Você está na página 1de 10

0263–8762/06/$30.00+0.

00
# 2006 Institution of Chemical Engineers
www.icheme.org/cherd Trans IChemE, Part A, July 2006
doi: 10.1205/cherd.05221 Chemical Engineering Research and Design, 84(A7): 543– 552

REVIEW PAPER

SAFETY AND RUNAWAY PREVENTION IN BATCH


AND SEMIBATCH REACTORS—A REVIEW
K. R. WESTERTERP1 and E. J. MOLGA2
1
Ramon Llull University, Institut Quimic de Sarriá, Barcelona, Spain
2
Warsaw University of Technology, Chemical and Process Engineering Department, Warsaw, Poland

I
n this paper a review of recent developments on safety and runaway prevention in batch
and semibatch reactors is given. To prevent thermal runaways in chemical reactors
usually three lines of defense have to be considered: choice of the right operating con-
ditions, early warning detection system and system to handle running away reactions. Because
of its too specific nature, the last line of defense is not discussed in this review, while the
former two are considered in detail. Firstly the capability and applicability of the reaction
calorimeter for safety assessments are reviewed, then fundamental safety criteria for safe
operation of batch and semibatch reactors are reported and discussed. A review of the most
useful and efficient methods for a correct choice of safe operating conditions is split into
two cases: one, when the reaction kinetics are known, and when kinetic information is not
available. Also several promising recent developments of an early warning detection
system for the on-line detection of unexpected situations leading to a thermal runaway are
presented.

Keywords: safety of chemical reactors; criteria of thermal runaway; reaction calorimetry;


early warning detection of runaway; batch and semibatch reactors.

INTRODUCTION of the reactions executed daily do not have a heat effect


large enough to cause a runaway, so that the chance of a
In this paper we will review the developments in the last runaway, nevertheless, is not hypothetical.
25 years in the study of safety and runaway prevention of To prevent reactor incidents and runaways in general
batch and semi-batch reactors. We will do so from the three lines of defense have to be considered in the design
view point of the science of reaction engineering and there- and operation of a (semi-)batch reactor. The first line of
fore, we will leave a number of relevant studies untouched defense is the choice of the right operating conditions.
that belong to the fields of chemistry, mechanical engineer- This refers to the degree of mixing, the cooling capacity,
ing and/or organizational sciences. In Table 1 we give stat- the choice of the coolant temperature, the dosing rate of
istical data on the prime causes of batch reactor incidents as the reactant(s) and so on. In the course of the years many
they have been reported in three different studies (Verwijs, papers have been written on the choice of the operating
1994). We can observe that at least more than 60% of the conditions; moreover, we should realize that the studies
incidents could probably be avoided, if a proper design of on tubular reactors also can be translated to batch reactors.
the reactor plant and choice of the safe operating conditions Most recommendations are based on the knowledge of the
has been performed. The number of incidents per month kinetics of the reaction. Regretfully batch reactors are
seems to increase in time according to Table 1, this is not almost exclusively used in the fine chemicals industry,
the case in reality: the numbers of reports and methods of where usually time and money are serious constraints for
reporting have increased and improved very much over the costly and time-consuming study of reaction kinetics.
the period from 1960 to 1990. We may feel that a For highly exothermic reactions the use of inaccurate kin-
number of two incidents per month for a few thousands etic information often leads to a too prudent selection of
of reactors considered in the surveys leads to a rather operating conditions with low reactor productivities. In
small percentage of incidents. However, more than 80% the last decennia the Reaction Calorimeter has become
widely accepted as a tool for the selection of operating con-

ditions, so that we will also dedicate some space to the dis-
Correspondence to: Professor E. J. Molga, Warsaw University of Tech-
nology, Chemical and Process Engineering Department, ul. Warynskiego
cussion of this piece of very useful laboratory equipment.
1, 00-645 Warsaw, Poland. The second line of defense is an early warning detection
E-mail: molga@ichip.pw.edu.pl system. As is clear from Table 1, a correct choice of

543
544 WESTERTERP and MOLGA

Table 1. Prime causes of batch reactor incidents (Verwijs, 1994). entitled Reaction Calorimetry, in a series of contributions
1962–1984 1962–1987 1986–1990 the expanding role of reaction calorimetry is demonstrated
and its theoretical principles, methods of modelling as
References well as areas of application are indicated and widely
discussed—see a.o. Landau (1996), Rowe (1996) and
Nolan and Barton and Zaldivar et al. (1996). Following Landau (1996), three
Incident cause Barton (1987) Nolan (1989) Etchells (1993) basic areas of application of reaction calorimetry can be
Thermo-reaction 21.4% 20.1% 14.8% distinguished: Process Safety, Process Development and
chemistry Basic Research. All of them contribute significantly to
Raw material quality 7.9% 8.9% 9.8% the evaluation of safety aspects of batch and semibatch
Maintenance/other 22.3% 21.3% 22.2% reactors by thermal and kinetic analysis of chemical reac-
factors
Temperature control 22.2% 18.9% 13.9% tions. In a more recent review by Zogg et al. (2004), the
Loss of 9.5% 10.1% 13.1% most important calorimetric principles are presented and
mixing/agitation discussed—including heat flow, heat balance, power
Mischarging of 16.7% 20.7% 26.2% compensation and the Peltier principle—as well as a
reactants classification and a comparison of the available reaction
Incident rate 0.46 0.54 2.03
(#/month) calorimeters. Among numerous relevant papers on the
contribution of reaction calorimetry to safe operation of

