You are on page 1of 10

A Full-Scale Experimental and

Modeling Investigation of Dust


Explosions in a Roller Mill
Kees van Wingerden,a Geir Pedersen,a Brian Wilkins,a Mogens Berg,b
and Niels Otto Findsen Nielsenc
a
GexCon AS, Fantoftvegen 38, 5072 Bergen, Norway
b
Vattenfall A/S, Støberigade 14, DK-2450 Copenhagen SV, Denmark
c
DONG Energy AS, Kraftvaerksvej 53, Skaerbaek, Fredericia 7000, Denmark

Published online 22 December 2010 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/prs.10434

Combustion of biomass is becoming an increas- INTRODUCTION


ingly important energy source. This is especially true Combustion of biomass is becoming a more and
for wood pellets, as several power companies have more important energy source. Power companies are
decided to use this fuel instead of coal. In this pro- using biomass as a fuel and especially straw and
cess, the wood pellets are ground in big mills and, wood pellets replacing coal. Before the combustion
from there, pneumatically transported to the burners process, biomass (and formerly coal) is ground in a
in the boiler where they are consumed. The grinding roller mill and from there pneumatically transported
of biomass and coal represents an explosion hazard, directly to the burners in the boiler.
which can potentially result in considerable damage The grinding of biomass (and coal) represents an
to the power plant upon propagation of the explosion explosion hazard, which can potentially result in con-
from the mill into other parts of the installation. siderable damage to the power plant upon propaga-
An experimental investigation was performed aim- tion of the explosion from the mill into other parts of
ing at understanding explosion loads that can occur the installation. Energy supplied to the coal and bio-
due to an accidental dust explosion in a mill under mass by the grinding process in the mill itself will
realistic operating conditions. The tests were per- apart from the crushing of the biomass also result in
formed in a roller mill (capacity 36 metric tons coal/ a heating of the material. This may eventually lead to
h; internal volume 23.2 m3) connected to a coal smoldering combustion in the mill. The smoldering
feeder located on top of the mill. combustion may cause ignition of a dust cloud of
Test conditions were predefined using the compu- fine powder in the mill volume. Alternatively, foreign
tational fluid dynamic-based dust explosion simula- metal objects (either present in the coal or the bio-
tion tool DESC. DESC was used to estimate the flow mass or parts of the plant upstream of the mill that
conditions inside the roller mill during normal oper- have come loose) entering the mill may due to over-
ating conditions. In the full-scale tests, these flow con- heating and generation of mechanical sparks cause
ditions have been simulated injecting the dust pneu- ignition of the fuel.
matically. Hence, the grinding of coal and biomass repre-
In total, 25 tests were performed, and a number of sents a significant explosion hazard potentially result-
parameters were varied, including dust type (wood ing in considerable explosion overpressures.
dust and coal dust), dust concentration, and dust The overall objective of the present experimental
cloud location. Ó 2010 American Institute of Chemi- project is to understand the potential damage that
cal Engineers Process Saf Prog 30: 87–96, 2011 can occur due to an accidental dust explosion in a
Keywords: dust explosions; experiments; modeling; roller mill under realistic conditions. Experiments
roller mill; explosion effects similar to those presented in this article have never
been performed at this scale.
The actual roller mill (a ‘‘Loesche LM 18’’ roller
Ó 2010 American Institute of Chemical Engineers mill, grinding capacity 36 tons coal/h) was part of a

Process Safety Progress (Vol.30, No.1) March 2011 87


Figure 1. Picture showing the coal mill as erected at the test site. [Color figure can be viewed in the online
issue, which is available at wileyonlinelibrary.com.]

