CHEMICAL REACTION HAZARDS : AN INTEGRATED APPROACH

N. Gibson*, R.L. Rogers*, T.K. Wright*.

A procedure is presented that integrates process design,

hazard evaluation and safety measure specification and
implementation into a comprehensive system for the
avoidance of chemical reaction hazards.
KEY WORDS: Chemical Reaction Hazards, Reactor Safety,
Vents, Process Control.

1. INTRODUCTION

The manufacture of chemicals can produce hazardous situations if

chemical reactions are uncontrolled (1).

Although the number of major incidents arising from uncontrolled

runaway reactions is small relative to the number of process operations
carried out in the chemical industry, they do indicate the need for
constant vigilance and the adoption of systematic procedures to maintain a
safe manufacturing situation.

The essential stages of the procedure are

(1) Definition of the process/operating conditions/plant design.

(2) Characterisation of the process with respect to chemical reaction

hazards.

(3) Selection and specification of safety measures.

(4) Implementation and maintenance of safety measures.

It is an oversimplification to consider that the evaluation

procedure can be based solely on some sequence of chemical testing. The
characterisation of the process must take account of the other three
considerations and interact with them.

Furthermore the procedure must be applicable to hazard assessments

carried out (a) during initial (research) development work, (b) prior to
transfer to pilot plant scale (c) before full scale manufacturing is
established and (d) when modifications to the process/plant are undertaken.
Factors important at each stage are shown in Figure 1.

This review considers the interaction between process/plant

design, process characteristics, identification of risk and the
specification of safety measures to produce a procedure that will ensure
safety in manufacture.

* Hazard and Process Studies, Process Technology Department,

I.C.I. PLC (Organics Division), Manchester, England.

61
 IChemE SYMPOSIUM SERIES No. 102

2. FACTORS INFLUENCING THE PROCEDURE

The procedure must take account of:

(1) evidence from previous incidents - their cause and means of

preventing them.

(2) the intrinsic link between process definition and the validity of
hazard evaluation.

(3) the options that are available in the design of safety measures.

The chemical industry uses a diverse range of chemical reactions

in a multiplicity of plant types. Uncritical evaluation procedures can
lead to an open-ended commitment to process testing - an impractical
situation that does not serve the cause of operating safety. Each of the
above must be examined to establish the boundaries to an investigation that
will lead to safe manufacture without unnecessary constraints on operating
conditions, plant design or production.

2.1. Evidence from Incidents

Incidents occur when uncontrolled or runaway exothermic

reactions are allowed to develop. The factors leading to this situation
vary from incident to incident and the kinetics of the runaway reaction
depend very much on the chemistry of the particular process.(25) It is
however possible to identify general trends that highlight the essential
features of a chemical reaction hazard testing procedure.

Once the details of the exothermic activity have been identified,

it is rarely, if ever, found to involve novel, hitherto unknown, chemistry.
Frequently however it is a "side reaction" that does not play a significant
role in the normal process but that has been accelerated by the attainment
of temperature, time or concentration conditions outside the values present
in the normal process. The test procedure must be capable of detecting
and quantifying not only the main reactions, but also any side reactions
that could develop under all expected process conditions. This in turn
demands that the process description specifies the normal variations in
process conditions that can be expected in full scale manufacture.
Furthermore the plant manager must recognise that both the process
specification and the hazard evaluation have boundaries, if operating
conditions move outside these limits then hazardous situations may develop
that have not been identified in the hazard evaluation.

Although certain chemicals and reaction types (e.g. nitrations,

diazotisations) are recognised as generally more dangerous than others
(e.g. sulphonations) there are exceptions, it is necessary therefore to
examine every process under development. The testing procedure must be
sufficiently simple for a large number of processes to be assessed.

Once the sources of risk have been identified their quantification

must provide data that enable safety measures to be specified.
Furthermore these must be compatible with the engineering, production,
economic and commercial criteria for the process.

62
 IChemE SYMPOSIUM SERIES No. 102

The last, but not least important lesson to be learned from the
incidents is that the basis of safety must be recognised by the plant
operators, and must be implemented and maintained.

2.2. Process Definition and Hazard Evaluation

The degree of safety achieved from a hazard assessment is directly

related to the range of process operating conditions and plant design
features considered in its preparation. It is important therefore that
the process definition be sufficient to produce the desired level of safe
operation.

Four levels of process definition can be identified.

Level 1: Process Definition with Fixed Parameters

Process descriptions, particularly at early stages of development,

often include specific values for such parameters as temperature, reactant
concentrations, time, etc. The hazard assessment can only cover a process
operating with these fixed values.

