Xvii Paper 44

SYMPOSIUM SERIES No.
149 © 2003 Crown Copyright
PRESSURE RELIEF OF LIQUIDS CONTAINING

SUSPENDED SOLIDS
Derick McIntosh, Simon Waldram (Hazard Evaluation Laboratory Ltd, 50 Moxon Street,
Barnet, Herts. EN5 5TS) and Janet Etchells (Health and Safety Executive,
Magdalen House, Stanley Precinct, Bootle, Merseyside L20 3QZ)
© Crown Copyright 2003. Reproduced with the permission of the Controller of Her
Majesty’s Stationery Office.
Abstract
The multi-phase venting of vapour, liquid and solids has been studied
experimentally on the 1 and 10 litre scales. In non-reacting systems, the
depressurisation prof iles of superheated water or water-glycerol mixtures were
studied on their own and with added glass particles. The particles were both solid
and hollow, with specif ic gravities both greater and less than 1. Similar
depressurisation experiments were made during the runaway reaction of acetic
anhydride and water, both with and without solids. Relief set pressures were
between 3 and 5 bara. Solids concentrations were up to 16% v/v. Nozzle to
median solid diameter ratios were between 6 and 500. Experimental design
techniques were used to study the ef f ects of many factors ef f iciently.
Depressurisation prof iles with and without solids present were compared: in
general the solids had little statistically signif icant ef f ect. There was limited
evidence that the less dense solids could increase depressurisation rates slightly.
Tests with runaway reactions have highlighted some dif ficulties in comparing
systems with and without solids present. In all cases, the vented f luids were less
concentrated in solids than those in the reactor. This may have important
implications for design and sizing of pressure relief vents.
INTRODUCTION
The field of two-phase liquid/vapour venting has been the subject of considerable research in
recent years1. The situation will often arise, however, when there are solids present in the
discharge stream. The solid phase might be a heterogeneous catalyst, (e.g. a platinum group
metal on a porous carbon support particle, for a hydrogenation reaction,) a partially dissolved
reactant or a solid product that is crystallizing as the reaction proceeds. There is little guidance
about how to allow for the presence of such solids when sizing a pressure relief system,
although some initial work has been carried out by Beyer and Steinbach2. Some preliminary
information, based on recommendations of The Design Institute for Emergency Relief Systems
(DIERS), is given in reference 3. However this has not been validated experimentally and the
authors point out that these methods may not apply if the solids are not carried over at the same
velocity as the liquid. There may also be problems if the solids affect the flow from the reactor
or cause fouling in the relief system. HSE, in collaboration with a consortium of companies has,
therefore, sponsored a project to investigate the problems further on the laboratory scale, and to
identify the main issues involved. This paper describes the main findings, which will be
published in an HSE Research Report4.
559
SYMPOSIUM SERIES No. 149 © 2003 Crown Copyright
EXPERIMENTAL PROGRAM
An experimental programme was devised to examine the effects of the addition of
suspended solids on a two-phase vented system. To examine the effects of the solids, the
venting profiles from tests with the addition of solids were compared to those without.
Initially non-reacting systems were examined, using superheated water or water/glycerol
liquid mixtures, and then reacting systems were examined, using acetic anhydride and
water. Experimentation was carried out on both the 1 litre and 10 litre scales.
There are a vast number of potential variables that could be studied in this project, e.g.
solid concentration, solid diameter, solid density, nozzle diameter, fill level, stir rate, relief
pressure etc. In these circumstances it was important to plan experiments, execute the
experimental work and process and analyse the results in a structured and efficient manner.
To achieve this, factorial experimental design techniques have been used: see reference 5
for an introduction to this topic. When the effects of many factors are to be examined then
very large experimental programmes are required. In these cases fractional factorial designs
(e.g. half or quarter) can be used to reduce the required number of experiments (e.g. by
factors of 2 or 4). However, a compromise must be made: such designs lose the ability to
discriminate clearly between the effects on the response of combinations of factors.
As an example, a full level factorial design involving 6 factors would require 26, or 64,
experiments. Statistical analysis of the results enables the effect on the response variable to
be estimated not only for every factor, but also for all possible combinations of factors. A