power plant and was made available by DONG carded ‘‘Loesche LM 18’’ roller mill from a Danish
Energy and Vattenfall. A ‘‘dummy’’ coal feeder was power plant. The grinding capacity of the mill was 36
attached to the roller mill to improve the realism of tons coal/h.
the experiments. The mill was dismantled and transported to the
The experimental program aimed at reproducing test site. The basic parts of the mill are shown in Fig-
realistic conditions as closely as possible. As normal ure 1 and consist of a mill foundation, the mill cham-
operation of the mill during the experiments was not ber, the sieve, and ‘‘the top’’ with coal feeder pipe
possible, realistic operating conditions were simulated and coal dust pipes.
by pneumatic injection of the dust from pressurized The internal volume of the mill is formed by the
vessels. The delay time between start of injection and lower, cylindrical part of the mill (the mill chamber,
moment of ignition was used as a parameter to volume 8.4 m3) and the upper conical part (con-
obtain realistic turbulence conditions. The choice of taining the classifier, volume 12.0 m3). The coal
the delay time was based on simulations performed feeder pipe and the coal dust pipes mounted at the
using the computational fluid dynamics (CFD)-based top of the mill represent an additional volume of 2.8
dust explosion simulation code DESC [1]. m3 resulting in a total internal volume of the mill of
First, DESC was used to describe the flow condi- 23.2 m3.
tions inside the mill during normal operation. Second, In the bottom of the mill chamber, the mill table is
DESC was used to determine the most appropriate located (see Figure 2). Under normal operating con-
delay time between injection and ignition when using ditions of a coal mill, the mill table is rotating, but in
pneumatic injection of the dust using the turbulence this investigation, the mill table was fixed. The mill
intensity as the main parameter. table forms the bottom of the internal volume of the
This article describes both the experiments and mill. The mill table has a diameter of 2,000 mm.
numerical simulations performed. There is an open clearance (referred to as throat in
this document) of 70 mm between the table and
DESCRIPTION OF THE EXPERIMENTAL SET-UP the surrounding wall (see Figure 2). Inside the throat,
several small flow straighteners are mounted. The net
Roller mill cross-section of the throat is 0.42 m2.
The experiments were performed at the GexCon In the centre of the top of the mill, a vertical coal
explosion test site at the island of Sotra, close to the feed pipe is mounted. This pipe connects the coal
city of Bergen, Norway. The actual roller mill is a dis- feeder with the mill.

88 March 2011 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.30, No.1)
Figure 2. Pictures of the mill. Left top: Inside of the mill with the two rollers. The ignition source can be seen
between the rollers. Right top: The mill table and the throat. Bottom left: The six coal dust pipes with adjustable
dampers. Bottom right: The ‘‘coal feeder’’ with the ‘‘opening to the coal silo.’’ One of the dust dispersion systems
can be seen in the far end of the feeder. [Color figure can be viewed in the online issue, which is available at
wileyonlinelibrary.com.]

In the top of the mill there are three coal dust out- Also parts of the original pair of primary air ducts
lets of 550 mm diameter. After a 908 bend, each of were fixed to the mill foundation underneath the mill
these outlets end up in two pipes of smaller diameter table. The ducts had a cross-section of 0.85 m 3 0.85
(coal dust pipes with diameter d 5 384 mm; see Fig- m and had a length of 6 m. The ends of the ducts
ures 1 and 2). The length of the 384 mm pipes were could be partly blocked and provided with polyethyl-
2 m. In real power plants, the pipes lead the fine ene plastic sheets as venting devices.
dust produced in the mill to the burners and could
be up to 60 m long. To simulate the resistance of the
pipes a diaphragm was installed at the end of the Dust Dispersion System
pipe blocking the cross-section by 80%. To generate a well-defined explosive dust atmos-
To simulate the effect of a coal feeder, a ‘‘dummy pheres parts of a traditional explosion suppression
feeder’’ was installed at the top of the mill (see Figure system were used. The parts were provided by
2). The dummy feeder was a rectangular vessel of 5.1 INCOM Explosionsschutz AG, Switzerland.
m3 capacity (5.2 m long, 1.1 m wide, and 0.9 m Two to six 20-l suppression containers were used
high). The dummy feeder was connected to the mill in parallel during each test [a maximum of five of
via the coal feed pipe. At the top of the opposite end them mounted onto the mill (three in the top part
of the feeder, an opening (0.27 m2) was installed to and two in the cylindrical part (milling room)] and
simulate the connection to a coal silo. The opening one to inject dust into the coal feeder). Figure 1
size could be varied to simulate the effect of a com- shows some of the suppressor containers as mounted
pletely or partly filled coal silo. and installed on the coal mill. The flammable dust to

Process Safety Progress (Vol.30, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2011 89
Figure 3. Implemented model of the coal mill with monitor points applied. [Color figure can be viewed in the
online issue, which is available at wileyonlinelibrary.com.]