It will not consider variations in the process conditions that

would be allowed to occur in full scale operation and not be considered to
be abnormal (e.g. small changes in temperature, concentration, batch hold
times). Project definition at this level will rarely produce a hazard
assessment giving an acceptable level of safety.

Level 2:
Process Definition Including Normal Variations in Operating Parameters

This level of process definition recognises that in actual

operation the process conditions will vary. It defines the range of
values over which each parameter will be permitted to change without
corrective action being taken i.e. the process as normally operated.

Certain of these variations are well recognised e.g. the

temperature of a batch varying by + 10°C. Others are less well recognised
and incidents have occurred because the significance of variations in basic
parameters that are an accepted part of the manufacture are not included in
the process definition and consequently not covered by the hazard
evaluation. For example, the hold time at elevated temperature for
analysis of product quality that is normally one hour extending to twelve
hours at weekends. This is a situation in which "side reactions"
virtually dormant over one hour at the elevated temperature can accelerate
exponentially to a dangerous level in the increased time.

Hazard assessment based on level 2 process definition should

adequately protect a process operating normally.

Level 3:
Process Definition Including Non-Specific Fault Conditions

Certain failure situations, although not common, are known to

occur in chemical processing. Examples are agitator failure, loss of
cooling, fracture of an internal coil. These are not specific to

63
 IChemE SYMPOSIUM SERIES No. 102

individual processes and the effect of them can be included in the hazard
assessment without additional detailed process definition.

Unless the plant design is such as to eliminate them then the

effect of such failures on process stability and the consequences of any
subsequent runaway situation has to be included in the hazard evaluation.

Level 4:
Process Definition Including All Conceivable Abnormal Situations

It is possible to postulate a large number of abnormal conditions

that could conceivably cause exothermic activity. Examples are:
contamination of the batch by a reactive chemical used in a neighbouring
process, variations in raw material quality, the possibility of a general
fire overheating the reactor. Unconstrained, this approach can lead to an
open-ended commitment to testing.

The techniques of Hazard and Operability that can provide guidance

as to the probability and consequence of any abnormal situation are a means
of determining the additional abnormal situations that need to be
considered in the hazard assessment.

It is considered that Level 3 is the minimum standard that leads

to an acceptable level of safety in the majority of processes.

The process/plant definition should include

(a) definition of the process/plant conditions including all

known/expected variations in the process parameters (e.g.
temperature ranges, concentration variations, hold times, etc.).

(b) details of operations (e.g. cooling, agitation, pumping, etc.)

that are not protected by high integrity trips.

Where necessary this assessment should be expanded to cover the

maloperatlons, etc., that Hazard and Operability Studies indicate could
realistically occur in the process.

2.3. Data Required to Design Safe Systems

Safety can be achieved by:

(a) Preventative Measures: process control, instrumentation, etc., to

prevent the initiation of an uncontrolled reaction.

(b) Protective Measures: containment, reactor venting, crash cooling/

drown-out, or reaction inhibition to protect against the
consequences of a runaway reaction.

A simplistic representation of chemical activity is shown in

Fig.2.

AB : normal process - balance maintained between heat generated

by chemical reaction and plant cooling.

64
 IChemE SYMPOSIUM SERIES No. 102

B : start of uncontrolled reaction.

BC : runaway reaction - process conditions cannot maintain

equilibrium situation.

For safety to be based on process control the reaction condition

must remain in AB. A hazardous situation will occur if process conditions
(e.g. reactant concentrations, feed rates, temperatures, hold times) are
intentionally or unintentionally changed to an extent that condition B is
attained and a runaway reaction is initiated.

The procedure must provide information on

(1) sensitivity of the process to changes in conditions e.g. safety

margin between plant operating temperature and exotherm onset
temperature.

(2) minimum temperature under plant conditions at which uncontrolled

exothermic activity will start.

Safety based on protection requires data on the "kinetics" of the

runaway reaction. Containment requires only the maximum pressure generated
by the reaction. For reactor venting, crash cooling and reaction
inhibition, both the rate and magnitude of pressure, temperature and gas
evolution changes need to be measured. Furthermore they must be measured
using techniques that can simulate the worst runaway situation that can
occur on the plant. If agitation ceases for example, the reactants layer
and then the agitator is restarted; the reactants are intermixed at
concentrations and rates not simulated in Figure 2. The rates of increase
in pressure, temperature and gas evolution can be an order of magnitude
greater than that predicted by a technique that merely initiates runaway by
continuously increasing the temperature of the reaction. The
experimentation used to characterise a process must take account of the
basis for safe operation that can be applied to the process and the plant.

3. CHARACTERISATION OF PROCESS FOR CHEMICAL REACTION HAZARDS

A typical investigative sequence is shown in Figure 3.

