Reactors, Kinetics and Catalysis
Reactor Operating Procedures for Startup
of
Continuously-Operated Chemical Plants
J.
W. Verwijs,
P.
H.
Kosters, and
H.
van
den Berg
Process Development
&
Control Dept., Dow Benelux N.V.,
4530
AA Terneuzen, The Netherlands
K.
R.
Westerterp
Chemical Reaction Engineering Laboratories, Dept. of Chemical Engineering,
Twente University,
7500
AE Enschede, The Netherlands
Rules are presented for the startup of an adiabatic tubular reactor, based
on
a
qualitative analysis of the dynamic behavior
of
continuously-operated vapor- and
liquid-phase processes. The relationships between the process dynamics, operating
criteria, and operating constraints are investigated, since a reactor startup cannot
be isolated from an entire plant startup. Composition control
of
the process material
is critical to speed up plant startup operations and to minimize the amount of offgrade
materials. The initial reactor conditions are normally critical for a successful startup.
Forprocess conditioning, aplant should have an operating mode at which the reactor
can be included
in
(z
recycle loop together with
its
feed system and downstream
process section. Experimental data of an adiabatic tubular reactor startup and
thermal runaway demonstrate some operationalproblems when such an intermediate
operating stage is missing. The derived rules are applied to an industrial, highly
heat-integrated reactor section, and the resulting startup strategy is summarized in
an elementary-step diagram.
Introduction
The chemical industry has put much effort on process op-
timization, resulting in increased process yields, reduced energy
consumption, decreased environmental pollution, and im-
proved product quality. To support these activities, extensive
research has been carried out in the area of heat exchanger
network design for energy integration. In the classic concept,
energy integration
is
the final step in developing the process
structure, but the process conditions as such remain un-
changed. New approaches manipulate also the operating con-
ditions to obtain a more efficient energy utilization (Westerberg,
1
992).
Another trend in chemical engineering is to integrate more
than one process function in a single piece of equipment, like
reaction, mass, heat
or
momentum transport operations. This
multifunctional equipment can have a significant impact on
the entire process structure, as was demonstrated for a meth-
anol plant design by Westerterp
(1992).
The process structure
was drastically simplified by selective methanol absorption in
Correspondence concerning this article should he addressed to
K.
R.
Westerterp.
the reactor section, which increased the raw material conver-
sion to almost
100%
and eliminated the need for material
recycles.
The tight coupling of process functions may make the entire
plant more difficult to control (Amundson,
1988).
Therefore,
integrated processes ask
for
a combined process and control
design approach. An actual industrial case according this strat-
egy is reported by Bouwens and Kosters
(1992).
They concluded
from the design of a highly integrated heat exchange section
that controllability analysis should cover complete process net-
works, because the best control structure design for a heater
unit investigated as
a
stand-alone system appeared to be un-
stable when operated in combination with other process units.
The integration of process functions may complicate the
plant startup or shutdown operations too. Unfortunately, the
required dynamic process models describing a plant startup
or
shutdown are necessarily much more complex, and costly, than
a
model which covers the range of conditions encountered
during normal operations only (Wolff et al.,
1992).
Never-
theless, fundamental research in nonsteady-state process de-
AIChE
Journal
148
January
1995
Vol.
41,
No.
1
sign, process control, and operating procedure synthesis is
strongly needed, because most process related incidents occur
when a chemical plant is not operating at a steady state
(Amundson, 1988). Two types of process knowledge are nec-
essary to study nonsteady-state plant operations (Stephano-
poulos, 1988):
Procedural knowledge,
representing the operating strategy
employed by plant personnel and process control systems to
run the process
Declarative knowledge,
based on first principles, char-
acterizing the dynamic behavior of process units.
A
dynamic process model should answer questions about
process behavior and operating limits, while a representation
of
the operating procedures should answer questions about
“how” and “when” to take some operational action. Note
that modern process control systems may perform several tasks,
like analog control to maintain the process state variables at
their target values, on/off control
to
establish material routing
through the process, interlocking to ensure process safeguard-
ing, and sequence control
to
guide the process through a series
of operating phases (Slijk, 1985). Well-defined operating tar-
gets, operating constraints, and operating procedures are nec-
essary to perform these tasks automatically. The relationship
between the above mentioned elements and the dynamic op-
eration of a process is shown in Figure
1.