half factorial design would only involve 32 experiments, and hence half of the experimental
effort, without significant loss of response data. Additionally, replicate centrepoint
experiments are commonly included to give an indication of the experimental error, and this
is used in the statistical analysis.
Planning such experiments, and interpreting the results from them, can be tedious and
complex but the use of standard software packages avoids many of these difficulties. In this
project the “Design Expert” software was used, see http://www.statease.com.
APPARATUS
A schematic diagram of the apparatus is shown in figure 1. The test rig consisted of a
reactor vessel, a vent line and a catch tank. The 10 litre reactor was a purpose built, baffled,
stainless steel vessel, rated to 20 bar. Thermocouples, a pressure transducer, a bleed valve, a
magnetic drive coupling, an overpressure relief valve and the vent line were incorporated
into the reactor top plate. Four electric rod heaters were used and entered through the
bottom plate. Two Rushton type impellers and a baffle system were used for the agitation of
the mixtures.
The reactor was connected to the catch tank via a 12.7 mm (½”) vent line that
incorporated a pneumatically operated ball valve. When open this gave full bore
unobstructed flow. The vent line was connected to the reactor via a fitting, designed to hold
a variety of different sized nozzles. The automatic ball valve was sited just downstream of
the nozzle. The catch tank was placed on a balance to allow continuous measurement of the
vented mass, as a function of time.
For the tests with reacting systems, it was important to minimise discharge of vapours.
The vented catch tank was replaced with a sealed vessel and downstream sparged quench
560
tank. To aid condensation of the vapours, approximately 8 kg of water were placed in the
catch tank prior to the test. This worked well and there was very little mass loss from the
system overall. A solenoid bleed valve, fitted between the catch tank and the quench tank,
was opened after venting to prevent the cooling vapours in the catch tank creating a vacuum
and sucking back liquid from the quench tank.
Control and data logging were achieved using HEL software. This allows data
acquisition at a maximum logging rate of 10 points per second. Other than during the
venting stage of the experiment, the data was only logged every 20 seconds.
Emergency
Vent
relief valve
Manual relief
Pneumatically valve
operated ball valve
T
P
Nozzle
holder
Catch Tank
Baffled Reactor
Balance
Figure 1. Schematic diagram of the 1 and 10 litre test apparatus
EXPERIMENTAL PROCEDURES
The same general procedure was used for all the tests using non-reacting fluids:
• Charge the materials and seal the reactor.
• Pull a vacuum in the reactor vessel1
• Allow the reactor contents to reach the relief pressure/temperature.
• Turn off the heating and wait 6 seconds.
• Open the relief valve and vent the reactor until the temperature dropped to 101°C or the
pressure reached 1.05 bara.
• Close the relief valve and allow the reactor to cool.
The 6 second delay was incorporated to allow the data logging rate to be changed to
the maximum rate without any delays caused by the software controls. Once initiated, the
experimental procedure was fully automated and computer controlled to allow reproducible
experimental sequences.
1
Note that in the case of the 1 litre tests, a vacuum was not drawn, rather the vessel was heated with the vent
valve open until the liquid reached its boiling point. In this manner the water was degassed and the reactor was
purged of air.
561
Non-reacting test series were carried out by depressurising water and water/glycerol
mixtures, both with and without the addition of solid particles. Glycerol was added to vary
the liquid density (liquid to solid density ratio) and the liquid viscosity. The solids used
were spherical glass particles in size ranges from 4–45 µm to 250–425 µm. Additionally,
low density hollow glass particles were used. These have a density of 0.6 kg m–3 and are in
the size range 0–65 µm. This range is not too dissimilar to the 4–45 µm solid particles, and
comparisons have been made between the results from tests using these two sets of particles.
In the case of reacting systems, using acetic anhydride and water, the acetic anhydride
was added to the reactor and heated to 50 °C under sealed conditions. The water was then
added, the reactor was re-sealed, the mixture heated to 80°C and the heaters were turned
off. The exothermic reaction was then allowed to proceed naturally until the venting
condition of 3 bara was reached. At this point the reactor vent valve was opened and the
reactor vented to the catch tank. The reactor was re-sealed when the pressure dropped to
1.05 bara.
TESTING ON 1 LITRE SCALE