An essential first stage is to exclude from manufacture any

materials/reaction masses/residues that could have detonation or
deflagration explosive properties.

Once this has been done the procedure

(a) examines the normal process situation

(b) determines the initial exotherm temperature and

(c) quantifies the characteristics of the runaway situation.

3.1. Explosibility Screening

The following procedure has been used to exclude detonating and

deflagrating explosive materials.

65
 IChemE SYMPOSIUM SERIES No. 102

3.1.1. Consideration of Chemical Constitution

Groups such as nitrate ester, aromatic nitro and nitramine are

closely linked with explosibility; azo, azide, nitrozo, peroxide and
acetylene groups can form part of explosive structures. No all embracing
guide can be provided but materials containing these or similar groups
should be tested for detonation and deflagration properties.

3.1.2. Calculation of Oxygen Balance

For oxidation reactions the Oxygen Balance enables a material

to be compared to those of known explosive potential.

The basic equation is

Typical values for materials of recognised instability are

nitrobenzene (-163), glycerol trinitrate (+3.5), dinitrotoluene (-114).
It is recommended that materials with an oxygen balance more positive
than minus 200 should be tested for explosibility.

3.1.3. Explosibility Tests

A number of tests, developed by the explosives industry, to

identify materials with detonating and deflagrating explosive properties
are available. The most appropriate for the explosibility screening
of chemicals are Pressure-Time Test, Trauzel Lead Block Test and Koenen
Tube Test. These have been reviewed by Cutler (2).

Materials giving a positive result in these tests normally require

specialised manufacturing facilities.

Certain materials not classified as explosives can decompose very

violently when subjected to localised heating. Gibson and Harper (3)
have shown for example that, under confinement, the propagating decomposition
of o-anisidine nitrate can generate pressures up to 500 bar at a rate of
5-6 k.bar s - 1 .

Wright and Butterworth (4) have developed a small scale (3 g)

test that enables pressure effects to be determined for reaction masses and
powders.

A Carius Tube (Figure 4) containing the sample, is heated by

placing it inside a circular oven. The temperature of the sample can be
measured by the bottom entry thermocouple and the pressure by a transducer
mounted on top of tube but outside the oven. Predecomposition gases and
volatiles can be released via a side arm.

The teat can detect compounds that undergo high rate decomposition
and quantify the conditions necessary to support high rate decomposition.

66
 IChemE SYMPOSIUM SERIES No. 102

It can measure

(1) the onset temperature for exothermicity (no allowance for scale
factors).

(2) the onset pressure - provided predecomposition gases and volatiles

are vented prior to onset.

(3) rate of decomposition as indicated by rate of pressure rise. The

measured pressures depend on test cell volume and geometry.
Dangerous pressures are avoided by allowing the tube to burst
safely inside the oven enclosure.

Temperature and pressure traces for dinitrotoluene, a material not

classified as an explosive but capable of high rate decomposition are shown
in Figure 5.

On the 3 g scale with a 2°C/min ramp rate self heating is evident

at ca. 250°C but the high reaction rates associated with propagative
decomposition do not occur until the sample temperature is well above the
atmospheric boiling point of 300°C.

Using normal rates of data capture the high rate event starting
at 350°/120 psi is not recorded in Fig. 5a. The system is also provided
with a transient recorder triggering at 300 psi to capture the high rate
event. For a 0.1 sec event time the system is arranged to record 0.09
sec. before 300 psi and 0.01 sec. after 300 psi as shown in Fig. 5b. Rise
time between 200-400 psi is 6 millisecs.

Applying a heat source to dinitrotoluene without confinement will

not produce propagation/deflagration - the material will vaporise.
Propagation would be expected (and has occurred) in heated pipelines which
can confine the material.

Experience has shown that compounds that propagate decomposition

(detonation/deflagration) are characterised in this test by a rise time of
less than 50 m.s. between 200-400 p.s.i.

The test identifies materials that have detonating or deflagrating

properties akin to explosive materials, identifies materials capable of
high rate decomposition and indicates semi quantitatively the consequences
of decomposition. Materials with uncontrollable decomposition
characteristics can be excluded from manufacture or kept in the more stable
damped form.

3.2. Characterisation of the Normal Process

Process development requires a knowledge of:

(a) the kinetics (particularly the equation linking reaction rate with
concentration) of the desired reactions.

(b) the optimum reaction conditions to maximise yield/quality.

(c) heats of reaction and rate of heat generation.

67
 IChemE SYMPOSIUM SERIES No. 102

(d) plant cooling characteristics (e.g. heat loss from plant,

evaporation, reflux, etc.).