Table 1. Properties of coal dust.


be used in a test was added, equally distributed
between the containers. The containers were charged
with pressurized air to 20 barg. In most of the experi- Coal dust
ments, 1–3 kg of dust was filled into the containers. Property Value
These amounts of dust correspond to dust concentra-
Median particle size diameter (lm) 55–90
tions varying from 250–500 g/m3.
Moisture content (% w/w) 2.1–2.3
Maximum explosion pressure (barg) 7.0–7.8
Diagnostics KSt-value (bar m/s)* 118–145
The main parameter measured was the overpres-
sures generated within mill and coal feeder during *The KSt-values and maximum explosion pressure val-
the explosion tests. A total of 10 piezoelectric pres- ues are based on a limited number of experiments
sure transducers were installed. The transducers were performed for each sample. A full test according to
located as follows: two transducers in the primary air European standards [2,3] can give different values.
ducts (transducers 2 and 3), one in the room below
the milling table (transducer 1), two in the cylindrical
part of the mill (mill chamber; transducers 4 and 5), tion throughout the course of the project due to in-
two in the top part of the mill (transducers 6 and 7),
herent reproducibility of the laboratory tests and vari-
one transducer in the coal outlet (transducer 8), and
ation in the raw material itself. Hence, the properties
finally two in the dummy coal feeder (transducers 9 in Table 1 are presented as a range. It is emphasized
and 10). The exact locations of these transducers
that the dust concentration at which the maximum
coincide with the monitoring points M1, M2, M3, M6,
explosion pressure and KSt-value occurred, varied
M8, M12, M14, M15, M16, and M19 in Figure 3.
between 500 and 750 g/m3. The particle size is deter-
In addition to pressure measurements, video
mined by sieving.
recordings were made for each test.
Several wood dust samples were used during the
tests with wood dust. The wood dusts involved in the
Dusts Used in the Experiments experimental campaign had two main sources of ori-
The dust explosion experiments in the roller mill gin. The properties of the wood dust of the first
were performed with coal and wood dust. The prop- source are given in Table 2. The properties of the
erties of the coal dust used are presented in Table 1. wood dust of the second source appeared to vary
Several coal dust samples were tested during the test considerably from packaging unit to packaging unit,
program regarding particle size, moisture content, and hence, the properties of the dust were deter-
and explosion effect properties showing some varia- mined for each test performed. On the basis of the

90 March 2011 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.30, No.1)
Table 2. Properties of wood dust.

Wood dust sample Wood dust Wood dust Wood dust Wood dust
from first from second from second from second
Property source source (class I) source (class II) source (class III)
Median particle size diameter (lm) 110 60–90 95–105 160–250
Moisture content (% w/w) 5.4 5.6–6.0 5.5–5.8 6.3–6.7
Maximum explosion pressure (barg) 7.3 7.6–7–7 7.3–7.6 6.5–7.2
KSt-value (bar m/s) 77 112–122 88–92 31–55

Table 3. Summary of experiments performed.

Dust cloud location


Maximum
Test Dust cloud concentration (g/m3) KSt-value pressure
no. Milling room Upper mill part Coal feeder Dust type (bar m/s) observed (barg)
1 119 – – Wood dust (source 1) 77 –
2 119 – – Wood dust (source 1) 77 –
3 238 – – Wood dust (source 1) 77 –
4 119 127 – Wood dust (source 1) 77 –
5 119 84 – Wood dust (source 1) 77 0.08
6 238 – – Wood dust (source 1) 77 0.11
7 119 127 257 Wood dust (source 1) 77 0.11
8 238 – – Coal dust 118–145 0.04
9 238 253 – Coal dust 118–145 0.08
10 238 169 333 Coal dust 118–145 1.46
11 238 253 – Wood dust (source 2, class I) 112–122 1.76
12 238 253 – Wood dust (source 2, class I) 112–122 1.90
13 238 84 – Wood dust (source 2, class III) 31–55 –
14 238 253 – Wood dust (source 2, class III) 31–55 1.08
15 238 253 – Wood dust (source 2, class III) 31–55 0.5
16 238 169 – Wood dust (source 2, class III) 31–55 0.39
17 238 253 – Wood dust (source 2, class II) 88–92 1.14
18 238 253 – Wood dust (source 2, class II) 88–92 1.58
19 238 253 – Wood dust (source 2, class II) 88–92 1.68
20 238 253 – Coal dust 118–145 0.19
21 357 380 – Coal dust 118–145 0.20
22 119 84 – Coal dust 118–145 0.18
23 238 253 303 Coal dust 118–145 1.65
24 238 253 303 Wood dust (source 2, class I) 112–122 1.87
25 119 169 303 Wood dust (source 2, class I) 112–122 1.59