Data obtained from Health and Safety Executive records (HSE, UK). chemical reactors, e.g., the following publications should
be mentioned: Ubrich et al. (2001) and Gesthuisen et al.
(2005). A few commercial applications of standard heat-
operating conditions is not enough for the safe running of a flow reaction calorimeters are available on the market for
reactor section in the plant. Human errors like mischarging both laboratory and research purposes. To avoid promoting
of reactants, maintenance troubles, agitator break down and any particular producer the reader is referred to the papers
so on always may occur. Therefore, it would be a god send cited above as well as to the webpage HarsNet (HarsBook).
if we could provide industry with a piece of equipment or a In principle each heat-flow reaction calorimeter
procedure, that can give an early warning to the operator to measures the heat evolution with time of a reaction under
anticipate a runaway to occur within a shorter period, say isothermal, isoperibolic or adiabatic conditions. A sche-
for example, 20 to 200 min in advance of reaching the matic diagram of a typical bench-scale reactor is shown
maximum runaway temperatures. In this field in the last in Figure 1.
ten years great advances have been made. The principles of reaction calorimetry are based on the
The third line of defense is a suitable system to handle determination of mass and energy balances carried out
running away reactions. Such systems depend on the reac- over the reaction vessel. In the general case—taking into
tants and reactions at hand. If vapors are developed, we account the behaviour of a semi-batch reactor, even with
need a blow down system with catch tanks; or in another evaporation of solvent or reactants—the molar balance
case, reactor contents must be discharged to refrigerated for each reactant in the reacting mixture has to be formu-
cellars loaded with ice to stop the reaction by freezing. In lated, including the molar flow rates due to evaporation
practice, many different approaches can be found. We will and refluxing of certain components. Their flow rates
not discuss this aspect in this paper, because the measures have to be determined from a heat balance around the con-
to be taken to handle a reaction mass, that is heating up denser as well as from a mass balance around the reflux
towards a runaway, are too specific for each particular case. collector (see Figure 1).
In this paper we will discuss first the capabilities of the The reaction heat-flow rate qR(t) can be found easily
reaction calorimeter, then the criteria based on known kin- using only temperatures measured inside the reactor and
etics to operate a batch reactor safely and after that the in the coolant jacket. Values of parameters such as the
methods developed to achieve safe operation without overall heat transfer coefficient U and the specific heat
knowing reaction kinetics. At the end we will dedicate capacity of the reaction mixture CP, can be determined in
some space to the promising recent developments of an a separate series of calibration experiments.
early warning detection system. An accurate estimation of heat flow through the reactor
wall is essential for this type of reaction calorimeters and
can be found from the following relationship:
THE REACTION CALORIMETER
The reaction calorimeter has been developed and now is qflow ¼ UAðTR  TA Þ (1)
in full use at almost every process development laboratory
in fine chemicals industry. In commercial applications of where TA is a corrected jacket temperature, Tc, to take into
reaction calorimetry the operation of batch and semibatch account the heat capacity of reactor walls (Zaldivar et al.,
reactions is supported in many ways apart from controlled 1996; Molga, 1997).
heating and cooling, stirring and mixing, distillation and Having determined the power generated due to the reac-
refluxing, dosing of reactants and measurement and control tion progress, qR(t), the total heat effect, QT, can be calcu-
of process parameters (pressure, pH, turbidity, chemical lated as well as the reaction enthalpy:
content by FTIR analysis and so on).
An extensive review of the application of the reaction Ð tF
calorimetry is given in the special issue of Thermochimica QT 0 qR (t)dt
DHA ¼ ¼ (2)
Acta 289, 1996, edited by R.N. Landau. In this issue DnA DnA

Trans IChemE, Part A, Chemical Engineering Research and Design, 2006, 84(A7): 543– 552
SAFETY AND RUNAWAY PREVENTION 545

Figure 1. Schematic diagram of a typical bench-scale (semi-)batch reactor or a reaction calorimeter: 1—basic unit containing reactor control and regulation
systems, 2—reactor vessel with jacket, head assembly and stirrer, 3—dosing pumps, 4—electronic weights, 5—computer for controlling the system and
acquisition of data, 6—cryostat, 7—distillation and reflux set.

Then, for a single reaction, the conversion rate of the com- . Process Development (Schmidt and Reichert, 1988; am
pound A can be estimated as a function of the reaction Ende et al., 1996; Ferguson and Puga, 1997; Erwin
time: et al., 2001; Clark et al., 2001).
. Basic Research (Nomen et al., 1997; Stoessel, 1997;
qR ð t Þ Singh, 1997; LeBlond et al., 1996; Ubrich et al., 1999;
RA ð t Þ ¼ (3) Nogent and Le Tacon, 2002).
DHA

and further the conversion of the compound A as:


Ðt CRITERIA FOR SAFE OPERATION AND RUNAWAY
qR ðtÞdt PREVENTION OF (SEMI-)BATCH REACTORS, BASED
XA ð t Þ ¼ 0 (4) ON KNOWN KINETICS
QT
Runaway prevention has drawn the attention of chemical
Based on the mass balance and the stoichiometric reaction engineers for many decennia. We refer to the early work of
equation as well as using the conversion rate, the concen- Semenov and others already before the Second World War,
tration of the ith reactant in the reacting mixture can be specially in the field of combustion engineering, and also to
found as a function of the reaction time. the work of Barkelew (1959). He demonstrated that for first
So, based on data for the reaction rate and the reactant con- order reactions in a batch reactor the reactor remains ther-
centrations, both as a function of time, the reaction kinetics mally stable provided the temperature difference between
can be studied and the safety of the process can be analysed. coolant and reacting mixture is not higher than:
As can be seen from equations (2) – (4), even with a single TR  Tc  RTc2 =E.
run the reaction enthalpy can be determined, while by carry- Provided this condition is fulfilled a runaway will not
ing out the reaction at isothermal conditions at least at two occur. Later Van Welsenaere and Froment (1970) demon-
different temperatures, the activation energy can be strated this also holds for a temperature difference smaller
obtained. So, with the reaction calorimeter the parameters than RTR2/E and that an even somewhat higher (TR 2 Tc)
DHR, E, DTad and Cp, required for process safety assessment, can be applied without a runaway. In the period under con-
can be estimated. They can be directly used for safe design- sideration Balakotaiah et al. (1995) demonstrated that the
ing of the reactor performance on an industrial scale. above criterion must be corrected for the heat taken up
Apart from the references mentioned before in this by the mass of the reactor vessel. They introduced the
section, some other representative examples of practical ratio of the heat capacity of the reactor vessel and its con-
applications of reaction calorimetry are listed below. tents compared to that of the reactor mixture only and
They have been here arbitrarily grouped according to the called this arbitrarily a Lewis number
categories proposed by Landau (1996), although all of
them can be directly related to safety of processes: ðmcp ÞR ðm cp Þw
Le ¼ ¼1þ
. Process Safety (Zaldivar et al., 1992; Schuler and ðmcp Þf ðm cp Þf
Schmidt, 1992; Regenass, 1997a, b; Serra et al., 1997;
Lerena et al., 1996; De Filippis et al., 2002; Molga This factor, called also thermal inertia, had been used ear-
et al., 2004). lier—e.g. see the paper by Townsend and Tou (1980),