(e) gas evolution rates and scrubber capacity.

Definition of the normal process is however insufficient to ensure

safe manufacture. Account must also be taken of the effect of
manufacturing change (e.g. variation in reactant concentrations, addition
times, temperature profiles, hold times, etc.) on heat generation and gas
evolution. The margin of safety between the "normal process" conditions
and those that initiate a runaway situation must be established. The
effect of potential mal-operations identified in Hazard and Operability
Studies on reaction safety has to be established.

Isothermal Reaction calorimetry (5) has been developed to

simulate in the laboratory the behaviour of full scale isothermal semi-
batch chemical processes under normal and maloperation conditions.
Adiabatic calorimetry can be used to simulate "all in" batch processes
with self heating (6).

The Reaction Calorimeter has to satisfy a number of complex

requirements:

(1) the heat generation data must be obtained whilst simulating the
full scale reactant addition rates, batch temperature and time
and the desired process conditions (e.g. stirring, distillation,
boiling under reflux, etc.).

(2) other sources of heat flow must be included in the heat equation
e.g. energy input from the stirrer, energy loss in a condenser.

(3) the effect of changes in physical properties during the reaction

(e.g. viscosity, specific heat, precipitation).

(4) the effect of change in cooling through the reactor wall due to
changes in the wetted area and inner wall fouling.

(5) the heat flux has to have a detection limit of the order of
+0.1 Watts.

Reaction Calorimeters and their use to prescribe safe

manufacturing conditions have been described by Regenass (5),Riesen (7),
Hub & Kupr (8), Stockton et Alia (9) and Wright and Butterworth (10).

All the techniques provide similar information but the methods

used to obtain it differ in detail.

In the design originating in Ciba-Geigy and now marketed by

Mettler Instrumente AG a stirred tank reactor is surrounded by a jacket in
which a heat transfer fluid is circulated at a very high rate. A cascaded
controller adjusts the temperature of the circulation loop so that the heat
transfer through the reactor wall equilibrates the heat evolution in the
reactor. Injection of thermostatted hot or cold fluid is used to adjust
the temperature in the loop.

68
 IChemE SYMPOSIUM SERIES No. 102

The rate of heat transfer (which equals the rate of heat

evolution) is related to the observed temperature difference T between
the jacket fluid and the reaction mixture by the equation

where the calibration factor Fc is the product of U, the overall heat

transfer coefficient, and A, the active ( = wetted) heat transfer area.
Because both A and U depend on the reactor contents and on the stirring
conditions, specific calibration is required. This is done by producing a
known heat input rate to the reaction mixture by means of an electric
heater. The need of frequent calibration is of some inconvenience as
compared with heat balance calorimeters. On the other hand, the method
chosen permits the use of an uninsulated glass reactor and thus allows
visual observation of phase changes, colour changes and mixing conditions.
This is a distinct advantage for process development work.

In the design described recently by Stockton et Alia (9) all

except the base of the reactor is surrounded by an "adiabatic shield";
this maintains the temperature of the air space surrounding the reactor at
the same temperature as the reactants in the inside of the vessel. Heat
flux only occurs through the flat base of the reactor - this minimises
problems associated with changes in wetted area.

A second problem can arise with the design of a sufficiently fast

response cooling system. Wright and Butterworth (10) have overcome this
by maintaining a constant temperature difference between the calorimeter
contents and the cooling jacket. Essentially the reactant mass is heated
by an electrical heater. As the reaction generates chemical power the
electrical power is correspondingly reduced. Measurement of the
electrical power reduction provides quantitative data on chemical power
generation rates.

A fourth variation on design has been described by Hub and

Kupr (8) and defined as the Heat Balance Method. The test cell
incorporates a condenser and reactions under reflux can be studied.

No one technique is intrinsically superior to all the others.

Each has its advantages and disadvantages. However all can be used to
provide the data on the normal process that is required to evaluate
potentially hazardous process conditions.

A classical experiment to demonstrate the use of isothermal heat

flow calorimetry is the nitration of toluene. In the example shown in
Fig.6 nitrating acid containing 1.14 mol nitric acid is added uniformly
to toluene 0.9 and at 60°C. The calorimeter indicates instantaneous power
output during the addition and work off, for comparison with plant cooling
capacity. By measurement of the area under the power/time curve the total
heat of reaction is calculated to be 35 Kcal/ml. toluene. Power output
profile in this particular case is square indicating rapid reaction and
freedom from significant accumulation problems provided the system is
agitated.

69
 IChemE SYMPOSIUM SERIES No. 102

3.3. Identification of Minimum Temperature for Runaway Reaction

Safety can be achieved by ensuring that a reaction mass does not

attain the temperature at which an uncontrolled reaction is initiated.