Some studies have been published on formal methods for
operating procedure synthesis. Fusillo and Powers (1987,1988)
developed a modified meandends planning technique to de-
Operating Targets Operating Constraints
+ie
Production Capocity
)c
Product Quality
m
Product Mix
*
Process Performance
*
Process Safety
m
Process Controllability
*
a
t
termine intermediate operating states in between shutdown and
the normal production state. These intermediate operating goals
are based on the existence of stationary states at which the
system can remain until the next operational action can be
taken.
A
constraint guided strategy searches for a feasible
sequence
of
actions to drive the process from shutdown
to
the
production state. Lakshmanan and Stephanopoulos (1988a,b,
1990) developed a similar approach, based on a hierarchical
object-oriented modeling technique in combination with a non-
linear planning method. By applying this methodology, it is
possible to include mixing constraints in the synthesis problem
to avoid the formation
of
undesirable or potentially dangerous
mixtures. Aelion and Powers (1991a,b) developed a strategy
for the retrofit synthesis
of
flowsheet structures and operating
procedures. If no effective operating procedure can be gen-
erated, flowsheet structure modifications are proposed by the
operating procedure planner.
The above mentioned research
is
focused mainly on the
development of computerized operating procedure synthesis
techniques, and showed some promising results. However, the
operating constraints as such are assumed to be known
apriori,
while
to
our best knowledge the problem
of
determining con-
venient startup criteria is not addressed yet.
Also, almost no studies are published about process dynam-
ics and operating procedures of industrial reactor systems dur-
ing a plant startup. In a preceding study, the behavior of an
industrial adiabatic tubular reactor during the startup is de-
scribed, together with the impact of a failure
of
the feed-pump
Operating Procedures
Y
Process Decomposition into
Unit
Operations
m
Sequence
of
Operat ional
Act
i
ons
*
Intermediate Stationary
States
Process Contro
I
~
*
Analog Control
+
+ie
On/Off
Control
I
t
I
4
m
Sequence Control
rn
Saf eguard
i
ng
Dynamic Operation
m
Start-up
x
Capacity Rate Chonges
m
Product Changeovers
m
Shutdown
Figure
1.
Relationship between process control elements and nonsteady-state process
operation.
LiChE
Journal January 1995
Vol.
41,
No.
1
149
of one of the reactants (Verwijs et al., 1992). It is shown that
for an exothermal adiabatic tubular reactor
a
much higher
initial temperature is required, compared to the reactor inlet
temperature at normal steady-state conditions, to ensure high
reactant conversion levels during the startup. Another impor-
tant conclusion is that a reactor startup cannot be studied
effectively without taking into account the operational aspects
of a plant startup.
This article defines some rules for the startup of tubular reactor
systems in continuously-operated chemical plants:
First, some axioms are defined to determine convenient
startup criteria.
Secondly, an entire plant has to be unraveled into smaller
units, which can be operated stand-alone during the startup.
Some rules for process decomposition are described, together
with some guidelines for sequencing the startup
of
individual
process units. The rules are based on an analysis
of
operational
problems in vapor and liquid phase processes, but can be
extended potentially to polymer and solid processing units.
Thirdly, the problem is addressed
of
how and when to
put an exothermal adiabatic tubular reactor system into a pro-
duction state. Industrial examples are used (a) to demonstrate
some points regarding the response
of
operating personnel to
dynamic reactor operations, and (b) to show some shortcom-
ings about reactor conditioning operations which are incor-
porated in many industrial designs.
The derived rules are applied to an industrial, highly heat-
integrated reactor section, and the resulting startup strategy is
described in a step-by-step procedure.
Deviations from Operating Targets
In general, the steady-state conditions
of
a continuously-
operated chemical plant have been proven to be safe, and the
observation of a few key state variables is often sufficient for
experienced operating personnel to control the process status.
Unfortunately, process data generated during the startup can
confuse even the best operators, due to control loop interac-
tions, inverse process responses, or time delays in process and
measurement responses. Ultimately, this may lead to operating
errors, due to an improper understanding
of
the process dy-
namics (Stephanopoluos, 1988).
Hazardous situations may result from running a chemical
plant outside the range
of
the normal operating conditions.
Therefore, process hazard and operability studies are widely
used in the chemical industry to improve process design, con-
trol and safeguarding. In these,
so
called, HAZOP studies a
systematic, qualitative analysis is carried out on impossible
deviations from the desired operating conditions, the causes
are identified, and their consequences are evaluated (Lees,
1991).