Trans IChemE, Part A, Chemical Engineering Research and Design, 2006, 84(A7): 543 –552
546 WESTERTERP and MOLGA

where its influence on runaway trajectories as well as on In general the batch reactor can be applied only for reac-
thermal stability is also discussed. So, the acceptable adia- tions with a low adiabatic temperature rise. For more
batic temperature rise for the batch wise executed reaction exothermic reactions the semi-batch reactor is used, in
can be a factor Le higher. In practice it is difficult to esti- which one or more reactants are loaded into the reactor
mate the value of Le: for instance the agitator will have and the balance of the recipe items are fed with a desired
the same temperature as the reaction mixture, but the mass rate profile (usually but not always linear with time)
wall of the reactor not so, because it will have a tempera- over a defined period. Doing so, we can keep the heat evol-
ture somewhere in between TR and Tc. These authors con- ution by a very exothermic reaction in hand by the dosing
sidered a well mixed batch reactor with a homogeneous, rate of the second reactant.
single and irreversible, exothermic reaction A ! B, with After the Seveso incident in 1976 much research was
a rate given by r ¼ k(T)cAn . Assuming that the reaction started to understand what the real cause of the runaway
heat generated is removed only by cooling, they gave the in a semi-batch reactor may have been. The authors Hugo
following operation equation, which can be easily derived and Steinbach (1985, 1986) deserve the honour for the
from the set of general mass and heat balance equations observation that an accumulation of the dosed component
after introducing dimensionless variables and parameters: at too low reactor temperatures is the cause of the runaway
in homogeneous semi-batch reactors. They presented a line
Le du að u  u c Þ in a diagram, which separates the region with a runaway
¼1 (5) from the region with a smoothly operating semi-batch
B dX Bð1  X Þn exp½u=1 þ u=g
reactor with only a small harmless accumulation of the
It holds for the reactant conversion 0  X , 1 and after dosed reactant. They did so for modified Stanton numbers
integration with u ¼ 0 at X ¼ 0 gives the reaction paths of 5, 10 and 15, where this number reads as
in the temperature-conversion plane. Particularly useful St ¼ (U A tdos =V r cP ). They also determined empirically
can be an observation of the reactor temperature trajectory from their calculations that—in their range of Stanton
at assumed operating conditions in relation to the adiabatic numbers—in the boundary between the two regions the
trajectory. A typical plot is shown in Figure 2. Damköhler number Da ¼ kcBo tdos must be equal to or
Because in many practical cases the adiabatic tempera- higher than the modified Stanton number.
ture is unacceptably high as the limiting temperature, Bala- In their paper Hugo and Steinbach (1986) derived
kotaiah et al. (1995) referred to the work of Adler and Enig empirical relations for the operating conditions in a homo-
(1964) to have a sharp and realistic criterion for the preven- geneous semi-batch reactor. They found that the thermal
tion of runaway. At some special conditions—i.e., for uc ¼ reaction number B ¼p DTad =R TR2 ) must
(Effiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 be smaller than
0 and g ¼ 1—the results are presented in an operating dia- the limit value B ¼ St þ 2G þ 1:2 , where G ¼ mA/m
gram, in which the lines separating runaway and safe is the ratio of the mass of the reactant A to the final mass
regions are plotted in a plane with the coordinates a/B of the reactor content, which represents the volume
and B/Le (see figure 2 in the cited paper). increase due to the dosage of the second reactant, and
Heiszwolf and Fortuin (1996) executed an experimental that the coolant temperature must be higher than the
study with a first order reaction and demonstrated that the limit value Td, which can be found from the following
predictions of Adler and Enig (1964) are on the dot to pre- equation:
dict for a certain set of operating and design conditions at
what reactant concentration in the feed a runaway will DaðTd Þ ¼ St (6)
occur in a batch reactor with a homogeneous reaction
mixture.
Specially their last relation is very simple.
Hugo et al. (1988) studied the danger of a runaway after
an agitator breakdown, because after a breakdown the reac-
tion proceeds adiabatically. They demonstrate that the most
dangerous situation prevails if the breakdown occurs at the
moment of just reaching the stoichiometric dosage of the
second reactant, because at that condition the concen-
trations of the two reactants reach such values that the reac-
tion rate of a second order reaction just has its highest
value, that is the highest value of the product cA  cB.
They found that the final temperature of the adiabatic reac-
tion mixture will exhibit a minimum value as a function of
the reaction temperature chosen in a particular case, pro-
vided the value for B is larger than 2.
Steensma and Westerterp (1990, 1991) developed full
safety diagrams for the same range of cooling numbers
for heterogeneous liquid –liquid semi-batch reactions and
for reactions both in the dispersed phase as well as in the
continuous phase. They have studied slow reactions and
Figure 2. Temperature trajectories calculated with the operation equation
[Equation (5)] according to the criterion by Balakotaiah et al. (1995). Cal-
also fast reactions, where the mass transfer is enhanced
culations carried out at: Le ¼ 1, B ¼ 2, n ¼ 1, uc ¼ 0, g ¼ 20 for different by the chemical reaction. They show that also for chemi-
values of dimensionless external heat transfer coefficient a. cally enhanced reactions, where the supplied reactant is

Trans IChemE, Part A, Chemical Engineering Research and Design, 2006, 84(A7): 543– 552
SAFETY AND RUNAWAY PREVENTION 547