Several series of tests were carried out on the 1 litre scale. Water and water/glass mixtures were
charged to the reactor and heated to a temperature of 152°C (5.06 bara) before venting. The data
obtained showed that there was very little difference in the depressurisation profiles between
700 ml of water and 700 ml of water plus up to 16% by volume of solids. Indeed, the profiles
from tests with the same volumetric fill levels were very similar. There was limited evidence
that the presence of the hollow particles resulted in a very slightly quicker depressurisation.
Two full factorial test series, with 4 factors, using a 2 mm and a 5 mm nozzle, were
performed. The data analysis on the times taken for the pressure to drop in small increments
throughout the depressurisation showed that there was no statistically significant effect due
to the addition of solids.
A potential problem that was identified on the 1 litre scale was that the repeatability of
the test was poor, particularly when comparing many identical tests carried out over an
extended time period. Tests on the 10 litre scale were more reproducible. For this reason,
experimental work concentrated on the 10 litre scale.
NON-REACTING EXPERIMENTS ON THE 10 LITRE SCALE

A series of tests were carried out by depressurising water and water/glycerol mixtures, both
with and without the addition of solid particles. A 6 factor, half factorial experimental
design was used. The following factors were examined: solid concentration, solid diameter,
nozzle diameter, fill level, stirring speed and glycerol concentration. The relief pressure was
fixed at 5 bara.
In order to examine the effects of the particle density, a further series of tests were
planned to complement those already performed. The low density glass has a similar
diameter range to the 4–45 µm solid glass. Many of the tests already run in the previous
series could be incorporated into a new experimental design, so that only 4 new tests had to
be performed, all of these were with the low density glass particles. A 5 factor, half factorial
562
design was analysed, with the factors being: solid concentration, solid density, fill level, stir
speed, and nozzle diameter. The effect of glycerol was not studied in this series.
Some factors, e.g. fill level and nozzle diameter, are already known to affect the
depressurisation. As this project is aimed at investigating the effects of solids, the prime
interest was to identify those experiments in which solid diameter, or solid concentration,
combined with the other factors affected any aspect of the depressurisation profile.
The response variables chosen for the analysis of this data are the times taken for
incremental depressurisation, in steps of 0.1 bar down to a pressure of 4 bara, then in 0.2 bar
steps to 3 bar, and thereafter in 0.5 bar steps. The results showed that the nozzle diameter
has an effect on the times throughout the depressurisation, and the fill level has an effect
down to 2.5 bara. This would be expected. Additionally, almost throughout there was the
effect of an interaction between the fill level and the nozzle diameter. The stirrer speed was
identified as having an effect in the early stages, down to 4.7 bara.
The first design showed that the solid concentration had an effect on the initial
depressuriation, with the incremental depressurisation times between 4.5 bara and 3.0 bara
reducing with increasing solids concentration (ie increasing the solids, reduced the time for
depressurisation). This may be simply because at a fixed reactor fill level the liquid contents
reduce at increased solid concentration, and the amount of vapour produced as the reactor
contents cool to atmospheric conditions will therefore become smaller at high solids
concentration. This may affect the transition point between single and multi-phase flow.
Otherwise the presence of the solids had no statistically significant effect on the
depressurisation.
It is interesting to note that there was no effect on any of the response variables due to
the amount of glycerol added. This suggests that the viscosity and density difference
between the water and water/glycerol mixtures had no effect on depressurisation for the
range of variables studied.
The analysis of the second experimental design showed similar findings to the previous
design in that neither the solid concentration nor the solid density had a statistically
significant effect on the response variables. Fill level and nozzle diameter had the main
effect on the response factors studied.
REACTING SYSTEM