Maintaining process performance, like product quality,
product consistency and process efficiency, is also critical for
a successful plant operation. An important trend in today’s
industrial practice is to define product quality no longer as
“high purity” or “within certain specification limits,” but
rather as “low variability around a specification target” (Vil-
lermaux, 1991). This is because there are always some losses
associated with deviations from given target values (Roy, 1990).
Consequently, the process state variables should be kept at the
desired nominal targets, and their variability should be de-
creased by making the appropriate changes in the operating
conditions and/or strategies to improve process performance
and product quality (Saraiva and Stephanopoulos, 1992).
It can be concluded from the above mentioned arguments
that process safety and performance are determined by how
well a continuously-operated chemical plant can be operated
at the target values of the state variables. Process performance
requirements ask for process control within tight operating
constraints, while hazardous situations may result from (large)
deviations from the required process conditions. Fortunately,
steady-state process operation adhering
to
performance con-
straints will generally also comply with process safety con-
straints simultaneously.
This rule is used as a leading principle
in HAZOP studies. It was proved quantitatively for cooled,
fixed-bed, tubular reactors by Westerterp et al. (1984), Wes-
terterp and Ptasinsky (1984), Westerterp and Overtoom (1985),
and Westerink and Westerterp (1988). They demonstrated for
this reactor type that
a
thermal runaway will not occur if the
selectivity and conversion criteria are adhered to.
Operating constraints
From an operational point of view, three categories of proc-
ess constraints can be defined:
Process safety constraints
are set by the design pressure
and temperature of equipment, the relief-pressure
of
safety
devices, the maximum allowable temperature differences in
heat exchanger equipment, the chemical and/or physical nature
of process materials, and environmental safety. Normally, a
process is safeguarded for exceeding these constraints by in-
terlock systems based on temperature, pressure or
flow
meas-
urements, and by safety devices. Exceeding of safety constraints
may lead to immediate process shutdowns.
Process controllability constraints
are determined by op-
erating capacity turndown ratio’s of process equipment with
respect to the design capacity, control valve ranges, control
loop interactions, and process control stability. For a
well-
designed
process, the controllability constrained range of op-
erating conditions is more confined than the range of possible
operating conditions set by safety constraints. Violation
of
controllability constraints may lead to process upsets and
ul-
timately to hazardous situations.
Process performance constraints
are determined by prod-
uct quality, product consistency and process efficiency re-
quirements. Usually, these are tight operating constraints.
Economic losses result from violating performance constraints.
The startup of a continuously-operated chemical plant is a
complex operational task, and known to be critical with respect
to process safety. Consequently, operating personnel and proc-
ess engineers are focused, intuitively, on running a chemical
plant within its safety constraints. Nevertheless, uncontrolled
process upsets are quite easily obtained during the startup,
since equipment and control loops are designed and tuned
usually for steady-state operations only (Amundson et al.,
1988). To avoid these upsets, and to ensure a safe operation
during startup, the process should be limited to the control-
Iability constraint range
of
operating conditions:
By a proper plant design. For example, if the composition
of
a mixture of two chemicals should be restricted to certain
limits, it may be necessary to install parallel control valves in
the feed line, upstream
of
the mixer, to avoid an overdosing
150
January
1995
Vol.
41,
No.
1
AiChE
Journal
of one chemical at low production rates, even in case of a loss
of control.
By appropriate operating procedures. As discussed above,
process operation adhering to controllability constraints will
generally also comply with safety constraints simultaneously.
Hence, operating procedures should be designed with respect
to criteria founded on process controllability considerations
instead of on safety constraints only.
Convenient criteria
for
startup operations
Generally, the most important controlled process variables
are the temperature, pressure, inventory (level), material com-
position and flow. Fortunately, the flow and process inventory
variables can be controlled usually at their intermediate targets
during all phases of the plant startup operations (inventory
control). The impact of controlling the material composition
too (material balance control), as
of
the moment when equip-
ment is filled with material from storage, can be explained
qualitatively by focusing on the dynamic behavior of a process
system.