already consumed in the film around the drops, still a caution about blindly applying of each safety criteria to
dangerous accumulation can occur in its own phase, caus- reactions that may be autocatalytic, is very important.
ing a runaway.
In their safety diagrams they plotted the Reactivity, Ry,
as a function of the Exothermicity, Ex, and surrounded SAFE OPERATION OF (SEMI-)BATCH REACTORS
the region of runaways by a boundary line. These dimen- WITHOUT KINETIC INFORMATION
sionless groups are defined as:
Gygax (1988, 1990) discusses in two concise papers the
DTad,o E=R available methods to determine the danger of runaways in
Ex ¼ 2 (7) batch and semi-batch reactors and how we can use chemi-
Tc ðRH þ U  Da=1Þ1 cal engineering principles to cope with these dangers.
(nA =nB )mDac Stoessel and Ubrich (2001) for homogeneous reactors
Ry ¼ (8)
ðRH þ U  Da=1Þ1 demonstrate how the operation of a semi-batch reactor
can be improved by modulating the feed rate to a semi-
Further regions with no accumulation, harmless reactions batch reactor, proving that the reactor productivity can be
and insufficient ignitions are distinguished. The boundary sharply increased. They also demonstrate how in their
diagrams (see Figure 3) depend on the value of the cooling case the appropriate information can be obtained with a
number—which is similar to the modified Stanton number reaction calorimeter and without knowing the kinetics of
of Hugo and Steinbach (1986)—and on whether the reac- the reactions involved. They illustrate their method with
tion takes place in the continuous phase or in the dispersed an example of a set of consecutive reactions.
phase in the reactor. Westerterp and Molga (2004a) demonstrate that the
Van Woezik and Westerterp (2000, 2002) publish exper- Cooling number is the key parameter, that determines the
imental work on a very exothermic nitration reaction difference in behavior of the large scale, industrial reactor
system with consecutive reactions and with a secondary and the laboratory reactor. The Cooling number is defined
heat effect, which is 3.25 times higher than that of the as:
desired reaction. They determined very accurately the kin-
etics of their consecutive reaction scheme. They also deter- ðU AÞo tdos
Co ¼ (9)
mined the interfacial areas in the reactor. They conclude 1ðV rcP Þo
that the set of differential equations as used by Westerterp
and coworkers to describe the behaviour of such reactors They indicate how this Co number must be determined
are confirmed accurately by their experiments. For safe for laboratory reactors as well as for industrial reactors.
operation much higher Cooling numbers are required than Further they elaborate on the concept of inherently safe
in the previous work (Steensma and Westerterp, 1990, reactor operation and how this must be achieved in prac-
1991) and they introduce the concept of the inherently tice. See Figure 3 for the indication of the inherently safe
safe operating region, in which the heat evolution is still operation region in a safety diagram for a slow liquid –
harmless and the accumulation of the dosed reactant liquid reaction in the continuous phase.
remains very small (see Figure 3). They determine the full range of values of Cooling num-
Maestri and Rota (2005a and 2005b) calculate safety dia- bers for all liquid – liquid reactors of all sizes in a laboratory
grams for kinetics different from those used by Hugo, Wes- or in an industrial plant. They also determine the maximum
terterp and coworkers. The diagrams are influenced indeed of the Exothermicity values below which, and the mini-
by a difference in kinetics, regretfully they do not discuss mum values of the Reactivity above which, never a run-
when and where their kinetics occurs in practical semi- away can occur. They state that the necessary information
batch reactor operations. Hugo (1990) determined that a can be obtained in a few days in a laboratory in a reaction
description by second order reaction always gives satisfac- calorimeter, provided the calorimeter is used in the
tory results, except for autocatalytic reactions. A special semi-batch mode. They argue that their method for
liquid –liquid reactions can be used for single as well as
for multiple reactions of any reaction scheme with the
exception of autocatalytic reactions. They also discuss
how incidents like agitator breakdown can be handled with-
out any danger of a runaway. The elaborated practical pro-
cedure to determine the inherently safe operating
conditions in semi-batch reactors is summarized in
Table 2 (Westerterp and Molga, 2004b). Westerterp and
Molga (2004a) in their paper give data on the practical
values of the dimensionless groups of the Reactivity, the
Exothermicity and the Cooling numbers for both laboratory
and industrial reactors. We should realize that laboratory
reactors usually have much higher Cooling numbers than
industrial reactors. Specially, for heterogeneous liquid –
liquid and gas – liquid reactions Cooling numbers necess-
arily are high in order to achieve good dispersions with
Figure 3. Inherently safe operating conditions for slow reaction in the con- large interfacial areas. In order to simulate industrial
tinuous phase. conditions often reaction calorimeters are equipped with

Trans IChemE, Part A, Chemical Engineering Research and Design, 2006, 84(A7): 543 –552
548 WESTERTERP and MOLGA

Table 2. Procedure for the determination of safe operating conditions

Method used to determine

Value (or group) determined In laboratory reactor In industrial reactor

I. The Exothermicity criterion for a single reaction


DTad,o Measured or calculated
E Measured in reaction calorimeter
or DSC
RH Calculated
1 Calculated
Co Measured (cooling curve for Measured (cooling curve for
originally loaded batch) originally loaded batch)
Ex Calculated as a function of tdos, and
at realistic Tc
Yield For Ex  Exmin check for the reactor
productivity at the chosen tdos

II. The Reactivity criterion for a single reaction


Co Measured (cooling curve for batch originally loaded)
Tta ¼ f (t) Calculated Calculated
RyQFS,min Calculated Calculated
Tc Tc,ind - see below Eq. T1

III. Multiple reactions


See original article. The following holds:
(a) for parallel reactions an apparent activation energy should be determined, then the procedure above can be directly applied;
(b) for consecutive reactions the following method should be applied:
Heat evolution curves Measured as a function of T
Yield Measured
Ex For the first reaction, the second reaction or for the sum of heat effects
E1/E2 Estimated
Ex criterion Calculated
RyQFS criterion Calculated Calculated
"  #
2
RTc,lab RyQFS, min ind
Tc,ind ¼ Tc,lab þ ln   (T1)
E RyQFS, min lab

stirrers with low agitation speeds to reduce the Cooling operation, because we must provide sufficient time for the
numbers. This helps indeed for homogeneous reactions, plant operator or an automatic protection system to take
but is unacceptable for heterogeneous reactions leading to appropriate protective countermeasures to stop a thermal
phase separations. excursion. Such countermeasures may be a pressure
relief, full cooling or quenching (i.e., addition of inhibitor
or cold inert liquid as well as dumping of the reactor
EARLY WARNING DETECTION SYSTEMS content into a cold catch tank). At the same time to
As has been mentioned before, even a correct choice of avoid false alarms, a warning system must be able to pre-
reactor operating conditions does not totally protect the cisely distinguish between dangerous and non-dangerous
plant against a thermal runaway. So, apart from the off- reactor states. So, an essential element of any EWDS is
line activities, which help to define safe operating a robust safety criterion, which works on-line and is
conditions, also on-line prevention measures are necessary practically independent of the actual process carried out
to detect any unexpected situation leading to a runaway in the reactor.
scenario (see Figure 4). Among others under the on-line Numerous criteria for the safe operation of batch and
safety measures as listed in Figure 4, an early warning semibatch reactions have been developed and can be
detection system (EWDS) is indispensable to detect found in literature, some of them are discussed in this
and evaluate unexpected dangerous situations, which paper in the sections above. The sensitivity towards the
may occur in batch and semibatch reactors e.g., due to a initial and the operating parameters has been extensively
failure of the cooling or stirring systems or to a human studied—(see Varma et al. 1999)— and a new series of
mistake. safety criteria based on parametric sensitivity has been
The development of an efficient and robust EWDS for developed (e.g., see Morbidelli and Varma, 1986, 1988,
batch and semi-batch reactors is very difficult due to the 1989; Vajda and Rabitz, 1992; Shukla and Pushpavanam,
complexity of the processes carried out in such reactors 1994). Unfortunately, these criteria can not be used for an
and their strongly nonlinear dynamics. The main problem on-line early warning detection of a runaway. A detailed
of a detection of dangerous situations in discontinuous discussion of the applicability of the existing safety criteria
reactors is the fast recognition of deviations from safe for an early detection system can be found in a series of