Several properties were desirable when choosing the reacting system. The reaction needed
to be proceeding at a relatively high rate during venting so as to promote multiphase flow
and to test venting under demanding conditions. Additionally, the 10 litre vessel has a high
thermal inertia (φ factor) and therefore a high fraction of the heat of reaction would be
absorbed by the vessel itself during slow periods of an exotherm.
The reaction of water and acetic anhydride to produce acetic acid was eventually
chosen for several reasons – in particular, the reaction mechanism is relatively simple, and
the flammability hazards are minimised by using an aqueous reaction. The stoichiometry of
this reaction is 1:1, but examination of vapour pressure data indicated that a slight excess of
water would generate a higher pressure, and hence the reaction was run with a 3:2
water:acetic anhydride mole ratio. Perhaps more importantly, this vigorous exothermic
reaction has been involved in a several publicly reported incidents6,7,8. Leigh and
563
Krzeminski8 quote an incident where water had entered a storage tank containing acetic
anhydride and 15% acetic acid at ambient temperature. The resultant overpressure and
vessel rupture killed one person and injured 20 more.
This series of tests was initially planned as a factorial design, with factors of fill level,
glass diameter and glass concentration. The high, low and centrepoint values used in the
factorial design are given in table 1. Note that reactant volume and glass concentration are
independent factors. Combinations of these at high, centrepoint and low values will
determine the actual reactor fill level.
A 10.5 mm diameter constriction was installed on the reactor. This was chosen following
small scale PHI-TEC tests, and was predicted to give significant overpressure during the
venting under the conditions of the reaction. A relief set pressure of 3 bara was used.
In order to give the same thermal inertia, as tests run with glass particles, the tests
run at 0% suspended solids concentration were, in fact, run with 15% v/v of 3 mm
diameter glass beads present. The diameter of these beads was so large as to ensure that
they always settled out in the bottom of the reactor and were not entrained in the vented
flow. This was thought to be a good method of keeping the heat capacity of the system
constant and at the same time the glass ballast would not affect the two-phase venting
rate from the reactor.
Additionally, two tests (one at each fill level) were run using the hollow glass beads
(density 0.6 kg m–3) for comparison with the tests with the 4–45 µm solid particles (density
2.5 kg m–3).
RESULTS
Table 2 gives the conditions for each test. The test numbers in table 2 have been sorted in
terms of reactant volume and have been given a letter designator for clarity.
The main conditions of the tests are given in Table 2. Figures 2 and 3 give the
respective pressure and temperature profiles during venting for the tests using 5000 ml of
reactants +15% v/v solids. It can be seen that the tests A and B (both with 3mm glass
beads), which had two-phase flow gave very much higher overpressures than the tests with
three-phase flow: tests C and D resulted in virtually no overpressure. This was not expected,
as in the non-reacting systems, the presence of solids had very little effect on blowdown
profiles.
Figure 4 shows the self-heat rate profiles for the tests starting from the temperature at
which the electrical heating was turned off (80°C). It is clear that tests A and B (with 3 mm
glass beads as thermal ballast) reached much higher heat-rates before venting commenced at
3 bara. As the test recipes are exactly the same, there must be an effect due to the diameter
of the solids. Transient heat transfer calculations confirm that the large glass beads absorb
less of the heat from the reaction, due to their smaller external surface area and low thermal
diffusivity9. This will become increasingly apparent at high runaway reaction rates. The fact
that glass is a relatively poor conductor of heat will also have an effect, with smaller glass
particles heating up more quickly and more uniformly than larger particles.
A similar effect is likely when φ factors are considered: at very high self-heat rates the
observed φ factor may be much closer to unity than expected from simple calculation,
564
simply due to the fact that the heat transfer to the reactor vessel cannot occur at a
sufficiently high rate for the reactants and vessel to be at an essentially identical
