Preprint, 11th IFAC Symposium on Dynamics and Control of Process Systems,

including Biosystems
June 6-8, 2016. NTNU, Trondheim, Norway

Control of an exothermic packed-bed tubular reactor

Isrrael Nájera *, Jesús Álvarez *, Roberto Baratti ** and Cesar Gutiérrez*

* Dep. de Ingeniería de Procesos e Hidráulica,
Universidad Autónoma Metropolitana UAM-I, Apdo. 55534, 09340 México, D.F. Mexico
** Dip. Ingegneria Meccanica, Chimica e dei Materiali,
via Marengo, 2 09123 Cagliari, Italy

Abstract: The problem of robustly controlling a highly exothermic gas-phase packed-bed tubular reactor
on the basis of feed and reactor temperature measurements is addressed. First, advanced nonlinear control
yields a detailed model-based output feedback (OF) control design with passivity and observability
solvability conditions, sensor location criterion, and simple tuning. Then, the behavior of the advanced
controller is recovered with a simplified model-based realization that amounts to an industrial temperature
tracking controller with feedforward (FF) dynamic setpoint compensation. The advanced and industrial
controllers are formally connected. The approach is illustrated and tested with a representative case
example through numerical simulations.
Keywords: packed bed tubular reactor, staged model, feedforward control, output-feedback control,
setpoint compensation, passive control, PI control.

applied to assess solvability and draw a detailed model-based
1. INTRODUCTION
nonlinear geometric OF control solution. On the basis of
An important class of industrial processes (like oxidation of passivity and observability properties in the light of industrial
ethylene, naphthalene, vinyl acetate synthesis and control requirements, the behavior of the advanced nonlinear
hydrogenation) take place in highly exothermic gas-phase OF controller is recovered with a linear PI temperature
catalytic fixed packed bed (FPB) tubular reactors. These controller equipped with model-based (preprogrammed)
reactors: (i) consist of a compact, immobile stack of randomly setpoint compensation. The methodology is illustrated and
arranged catalyst pellets that are bathed by the (gas) reactant tested with a case example through numerical simulations.
fluid, which reacts over the (interior or exterior) catalyst
2. CONTROL PROBLEM
surface (Rase, 1990; Jakobsen, 2008), (ii) are relatively easy
to maintain, and produce high per-volume conversion, but (iii) Let us consider the jacketed FPB tubular reactor, depicted in
rises a difficult control problem due to complex nonlinear Fig. 1, where a gas stream is fed (with molar flow ,
dynamics (spatially distributed, multiplicity, parametric temperature , and molar composition ) and converted into
sensitivity, nonminimum phase characteristics, limit cycling, product (at reactant composition and temperature ) by
and structural instability) (Jensen et al., 1982; Jorgensen, means of a reaction rate ( .
1986).
Basically, FPB reactors are controlled by adjusting the heat
extraction rate according to a temperature PI controller driven
by a sensor at the most sensitive axial location (Bashir et al.,
1992; Jaisathaporn et al., 2004; Del Vecchio et al., 2005). The
effect of measured disturbances are compensated by adjusting
the temperature setpoint with a diversity of procedures that
range from manual to model-based (Brosilow & Joseph,
2002). Fig. 1. FPB tubular reactor and its control scheme

The PI controllers are robust and cheap, but their design and Under standard assumptions (ideal gas, only axial spatial
supervision rely heavily on per-reactor characteristics and distribution, quasi-static gas phase concentration, low pressure
experience. The related setpoint compensation is done with drop, negligible mass axial dispersion and heat radiation, as
heuristic procedures (Aguilar-López, 2007; Akamatsu et al., well as negligible intra-solid and gas-solid transport
2000). This motivates the development of systematic ways to resistance) (Hlavacek, 1970; Rase, 1990), the reactor
design robustly stabilizing temperature controllers with dynamics are given by the spatially distributed dynamic mass
feedforward-based disturbance rejection capability. and heat balances
0 , , 0 1; 1, (1a-b)
In this study, the problem of robustly controlling an ∂ , (1c-d)
exothermic FPB tubular reactor is addressed by combining 0 1; , , ∈
advanced and industrial control ideas. Advanced control is
with boundary and initial conditions