First, the response time to a feed
flow
rate and/or com-
position change, is determined by the average residence time
and the residence time distribution
of
the material in a process
unit. The average residence time is controlled by the total feed
rate into the system and the total volume of the process in-
ventory. The residence time distribution depends on the mixing
characteristics of the installed process equipment, like the plug-
flow behavior
of
a tubular reactor or the well-mixed tank
behavior in a distillation column sump. The response time to
a step change in the process inlet conditions will vary between
one to more than several orders
of
magnitude
of
the average
residence time to reach the new operating conditions. For ex-
ample, liquid-phase tubular reactors may have a response time
close to the average residence time of the system. On the other
hand, the response time of systems with a large thermal ca-
pacitance compared to that of the total feed into the system,
similar to exothermic vapor-phase reactions in fixed-bed re-
actors, can be several orders of magnitude greater than the
residence time of the gaseous feed itself. In practice,
man@-
ulated state variables are ramped up
(slowly)
to their new
target values,
and a change in process conditions will take even
longer than the time span mentioned before. Secondly, several
recycle streams may be present in
a
chemical plant, because
per pass only a certain proportion of the raw materials is
converted into products; the rest has to be recovered and re-
cycled. Processes with recycle streams are quite common, but
their dynamics are poorly understood at present (Luyben, 1992).
Nevertheless, some key conclusions from studies on the rela-
tionship between the process structure and its dynamics are
that (a)
recycles increase theprocess sensitivity
to
disturbances,
and (b)
the response time
of
recycle processes
is
substantially
longer than the response time
of
the forwardpath alone
(Denn
and Lavie, 1982; Kapoor et al., 1986).
As a consequence of this dynamic process behavior, upset
conditions for the material composition may be sustained for
a long period in the process. Therefore, the overall plant startup
time will increase significantly in case the material composition
has to be corrected, either by a mismatch
of
the charged ma-
terial or by a process upset. On the other hand, a plant startup
can be accelerated by restricting the material composition, as
AIChE
Journal January
1995
close as possible,
to
its (steady-state) target value when the
process units are charged with material from storage. Mini-
mizing the total volume of the process inventory will have a
similar effect. Additionally, the amount of offgrade materials
produced during the startup may be decreased too.
These aspects can be illustrated by focusing on a distillation
column which is running out
of
specification on its distillate
product. Depending on the reflux drum holdup time, the reflux
ratio, and the distillate to feed ratio, the distillate specifications
may be obtained much faster by dumping the offgrade contents
of
the reflux drum into the distillation column or into a storage
tank, followed by filling up the reflux drum again with on-
specification product, either by distillation or from storage,
than bringing the offgrade material on specification by a con-
tinued operation.
Process Decomposition into Unit Operations
Decomposition of an entire plant into smaller subsystems is
necessary to decrease the complexity of the plant startup op-
erations. Additionally, it is
a
means to focus the attention of
operating personnel
to
a particular plant segment where the
next operational action is taken (Aelion and Powers, 1991a).
Some generic criteria to determine the boundaries of a sub-
system within an entire process system are reported by Fusillo
and Powers (1987) and Aelion et
al.
(1991~). The following
sections describe some process decomposition rules in rela-
tionship with the startup of continuously-operated reactors.
The rules are based on an analysis
of
operational problems in
vapor- and liquid-phase processes, but can be extended po-
tentially to polymer and solid processing units.
The nonheat-integrated process in Figure 2 will be used to
illustrate the various steps of the analysis. In this process, the
products are formed by an exothermal, liquid-phase reaction
between reactant A and
B.
Reactant A is fed from storage into
a
buffer tank, which receives recycle material also from a
recovery unit. Reactant A is fed from the buffer tank into the
reactor, together with reactant
B
which is fed from storage.
The reactor system consists of a feed mixer, a preheater and
an adiabatic reactor vessel. Reactant A is fed in e.xcess, because
reactant B should be totally converted at the reactor exit. A
breakthrough
of
reactant
B
in the reactor effluent is not al-
lowed for process safety and operability reasons. The excess
amount of reactant A is recovered and recirculated. The crude
product is discharged from the recovery unit into the product
refining unit. The finished products are routed to storage.
Reversible and irreversible unit operations
The usual way to bring a process unit into operation is to
charge the required amount
of
materials into the system, as
soon as purging and leak testing are finished. Subsequently,
the process unit is driven in steps to predefined operating
conditions at which it can be integrated operationally with
neighboring process units.
Fusillo and Powers (1987) developed
a
so-called
stationary
state
concept to determine stable intermediate operating states
in between shutdown and the normal production state at which
a process system can remain until the next operational action
can be taken. The presence of
simultaneous inverse operations
within the system boundaries, like heating vs. cooling or sep-
aration vs. mixing, indicate the possibility
of
a stationary state.