temperature at any time. This means that, during fast thermal runaways, the phi factor may
approach unity irrespective of the thermal mass of solids that may be present or of the
reactor vessel itself. As a consequence, large diameter inert solids result in a much more
vigorous reaction compared with the same mass of smaller glass particles.
Table 3 gives the mass balance information and maximum temperature and pressure
data for the tests with 5000 ml of liquid. Again the effects of the glass diameter can be seen.
The finest particles, which also result in the lowest heat rate due to reaction, result in a
greater mass retained in the reactor. The lower heat rate and hence lower rate of vapour
generation should give a lower superficial velocity and would be expected to result in less
carryover.
Similar observations can be seen in figures 5 to 7 for the tests with the higher fill level.
The data is also summarised in table 4. It is interesting to note from figures 3 and 6 that
there is slight variation between the temperatures at the start of venting even though the
pressure was identical each time. This could be attributable to experimental errors.
However, the lower fill level does appear to result in slightly higher relief temperatures.
One possible explanation is that the density of the acetic anhydride and acetic acid mixture
falls with increasing temperature and the compression of the head space gas may be a
significant factor. At higher fill levels the compression effect would be greater and venting
would occur slightly earlier in the runaway. The data from a test with 6500 ml of reactants
but without glass solids is also shown, and the significant effect of adding glass can be seen
on the reaction rate.
The data for the repeat centrepoint tests are shown in figures 8 to 10 and in table 5.
These tests show excellent repeatability in terms of blowdown profiles and self heat rates:
the lines on the graphs virtually overlap each other.
NOTE ON MIXING
Table 6 shows the mass balance data for the glass particles in each test. In the case of an
ideally mixed system, with no slip between the solid and liquid phases, it could be
expected that the glass concentration in the reactor and the vented fluid would be the
same. Table 6 gives the fraction of the initial liquid and glass masses that were drained
from the reactor. In every case, there is an increased concentration of solids in the
reactor following venting. It is likely that there was some accumulation of solids
towards the bottom of the reactor with the solid glass, particularly with the larger solids.
The nature of the reactor vessel makes it impossible to make a visual inspection and
qualitative judgement of the stirring efficiency. However, based on visual observation
on the 1 litre scale, at similar tip speeds, it was believed to be good at the centrepoint
and high stirring rates.
It is very interesting to note the data from the tests using the hollow solids. From previous
observations on the one litre scale, these particles are very easily mixed with the liquid, and
float when not agitated. Therefore, with poor mixing, any increase in solid concentration would
be towards the top of the reactor. It would be expected that as this will vent first, and that a
greater fraction of the glass would be vented than that in the reactor as a whole. If the mixing
565
was always perfect, then an equal fraction of solid and liquid would remain in the reactor and
the catch tank. The data in table 6 shows that this is not the case, and a greater fraction of the
liquid was vented. It is clear therefore that the liquid is vented preferentially even though the
solids are lighter than the liquid. If the solids are taking part in the reaction, then this will need to
be considered in the design of the relief system, particularly if it means that the reactants could
be concentrated, either in the reactor, or in the downstream containment or disposal system.
CONCLUSIONS
From the testing and analysis described above the following can be concluded.
A. Non-reacting tests
a. Pressure and temperature profiles versus time during venting were in general not
influenced to a statistically significant extent by the presence of solids.
b. There is limited evidence that, under some circumstances, the presence of solids
can increase depressurisation rates, particularly at intermediate times. It is likely
that the presence of solids may promote even more vapour bubble nucleation and
promote bubbly or homogeneous rather than churn turbulent flow.
c. During venting, liquid is discharged preferentially to the solids. This was observed