properties found, three subclasses of the wood dust was always effected at the most likely position for
from the second source were defined class I, II, and ignition sources in the roller mill, viz. at the milling
III. Also these properties have been summarized in table (the ignition source was located 0.65 m above
Table 2. As for the coal dust, it is emphasized that the center of the milling table). The ignition source
the wood dust concentration at which the maximum itself was a pair of 5-kJ chemical igniters (producer:
explosion pressure and KSt-value occurred varied Frau Sobbe). A constant delay time of 900 ms
between 1,000 and 1,500 g/m3. between dust injection and ignition was applied. The
delay time was based on DESC simulations as
EXPERIMENTAL PROGRAM reported below. The turbulence intensity after a delay
The overall aim of the experimental program was time 900 ms is a representative for the turbulence
to determine the maximum possible explosion load intensity prevailing under normal maximum load
due to a biomass dust explosion in the mill. The operating conditions.
wood dusts investigated are expected to be represen- The most important parameters that were varied
tative for biomass dusts encountered in practice. The include, dust type (wood dust and coal dust), wood
tests with coal dust were performed as a reference as dust reactivity, dust cloud size, and dust cloud loca-
coal is still applied as a fuel in power plants. Ignition tion. Also the size of the (vent) opening on top of

Process Safety Progress (Vol.30, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2011 91
Figure 4. Turbulence intensity in the roller mill under realistic conditions shown for different cross-sections
through the mill. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.
com.]

the coal feeder (see Figure 2) was varied. Dust con- Geometry
centrations were limited to concentrations below 400 A model of the coal mill was implemented into
g/m3 due to the explosion effects seen at these very DESC shown in Figure 3. Monitor points were
dust concentrations during the experiments. At higher included to monitor pressure at different locations.
concentrations, the explosion loads were expected to Among the chosen, monitor points were those loca-
cause damage to the mill. tions where pressure transducers were installed in the
A total of 25 tests were performed. A summary of actual experiments, the positions of the dust disper-
the test parameters in each of these tests is given in sion nozzles, and the ignition source.
Table 3.
Simulations Performed to Determine an
Ignition Delay Time Simulating Realistic
EXPLOSION SIMULATIONS PERFORMED WITH THE CFD-TOOL DESC Operating Conditions
DESC is a simulation tool based on CFD that can The explosion experiments could not be per-
predict the potential consequences of industrial dust formed under realistic conditions with the mill pro-
explosions in complex geometries. The DESC code is ducing fine biomass dust or coal dust with air enter-
based on the existing CFD code FLACS (FLame ing the mill via the primary air ducts to transport the
ACceleration Simulator) for gas explosions. The mod- ground dust up via the classifier in the upper part of
eling approach adopted entails the extraction of the mill and next into the pipes running toward the
combustion parameters from pressure–time histories burner. The DESC code was used to run simulations
measured in standardized 20-l explosion vessels. A to determine the flow conditions in the mill. Accord-
detailed description of the DESC code has been given ing to information on the mill, a total flow rate of
by Skjold [1]. 70,000 m3/h through the mill prevails when produc-
DESC simulations were performed to select experi- ing 36 tons/h of fine coal dust. The flow conditions
mental conditions. Simulations were performed both prevailing in the mill were investigated for these
before the experiments were performed and during operating conditions. The results are shown in Figure
the experimental program. 4. Turbulence intensities during operational flow con-

92 March 2011 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.30, No.1)
Figure 5. Turbulence intensity in the coal mill with dust dispersed from six 20-l containers at t 5 900 ms, same
cross-sections as in Figure 4. [Color figure can be viewed in the online issue, which is available at wileyonline-
library.com.]

ditions vary between 0.5 and 2.0 m/s. Higher turbu- of 400 and 700 g/m3 were applied (compared to the
lence intensities prevail in the primary air ducts (3.0– coal dust concentration of 514 g/m3 when running
3.5 m/s) and considerably lower in the coal feeder under full load conditions).
(0 m/s). Figure 6 shows an example of pressure–time his-
Simulations using dust injection were performed tories obtained when dispersing coal dust in the mill-
using the injection points as applied in the experi- ing room and the upper part of the mill (classifier)
ments. The simulations involved different delay times. only. The simulations concern effect of length of the
On the basis of these simulations, a 900-ms delay burner pipes (2–10 m) and effect of partial blockage
time was chosen for the experiments. Figure 5 of these pipes (varied from 100% open to completely
presents the turbulence intensity fields after 900 ms. closed for 2-m long burner pipes). Direct comparison
In the upper part of the mill (classifier), in the pri- between the blind code predictions and experiments
mary air ducts as well as in the outlet pipes turbu- performed is in general not possible, because the
lence intensities are somewhat lower than when experiments were performed with leaner coal dust
applying realistic flow conditions. The turbulence in- mixtures. However, simulation number 010203 can be
tensity in the coal feeder is considerably higher than compared with the results of test number 21, where
seen during operational flow conditions. conditions are very close. The maximum pressure in
the experiment was 0.2 barg. The simulation predicts
slightly higher pressure (0.22 barg).
Explosion Simulations Performed
Explosion simulations were performed using the
DESC simulator. These explosion simulations were
performed before the performance of the experi- EXPERIMENTAL RESULTS
ments and concerned coal dust only. Combustion A summary of the maximum pressures obtained in
data were extracted from 20-l sphere tests for a coal each of the tests is included in Table 3. For those
dust with a maximum explosion pressure of 7.4 barg tests, where no pressure value is indicated, the test
and a KSt-value of 119 bar m/s. Dust concentrations failed.

