strong detonation phenomenon. Four factors could force tran-
1 2 kHz
sition from deflagration to detonation. Venting, a higher ig-
nition energy, initial turbulence, or greater run-up distance
(larger vessels) could each initiate the transition to detonation.
In case of transition, the PM and (dP/dt)m values would be
much higher and extreme caution should be used when using
this data for scale up.
Simulating the effects of turbulence on flammability at proc-
0 4 I
ess conditions was performed in the modified 35 liter flam-
4 5 6 7 8 mability system as shown in Figure 3. The six flat blades were
TIME (after ignition). MIIIIsecOnds 2 “ wide and the overall fan diameter was 12”. The ‘fan Reyn-
olds number’ NRe,tan, is calculated as follows:
FIGURE 2(b). Hydrogen detonations.
= ND2/(v60), (N = fan speed, rpm, D = fan dia., ft,
NRe,fan
v = kinematic viscosity, ft2/sec)

-
simultaneously at 1.1, 3.6, and 46 KHz. These three oscil- Turbulent burning velocities and turbulent flame propaga-
lation frequencies represent three separate phenomenon. The tion was measured at the University of Leeds by Bradley et
1.1 KHz oscillations are acoustic vibrations due to pressure al. in equipment that utilized high speed schlieren photography
wave feedback on the flame front from the vessel walls. A 3.6 and laser doppler anemometry. These measurements were cen-
KHz oscillation was first seen on the oscilloscope and was tered around the quantification of two fundamental units, the
expected that this might be a stroboscope effect from a higher rms turbulent velocity u’, and the integral length scale of
frequency due to the analog to digital data conversion. This turbulence L, and the use of a two eddy theory of turbulent
stroboscope phenomenon is referred to as “Aliasing”, “fold- combustion. The rms turbulent velocity u‘, is a measure of
back”, or the “Nyquist frequency” and can occur when con- the fluctuating velocity on a microscale and can be related to
verting an analog to a digital signal. The Nyquist criteria states macroscale mean velocities in a pipe (or fan tip speeds) by
that at least 2 points/cycle need to be taken in order to digitally correlations below. The integral length scale of turbulence L,
reproduce an analog signal. When the Nyquist criteria is not is a measure of the larger fundamental eddy size seen in tur-
obeyed and the sampling rate is near 1 point/cycle or less a bulence for which there is no fundamental change in the lam-
lower frequency “alias” signal will be generated and is only inar burning velocity u,. The smaller vortex type dissipative
a shadow of the true higher frequency. Further experiments eddy is where the chemical reaction rate is increased and results
were run at the fastest sampling rate possible on the oscillo-
scope and a 46 KHz frequency was consistently observed. This
higher frequency was found to be the natural frequency of the F n Speed
Lower Flammable Limit of Methane vs. Fan Speed
pressure transducer. RPM

Smoothing over the appropriate number of points can be


used to effectively “filter” out noise or oscillations in the raw
data. Higher frequencies could be “filtered out” this way and
allow lower frequencies to be clearly seen. The Pm criteria was
chosen as the pressure that the wall of a process reactor or
pipe would “see” assuming a natural frequency of 1 KHz.
This is the (1 KHz “Filter”) Pm criteria used for the results
-
2.m

l.5W 1
4 35 Liter Resum (zero probability)
Slope = Moo R P W X LFL shm
1
Bradley 1979 (a0X lgn)

shown in Figure 1. This means that the raw data shown in Comlallon Factor
N35MB exp. = 0.90
Figure 2a would “look like” Figure 2b to a vessel with a nat. 5w
freq. of 1 KHz. The high frequency oscillations picked up by
the pressure transducer are so fast that the wall of a 1 KHz o j . : . : . : . :’. : l . ; , : . ; . : .1
0 1 2 3 4 5 6 7 8 9 10
nat. freq. vessel cannot respond or stress that fast. Therefore Lower Flammable Llmil. X CH4
it is imperative when using this data to design vessel stress
levels that the nat. freq. and stress response time of the system FIGURE 4. Lower flammable limit of methane vs. fan
or vessel be known in order to apply the relevant data. speed.