Trans IChemE, Part A, Chemical Engineering Research and Design, 2006, 84(A7): 543– 552
SAFETY AND RUNAWAY PREVENTION 549

can be found in the paper by Strozzi et al. (1999), while


further theoretical developments as well as practical appli-
cations and considerations are published in a series of
papers by Strozzi, Zaldivar and their coworkers, which
report on the results obtained in an EU funded project
AWARD (2001 – 2005) (see Zbilut et al., 2002; Bosch
et al., 2004c; Zaldivar et al., 2005, 2003; Bosch et al.,
2004 a, b, d), respectively. Based on the results of the
studies carried out within the frame of the AWARD
project, several prototypes of an Early Warning Detection
System for batch and semibatch reactors have been
produced (for details see AWARD, 2001 – 2005).
In the mentioned method a new runaway criterion has
been defined and the onset of loss of control has been ident-
Figure 4. Prevention measures developed to avoid situations that may lead ified at the point where the divergence of the reactor beha-
to loss of thermal control and runaway scenario, see also Zaldivar (1991a) vior becomes positive on a segment of the reaction path:
as well as Westerterp and Molga (2004a).
div Fðu, X1 , X2   Þ . 0 (11)

papers by Strozzi and Zaldivar (e.g., see Strozzi et al., The research group led by Strozzi and Zaldivar developed a
1999; Zaldivar et al., 2003). The general conclusion arising method to calculate the divergence on-line, using one
from this analysis indicates that most of the existing safety representative measurable variable of the system. State
criteria can not be directly used for an EWDS, because reconstruction techniques and a theory of embedding
either they work off-line or they are model-based; so have been applied for this purpose. The method based on
expensive and time consuming kinetic studies are required, a delayed temperature vector turned out as the most
which are seldom applied to processes carried out in batch robust one with respect to noised experimental data, but
and semi-batch reactors. also pressure measurements have been successfully tested
A simple on-line safety criterion has been based on (e.g., see Bosch et al., 2004d and a). The divergence cri-
measurements of the reactor and cooling jacket tempera- terion developed has been tested and successfully validated
tures and some their derivatives: for the recognition of a for different types of reactors (batch, semibatch and CSTR),
dangerous situation it is enough to check the following different operating modes (isothermal, isoperibolic and
expressions (Hub and Jones, 1986): adiabatic) as well as for different schemes of reaction
kinetics—a.o. for simple and complex (parallel, consecu-
d2 TR tive, mixed and equilibrium) reactions, homogenous liquid
.0 (10a) and heterogeneous liquid –liquid reactions, polymerization
dt2
reactions and so on (see the references cited above).
Several methods employing the state reconstruction tech-
and nique with use of one or more temperature sensors have
been developed and successfully tested for early detection
dðTR  Tc Þ of runaway (see Bosch et al., 2004b; Zaldivar et al.,
.0 (10b)
dt 2005). It has been found that for all considered batch and
semi-batch runs only one temperature sensor is usually suf-
This criterion does not require any detailed information on ficient to reconstruct the divergence with a reasonably
the process carried out in the reactor and it can be also accuracy, so to detect runaway.
applied for the detection of thermal runaway in other equip- The developed method of runaway detection, employing
ment, e.g., in storage tanks. Its main disadvantage is that only one temperature sensor, simplifies considerably an
the noise, always measured in experimental signals, is application of the EWDS in industrial practice. In case of
amplified and significantly affects the results of evalu- one single temperature sensor we have to be sure that the
ations, so false alarms may be triggered. A very useful dis- measured temperature is representative for the entire reac-
cussion of practical tests of this method as well as the tor, i.e., that perfect mixing has been achieved. In case of
results obtained is given by (Zaldivar, 1991b). failure of the stirring system significant non-uniformities
Strozzi and Zaldivar developed a new method for asses- of temperature and reactants concentrations inside the reac-
sing the thermal stability of batch and semibatch chemical tor will occur, so a proper location of the required number
reactors (e.g., see Strozzi and Zaldivar, 1994a; Strozzi of the temperature sensors is essential. Computational fluid
et al., 1994b; Alos et al., 1996a,b). In their method the dynamics (CFD) have been used to investigate this EWDS
Lyapunov exponents were used to calculate the sensitivity method and to identify ‘hot-spots’ appearing in the reaction
and to formulate a new safety criterion. An essential advan- mass (see Rudniak et al., 2004; Molga et al., 2004;
tage of this method, in comparison to the previous ones, is Milewska et al., 2005). The CFD models of stirred tank
that it can be applied for on-line runaway detection. This reactors for bench and industrial scale reactors and also
was made possible by applying a phase-space reconstruc- for both homogeneous and heterogeneous reaction systems,
tion technique and resulted in the fact that the divergence enable a determination of local and instantaneous values of
of the system can be evaluated from the measurement of reactor temperatures and reactant concentrations in badly
one system variable only. The background of this method mixed vessels.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2006, 84(A7): 543 –552
550 WESTERTERP and MOLGA