The minimum temperature for runaway reaction depends not only upon
the kinetics of the reaction but also on the rate of loss of heat that will
occur from the full scale reactor. Heat loss from batch reactors depends
on the size of the reactor, agitation, condition of the vessel jacket.
Typical values for stirred vessel with filled jackets are
0.04-0.08 W . k g - l o C - 1 for a 5000 gallon vessel.

Experimental techniques to achieve test sensitivities appropriate

to these values require accurate, stable temperature control and high
levels of detector sensitivity.

Laboratory testing has developed on two levels:

(1) screening tests that enable a large number of reactions to be
investigated quickly with respect to the possibility of exothermic
activity and the approximate value of the minimum temperature.

(2) adiabatic tests that simulate plant scale heat loss conditions and
provide data on self-heating rates directly relevant to full scale
manufacture.

3.3.1. Screening Methods

Ideally the screening test should satisfy the following criteria:

(1) be capable of monitoring starting materials, samples of the

reaction mass as the reaction progresses, final reaction mass,
isolated final product and distillation residues.

(2) provide data from measurements on the small samples (grammes)

available during product development.

(3) clearly identify exothermic activity.

(A) provide some guidance as to the size of the exotherm - ideally a

closed test cell to avoid effect of evaporation.

(5) provide some guidance on rate of heat generation.

(6) detect production of gas/volatiles.

(7) detect high rate of events.

(8) detect induction period phenomena leading to time dependent onset.

3.3.1.1. DSC/DTA

DSC/DTA are well known thermo-analytical techniques. When used

for the study of exothermic decomposition the following lead to certain
limitations:

70
 IChemE SYMPOSIUM SERIES No. 102

(1) test condition is essentially isothermal whereas under plant

conditions, decomposition occurs in a situation corresponding
more to the adiabatic limiting case.

(2) sensitivity of the test used in the traditional manner is

relatively low (1-5 W k g - 1 ) .

(3) measured onset temperature is a function of the sample heating

rate. The temperature determined for p-nitroanlline increased
from 210°C to 300°C when the heating rate moved from
0.1 K min - 1 to 5 K min - 1 (11). Duch et alia (12) have
confirmed this effect.

(4) non-sealed cells can produce errors due to evaporation losses.

(5) small sample size can lead to it being unrepresentative of plant

material.

(6) violent decompositions can cause damage to the equipment.

The limitations and advantages of DSC/DTA are discussed in detail

by Hentze (11), Schulz, Pilz and Schacke (13) and Duch et alia (12).
Certain of the above limitations can be overcome by the use of sealed
pressure cells and by control of the sample heating mode (low rates,
isothermal "stepping" of the sample, etc.) but there still remains the high
rate of heat loss from the sample, relatively low measurement sensitivity
and absence of gas evolution data. Nonetheless these techniques do
provide a display of exothermic activity and provide some indication of
heat generation characteristics under isothermal conditions.

3.3.1.2. Linearly Ramped Closed Vessel Tests - CV Test

These tests are variants on the ASTM 76-79 test. The sample, in
a container fitted with an internal thermocouple and transducer, can be
heated with a linear ramp up to 500°C. Fenian (14) comparing this test
with the ARC found that the onset temperature for the CV Test lay in the
range 59°C above to 8°C below the comparable ARC result with a mean
value of 21 o C above the ARC results.

3.3.1.3. I.C.I. Sealed Tube Test

The I.C.I. test uses a glass Carius tube (fitted with re-entrant
thermocouple and pressure transducer (Figure 4)). In the screening test
10-20 g sample are sealed in the tube which is then heated at 2°C min - 1
up to 400°C or up to a set pressure cut off. Temperature and pressure are
monitored continuously. The onset temperature can be read directly from
the trace - Figure 7. The test sensitivity is 3-10 W kg-1 .

The test is primarily used as a screening test - it provides an

indication of exothermic activity, the minimum temperature under the test
conditions, vapour pressure-temperature data.

71
 IChemE SYMPOSIUM SERIES No. 102

3.3.1.4. Commercial Calorimeters in Programmed Mode

Commercial adiabatic calorimeters (e.g. ARC, Sikarex) can carry

out screening tests.

The Heat-Wait-Search operational logic of the ARC (15) is shown in

Figure (8). The sample is heated to a desired starting temperature and
held for a time to establish thermal equilibrium. A rate search is
performed to detect exothermic activity. If none is detected at a
-1
selected threshold level (normally 0.02°C m i n ) the temperature is
increased and the sequence repeated. The temperature at which exothermic
activity is eventually detected is a measure of the onset temperature.