Process Safety Progress (Vol. 14, No. 1) January 1995 23


Table 1 35 Liter Fan Speed Correlation to Pipe Flow Reynolds Numbers, Re
Ns5 NB UI L v = 16.3 x 1 0 - ~ ~=8.62~10-’ v= 1.29~
rPm rPm ft/s ft Rl, R e x lo-’ RL R e x lo-’ R L. R e x lo-’
1500 1667 4.101 0.0328 813 323 1540 457 10230 1280
1000 1111 2.734 0.0288 483 244 913 344 6074 965
500 556 1.367 0.0253 212 156 400 220 2664 616
100 111 0.273 0.0224 38 61 71 86 472 24 1

in a turbulent burning velocity u,. These units are used to define process turbulence having Reynolds #’s up to 1.3 million. Tur-
the turbulent Reynolds number R, as: bulent Reynolds numbers corresponding to typical pipe flows
can be simulated by relatively low fan speeds for most process
R, = u‘L/v conditions.
The effects of turbulence on LFL, COC, or any other com-
R, is used as a measure of the fundamental turbulence pres- position limiting condition is misleading in nature because of
ent and can be correlated to the classical pipe flow Reynolds the danger of flame kernel ignition in a quies;centzone followed
number Re. The 35 liter experimental goal was to reproduce by the much higher turbulent flamespeeds and rates of pressure
the effects of turbulence on the lower flammable limit (LFL) rise. Bradley’s work shows a peak value of - 20 for the u,/u,
of methane vs. fan speed published by Bradley. The results of ratio and is useful as a vent sizing parameter for worst case
these experiments are shown in Figure 4 where a “zero-prob- turbulence effects. Further experiments on the u,/uI ratio for
ability” of propagation limit (ASTM) figure is utilized in the compositions closer to stoichiometric or several Yo away from
35 liter system. The criteria used by Bradley is “based upon the limit (process upset conditions) are recommended for better
an 80% probability of ignition and a minimum of 20 sparks definition of this parameter.
at each set of conditions”. These different criteria explain the Freon 22, 134a, and 123 were being considered as alternative
relative shift in absolute value of the LFL when plotted together refrigerants in a chlorine liquefaction process application.
in Figure 4. Flammability testing in chlorine was needed to confirm cal-
The slopes of the fan speed vs. LFL plots are what represent culated maximum pressures, flammable limits, and the rates
equivalent effective levels of turbulence for correlation between of pressure rise for vent sizing and/or containment. Calcu-
the two systems. The experimental correlation factor (35 liter lations and previous experiments indicated that R134a would
fan speed/Bradley fan speed) is found to be N,,/N,exp = 0.90 be flammable in chlorine and have higher compression ratios
and compares effective levels of turbulence in the Bradley
system to the modified 35 liter system. The classical pipe flow
Reynolds number RE is defined as:
PmlPi
Compression Ratio (PmlPi) vs. Freon Vol%
Re = V d/v (V = mean velocity in pipeline, ft/sec, in Chlorine @ 100 psia & 38 C

d = pipe diameter, ft)

The turbulent Reynolds number RL is related to the classical


Reynolds number Re by: R22.
CHCLF2 calculated

y \ expenmental
R,= 5.927 x 10-8(Re)’.84

From these equations and the correlation factor (Ns5/


NB = 0.90), values of Re were calculated for selected fan speeds
and three kinematic viscosities presented in Table 1 .
10 20 30 40 50 60
The three kinematic viscosity values correspond to methane- Freon Yo1 % i n Chlorine
air @25”C & 1 atm, and process conditions @220°C & 97
psia, and 235°C & 354 psia. The 35 liter system simulates FIGURE 6. Compression ratio (PmlPi) vs. freon vol% in
chlorine @ 100 psia & 38C.

R22 30-57
R134A 18.38

- “22

-” /A”

Non Flammable mixlYles

CHLORINE HYDROGEN CHLORIDE

FIGURE 5. Modified 35L for freons in chlorine. FIGURE 7. Flammability of freons in chlorine.