In off-line applications the results from CFD simulations reactor temperature. In this area a very promising approach,
can be directly converted into a spatial map of the local developed and tested by Zaldivar, Strozzi, Westerterp and
divergence of the system so, a detection of local ‘hot their collaborators, must be noticed. From this approach,
spots’ and their propagation into a global runaway may initially based on the divergence criterion and then sub-
be indicated. It can be particularly important for a so- jected to several modifications, a universal and robust
called ‘computer experimentation’, during which different EWDS has been developed. In these transformations the
types of failures occurring in industrial reactors can be state space reconstruction has been employed, so the detec-
safely investigated. However, it must be clearly distin- tion of runaway with the developed EWDS has become
guished, that for industrial reactors of complex geometry possible by on-line measurement of one single operating
local CFD predictions are difficult for experimental verifi- parameter such as the reactor temperature or the reactor
cations, so the results of CFD simulations should be treated pressure, provided the latter can be considered as a variable
as preliminary indications. representing the state of the reactor.
The presented review can be concluded with the follow-
ing corroborations, which are particularly useful and sig-
SUMMARY AND CONCLUSIONS
nificant for the fine chemicals industry:
From statistical data, which report on the prime causes of
. Off-line safety assessments for batch and semibatch
incidents in batch and semibatch chemical reactors carrying
reactors, to determine inherently safe operating
out strongly exothermic reactions, the general conclusion
conditions, should be carried out on the basis of the
arises that a thermal runaway in these reactors is still a pro-
boundary diagram. This approach does not demand for
blem of great importance. So, a robust safety assessment is
detailed kinetic investigations, only a few measurements
always indispensable. Usually three lines of defense have
with a laboratory reaction calorimeter are needed.
to be considered to prevent thermal runaways in chemical
. For on-line runaway detection in reactors operated
reactors: choice of the right operating conditions, application
according to the state of the art, the EWDS system ela-
of early warning detection system and methods to handle
borated by Zaldivar, Strozzi and their collaborators
running away reactions. Because of its too specific nature,
should be employed. It is a robust early warning detec-
the last line of defense was not discussed in this review,
tion system, for which only an on-line, single point
while the former two have been considered in detail. The
measurement of the reactor temperature (and/or the reac-
choice of safe or inherently safe reactor operating conditions
tor pressure) is needed.
is usually performed in an off-line analysis. Two main
approaches can be here distinguished: the model-based
analysis in which a detailed knowledge of the reaction kin- NOMENCLATURE
etics is indispensable and the most recent approach based
on the boundary diagram, which is particularly useful for A E DHcAo VRo heat exchange surface area, m2
B¼   dimensionless adiabatic
the reacting systems with unknown reaction kinetics. In RTc mcp f Tc temperature rise
the last approach, based on the work of Westerterp and his ci concentration of the ith
collaborators which are published in the series of cited component, mol m23
papers, safety boundary diagrams are employed to dis- cp mass specific heat capacity, J kg21
tinguish the inherently safe operating conditions for batch K21
Cp molar specific heat capacity, J
and semibatch reactors with liquid and liquid –liquid react- mol21K21
ing systems, respectively. In these safety diagrams in a Co Cooling number, see equation (9)
plot of the coordinates Reactivity versus Exothermicity for definition
numbers, runaway and non-runaway regions can be easily Da ¼ kc cBotdos Damköhler number
E activation energy, J mol21
distinguished (see Figure 3). In these safety diagrams the Ex exothermicity number, see equation
boundary lines are still a function of the Cooling number. (7) for definition
The most relevant advantage of this approach is that even k reaction rate constant, m3 mol21
for the case, when a detailed reaction kinetics is not know, s21
the problem of safety assessments can be tackled in a few m mass, kg
m distribution coefficient (liquid–
days with use of the reaction calorimeter. liquid system)
Also the capability and applicability of the reaction calori- n number of moles, reaction order
meter for safety assessments have been reviewed. Several q heat flow, W
applications, useful particularly for direct process safety Q amount of energy, J
r reaction rate, mol m23 s21
improvement as well as for supporting assessments based R conversion rate, mol s21
on the use of boundary diagrams, have been indicated. R gas constant, J mol21 K21
Due to a failure of the cooling/stirring systems or a RH ¼ (rcP )d =(rcP )c ratio of heat capacities of the
human mistake, different unexpected dangerous situations, dispersed and the continuous phase
which can lead to a runaway scenario, may occur even in a Ry reactivity number, see equation (8)
for definition
reactor operated along a safe trajectory. Taking into St ¼ UAtdos =V rCp modified Stanton number
account that even a correct choice of safe operating con- tdos dosing time, s
ditions does not totally protect the reactor against a thermal T temperature, K
runaway, the recent developments of early warning detec- Tta ¼ Tc þ
1:05DTad,0
target temperature, K
tion system have been reported. Such a warning detection 1½RH þ Coð1 þ 1tÞ
system makes possible an early on-line prediction of the U overall heat transfer coefficient,
reactor states, which may lead to the uncontrolled rise of W m22K21

Trans IChemE, Part A, Chemical Engineering Research and Design, 2006, 84(A7): 543– 552
SAFETY AND RUNAWAY PREVENTION 551