The screening procedure for the Sikarex Calorimeter employs a

linearly ramped temperature scan.

3.3.2. Adiabatic Test Methods

The limitations of the screening tests with respect to simulation

of the plant conditions can be overcome by the use of adiabatic test
methods.

In the simple tests the temperature monitor may underestimate the

heat generation due to
(1) heat loss from the sample to its surroundings - this can be
minimised or eliminated by reducing the temperature difference
between the sample and its environment.

(2) heat loss from the sample to the test cell. A parameter phi is
used to characterise this effect.

phi= Thermal Capacity of Sample + Thermal Capacity of Test Cell

Thermal Capacity of Sample

This can be minimised by reducing the mass of the container or by

using a relatively large sample to reduce the ratio of the thermal capacity
of the test cell to sample.

Two techniques are used to obtain data under adiabatic or near

adiabatic conditions.

3.3.2.1. Adiabatic Calorimeters

The adiabatic condition is achieved by using the sample

temperature as the "set point" in the instrumentation. Once exothermic
activity has been detected the temperature of the oven is increased to
match the temperature of the sample. This minimises heat loss from the
sample/container to the environment but account still has to be taken of
the phi factor. An analysis of the importance of phi has been given by
Townsend (15) and recently discussed in the ARC Newsletter (16). For
direct simulation of plant situations phi values of 1.0-1.5 are required.
In practice phi values up to 3 are used and the experimental results
adjusted mathematically to correspond to a phi = 1 situation. Changes in
reaction mechanisms are usually small at the levels of activity present

72
 IChemESYMPOSIUM SERIES No. 102

when the exothermicity is first detected so the correction is usually

reasonable. It can be invalid however if complex chemical or mass
transfer mechanisms are involved (17).

3.3.2.2. Dewar Techniques

The use of Dewar vessels to simulate low heat loss situations are
well established (18).

The construction of the Dewar vessel leads to low heat loss and
phi factors. Rogers and Wright (6) have developed a simple adiabatic
Dewar system that further reduces heat loss and reduces the need for
accurate temperature control. The phi factors are in the range 1.05-1.5
depending on the thermal capacity of the reaction mass. The inherent low
heat loss enables temperature drifts as low as 0.1 K h r - 1 (2 K per day)
to be obtained from a control system that maintains a sample/oven
temperature differential within 1 K. To achieve corresponding stability
in small scale adiabatic calorimeters a drift free temperature differential
within 0.01 - 0.1 K has to be achieved.

3.3.3. Comparison of Onset Temperatures obtained from

Different Tests

Typical sensitivity data for commercial equipment has been given

by Schulz et alia (13).

Onset temperatures obtained from the different test methods are

available in the literature.

73
 IChemE SYMPOSIUM SERIES No. 102

Comparisons of the above type show that, provided a slow

temperature scan rate is used, the screening tests satisfactorily indicate
the possibility of exothermic activity and provide guidance on the onset
temperature. It must be emphaised that due to the limitations of the
screening tests a minimum safety factor of 60-100°C be required. If the
temperature differential between operating and test result is less than
this then the onset temperature must be confirmed using an adiabatic
method.

3.4. Characterisation of the Runaway Reaction

Characterisation of the runaway reaction requires data on the rate

and magnitude of temperature, pressure and gas evolution changes that could
occur in the plant situation. The pattern of a runaway reaction will tend
to take one of three forms:

(a) an accelerating reaction that is linked by a continuous function

to the normal process reaction rates as displayed in Figure 2,
e.g. caused by inadequate cooling.

(b) a step change in which the reaction increases markedly over a

short period of time e.g. caused by restoration of agitation in a
two phase system.

(c) a combination of (a) and (b).

74
 IChemE SYMPOSIUM SERIES No. 102

The progression of the reaction will be determined by the fault

condition that causes it.

The experimental techniques must not only be able to simulate the

normal process and plant operations but also be sufficiently flexible to
allow moaloperations (e.g. loss of cooling, loss of agitation, incorrect
reactant additions, etc.) to be simulated at any stage of the process.

Adiabatic calorimetric techniques must be used.

3.4.1. Commercial Calorimeters : ARC

Commercial calorimeters as typified by the ARC can be operated

under adiabatic conditions. Data can be generated on adiabatic reaction
pressure and temperature parameters, kinetic data, reaction rate constants
and time to maximum reaction rate (15). Gas evolution cannot be directly
measured but can often be inferred from pressure data. Wilberforce (20)
and Coates (21) have described the application of the ARC in thermal
hazards evaluation.

It can successfully provide adequate data on the parameters listed

above. A practical limitation however is the small size of test cell
(10 g maximum) that precludes the addition of reactants as the reaction
proceeds.