Process Safety Progress (Vol.30, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2011 93
Figure 6. Pressure development as simulated with DESC for a coal dust cloud (dust concentration 400 g/m3)
present in the milling room and upper part of the mill (classifier area). Simulation 011000 concerns burner
pipes of 10 m (fully open). Simulations 010200 to 010204 concern burner pipes with a length of 2 m where the
opening of the burner pipe end is gradually blocked from 0 to 100% in steps of 25%. [Color figure can be
viewed in the online issue, which is available at wileyonlinelibrary.com.]

Effect of Dust Cloud Location


The effect of dust cloud location was investigated
effectively as an investigation of dust cloud size. An
example is given in Figure 7 for three tests (test nos.
8, 9, and 10) and involving coal dust. The dust cloud
was gradually increased from being present in the
milling room only (test 8), via being present in the
entire mill (test 9) up to filling both the mill and the
coal feeder in test 10. Pressure time histories meas-
ured in the mill show a very big difference between
the cloud being present only in the mill (tests 8 and Figure 7. Effect of cloud size on explosion pressure
9) and also being present in the coal feeder. In the development in roller mill. Measurement at pressure
latter experiment, the pressure in the mill is more location in mill (P4). Test 8: cloud in milling room,
than 20 times higher than in experiments 8 and 9. An test 9: cloud in entire mill, and test 10: cloud in mill
explanation for this behavior can be given by study- and coal feeder. [Color figure can be viewed in the
ing the pressure–time histories of the same experi- online issue, which is available at wileyonline
ments measured in the coal feeder (see Figure 8). library.com.]
The pressure rise in the coal feeder occurs much
faster, and a maximum pressure is reached much ear-
lier than in the mill. When the explosion in the coal
feeder is completed, the explosion in the mill starts again due to the explosion pressure generated in the
becoming more violent and reaches a maximum pres- mill and communicated to the coal feeder.
sure considerably later than reached in the coal The ignition point in the mill is directly under-
feeder. The pressure in the coal feeder increases neath the coal feeder pipe. As a result, the flame

94 March 2011 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.30, No.1)
Figure 8. Effect of cloud size on explosion develop- Figure 10. Effect of dust type: Maximum overpressure
ment in roller mill. Measurement at pressure location in the mill due to ignition of clouds of coal dust (test
in coal feeder (P10). Test 8: cloud in milling room, 9) and wood dust type class I (test 11) filling the
test 9: cloud in entire mill, and test 10: cloud in mill entire mill.
and coal feeder. [Color figure can be viewed in the
online issue, which is available at wileyonlinelibrary.
com.]
conditions. In these tests, the mill is entirely filled
with similar concentrations of respectively coal dust
or wood dust (class I; second source). Figure 10
shows that these dust clouds behave very differently:
whereas wood dust causes a very strong explosion,
coal dust causes a low pressure generating mild
explosion. The different behavior may be due to dif-
ferent combustion characteristics at the dust concen-
tration investigated or different dust cloud dispersion
properties.