U ¼ (U A)0/(r cP VR k cB)0 dimensionless cooling capacity Erwin, S., Schultz, K., Moritz, H.U., Schwede, C. and Kreber, H., 2001,
V volume, m3 Increased reactor performance versus reactor safety aspects in acrylate
X conversion copolymerization, Chem Eng Techn, 24: 305– 311.
Etchells, J.C., 1993, Prevention and control of exothermic runaway, Con-
Greek symbols ference Proceedings Assessment and control of Chemical Reaction
UA dimensionless external heat transfer Hazards, IBC London, 8–9 December 1993, London.
a¼ n1 Ferguson, H.D. and Puga, Y.M., 1997, Development of an efficient and
ðmcP Þf kðTc ÞcAo coefficient
safe process for a Grignard reaction via reaction calorimentry, J
g ¼ E=RTc dimensionless activation energy Therm Anal, 49: 1625– 1633.
DH reaction enthalpy, J mol21 Gesthuisen, R., Kramer, S., Niggemann, G., Leiza, J.R. and Asua, J.M.,
DTad adiabatic temperature rise, K 2005, Determining the best reaction calorimetry technique: theoretical
1 relative volume increase development, Comp Chem Eng, 29: 349–365.
E T  Tc Gygax, R., 1988, Chemical engineering for safety, Chem Eng Sci, 43:
u¼ dimensionless temperature 1759–1771.
RTc Tc
n stoichiometric coefficient Gygax, R., 1990, Scaleup principles for assessing thermal runaway risks,
t ¼ t=tdos dimensionless time Chem Eng Progress, Feb. 1990, 53–60.
HarsNet, http://www.harsnet.de.
Subscripts Heiszwolf, J.J. and Fortuin, J.M.H., 1996, Runaway behaviour and para-
0 initial metric sensitivity of a batch reactor—an experimental study, Chem
c jacket, cooling, at cooling Eng Sci, 51: 3095–3100.
temperature Hugo, P. and Steinbach, J., 1985, Practically oriented representation of
f fluid (liquid reactor content) thermal safe limits for an indirectly cooled semi-batch reactor, Chem
F final Ing Tech, 57: 780– 782.
flow flow through the reactor wall Hugo, P. and Steinbach, J., 1986, A comparison of the limits of safe oper-
QFS quick onset, fair conversion, smooth ation of SBR and CSTR, Chem Eng Sci, 41: 1081–1087.
temperature profile Hugo, P., Steinbach, J. and Stoessel, F., 1988, Calculation of the maximum
R reactor, reaction temperature in stirred tank reactors in case of a breakdown of cooling,
w reactor walls and inserts (stirrer and Chem Eng Sci, 43: 2147–2152.
baffles) Hugo, P., 1990, private information.
Hub, L. and Jones, J.D., 1986, Early on-line detection of exothermic reac-
tions, Plant Oper Prog, 5: 221– 229.
REFERENCES Landau, R.N., 1996, Expanding the role of reaction calorimetry, Thermo-
chim Acta, 289: 101–126.
Alos, M.A., Zaldivar, J.M., Strozzi, F., Nomen, R. and Sempere, J., 1996a, LeBlond, C., Wang, J., Larsen, R.D., Orella, C.J., Forman, A.L., Landau,
Application of parametric sensitivity to batch safety: theoretical and R., Laquidara, J., Sowa, J.R. Jr., Blackmond, D.G. and Sun, Y.K., 1996,
experimental studies, Chem Eng Technol, 19: 222–232. Reaction calorimetry as an in-situ kinetic tool for characterizing com-
Alos, M.A., Strozzi, F. and Zaldivar, J.M., 1996b, A new method for asses- plex reactions, Thermochim Acta, 289: 189 –207.
sing the thermal stability of semibatch processes based on Lyapunov Lerena, P., Wehner, W., Weber, H. and Stoessel, F., 1996, Assessment of
exponents, Chem Eng Sci, 51: 3089–3094. hazards linked to accumulation in semi-batch reactors, Thermochim
Adler, J. and Enig, J.W., 1964, The critical conditions in thermal explosion Acta, 289: 127–142.
theory with reactant consumption, Combust Flame, 8: 97 –103. Maestri, F. and Rota, R., 2005a, Thermally safe operation of liquid-liquid
AWARD, 2001–2005, Advanced Warning and Runaway Disposal, the EU semibatch reactors. Part I: Single kinetically controlled reactions with
funded project (G1RD-CT-2001-00499), http://www.arpconsortium.org/ arbitrary reaction order, Chem Eng Sci, 60: 3309–3322.
AWARD.htm. Maestri, F. and Rota, R., 2005b, Thermally safe operation of liquid-liquid
Balakotaiah, V., Kodra, D. and Nguyen, D., 1995, Runaway limits for semibatch reactors. Part II: Single diffusion controlled reactions with
homogeneous and catalytic reactors, Chem Eng Sci, 50: 1149–1171. arbitrary reaction order, Chem Eng Sci, 60: 5590–5602.
Barkelew, C.H., 1959, Stability of chemical reactors, Chem Eng Progr Milewska, A., Rudniak, L. and Molga, E., 2005, CFD modelling and diver-
Symp Ser, No. 55(25), 37–46. gence criterion for safety of chemical reactors, in Puigjaner, L., Espuna,
Barton, J.A. and Nolan, P.A., 1989, Incidents in the chemical industry due A., (eds), Series: Computer-Aided Chemical Engineering, 20A, 259 –
to thermal runaway chemical reactions, in Conference and Exhibition 264, (Elsevier, Amsterdam).
on Techniques for Assessment of Chemical Reaction Hazards, 5 –6 Molga, E., 1997, Heat exchange in the reaction calorimeter, Inz Chem
December 1989, London Press Centre, London EC1, IBC Techn. Proc, 24: 38 –45.
Press, 1–17. Molga, E., Rudniak, L., Machniewski, P. and Milewska, A., 2004, Appli-
Bosch, J., Kerr, D.C., Snee, TJ., Strozzi, F. and Zaldivar, J.M., 2004a, Run- cation of CFD for modelling of chemical reactors: safety aspects, Inz
away detection in a pilot-plant facility, Ind. Eng Chem Res, 43: 7019– Chem Proc, 25: 1341–1346.
7024. Morbidelli, M. and Varma, A., 1988, A generalized criterion for para-
Bosch, J., Strozzi, F., Zbilut, J.P. and Zaldivar, J.M., 2004b, On line run- metric sensitivity, Chem Eng Sci, 43: 91 –102.
away detection in isoperibolic batch and semibatch reactors using the Morbidelli, M. and Varma, A., 1986, Parametric sensitivity in fixed bed
divergence criterion, Comput Chem Eng, 28: 527 –544. reactors: the role of interparticle transfer resisance, AIChE J, 32:
Bosch, J., Strozzi, F., Lister, D.G., Maschio, G. and Zaldı́var, J.M., 2004c, 297–306.
Sensitivity analysis in polymerization reactions using the divergence Morbidelli, M. and Varma, A., 1989, A generalized criterion for para-
criterion, Trans IChemE, Part B, Proc Safety Environ Protect, metric sensitivity: application to a pseudohomogeneous tubular reactor
82(B1): 18–25. with consecutive or parallel reactions, Chem Eng Sci, 44: 1675–1696.
Bosch, J., Strozzi, F., Snee, T.J., Hare, J.A. and Zaldivar, J.M., 2004d, A Nogent, H. and Le Tacon, X., 2002, The differential reaction calorimeter: a
comparative analysis between temperature and pressure measurements simple apparatus to determine reaction heat, heat transfer value and heat
for early detection of runaway initiation, J Loss Prev Process Ind, 17: capacity, J Loss Prev Proc Ind, 15: 445–448.
389– 395. Nolan, P.F. and Barton, J.A., 1987, Some lessons from thermal-runaway
Clark, J.D., Shah, A.S., Peterson, J.C., Grogan, F.M. and Camden, S.K., incidents, J Hazard Materials, 14: 233– 239.
2001, Application of reaction calorimetry toward understanding the Nomen, R., Sempere, J. and Serra, E., 1997, A comparison of calorimetric
large scale chemistry of ethyl diazoacetate, Thermochim Acta, 367– measurements by using different reaction calorimetres, J Therm Anal,
368: 75–84. 49: 1707–1713.
De Filippis, P., Giavarini, C. and Silla, R., 2002, Thermal hazard in a batch Regenass, W., 1997a, The development of stirred tank heat flow calorime-
process involving hydrogen peroxide, J Loss Prev Process Ind, 15: try as a tool for process optimisation and process safety, Chimia, 51:
449– 453. 189–200.
am Ende, D.J., Clifford, P.J. and Orthrup, D.L., 1996, The role of reaction Regenass, W., 1997b, The development of heat flow calorimetry as a tool
calorimetry in the development and scale-up of aromatic nitration, Ther- for process optimization and process safety, J Therm Anal, 49: 1661–
mochim Acta, 289: 143–154. 1675.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2006, 84(A7): 543 –552
552 WESTERTERP and MOLGA