Copyright © 2016 IFAC 278

IFAC DYCOPS-CAB, 2016
June 6-8, 2016. NTNU, Trondheim, Norway

0: 0, , (1e-f) 3.1 Staged model

1: ∂ 0; 0: ,0 (1g-h)
where Following studies that use staged models to describe tubular
, ⁄ , reactors (Deckwer, 1974; Badillo-Hernandez et al., 2013;
/ , / , / , / , / Nájera et al. 2015), the application of spatial finite differences
⁄ , / , / (with domain nodes and two boundary ones) to the
distributed reactor (1) yields the pair of static-dynamic spatial
∆ ⁄ , ⁄ , ⁄
difference equations
, , ⁄ , ,
1 ⁄ , ⁄ 0 ∆ , , 1 ; (3a-b)
∆ ∆ , ; (3c-d)
, [or , ] is the composition (or temperature) time- 0: ; ∆ (3e-f)
varying spatial profile, is the time, is the axial length, is
1: ∆ 0; (3g-h)
the (nearly constant) feed composition, is the (Peclet
0: 0 ; ∈ 1, … , (3i-j)
inverse) heat dispersion number, is the heat transfer
parameter, is the molar feed rate, , is the reaction rate, where
and is the Damköhler number. The exit gas composition ,
( ) is the regulated output, the cooling jacket temperature ( ) ∆ ∙ ∙ ∙ , ∆ ∙ ∙ ∙
is the manipulated input, the measured output is the ∆ ∙ ∙ 2 ∙ ∙
temperature ( ) at axial location to be determined, and the
In compact notation, eq. (3) is written as
feed temperature is the measured disturbance ( ).
0 , ; (4a-b)
As case study, let us consider reactor (1) with an irreversible
first-order exothermic reaction , with Arrhenius , , , , 0 ; (4c-d)
temperature dependency (2a), the parameters (2b) and nominal , , dim dim (4e-g)
inputs (2c): where
, , 1/ exp ⁄ (2a)
,…, , ,…,
ln 25, 50, 0.4, 3 (2b)
̅ 1, 0.57, ̅ 1.75, ̅ 2 (2c) The unique solution [ ] for the composition vector
of eq. (4a) followed by substitution in the heat balance (4c)
This corresponding reactor (1): (i) has the single stable steady-
yields the -dimensional staged model
state profile pair ̅ , ̅ shown in Fig. 2 (drawn with a
200-staged model plus interpolation), and (ii) exhibits strong , , , 0 , dim (5a-b)
parametric sensitivity with temperature hotspot at length ; ; (5c-e)
0.15 . with dynamic (or quasi-static) temperature (or composition
1.0 2.20
) sequence, where
c( s )
0.8  ( s) , , , , ,
2.15
In Fig. 3 the van Heerden (van Heerden, 1958) diagram (heat
0.6 generated Qg and removed Qr versus 1st stage temperature )
c( s )  ( s) for Example (2) is presented for 10 and 200 stages,
0.4
showing that: (i) in both cases the staged model has one stable
2.10 steady-state (SS) ( ), and (ii) the SS with 10
0.2
approximates well (up to admissible deviation) the one with
200.
0.0 2.05
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
120 x N 100
s x N 10 6
N=100

Fig. 2. Steady-state composition-temperature profile pair. 100 Qg

Qr

The problem consists in designing an feedforward-output N  100

80 4
N  10
feedback (FF-OF) robust controller for reactor (1) that Qg Qg
60
manipulates the coolant temperature on the basis of the feed Qr Qr
2
input disturbance ( ) and output temperature ( ) 40 N=10

measurements, so that the exit gas composition is indirectly Qg

regulated about its prescribed value ̅, and the temperature 20 Qr

0

profile is regulated about a compensated setpoint ( ∗ ). 0

1.8 1.9 2.0 2.1 2.2 2.3
1
3. FF-SF CONTROL
Fig. 3. Reactor SS uniqueness according to van Herden
In this section nonlinear feedforward-state feedback (FF-SF)
diagram, on the basis of the (— 100 and — 10)-staged
robust control problem is addressed on the basis of a staged
model (5).
model of the distributed reactor (1).