for both naturally floating and sedimenting particles, i.e. both less and more dense
than the fluid in which they are suspended.
B. Reacting systems
a. Adding inert particles to a reacting system increases the phi factor and hence reduces
the reaction runaway rates. This effect is highly non-linear with particle mass.
b. Depressurisation of reacting systems containing inert solids was highly
reproducible on the 10 litre scale.
c. The temperature of inert particles suspended in a liquid whose temperature is
changing rapidly may lag behind the fluid temperature. This means that the
effective phi factor can change during the course of a fast runaway. For this
reason, large inert particles appeared to accelerate the runaway reaction relative
to the same mass of smaller particles and this can lead to larger overpressures
during venting.
d. Heat transfer limitations to the body of a large or massive reactor vessel during a fast
runaway may mean that the average temperature in the reactor body is much lower
than in the reacting fluid. In an analogous way to item c this can lead to a shifting
value of the phi factor as a reaction proceeds. The reaction may become much faster
than that anticipated from small scale studies at an analogous phi factor.
e. Direct comparison of the results from the reacting system tests with, and
without, solids is very difficult because of the change in the phi factor (and
hence reaction rate) and the ability of inert solids to accelerate a runaway
reaction, see items a, and c.
f. Production of vapour and preferential flow of the liquid (relative to the solid) in the
vent discharge will enhance the solid concentration in the reactor. This will then
alter the phi factor and hence the runaway rate. If the solids are participating in the
reaction this may also affect the reaction rate per unit volume.
566
As a general conclusion these preliminary studies show that, for the ranges of variables
studied, small diameter inert solids have little influence on the rates of depressurisation
achieved. Although it was outside the scope of this project, there are several issues that
would need to be carefully considered before calculating the required vent areas in the case
of three-phase flow. Any effects of concentration of the solids during venting must be
considered in selection of the calorimetric test methods used for relief line sizing. Solid
deposition and downstream fouling may also be issues.
REFERENCES
1. Fisher, H.G., Forrest, H.S., Grossel, S.S., Huff, J.E., Muller, A.R., Noronha, J.A.,
Shaw, D.A. and Tilley, B.J: “Emergency relief system design using DIERS technology,
The Design Institute for Emergency Relief Systems (DIERS) project manual,” 1992,
AIChE, NY. ISBN 0-8169-05668-1.
2. Beyer, R. and Steinbach, J: “Source term characterization for three-phase venting
scenarios.” Paper presented at the 10th International Symposium on Loss Prevention
and Safety Promotion in The Process Industries, Stockholm, June 19th –21st, 2000.
3. Etchells, J. and Wilday, J: “Workbook for chemical reactor relief system sizing,” 1998,
HSE Books. ISBN 0-7176-1389-5
4. McIntosh, R.D. and Waldram, S.P.: “Reactor pressure relief of fluids containing
suspended solids” HSE Contract Research Report, project reference
4187/R05.101.
5. Davies, L., Efficiency in research, development and production: the statistical design
and analysis of chemical experiments, 1993, Royal Society of Chemistry, ISBN 0
85186 137 7
6. Barton, J. and Rogers, R.L.: “Chemical reaction hazards,” second edition, 1997,
I.Chem.E. ISBN 0-85295-341-0
7. I.Chem.E, Industrial accidents database, 2002, available from http//www.icheme.org.uk.
8. Leigh, W.R.D., and Krzeminski, Z.S.: “The uncatalysed reaction of acetic anhydride
with water,” Chemistry and Industry, pp778-779, April 28, 1962
9. Welty, J.R., Wicks, C.E. and Wilson, R.E.: “Fundamentals of momentum, heat, and
mass transfer.” John Wiley and Sons, New York 1976. ISBN 0-471-93354-6
ACKNOWLEDGEMENT
This work was co-sponsored by the Health and Safety Executive, Great Lakes Fine
Chemicals and Syngenta. The inputs from Mr Graham Arthur (Syngenta), Dr Caroline
Ladlow (Ciba Speciality Chemicals), Dr Allan Timms (Great Lakes Fine Chemical
Company) and Ms Jill Wilday (Health and Safety Laboratory) are gratefully
acknowledged.
The views expressed in this paper are those of the authors and should not necessarily
be taken as those of the Health and Safety Executive.
567
Table 1. Variables and values used in factorial design for the reacting system
Variable High Centrepoint Low
Reactant volume (ml) 6500 5750 5000

