Sunteți pe pagina 1din 14

Journal of Process Control 17 (2007) 463476

www.elsevier.com/locate/jprocont

Constructive control of continuous polymer reactors


Jesus Alvarez *, Pablo Gonzalez
Universidad Autonoma Metropolitana-Iztapalapa, Depto. de Ingeniera de Procesos e Hidraulica, Apdo. 55534, 09340 Mexico, DF, Mexico
Received 16 November 2005; received in revised form 9 August 2006; accepted 26 September 2006

Abstract
The problem of controlling (possibly open-loop unstable) continuous free-radical solution polymer reactors with temperature, level
and ow measurements is addressed. The application of a nonlinear constructive control procedure, with emphasis on the attainment of
linearity, decentralization, robustness and model independency features, yields: (i) a measurement-driven control scheme with PI volume
and cascade temperature loops, a ratio-type feedforward free monomer controller, and a material balance molecular weight (MW) controller, and (ii) a closed-loop nonlocal stability criterion coupled with conventional-like tuning guidelines. The methodological developments connect the nonlinear geometric and constructive control design techniques with conventional-like schemes employed in industrial
polymer reactor control. The proposed approach is illustrated and tested with a representative example through simulations.
 2006 Elsevier Ltd. All rights reserved.
Keywords: Nonlinear control; Constructive control; Polymer reactor control

1. Introduction
A wide variety of materials and products are manufactured in continuous free-radical polymer reactors. Due to
the strong exothermicity of the reaction and the presence
of the gel eect, which causes reaction autoacceleration
accompanied by viscosity increase and heat removal capability decrease, the reactor exhibits strongly nonlinear
behavior, with asymmetric input-output coupling, multiplicity of steady-states, and parametric sensitivity [1,2].
At the cost of less productivity, the reactors can be operated in open-loop stable regime by choosing the conversion
suciently low. In industrial practice, these reactors are
controlled with volume and cascade temperature linear PI
loops, and the monomer content and molecular weight
(MW) are regulated by adjusting the monomer and initiator and/or transfer agent dosages via supervisory or advisory control schemes. If the reactor is operated with or
close to gel eect, a free monomer controller driven by

Corresponding author. Tel.: +52 55 5804 4958; fax: +52 55 5804 4900.
E-mail address: jac@xanum.uam.mx (J. Alvarez).

0959-1524/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jprocont.2006.09.007

an on-line measurement must be used [3]. Thus, the objective of an advanced joint process-control design scheme is
to attain an operation with the best compromise between
safety, operability, productivity, and quality in the light
of investment and operation costs.
The polymer reactor control problem has been the subject of extensive theoretical, simulation and experimental
studies, the related state of the art can be seen elsewhere
[4], and here it suces to mention that: (i) only parts of
the multi-input multi-output (MIMO) control problem
have been addressed, (ii) a diversity of techniques has been
employed, including linear PI decoupling [2] and model
predictive (MPC) [5], as well as nonlinear geometric [3,6
9], MPC [10], and calorimetric [11,12] control techniques,
and (iii) the controllers have been implemented with
open-loop [7], extended Kalman lter (EKF) [10], and
Luenberger [8,9] nonlinear observers. Even though valuable insight and understanding have been gained, the
consideration of these highly nonlinear interactive and
model-dependent controllers raises serious complexity
and reliability concerns among industrial practitioners.
Recently, a combined PI-inventory approach was used to
draw a control scheme with linear components and reduced