279
IFAC DYCOPS-CAB, 2016
June 6-8, 2016. NTNU, Trondheim, Norway

The number =10 of stages was determined as follows. First, a simpler robust version, according to the following rationale
the root-mean squared (RMS) error was calculated for the (Porru et al., 2015; Nájera et al., 2015):
(“almost” distributed approximation) 200-stage model over a (i) The dynamic element ( ∗ ) of the concentration regulatory
sample of typical parameter errors and with respect to the controller (7a) is replaced by its static counterpart (eq. 7a
nominal parameters. Then, the RMS error was computed 0), so that the resulting system is robustly passive with relative
for less ( 200) stages until the error with N stages degree equal to zero ( 0).
became comparable ( ) with .
(ii) Then, the dynamical error made in step (i) is compensated
In Fig. 4 are presented: (i) the stable temperature ( ) and by adding a first order lag (9a) with characteristic time ( ∗ )
composition ( ) sequences of the stable steady-state for similar to the one ( ) of the open-loop temperature response.
10, and (ii) the interpolated (continuous line) profiles of and
. The result is the redesigned FF controller (the complete
1.0
derivation is shown in the Appendix B)
∗ ∗ ∗
ci 2.15 ∗ , ∗ 0 ; , ̅ (9a-b)
0.8
i where , ̅ (defined in eq. B1e of Appendix B) yields the
dependency of the static assumption-based temperature
0.6
setpoint on the measured-given data , ̅ . The
ci i
2.10
corresponding, -passivity solvability condition, with 0
0.4
for the input-output pair for ̅, (Khalil, 2002), is
0.2 det , , 0 (10)
where , , is the Jacobian matrix (defined in eq. B1b
0.0 2.05 of Appendix B) of the -algebraic equations that result from
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
s imposing the static assumption on the primary concentration
Fig. 4. Temperature (□) and concentration (○) sequences regulatory controller (7a).
of the unstable steady-state for 10, and interpolated
profiles (—). 3.3 SF temperature tracking controller

3.2 FF concentration regulatory controller The objective is to track the time-varying setpoint temperature
∗
generated by the FF composition controller (9) by
The aim is to maintain the exit gas composition at its nominal manipulating the control input , according to the prescribed
value ̅ by adjusting control according to , i.e., dynamics (with adjustable gain )
∗
̅ (6) , (11)
The enforcement of this condition on the staged reactor (5) The enforcement of eq. (11) on the staged model (5) followed
yields the FF controller ( ∗ and ∗ are defined in eqs. A1a and by solution for yields the SF temperature tracking controller
A1-b of Appendix A) ∗
, , , (12)
∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗
, , ̅ , 0 , (7a-b) Eq. (12) is the unique solution for of the algebraic equation
∗ ∗ ∗ ∗ ∗ ∗
, ̅ ; , , ̅ (7c-d) ∗ ∗
, ,
where
where is the -th entry of vector field of the open-loop
∗ ∗ ∗ ∗
,…, , dim 1 temperature dynamics (5a).
∗ ∗ ∗
The corresponding -passivity conditions are
The corresponding -passivity conditions are
, 1 ⇔ 0 (13a)
∗
, 1⇔ ∑ ∏ 0 (8a) - R-stable, ∈ (13b)
- (7a) R-stable, (8b) Eq. (13a) says that the staged reactor (5) has 1 with
Eq. (8a) states that the -staged reactor model must have respect to the input-output pair , , and eq. (13b) states that
relative degree ( ) equal to one with respect to the input- the ( -1)-dimensional with respect to , must be
output pair , , and eq. (8b) says that the associated zero- stable, depending on the temperature sensor location .
dynamics ( ) must be robustly (R)-stable.
3.4 Cascade FF-SF dynamic nonlinear controller
The FF controller (7) is the dynamical inverse (Hirschorn,
1979) of the process and its dynamical component (7a) is the Assuming that the temperature state sequence and the
related with respect to the input output pair , (Isidori, disturbance are known, the combination of the FF
1988) ( - ). However, controller (7) is not robust because composition regulatory (9) and temperature tracking (12)
the relative degree condition (8a) is poorly met. This is so controllers yields the composition-temperature cascade
because the control is calculated from an ill-conditioned controller.
equation, as the related concentration is small. This obstacle is ∗ ∗ ∗ ∗
circumvented by replacing the poorly robust FF controller with ∗ , 0 ; , ̅ (14a-b)