Solid concentration (% v/v) 15 7.5 0*
Solid diameter (µm) 150–2502 70–110 4–45
*15% of 3 mm glass ballast was added in these cases
Table 2. Data from 10 litre blowdown tests with the reacting system
Test Reactant Glass

Number volume (ml) diameter (µm)
A 5000 3000
B 5000 3000
C 5000 4–45
D 5000 150–250
E 5750 70–110
F 5750 70–110
G 6500 3000
H 6500 4–45
I 6500 150–250
J 6500 250–4253
K 6500 0–65 (hollow glass)
L 5000 0–65 (hollow glass)
Note: the tests with 3000 µm glass should be treated as 0 % solids (i.e. 2-phase venting with
no solid carryover)
2
This value was chosen as the slightly larger particles (250–425 µm diameter) resulted in no solid carry
over.
3
This test gave no solid carryover and is included for comparison only
568
Table 3. Data obtained from acetic anhydride/water tests with 5000 ml of reactants
Liquid Liquid
Glass Glass retained retained
Glass carryover carryover in in Maximum Maximum
diameter to catch to catch catch reactor pressure temperature
Test (µm) tank (g) tank (ml) tank (g) (g) (bara) (°C)
A 3000 0 0 2425 2880 5.24 169

B 3000 0 0 2523 2045 5.07 168
C 4–45 44.69 17.9 1048 4050 3.05 145
D 150–250 20.36 8.1 2449 2736 3.08 145
0–65
L (hollow 69.78 116.3 1707 3508 3.09 144
glass)
Table 4. Data obtained from acetic anhydride/water tests with 6500 ml of reactants
Glass Glass Liquid Liquid

Glass carryover to carryover to retained retained Maximum Maximum
diameter catch tank catch tank in catch in reactor pressure temperature
Test (µm) (g) (ml) tank (g) (g) (bara) (°C)
G 3000 0 0 3479 2693 4.39 162

H 4–45 299.0 119.6 2989 3619 3.10 147
I 150–250 24.58 9.8 4190 2443 3.92 157
J 250–425 0 0 3471 2898 4.67 165
0–65
K (hollow 219.56 365.9 3811 3048 3.47 151
glass)
M No glass N/A N/A 3350 3571 7.68 195
569
Table 5. Data obtained from centrepoint cases
Glass Glass Liquid Liquid

Glass carryover carryover retained retained Maximum Maximum
diameter to catch to catch of catch of reactor pressure temperature
Test (µm) tank (g) tank (ml) tank (g) (g) (bara) (°C)
E 70–110 19.92 8.0 3116 2767 4.05 159

F 70–110 19.82 7.9 3063 2939 4.05 159
Table 6. Selected mass balance data from acetic anhydride/water tests
Fraction of
Initial Glass Glass Fraction of initial liquid
Glass glass carryover remaining initial glass charge
diameter charge to catch in reactor remaining in remaining in
Test (µm) (g) tank (g) (g) reactor reactor
A 3000 1875 0 1868 1.00 0.54