3.4.2. Adiabatic Dewar Techniques

Stainless steel Dewars of the type described by Rogers and Wright

(6) can be used to measure directly pressure, temperature and gas evolution
changes under heat loss conditions that simulate the plant situation.

The larger sample and test cell access make it possible to

manipulate the process conditions as the reaction continues. The data are
directly applicable to batch reactions up to 5000 gallons. The basic
pressure/temperature data are exemplified by that for Methanol/Acetic
Anhydride shown in Figure 9. The temperature-time curves can be rapidly
analysed to yield thermodynamic and kinetic data for simulation
purposes (6).

3.4.3. V.S.P. Apparatus

A recent addition to commercially available equipment is the

V.S.P. Apparatus (22). This, produced as part of the D.I.E.R.S. project
(23), is intended to produce the data required to design reactor relief
systems.

It consists essentially of a 120 ml metal test cell that is heated

by the oven enclosure and an electric heater wound on its outer surface.
The latter reacts to the temperature of the sample and minimises heat loss.
The phi factor of the cell is approximately 1.5. A novel feature is the
pressure control system that produces a pressure outside the cell equal to
that produced by the reaction inside the cell. This enables pressure
reactions to be studied in a relatively weak test cell (low thermal
capacity).

75
 IChemE SYMPOSIUM SERIES No. 102

When used with a closed cell, the equipment is essentially an

adiabatic calorimeter that can provide information on the reaction rate
under runaway conditions, the maximum pressure and temperature, the overall
heats of reaction and the vapour pressure/temperature relationship. An
open cell can be used to provide information specific to the design of
relief systems - flow behaviour of the discharging reaction mass,
identification of tempered and non-tempered reactions.

A number of organisations are testing this equipment but few

results are available in the open literature. In this symposium Gibson,
Maddison and Rogers (24) report comparative testing that indicates that
thermal data from the V.S.P. apparatus is in reasonable agreement with that
from Dewar Calorimetry.

The character of a runaway reaction depends on a number of

interacting factors (e.g. "kinetics" of the runaway reaction, autocatalytic
effects, influence of volatiles, heat loss from reactor, effect of
pressure, etc.). In assessing a process the "worse case" conditions must
be identified and simulated in the laboratory tests. No one system gives
unequivocal data that can be used without interpretation but provided this
is correctly done then data can be obtained that enables protective systems
to be designed.

4. SELECTION OF SAFETY MEASURES

Safe operation can be based on

(A) Process control that prevents conditions being attained under

which uncontrolled exothermic reaction will be initiated.

or (B) Process control to minimise the probability of a runaway

reaction combined with protective measures should such a reaction
occur.

Options are:

(1) Process control + containment.

(2) Process control + reactor venting.

(3) Process control + crash cooling/drown out.

(4) Process control + reaction inhibition.

The most appropriate safety measure depends on process detail -

including toxicity of products, magnitude and rate of the runaway
parameters and the practicality of implementing and maintaining the
safety measures.

Critical technical considerations for each are:

4.1. Process Control

- definition of minimum temperature at which uncontrolled

exotherm will start under plant conditions.

76
 IChemE SYMPOSIUM SERIES No. 102

- safety margin between operating temperatures and exotherm

temperature.

- monitoring and control systems to maintain temperature in

the safe region.

- maintenance of temperature should agitation or cooling fail -

e.g. stop feed of reactant, use solvent that boils at safe
temperature.

-control sources of risk external to process - e.g. addition

of wrong materials.

-specification of lower temperature limit to prevent

accumulation.

-reaction of two phase systems to agitation failure.

4.2. Process Control and Reactor Venting

-definition of worst case - i.e. conditions leading to maximum

rate of exothermic activity.

-establishment of "kinetics" of the runaway reaction.

- nature of discharge material - gas, liquids, solids.

-methods for calculating reactor vent area and discharge system

for the vented materials.

-safe discharge area - flammable and toxic hazards - dump tanks.

4.3. Process Control and Crash Cooling/Drown Out

- rate of temperature rise/heat generation after runaway

detected.

- time to hazardous pressure.

- availability of compatible cooling medium.

- relative thermal capacities of reaction mass and cooling

medium.

- plant design/operation to intermix reaction mass and cooling

medium and stop temperature rise before maximum permissible
pressure is attained.

4.4. Process Control and Reaction Inhibition

- availability of compatible reaction inhibitor.

- time to hazardous pressure.

77
 IChemE SYMPOSIUM SERIES No. 102

inhibitor efficiency.

plant design and operation to intermix reaction mass and

inhibitor and stop temperature rise before maximum permissible
pressure is attained.