Rowe, S.M., 1996, The role of a calorimetry in chemical plant safety: a Van Woezik, B.A.A. and Westerterp, K.R., 2000, The nitric oxidation of
chlorination reaction, Thermochim Acta, 289: 167 –175. 2-octanol. A model reaction for multiple heterogeneous liquid-liquid
Rudniak, L., Machniewski, P.M., Milewska, A. and Molga, E., 2004, CFD reactions, Chem Eng Process, 39: 521 –537.
modelling of stirred tank chemical reactors: homogeneous and hetero- Van Woezik, B.A.A. and Westerterp, K.R., 2002, Runaway behaviour and
geneous reaction systems, Chem Eng Sci, 59: 5233–5239. thermally safe operation of multiple heterogeneous liquid-liquid reac-
Schmidt, C.U. and Reichert, K.H., 1988, Reaction calorimeter—A contri- tions. The nitric oxidation of 2-octanol, Chem Eng Process, 41: 59– 77.
bution to safe operation of exothermic polymerisation, Chem Eng Sci, Van Welsenaere, R and Froment, G.F., 1970, Parametric sensitivity and
43: 2133–2137. runaway in fixed bed catalytic reactors, Chem Eng Sci, 25: 1503–1516.
Schuler, H. and Schmidt, C.U., 1992, Calorimetric-state estimators for Varma, A., Morbidelli, M. and Wu, H., 1999, Parametric Sensitivity in
chemical reactor diagnosis and control, Chem Eng Sci, 47: 899–915. Chemical Systems (Cambridge University Press, Cambridge, MA,
Serra, E., Nomen, R. and Sempere, J., 1997, Maximum temperature attain- USA).
able by runaway of synthesis reaction in semi-batch processes, J Loss Verwijs, J.W., 1994, Reactor start-up and safeguarding in industrial
Prev Process Ind, 10: 211–215. chemical processes, PhD thesis, Twente University, Enschede, The
Shukla, P.K. and Pushpavanam, S., 1994, Parametric sensitivity, runaway Netherlands.
and safety in batch reactors: experiments and models, Ind Eng Res, 33: Westerterp, K.R. and Molga, E.J., 2004a, No more runaways in fine chemi-
3202–3208. cal reactors, Ind Eng Chem Res, 43: 4585–4594.
Singh, J., 1997, Reaction calorimetry for process development: Recent Westerterp, K.R. and Molga, E.J., 2004b, Runaway prevention in liquid-
advances, Proc Safety Progress, 16: 43 –49. liquid semibatch reactors, Inz Chem Proc, 23: 2041–2050.
Steensma, M. and Westerterp, K.R., 1990, Thermally safe operation of a Zaldivar, J.M., 1991a, Fundamentals on runaway reactions: prevention and
semibatch reactor for liquid-liquid reactions. Slow reactions, Ind Eng protection measures, in Benuzzi, A. and Zaldivar, J.M. (eds). Safety of
Chem, 29: 1259–1270. Chemical Reactors and Storage Tanks, 19–47 (Kluwer Academic
Steensma, M. and Westerterp, K.R., 1991, Thermally safe operation of a Press, Dordrecht, The Netherlands).
semibatch reactor for liquid-liquid reactions. Fast reactions, Chem Zaldivar, J.M., 1991b, Early on-line detection of runaway initiation, in
Eng Technol, 14: 367 –375. Benuzzi, A. and Zaldivar, J.M. (eds). Safety of Chemical Reactors
Stoessel, F., 1997, Applications of reaction calorimetry in chemical engin- and Storage Tanks, 201–223 (Kluwer Academic Press, Dordrecht,
eering, J Therm Anal, 49: 1677–1688. The Netherlands).
Stoessel, F. and Ubrich, O., 2001, Safety assessment and optimization Zaldivar, J.M., Barcons, C., Hernandez, H., Molga, E. and Snee, T.J.,
of semi-batch reactions by calorimetry, J Therm Anal Calorym, 64: 1992, Modelling and optimization of semibatch toluene mononitration
61–74. with mixed acid from performance and safety view points, Chem Eng
Strozzi, F. and Zaldivar, J.M., 1994a, A general method for assessing the Sci, 47: 2517–2522.
thermal stability of batch chemical reactors by sensitivity calculation Zaldivar, J.M., Hernandez, H. and Barcons, C., 1996, Development of a
based on Lyapunov exponents, Chem Eng Sci, 49: 2681–2688. mathematical model and a simulator for the analysis and optimisation
Strozzi, F., Alos, M.A. and Zaldivar, J.M., 1994b, A method for assessing of batch reactors: experimental model characterisation using a reaction
thermal stability of batch reactors by sensitivity calculation based on calorimeter, Thermochim Acta, 289: 267 –302.
Lyapunov exponents: experimental verification, Chem Eng Sci, 49: Zaldı́var, J.M., Cano, J., Alos, M.A., Sempere, J., Nomen, R., Lister, D.,
5549–5561. Maschio, G., Obertopp, T., Gilles, E.D., Bosch, J. and Strozzi, F.,
Strozzi, F., Zaldı́var, J.M., Kronberg, A.E. and Westerterp, K.R., 1999, 2003, A general criterion to define runaway limits in chemical reactors,
On-line runaway detection in batch reactors using chaos theory tech- J Loss Prev Process Ind, 16: 187–200.
niques, AIChE J, 45: 2429– 2443. Zaldı́var, J.M., Bosch, J., Strozzi, F. and Zbilut, J.P., 2005, Early warning
Townsend, D.I. and Tou, J.C., 1980, Thermal hazard evaluation by an detection of runaway initiation using non-linear approaches, Comm
accelerating rate calorimeter, Thermochim Acta, 37: 1–30. Nonlinear Sci Num Sim, 10: 299 –311.
Ubrich, O., Srinivasan, B., Lerena, P., Bonvin, D. and Stoessel, F., 1999, Zbilut, J.P., Zaldı́var, J.M. and Strozzi, F., 2002, Recurrence quantification-
Optimal feed profile for a second order reaction in a semi-batch reactor based Liapunov exponents for monitoring divergence in experimental
under safety constraints. Experimental study, J Loss Prev Process Ind, data, Physics Letters A, 297: 173–181.
12: 485–493. Zogg, A., Stoessel, F., Fischer, U. and Hungerbuhler, K., 2004, Review:
Ubrich, O., Srinivasan, B., Lerena, P., Bonvin, D. and Stoessel, F., 2001, isothermal reaction calorimetry as a tool for kinetic analysis, Thermo-
The use of calorimetry for on-line optimisation of isothermal semi-batch chim Acta, 419: 1–17.
reactors, Chem Eng Sci, 56: 5147–5156.
Vajda, S. and Rabitz, H., 1992, Parametric sensitivity and self-similarity in The manuscript was received 30 September 2005 and accepted for
thermal explosion theory, Chem Eng Sci, 47: 1063–1078. publication after revision 15 February 2006.

Trans IChemE, Part A, Chemical Engineering Research and Design, 2006, 84(A7): 543– 552

Você também pode gostar