B 3000 1875 0 1870 1.00 0.38
C 4–45 1875 45 1782 0.95 0.76
D 150–250 1875 20 1806 0.96 0.51
E 70–110 1078.1 20 1042 0.97 0.45
F 70–110 1078.1 20 992 0.92 0.48
G 3000 2437.5 0 2421 0.99 0.39
H 4–45 2437.5 299 2078 0.85 0.52
I 150–250 2437.5 25 2354 0.97 0.35
J 250–425 2437.5 0 2362 0.97 0.43
0–65
K (hollow 585 220 346 0.59 0.44
glass)
0–65
L (hollow 450 70 383 0.85 0.66
glass)
570
6
Test A - 3 mm dia glass
Test C - 4-45 µm glass
Test D - 150-250 µm glass
Test B - 3 mm dia glass
5 Test L - hollow glass
4
Pressure (bara)
1
0 30 60 90
Time (s)
Figure 2. Pressure profile for venting of reacting system tests at low fill level
10 litre scale, 5750 ml charge (5000 ml reactants + 15% by volume glass), water and
acetic anhydride (mole ratio 1.5) 9 mm nozzle, 200 rpm stirring
170
Test A - 3 mm dia glass
160 Test D - 150-250 µm glass
Test L - hollow glass
150
Temperature (°C)
140
130
120
110
100
0 30 60 90
Time (s)
Figure 3. Temperature profile for venting of reacting system tests at low fill level
571
300
Test D - 150-250 µm glass
250 Test A - 3 mm dia glass
Test L - hollow glass
dT/dt (K/min) 200
150
100
50
0
80 100 120 140 160 180
Temperature (°C)
Figure 4. Heat-rate profiles of reacting system tests at low fill level

8
Test G - 3 mm dia glass
Test J - 250-425 µm glass
Test H - 4-45 µm glass
Test I - 150-250 µm glass
Test M - 6500 ml liquid, no solids
6 Test K - hollow glass
Pressure (bara)
0
0 30 60 90
Time (s)
Figure 5. Pressure profile for venting of reacting system tests at high fill level
572
200
Test J - 250-425 µm glass
180 Test M - 6500 ml liquid, no solids
Temperature (°C) Test K - hollow glass
160
140
120
100
0 30 60 90
Time (s)
Figure 6. Temperature profile for venting of reacting system tests at high fill level
800
700 Test J - 250-425 µm glass
Test M - 6500 ml liquid, no solids
600 Test K - hollow glass
500
dT/dt (K/min)
400
300
200
100
0
80 100 120 140 160 180
Temperature (°C)
Figure 7. Heat-rate profiles of reacting system tests at high fill level

573
5
Test E - 70-110 µm glass
Test F - 70-110 µm glass
Pressure (bara) 3
0
0 30 60 90
Time (s)
Figure 8. Pressure profile for venting of reacting system tests at centrepoint

170
160
150
Temperature (°C)
140
130
120
110
100
0 30 60 90
Time (s)
Figure 9. Temperature profile for venting of reacting system tests at centrepoint

574
160
140
120
100
dT/dt (K/min)
80
60
40
20
0
80 100 120 140 160 180
Temperature (°C)
Figure 10. Heat rate profiles for reacting system tests at centrepoint
575

Xvii Paper 44

Uploaded by

Copyright:

Available Formats

Xvii Paper 44

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Xvii Paper 44

Uploaded by

Copyright:

Available Formats

SYMPOSIUM SERIES No.

149 © 2003 Crown Copyright

PRESSURE RELIEF OF LIQUIDS CONTAINING

Figure 1. Schematic diagram of the 1 and 10 litre test apparatus

TESTING ON 1 LITRE SCALE

NON-REACTING EXPERIMENTS ON THE 10 LITRE SCALE

Variable High Centrepoint Low

Reactant volume (ml) 6500 5750 5000

*15% of 3 mm glass ballast was added in these cases

Test Reactant Glass

A 3000 0 0 2425 2880 5.24 169

Glass Glass Liquid Liquid

G 3000 0 0 3479 2693 4.39 162

Table 5. Data obtained from centrepoint cases

Glass Glass Liquid Liquid

E 70–110 19.92 8.0 3116 2767 4.05 159

Table 6. Selected mass balance data from acetic anhydride/water tests

A 3000 1875 0 1868 1.00 0.54

dT/dt (K/min) 200

Figure 4. Heat-rate profiles of reacting system tests at low fill level

Figure 7. Heat-rate profiles of reacting system tests at high fill level

Figure 8. Pressure profile for venting of reacting system tests at centrepoint

Figure 9. Temperature profile for venting of reacting system tests at centrepoint

You might also like