In addition to the technical considerations, the selection of the

most appropriate basis for safe operation must take account of:

(a) acceptability to the engineering and manufacturing functions with

respect to their compatibility with the design, construction,
operation, maintenance and economic requirements of the process.

(b) the essential features of the safety measures must be understood

by the manufacturing personnel.

(c) the safety measures must be fully implemented and maintained.

(d) the boundaries of the safety evaluation and the effect of changes
in plant construction or process operation.

5. THE INTEGRATED APPROACH - CONCLUSION

The evaluation of chemical reaction hazards and the specification

of realistic and practical measures that ensure manufacturing safety can
best be achieved by an approach that integrates process and plant
definition, process characterisation for the normal and runaway situation,
the selection, design, implementation and maintenance of safety measures.

The procedural details will depend on the infrastructure of the

organisation but the essential details for initiation, evaluation,
implementation and monitoring are shown in Figure 10. Fully implemented it
will ensure manufacturing safety.

Acknowledgment

The authors wish to acknowledge the value of discussions with

their colleagues in Hazards and Process Studies Group, I.C.I, in the
development of the ideas presented in this paper.

References

1. P.F. Nolan, J.A. Barton, Runaway reactions in batch reactors.

I. Chem. Eng. Symp. Ser. No. 85 p.13 (1984).

2. D.P. Cutler, Hazards in the Process Industries. I. Chem. E. Symp.

Ser. No. 97 p.133 (1986).

3. N. Gibson & D.J. Harper. I. Chem. E. Symp. Ser. No. 68 p.3/R:l

(1981).

4. T.I. Wright & C.W. Butterworth - to be submitted to next Chem.

Eng. Symp.

78
 IChemE SYMPOSIUM SERIES No. 102

5. W. Regenass A.C.S. Symp. Ser. 65 p.37-49 (1978).

6. T.K. Wright & R.L. Rogers. I. Chem. E. Symp. Ser 97 p.121 (1986).

7. R. Riesan "Exothermicity of the Desired Reaction - Reaction

Calorimetry". Conf. Control & Prevention of Runaway Chem. React.
Haz. Amsterdam, 11/12 Nov. 1986. Publ. IBC Tech. Serv.

8. L. Hub, T. Kupr. "Heat Balance Calorimetry & Automation of testing

procedures". Ibid.

9. G.W. Stockton, E. Donoghue, S.J. Ehrlich, D.H. Chidester,

R.S. Wayne, North American Thermal Anal. Conf. Sept. 1986.

10. T.K. Wright & G.W. Butterworth - This symposium.

11. G. Hentze. Thermochimica Acta. 72. 127-138 (1984).

12. M.W. Duch, K. Marcali, M.D. Gordon, C.. Heusler, G.J. O'Brien.
Res. & Dev. Divn. Publn. No. 582. E.I. Dupont De Nemours.

13. N. Schulz, V. Pilz, H. Schacke. I. Chem. E. Symp. Ser. 82 Bl

(1983).

14. W.J. Fenian. Plant/Operations Progress Vol. 3, No. 4 p.197

(1984).
1-3
15. D.I. Townsend, J.C. Tou, Themochimica Acta 37 ° (1980).

16. ARC Comments. Columbia Scientific Industries, Vol.4, No.3,

Nov. 1986.

17. D.I. Townsend. "Use of the ARC to study thermally unstable

chemicals and runaway reactions". Conf. on Control & Prevention
of Runaway Chemical Reaction Hazards. Amsterdam 11/12 Nov. 1986.
Publ. IBC Tech. Serv.Ltd.

18. T. Grewer. "Use of Dewar Calorimetry". Ibid.

19. N. Gibson. Chem. & Ind. 19 Mar. p.209 (1984).

20. J.K. Wilberforce, I. Chem. Eng. Symp. Ser.85 p.329 (1984).

21. C.f. Coates, W. Riddell. Chem. & Ind. 7 Feb. p.84 (1981).

22. Diers report: Bench Scale ERS Sizing Tools: Equipment Details
& Test procedures: FAI/84-4: Revised May 1984. A.I. Chem.E.

23. H.G. Fisher, Chem. Eng. Prog. 81 No. 8 Aug 1985.

24. N. Gibson, N Maddison & R.L. Rogers.

This symposium.

25. L. Bretherlck "Handbook of Reactive Chemical Hazards".

Butterworths (1985).

79
IChemE SYMPOSIUM SERIES No. 102

80
IChemE SYMPOSIUM SERIES No. 102

81
IChemE SYMPOSIUM SERIES No. 102

82
IChemE SYMPOSIUM SERIES No. 102

83