280
IFAC DYCOPS-CAB, 2016
June 6-8, 2016. NTNU, Trondheim, Norway

∗
, , , (14c) convergence dynamics. The related solvability conditions are
with -pasivity (10) and -passivity (13) conditions. (the Jacobian matrix is defined in eq. B1b of Appendix B):
, , 0; 0; ∈ (17a-c)
3.5 Solvability assessment
which in our case become (17) with 1. The sensor
The set of (10) and (13) -passivity solvability conditions locations for the secondary control (15c) and state estimation
of the FF-SF cascade controller (14) are (17c) (Fernandez et al., 2012) coincide.

det , , 0↔ 0, (15a) The main drawback of the dynamic FF-OF controller (16) is
, 1↔ 0, - (12a) R-stable ∈ (15b-c) its complexity: (i) highly nonlinear interaction, (ii) 2
ODEs (12 for our case study) to be on-line integrated, and (iii)
While conditions (15a) and (15b) are robustly met because the strong dependency on the staged model (5). The overcoming
heat exchange coefficient ( 0) is sufficiently large, of this applicability obstacle is the subject of the next section.
condition (15c) is robustly met only if the temperature sensor
is located adequately (at stage ). 5. SIMPLIFIED FF-OF CONTROLLER
As a suggestive sensor location criterion (to be ratified or In this section, the functioning staged model-based robust FF-
rectified by control functioning -in Section 7-) let us consider OF controller (16) is recovered with a simplified controller
two ones employed in previous studies on staged systems: (i) built according to passivity and observability properties.
the largest absolute value of the temperature gradient |∆ |,
commonly used in distillation column estimation (Fernandez 5.1 Secondary controller redesign
et al., 2012) and control (Porru et al., 2015), and (ii) the stage
Following the approach employed in previous
before the hot spot one with the lowest (absolute value |∆ |
Homopolymerization Reactor (Gonzalez and Alvarez, 2005)
of the temperature slope change) “concavity”, a heuristics
criterion employed to avoid thermal runaway in tubular and continuous exothermic reactors with isotonic kinetics
(Díaz-Salgado et al., 2012) studies, here the secondary
reactors (Bashir et al., 1992). In Fig. 6 are plotted the
component (14c) of the cascaded controller (14) is redesigned
temperature spatial sequence and its gradient, and concavity
sequences of the nominal SS, yielding that the sensor must be to make a simplified FF-OF cascade controller, according to
the following rationale:
located at stage 1 or 2.
Temperature spatial sequence Recall the -stage model (5) and express its -output dynamic
Temperature gradient spatial sequence
2.15
Temperature concavity spatial sequence
2.0 2.0 in the form
3 4
1.5 1.5
, , , 1 (18a-b)
i
2 5   i  2 i where
6 1.0 1.0 , , , , , ̅, 0 (19a-b)
1
7
is defined after (12), is an observable input, and ,
2.10 0.5
0.5
satisfies the matching condition (Sepulchre, 2011). The
8
9 10
elimination of the static nonlinear component (19a) in eqs. (18,
0.0
0.0
19) yields the simplified model for temperature control design,
0.0 0.1 0.2 0.3 0.4 0.5
s
0.6 0.7 0.8 0.9 1.0 with unmeasured-observable input .
Fig 6. Temperature, temperature gradient and concavity The enforcement of the tracking condition (11a) upon model
sequences of the staged reactor for 10. (18a-b) yields the controller (12) in the -dependent form
4. FF-OF CONTROL (18a), and a convergent estimate ̂ of input of model (19a) is
given by the reduced-order observer (20b) with adjustable (up
The combination of the cascade controller (14) with a to measurement noise) exponential convergence rate
geometric observer, with robust second-order detectability
∗ ∗
innovation structure (Fernandez et al., 2012), yields the robust ∗ / (20a)
FF-OF geometric controller , 0 0, ̂ (20b)
∗ ∗ ∗ ∗ The combination of control (20a) (with )̂ with observer
∗ , 0 ; , ̅ (16a-b)
(20b) yields the dynamic temperature tracking controller
, , , , (16c)
, , ;̂ 0 , 0 0 (21a)
∗ ∗ ∗
̂ , ̂0 ̂ (16d) ∗ / (21b)
, , , ∗, (16e) with one linear ODE, and similar behavior than the one (16c-
e) of its detailed model-based counterpart (16). This control
where
∗
has antiwindup protection because the -dynamics (21a) runs
, , ̂ 2 12, 1, 3 , ∈ 5,10 regardless of control saturation (Díaz-Salgado et al., 2012).
(or ), defined in eq. A2 of Appendix A, is the
5.2 PI temperature controller with FF setpoint compensation
proportional (or integral) nonlinear gain, (or ) is the
damping factor (or characteristic frequency) of the output The combination of the primary (16a-b) and redesigned
secondary (21) control yields the simplified FF-OF control:

281
IFAC DYCOPS-CAB, 2016
June 6-8, 2016. NTNU, Trondheim, Norway

∗ ∗
∗ , , ̅
( -feedforward) (22a-b) simulated measurement noise yielded (after a few iterations)
0 ( -feedback)
, 0(22c) the following gains for control (22) ( 0.5: open-loop
∗ ∗ ∗
∗ / (22d) characteristic time)
with: (i) two linear ODEs, and (ii) and similar behavior than , , , 5, 4
∗
the one of its detailed model-based counterpart (16) with
2 ODEs (12 for case example). In Fig. 8 are presented the behaviors of the standard (with fixed
setpoint) (24) and proposed (with setpoint compensation) (23)
In the absence of saturation, controller (22) has the PI form controllers when the reactor is subjected to a sequence of step
∗
∗
∗
, ∗
0 ∗
; , ̅ (23a-b) feed temperature ( ) changes, showing that the proposed
∗
∗
, ∗
(23c) controller (24): (i) robustly controls the reactor, (ii) reduces, in
comparison to its standard PI counterpart (25), by 95% the
where
variability of the exit composition ( ), and (iii) keeps
∗
, , / , 1/ bounded (below 5 K) the temperature excursions of the hot
spot ( ).
and (or ) is the proportional gain (or reset time). This Step changes Nominal values
signifies that the proposed controller (23) is an upgraded 1.8

version the conventional PI: e 1.7

1.6
∗
(24)
0.10
temperature control employed in industrial reactors, with the
upgrade consisting in: (i) conventional-like tuning guidelines c10 0.05
PI PI+setpoint compensation
coupled with closed-loop robust stability assessment, (ii) 0.00
2.20
antiwindup protection, and (iii) load measurement-based FF limit in hot spot temperature

dynamic setpoint compensation. 3 2.15

2.10
6. CONTROL IMPLEMENTATION 2.2

The FF-OF controller (22) (with FF-based setpoint  10 2.1

compensation) is tested and compared with its conventional 2.0
counterpart (25) (with fixed setpoint), for the case study (2) 2.2
with 10 and sensor at 1. The aim is to control y* 2.1
temperature peaks (not more than 5 degrees in the hot spot)
2.0
and regulate the exit concentration ( ̅ 0.06), in spite of 2.1
step feed temperature disturbances. c 2.0