464

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

model dependency [13]. However, the design is highly


dependent on process insight, and lacks formal connections with previous geometric control studies for polymer
reactors.
From the nonlinear control theory we know that [14
18]: (i) geometric controllers may be poorly robust, (ii)
optimal stabilizing state-feedback (SF) controllers are
inherently robust and passive (i.e., have relative degrees less
or equal to 1, and stable zero-dynamics) with respect to
some regulated output, and their measurement-driven
(MD) implementation requires closed-loop detectability
with respect to the measured output, and (iii) the robust
closed-loop nonlocal stability assessment can be handled
within an input-to-state (IS) stability framework [19]. Since
the analytic construction of an optimal controller requires
the solution of a HamiltonJacobiBellman nonlinear partial dierential equation, the direct optimality approach is
not a tractable procedure, and the removal of this obstacle
is a central aim of the constructive control method [1618].
In this work, the problem of controlling (possibly openloop unstable) continuous free-radical solution polymer
reactors with temperature, level and ow measurements is
addressed. First, geometric control is applied to characterize the solvability of the problem. Then, the application of
a constructive control procedure yields a MD control
scheme with: (i) decentralized linear PI volume and cascade
temperature loops, a ratio-type free monomer controller,
and a material balance MW controller, and (ii) reduced
model dependency, especially on the components that perform the stabilization task. An input-to-state (IS) stability
framework is employed to draw a closed-loop nonlocal
robust stability criterion coupled with conventional-like
tuning guidelines. The approach is illustrated and tested
with a representative example through simulations.
2. Control problem

v_ qm qs  em =qm r  q : fv ; zv y v v
1a
m_ r qm qm  qm=v : fm ; zm m
1b
T_ Dr  qm cm qs cs T  T e  U T  T j =C : fT ;
zT y T T

r fr v; m; T ; I; s;

Q Dr

U fU v; m; T ; T j ; s;

yj T j

zi i

2ab

C fC v; m; s

2cd

p fp v; m; s
ri ET I : fri T ; I

2e
2f

r0 cd ET I it T ; m; s;

r : f0 v; m; T ; I; s

2g

For the sake of simplicity, the function dependencies will


be occasionally omitted, and /(x1, . . . , xn) will be simply
written as /.
The states (x) are: the reactor (T) and jacket (Tj) temperatures, the volume (v), the unreacted monomer (m), solvent
(s) and initiator (I) masses, as well as the (number-average)
MW inverse i M 1
n and its polydispersity (p). The measured exogenous inputs (d) are: the reactor (Te) and jacket
(Tje) feed temperatures, and the solvent (qs) volumetric
owrate. The regulated outputs (z) are: the temperature
(T), the volume (V), the monomer content (m), and the
MW inverse (i). The measured outputs (y) are: the temperature (yT), the volume (yv), and the jacket temperature (yj).
The control inputs (u) are: the coolant volumetric owrate
(qj) through the jacket circuit, the exit owrate (q), the
monomer owrate (qm) and the initiator mass feedrate
(wi). The (number-average) MW inverse (i) state, with linear dynamics (1e) in i, is used for control simplication and
stability analysis purposes. In practice, one is interested in
the conversion (c) and the solid fraction (r) (q is the mixture density):
zr r p=vq

In vector notation, the reactor control system (1) is written as follows:


x_ f x; d; u; pr ;

y C y x;

z Czx

where
0

u q; qm ; qj ; wi ;

x v; m; T ; T j ; i; I; s; p ;
0

y y v ; y T ; y j ;

z zv ; zm ; zT ; zi

f fv ; fm ; fT ; fj ; fi ; fs ; fp ; d T e ; T je ; qs ;
 u; pr 0; y C y x; z C zx
f x; d;

1e

pr is the vector of model parameters,  denotes the steadystate nominal value of (), and x can be nonunique, either
stable or unstable. As it can be seen in (1), this system is
highly nonlinear and interactive. The denition of nonlocal
stability that underlies the present study is stated next.
Consider a nonlinear system with time-varying exogenous input de(t):

1f

e_ fe e; d e ;

1c

T_ j U T  T j  qj cj T j  T je =C j fj ;

1g
1h

where the polymerization rate (r), initiation rate (ri), freeradical generation rate (r0) heat generation rate (Q), heat
transfer coecient (U), heat capacity (C), and polymer
mass (p), are set by the nonlinear functions

zc c p=vq  s;

Consider the class of continuous stirred-tank reactors


where an exothermic free-radical solution homopolymer
reaction takes place. Monomer, solvent and initiator are
fed to the tank, and heat exchange is enabled by a cooling
jacket. Due to the gel eect [20], the reactor can present
steady-state multiplicity [1,2]. From standard free-radical
polymerization kinetics [1], and viscous heat exchange considerations [2,3], the reactor dynamics are described by the
following mass and energy balances (the model functions
are given in Appendix A):

i_ r=pi  it cd E=pI : fi ;
I_ ri wi  qI=v : fi

s_ qs qs  qs=v : fs
p_ r=pf2  r0 =r=ip  2r=r0 mw ig : fp

1d

e0 e0 ;

f e 0; 0 0

4a

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

The steady-state e = 0 is input-to-state (IS) stable [19] if


there is a KL-class (increasing-decreasing) function a and
a K-class (increasing) function c so that
jetj 6 aje0 j; t  t0 ckd e tk;