Venting Parameters
In tests nos. 17–19, the size of the opening on top
of the coal feeder (see Figure 2) was varied from
100% open in test 17, via 50% in test 18 to 13.5% in
test 19 (all tests using wood dust, source 2, class II).
The tests performed show a strong effect when
Figure 9. Effect of dust reactivity of wood dust reducing the opening from 100 to 13.5% (causing a
clouds filling the entire mill on the maximum over- peak pressure increase from 1.14 barg to 1.68 barg).
pressure generated in the mill. [Color figure can be This indicates that the influence of restricted venting
viewed in the online issue, which is available at as a result of the coal silo upstream of the coal feeder
wileyonlinelibrary.com.] needs to be taken into account when determining the
maximum dust explosion load on the roller mill. A
similar trend can be seen in the DESC simulations
reported in Figure 6.
shoots into the coal feeder causing a strong second- Figure 11 shows two plots of the pressure distribu-
ary explosion there. The explosion pressure is flow- tion in the mill in an experiment where a high explo-
ing back into the mill causing turbulence in the mill sion pressure was generated (test 10) and one where
and thereby enhancing the combustion rate in the the explosion was weak (test 9). The pressures are
mill. more or less similar throughout the entire mill for the
weak explosion of test 9. Only slightly lower pres-
Effect of Dust Reactivity sures are seen in the primary air ducts (P2, P3) and
In tests nos. 11–17, dust clouds of different wood the room underneath the milling table (P1). However,
dust reactivities (described as wood classes I, II, and in test 10, pressures in the primary air ducts and the
III) were generated filling the entire mill (not the coal room underneath the milling table are considerably
feeder) and ignited. Figure 9 shows that, as expected, lower than in the area above the milling table. This
the maximum pressure generated increases with the feature is seen in all weak and strong explosions.
reactivity of the dust described by the KSt-value The reason is the throat between the milling table
(taken as an average from each class; see Table 2). and the walls of the mill limiting flow from the upper
part of the mill.
Effect of Dust Type
A direct comparison of coal and wood dust is diffi- CONCLUSIONS
cult on the basis of the experiments performed. How- A comprehensive set of experiments was per-
ever, tests 9 and 11 were performed under similar formed in a roller mill aiming at determining the

Process Safety Progress (Vol.30, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2011 95
Figure 11. Pressure distribution in the mill in tests 9 and 10. [Color figure can be viewed in the online issue,
which is available at wileyonlinelibrary.com.]

maximum explosion load due to biomass and coal maximum explosion pressure in the mill. This indi-
dust explosions under realistic operating conditions. cates that the influence of restricted venting of the
Dust clouds were generated by pneumatic injec- coal feeder needs to be taken into account when
tion of dust. To obtain realistic operating conditions determining the maximum dust explosion load on
simulations were performed using a CFD-based dust the roller mill.
explosion code. A delay time between the moment The maximum pressure seen in all experiments
of starting dust injection and ignition of 900 ms was performed was 1.9 barg. In this test, a wood dust
found to be adequate in this respect. The ignition cloud (KSt-value 112–122 bar m/s; dust concentration
source position (close to the milling table) was held 250 g/m3) filling the entire mill was ignited (no
constant throughout the test program. dust cloud in coal feeder). The mill withstood this
Parameters investigated included dust cloud loca- pressure load without significant damage.
tion and size, dust cloud reactivity and dust type, and
venting parameters.
Explosions with coal dust resulted in low pressures
(max. 0.20 barg) when the dust cloud was filling the
mill partially or entirely (dust concentrations < 400 g/ ACKNOWLEDGMENTS
m3). If a dust cloud was generated in the coal feeder The assistance of Matthijs van Wingerden, Cathrine
as well considerably higher pressures (1.65 barg) Andersen, Knut Sømoe during the experiments, and
were seen. The reason for the much higher pressures Jon-Thøger Hagen performing the DESC simulations
is the explosion propagation between two intercon- is gratefully noted.
nected vessels. When dust explosions propagate
between items of equipment the peak pressure can
be significantly higher than would occur in a single LITERATURE CITED
enclosure. 1. T. Skjold, Review of the DESC project, J Loss Prev
As expected, an increase of dust cloud reactivity Process Indust 20 (2007), 291–302.
results in an increase of explosion loads. 2. CEN, European Standard EN 14034-1, Determina-
High explosion loads in the mill using wood dust tion of explosion characteristics of dust clouds—
as a fuel also when no dust cloud is generated in the Part 1: Determination of the maximum explosion
coal feeder when indicate that wood dust behaves pressure pmax of dust clouds, 2004.
differently compared with coal dust. This is most 3. CEN, European Standard EN 14034-2, Determina-
likely related to differences in the dust dispersion tion of explosion characteristics of dust clouds—
properties of the two dusts. Part 2: Determination of the maximum rate of
The tests performed show a strong effect of reduc- explosion pressure rise (dp/dt)max of dust clouds,
tion of the vent opening in the coal feeder on the 2006.

96 March 2011 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.30, No.1)

You might also like