6.1 The static FF component 1.9

0 10 20 30 40 50 60 70 80 90 100
t
The curve , ̅ of the static FF component (22b) was
calculated over a feed temperature interval of 20 , then the Fig. 8. Closed-loop behavior with PI temperature control and
map , ̅ was approximated by the regression-based step feed temperature disturbances, with (•••) and without (—
polynomial curve ) setpoint compensation.
, ̅ (25a) 7. CONCLUSIONS
0.228, 1.432, 0.056, 0.085(25b)
The problem of regulating temperature profile excursions and
In Fig. 7 are plotted the curves and . exit concentration for a gas-phase packed tubular reactor has
2.2
been studied. On the basis of a staged model, advanced control
 (d , z )
theory was applied to solve the nonlinear FF-OF robust control
ˆ(d , z )
problem, yielding passivity and observability solvability
2.1 conditions with sensor location criterion. Then, the behavior
ys of the advanced controller was approximated with a realization
built according to a simplified model tailored according to
passivity and observability properties. The result was a PI
2.0
temperature controller with dynamic setpoint compensation
driven by the measured feed temperature disturbance. The
1.6 1.7 d 1.8 1.9 setpoint compensator had a precomputed detailed model-based
Fig. 7. Static element , ̅ (22b) (—) of the setpoint FF nonlinear component followed by a first-order linear lag.
compensator (with 10), and its approximation The proposed approach was tested and illustrated with a case
, (25) (). example with numerical simulations.
REFERENCES
6.2 Control testing
Aguilar-López, R. (2007). Outlet Temperature Regulation for
The application of conventional-like tuning guidelines a Class of Plug Flow Chemical Reactor via Nonlinear
(Gonzalez and Alvarez, 2005; Díaz-Salgado et al., 2012) with Feedback. Int. J. Chem. React. Eng., 5(1).