Vol.
41,
No.
1
151
Reactant
B
Reactant
A
fi
I-J
re‘xtant
A
Mixerr
4
*
.
. .
. . . .
. . .
.
.
.
Preheoter
~
,
4
Reactant
A
retOVBrY
.
.
. . .
.
.
.
.
.
Reactor
b---;
-----,
storage
final
products
Figure
2.
Process scheme.
Systems with a large capacitance for
a
physical quantity, like
thermal energy or mass, may also have stationary states. A
distillation column running at total reflux conditions is an
example
of
a system exhibiting a stationary state due to the
presence
of
simultaneous inverse operations, being evaporation
in the reboiler vs. condensation in the overhead condenser
(Fusillo and Powers,
1987).
The concept of
simultaneous inverse operations
provides
also
a
key to distinguish between reversible and irreversible
unit operations:
A
reversible unit operation
is defined as a process system
that can be operated stand-alone, without any process streams
fed into or exiting the subsystem, due to the presence
of
inverse
operations. This yields the possibility of conditioning
a
process
system during startup operations. A prerequisite for the si-
multaneous operation of the inverse process functions is that
material can be recycled between them, like between the re-
boiler and condenser
of
a distillation column.
Irreversible unit operations
are started up by supplying the
appropriate feed into the system, and
production is started.
An example is the reactor system in Figure
2.
Other examples
are drying, filter, and centrifuge operations. Theoretically, it
is possible for the latter type
of
operations to suspend the
solids again in the solvent, but in practice the equipment to
do
so
is not installed. Generally, whether a process unit is
reversible or irreversible depends often on the type
of
opera-
tions present in the process, an example being
a
mixer; how-
ever, a separator is not installed. It can be determined by
thermodynamics also due to irreversible state transitions or
chemical reactions.
Distinguishing between reversible and irreversible unit op-
erations serves as a guiding principle in planning the startup
order
of
different process units. Reversible process units are
put into operation first. The sequence is controlled by their
heating and cooling sources. Units with process independent
heating or cooling sources, like aircoolers or steam reboilers,
are usually started up first. Finally, the irreversible units are
put into operation, and production is started. Generally, this
rule implies that product finishing sections are started up first
for vapor- and liquid-phase processes, followed by the reactor
feed preparation section, the reactant recovery section, and
the reactor section, respectively.
Startup
of
reversible unit operations
Severa
(1973)
presented a simplified operating procedure for
the startup of a crude and vacuum distillation unit of a refinery.
Some principal steps are:
Fill the distillation unit through the normal
flow
route
with raw material from the crude charge tank, as soon as
purging and leak testing are finished.
Establish a cold product circulation from the distillation
unit to the crude charge tank, with all streams being returned
into the crude charge tank.
Bring the entire unit to the required operating conditions
by putting the reboiler and condenser systems into service,
while recirculating ail products to the crude charge tank via
coolers in the rundown lines,
until
at1
products are running
reasonably well
on
specification.
Start production by routing all products into storage in-
stead of into the crude charge tank within
a
short period
of
time to avoid excessive changes in the composition of the crude
charge tank.
In the first step, the initial process conditions are established
with respect to the process inventory. In the second step, pumps,
level and flow controllers are put into service (inventory con-
trol). In the third step, the unit is driven to the normal operating
conditions, without being in a production state (simultaneous
material and heat balance control). Once reaching these, the
unit is put into a production state.
This policy can be used for many different types of reversible
unit operations. Of course, the strategy should be adjusted for
the particular unit characteristics, if necessary, however the
basics
being inventory control followed by simultaneous ma-
terial and heat balance control
will remain the same.
Two additional notes should be made on the startup of
reversible unit operations. First, during a plant startup,
op-
erating personnel is often focused on getting the entire process
into a production state. Nevertheless, it is profitable to put
(some) additional time in running the particular process units
strictly on the required operating targets before continuing with
the operational integration with neighboring units. The time
“spent” at this stage will be gained many times in the tail end
of
the startup operations. The impact
of
a
process upset is
limited for stand-alone operated units, and corrections can be
made relatively easily, because there are no interactions with
neighboring units. On the other hand, operational difficulties
may arise when process upsets propagate through a number
of process units, especially during startup operations when
152
January
1995
Vol.
41,
No.
1
AIChE
Journal