4b

tP0

kd e tk sup jd e tj
t

where a (or c) bounds the transient (or asymptotic) response. The (necessary and sucient) Lyapunov characterization of the IS stability property is given by [19]
a1 jej 6 V e 6 a2 jej; V_ a3 jej a4 kd e k
4c
where V is a positive denite radially unbounded function
and ai is a K-class function.
Our problem consists in designing a feedback controller
that, driven by the measurements (y and d), regulates the
reactor operation about a (possibly open-loop unstable)
nominal steady state x. We are interested in: (i) solvability
conditions with physical meaning, (ii) closed-loop stable
functioning, (iii) a systematic construction-tuning procedure, and (iv) the attainment, as much as possible, of
linearity, decentralization, robustness and modeling independency features.
3. Solvability assessment
Assuming that the detailed reactor model (1), (2) and its
state (x) are known, in this section the related nonlinear
(feedforward) FF-(state-feedback) SF control problem is
addressed with geometric [14] and constructive methods
[18]. The purposes are the identication of the underlying
solvability conditions and of the behavior attainable with
robust FF-SF control.
3.1. Relative degrees (RD) and zero-dynamics (ZD)
Regarded individually, the coolant ow-temperature
inputoutput pair qj  zT has relative degree (RD) equal
to 2, the associated controller (qj) requires the monomer
feed (qm) control time derivative, and consequently, the
4-input (u) 4-output (z) reactor system (1) does not have
RDs, meaning that the FF-SF problem cannot be solved
with static control [14]. From the application of the
dynamic extension procedure [14], the next proposition follows [the function f(x1, . . . , xn) is said to be xi-monotonic if
oxi f is of one sign]:
Proposition 1 (Proof in Appendix B). With the dynamic
extension (5a), the augmented reactor system (1, 5a) has
relative degree vector j (5b)
_ q_ m tq ; tqm
q;

5a

j jv ; jm ; jT ; ji 2; 2; 2; 2

5b

if and only if, in a compact set about the nominal operation:


m=v 6 qm ;

T je 6 T j

fT is T j -monotonic;

6ab
fi is I-monotonic: r

6cd

465

Physically speaking, the preceding conditions are always


met because: the free monomer is part of the reacting mixture [m < qmv ) (6a)], (ii) there is heat exchange between
the reacting mixture and the cooling jacket uid
[Tje < Tj ) (6b)], (iii) the heat exchange rate U(Tj)(T  Tj)
is uniquely determined by the jacket temperature
T j oT j U T  T j < 0 ) 6c, and (iv) the MW decreases
with the initiator content [oIfi0 ) (6d)].
The corresponding zero-dynamics (ZD) are given by
s_ skoq s; z; qs ; pr qs qs
p_

kor s; z; pr p

7a

 2 mwi

7b

where fi is dened in (1), and fr and fp are given in Appendix A


koq qo =v;

kor ro =po ;

 T ; loI s; s
ro fr v; m;

 s
po fp v; m;
 o
qo 1  em ro =qm qs =1  m=p
 T ;i; I; s 0 ) I loI s
f i v; m;
and loI s denotes the solution for I of the algebraic equation fi = 0.
Physically speaking, the ZD (7) represent the reactor
behavior [13] in perfect material balance control [21], with
the mass and the energy (u) delivered to the system exactly
balanced against the load demand z; d without gel eect,
or equivalently, without the sole potentially destabilizing
mechanism at play [3]. The ZD stability property is formally stated in the next proposition.
Proposition 2 (Proof in Appendix B). The ZD (7) are IS
stable, with respect to the exogenous input dz = (d,z,pr), if in
a compact set about the nominal operation:
skoq is s-monotonic; koq > 0;

kor > 0: r

8ac

In a practical reactor, the preceding conditions are always


met because: the residence time is strictly positive (8b), kor is
similar to koq r=p  q=v, and the presence of the solvent is
meant to attenuate or dominate the destabilization by gel
eect, or equivalently, the solvent per se does not induce
steady-state multiplicity (8a).
For its use in subsequent stability analysis, let us recall
from the proof of the last proposition the ZD IS Lyapunov
function (LF) characterization:

V o e2s e2p =2 es s  s; ep p  p
9a
_V o ao es ; ep ; do bo es ; do P 0 8jes j 6 co kdo k
s
jep j P cop kdo k
9b
where the stabilizing (ao) and potentially destabilizing (bo)
dissipation rates, as well as the asymptotic gains cos ; cop are
given in Appendix B (B6), do is the exogenous input for the
ZD, and ea denotes deviation of a from its nominal value:
0
0
do dop ; dqs ; d_ qs ;

dop ev ; em ; eT ; ei ; e0pr

ea a  
a

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

466

3.2. FF-SF geometric control

(1cf), associated with the inputoutput pair (qj, wi) 


(yT, yi), in the following way:

From the enforcement of the linear, noninteractive, pole


assignable (LNPA) output regulation dynamics

T_ aT v; m; T ; T j ; sT j bT v; m; T ; T j ; r; s; qm ; T e ; qs 12a
T_ j aj T j ; T je qj bj v; m; T ; T j ; s
12b

ea fca xca e_ a xca 2 ea 0

a v; m; T ; i

10

the FF-SF nonlinear geometric dynamic controller follows


[the functions are given in (B4) of Appendix B]
q_ lgq v; m; T ; T j ; i; I; s; q; qm ; T e ; qs ; q_ s

11a

q_ m lgqm v; m; T ; T j ; i; I; s; q; qm ; T e ; qs ; q_ s

11b

wi lgwi v; m; T ; i; I; s; q; qm ; T e ; qs

11c

qj lgqj v; m; T ; T j ; i; I; s; q; qm ; T e ; T je ; qs ; T_ e ; q_ s

11d

From well-known results in geometric control [14,15],


the IS stability of the resulting closed-loop system is a consequence of the IS stability of the LNPA (10) and ZD (7).
This controller: (i) has two dynamic components (11ab)
and two static ones (11cd), (ii) has a dynamic-to-static
component cascade interconnection, (iii) requires the derivatives of two exogenous inputs (Te and qs), and (ii) the four
components depend on the entire state-input pair xu.
These dynamic, interaction and model dependency features
are depicted in Fig. 1.
3.3. Passive FF-SF control
From a constructive control perspective [16,18], the lack
of robustness is a major drawback of the geometric controller (11), because it is not underlain by a structure with
all the relative degrees equal or less than 1. Following the
constructive control approach, this high RD obstacle for
robustness is then removed by passivation via backstepping
[16,18]. For this aim, let us rewrite the four-state subsystem

i_ ai v; m; T ; sI  bi v; m; T ; r; i; s
I_ ki v; T ; qI wi

12d

p_ kp v; m; T ; r; s; I; ip cp v; m; T ; r; i; I; s

12e

12c

where
aT U =C;
ki kq E;

aj cj T je  T j =C j
ai cd E=p;

13ab

kq q=v

13ce

bT Dr  qm cm qs cs T  T e  UT =C
bj U T  T j =C j ;

bi kr i  it

13f
13gh

kp kr 2  r0 =r=i

13i

cp kr 2r=r0 mw i;

kr r=p

13jk

Introduce the LF [Vo has been dened in (9)]


V p V pv V pm V pT V pt V o ;
ej T j  T j ;
V

p
v

e2v =2;
V o cvs e2s

V pT e2T e2j =2

ei I  I 
p
m

e2m =2;

14a
p
i

e2i

e2i =2

cvp e2p =2

14b

where T j (or I*) is the jacket temperature (or initiator content) virtual control (i.e., setpoint).
From the enforcement of the dissipation rates
V_ pv k v e2v ;

V_ pm k m e2m

V_ pT k T e2T  k j e2j ;