282
IFAC DYCOPS-CAB, 2016
June 6-8, 2016. NTNU, Trondheim, Norway

Akamatsu, K., Lakshminarayanan, S., Manako, H., Takada, Rase, H. F., (1990). Fixed-bed reactor design and diagnostics:
H., Satou, T., Shah, S. L. (2000). Data-based control of an gas-phase reactions, Butterworth Publishers, USA.
industrial tubular reactor. Control Engineering Practice, Sepulchre, R., Jankovic, M., Kokotovic, P.V., (2011).
8(7), 783-790. Constructive Nonlinear Control, Springer, London.
Badillo-Hernandez, U., Alvarez-Icaza, L., Alvarez, J. (2013). van Heerden, C. (1958). The character of the stationary state
Model design of a class of moving-bed tubular of exothermic processes. Chemical Engineering Science, 8(1),
gasification reactors. Chem. Eng. Sci., 101, 674-685. 133-145.
Bashir, S., Chovln, T., Masri, B. J., Mukherjee A., Pant A., Sen
S., and Vijayaraghavan P., (1992). Thermal Runaway Appendix A. FF CONTROL AND GEOMETRIC
Limit of Tubular Reactors, Defined at the Inflection Point OBSERVER FUNCTIONS
of the Temperature Profile. Ind. Eng. Chem. Res., 31,
A.1 FF concentration regulatory controller (7)
2164-2171.
Brosilow, C., Joseph, B. (2002). Techniques of model-based
The right-hand-side expressions of (7a) and (7c) are defined as
control. Prentice Hall Professional. follows
Deckwer, W., (1974). The Backflow Cell Model-Applied to
∗ ∗ ∗ ∗ ∗
Non-isothermal Reactors, The Chemical Engineering , , ̅ , , , ̅ , (A1a)
Journal, 8, 135-144. ∗ ∗ ∗ ∗
, ̅ ,…, , ̅ (A1b)
Del Vecchio, E., Petit, N., (2005). Boundary control for an ∗
, ̅ ∗
industrial under-actuated tubular chemical reactor,
Journal of Process Control, 15, 771-784. A.2 FF-OF geometric controller (16)
Díaz-Salgado, J., Alvarez, J., Schaum, A., Moreno, J. A.
(2012). Feedforward output-feedback control for The proportional and integral nonlinear gains of the OF
continuous exothermic reactors with isotonic kinetics. geometric controller are defined as follows
Journal of Process Control, 22(1), 303-320.
Fernandez, C., Alvarez, J., Baratti, R., and Frau, A. (2012). , , , , , , , , (A2a)
Estimation structure design for staged systems. Journal of , , , , (A2b)
Process Control, 22 (10), 2038-2056. , , , , , (A2c)
Gonzalez, P. and Alvarez, J. (2005). Combined 2 1 , , 0,1 (A2d)
Proportional/Integral-Inventory Control of Solution where (or ) is the damping factor (or characteristic
Homopolymerization Reactors. Ind. Eng. Chem. Res., 44, frequency) of the output convergence dynamics.
7147-7163.
Hlavacek, V. (1970). Aspects in design of packed catalytic Appendix B. FF CONTROL REDESIGN
reactors. Industrial & Engineering Chemistry, 62(7), 8-
The enforcement of the static ( ∗ 0) assumption on eq. (7a)
26.
followed by substitution of ( ∗ , ∗
( , , yields the -
Hirschorn, R. M. (1979). Invertibility of Nonlinear Control
aequation algebraic set
Systems. SIAM Journal of Control and Optimization. 17
(2), 289-297. , , ̅ 0, , (B1a)
Isidori, A. (1989). Nonlinear Control Systems, Springer- where , , ̅ : , , , ∗
, , ̅ 0
Verlag, Berlin.
Jaisathaporn, P.,Luyben, W. L. (2004). Dynamic Comparison Provided the map is , -invertible, i.e.,
of Alternative Tubular Reactor Systems. Ind. Eng. Chem. , , , , ̅ 0 (B1b)
Res.,43 (4), 1003-1029.
Jakobsen, H. A. (2008). Chemical reactor modeling. for given ( , ̅), the solution for , of eq. (B1a) is
Multiphase Reactive Flows, Berlin, Germany: Springer- denoted by
Verlag. , ̅ ≔ , ̅ , , ̅
Jensen, K. F., and Ray, H. W. (1982). The bifurcation behavior and
of tubular reactors. Chem. Eng. Sci., 37 (2), 199-22. , ̅ , ̅ , , ̅ (B1c)
Jorgensen, S. B. (1986). Fixed bed reactor dynamics and , ̅ , ̅ ,…, , ̅ , , ̅ (B1d)
control-a review, Proc. IFAC Control of Distillation , ̅ ≔ , ̅ (B1e)
Columns and Chemical Reactors, 11-24.
Khalil H. K. (2002). Nonlinear systems, Prentice Hall, NJ. with being the -th entry of
Nájera, I. Álvarez, J., Baratti, R. (2015). Feedforward output- To compensate for the omission of the accumulation term
feedback for a class of exothermic tubular reactors, ( ∗ 0) in the calculation of the preceding static setpoint
Proceedings of the 9th IFAC ADCHEM, Whistler, approximation, let us add a first order lag (B1f) to (B1e) in
Canada, 48(8), 1075-1080. order to get the dynamic stable the setpoint compensator (with
Porru, M., Baratti, R., Alvarez, J. (2015). Energy Saving gain ∗ )
through Control in an Industrial Multicomponent ∗ ∗ ∗ ∗
Distillation Column Proceedings of the 9th IFAC ∗ , 0 ,; , ̅ (B1f-g)
ADCHEM, Whistler, Canada, 48(8), 1138-1143.

283