24 January, 1995 Process Safety Progress (Vol. 14, No. 1)


than R22, and R123 would possibly be nonflammable. Deter- 2. Chippett, S., “An Investigation of Vented Explosions at
mine of peak compression ratios (Pm/pi) and maximum rates Initially Elevated Pressures for Propane/Air Flames,”
(dP/dtm) was performed in the 35 liter system as shown in Project Report for NFPA, February 2, (1984).
Figure 5 . 3. Zabetakis, M. G., Flammability Characteristics of Com-
All R123 runs were negative and would not propagate a bustible Gases and Vapors, United States Bureau of Mines,
flame. The flammable limits for R22 & R134a were found to Bulletin 627, (1965).
be 30-57 & 18-38 + / - 3 vol Yo. Figures 6 & 7 show these 4. Aerstin & Street, Applied Chemical Process Design,
limits and compare the calculated and experimental compres- Plenum Press, New York and London, pp. 9-12, (1978).
sion ratio (Pm/Pi) values vs. Freon vol 070 in Clz. The exper- 5. Andrews, G. E., and D. Bradley, “Determination of Burn-
imental values compared very closely with the calculated ing Velocitv bv Double Ignition in a Closed Ves-
compression ratios and can be utilized in pressure containment se[”Combukion and Flame 50, pp. 77-89, (1973).
design calculations. 6. Andrews, G. E., D. Bradley, and S. B. Lwakabamba,
The flame speeds of R22 or R134a in chlorine are several “Turbulence and Turbulent Flame Propagation-A Crit-
orders of magnitude below detonation speeds and are not ca- ical Appraisal,” Combustion and Flame 24, pp. 285-304,
pable of running up to anywhere near sonic velocity. The (1975).
compression ratio Pm/Pi is the same as the factor “R” used 7. Abdel-Gayed, R. G., D. Bradley, and McMahon, “Tur-
in NFPA 69 for design of pressure containment. Although bulent Flame Propagation in Premixed Gases: Theory and
flammability was achieved in the 35 liter vessel, it is difficult Experiment,” The Combustion Institute, Seventeenth
to create these same conditions at the industrial process scale. Symposium (International) on Combustion, pp. 245-254,
Leak scenarios produce lean and rich zones that quench any (1979).
small flammable pockets. Leaks would also produce operating 8. Abdel-Gayed, R. G., K. J. Al-Khishali, and D. Bradley,
difficulties and warning signals long before the LFL concen- “Turbulent Burning Velocities and Flame Straining in Ex-
trations could be reached. The deflagration pressure may stop, plosions,’’ Proc. R. SOC. Lond., A 391, pp. 393-414,
overcome, or reverse the leak scenario. Liquid and metal heat (1984).
sinks and heat of vaporization cooling would quench localized 9. Abdel-Gayed, R. G., and D. Bradley, “Criteria for Tur-
reactions or combustion. bulent Propagation Limits of Premixed Flames,” Com-
bustion and Flame 62, pp. 61-68, (1985).
ACKNOWLEDGMENTS 10. ASTM E 681-79, Standard Test Method for Limits of
Flammability of Chemicals, (1979).
11. ASTM E 918-83, Standard Practice for Determining Lim-
Bruce Powers developed the 35 liter flammability system,
its of Flammability of Chemicals at Elevated Temperature
Howard Graves aided in system operation, and Charles
McDonald designed the ignition triggering. Milton Marks and Pressure, (1983).
12. Chomiak, Jerzy, “Dissipation Fluctuations and the Struc-
modified the smoothing function for the hydrogen detonation
ture and Propagation of Turbulent Flames in Premixed
work, and the freon in chlorine experimental work was per-
Gases at High Reynolds Numbers,” The Combustion In-
formed with the assistance of Rick Kwasny, Steve Froelicher,
and Bob Zondlak. stitute, Sixteenth Symposium (Intl) on Combustion, pp.
1665- 1673, (1977).
13. National Fire Protection Association, NFPA 69 Explosion
LITERATURE CITED Prevention Systems 1986 Edition.

1. Lewis and vonElbe, Combustion, Flames, and Explosions


of Gases, 2nd Edition, Academic Press, N.Y., N.Y ., This paper (8a) was presented at the AlChE Spring Na-
(1961). tional Meeting in Atlanta, GAYon April 18, 1994.

Process Safety Progress (Vol. 14, No. 1) January 1995 25

You might also like