15ab

V_ pi k i e2i  k i e2i

15cd

upon the reactor dynamics (1ab, 12), the nonlinear static


FF-SF passive controller follows:
q r  k m em qm k v ev qs  em r=qm =qm  m=v

16a

qm r  k m em mk v ev qs  em r=qm =v=qm  m=v


16b
_

16cd

qj T_ j  k j ej  bj  aT eT =aj

16e

T j

16f

wi I  k i ei ki I ai ei ;

I bi  k i ei =ai

bT k T eT =aT

or equivalently,
q lq v; m; T ; I; s; qs ;

qm lqm v; m; T ; I; s; qs

17ab

qj lqj v; m; T ; T j ; i; I; s; T e ; T je ; qs ; T_ e ; q_ s

17c

wi lwi v; T ; m; i; I; s; T e ; qs

17d

in a form obtained after performing the time derivations of


(16d,f). Take the derivative of the overall LF (14), recall the
ZD dissipation rate V_ o (9b), and apply standard Lyapunov
stability arguments [19,22] to draw the dissipation rate
Fig. 1. Dependency diagram of the nonlinear geometric (11) and passive
(16) controllers.

V_ p V_ pv V_ pm V_ pT V_ pi V_ o < 0
0
8jej P ap kdk; d d0 ; d_ 0 ; e0
d

pr

18

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

implying the IS stability of the closed-loop reactor (1) with


the passive controller (17).
Comparing with the geometric controller [(11) and
Fig. 1], the passive controller (17): (i) has only static elements, (ii) depends on the derivatives of the exogenous
inputs Te and qs, (iii) is considerably less interactive, and
(iv) is less model dependent, in the sense that the volume
(17a) and monomer (17b) controllers do not need partial
derivatives of the polymerization rate (fr) and heat transfer
(fU) functions. These dynamic and interaction features of
the passive controller are depicted in Fig. 1.

467

equipment characteristics and operating conditions provides a fundamental connection between reactor process
and control designs [12]. The a posteriori verication of
the control robustness property yielded a connection between passive and MPC polymer reactor control designs.
Finally, it must be pointed out that the IS stable reactor
dynamics (7), of the (exact model-based) nonlinear passive
controller (17) represents: (i) the behavior attainable with
any robust controller, and (ii) the recovery target for the
MD control design of the next section.
4. Measurement-driven control

3.4. Optimality and connection with MPC


From standard arguments in constructive nonlinear
control [16,18], the passive controller (17) is an optimal stabilizing controller with respect to the objective function
Z 1
J 0; 1
Lp x; d; up Ls x; d; up ; us  dt
19

In this section, the closed-loop behavior of the exact


model-based passive nonlinear controller (17) is recovered
with: (i) an interlaced control-observer design on the basis
of control model with suitable detectability structure, and
(ii) emphasis on the attainment of linearity, decentralization, and model independency features.

up q; qm ; T j ; I  ; us qj ; wi
Lp Lv Lm LT Li ;

Ls Lj Li

with primary (Lp) and secondary (Ls) Lagrangian terms


2
2
2
2
Lv e2v k 2
v fv ; Lm em k m fm ; LT eT k T fT ;
2
2
2
2

Li e2i k 2
i fi ; Lj ej k j fj ; Li ei k i fi ; f T aT T j bT aT ej ;

fi ai I   bi ai ei ; fv ; fm ;fT ; fi 0 : fpx ; fj ; fi 0 : fsx

The objective function J is meaningful, in the sense that the


regulated outputs and control deviations are eectively
penalized [16]. This is so because the primary (or secondary) control up (or us) and the function fpx (or fsx ) are in
one-to-one correspondence, and consequently, penalizing
fpx (or fsx ) is equivalent to penalizing up (or us). This optimality property veries the robustness property of the passive controller (17), and above all reveals a key connection
with MPC: the analytically constructed passive controller
(17) is equivalent to a particular choice of input-output
weighting scheme in an unconstrained model-based MPC
design over an innite receding horizon [23]. That optimal
controllers are robust with respect to the particular choice
of weighting function, is a well known fact in linear [24]
and nonlinear [16,17] control theory.
3.5. Concluding remarks
The solvability of the robust FF-SF reactor control
problem has been established with conditions (6, 8) that
bear physical meaning, and these conditions: (i) are generic
in the sense that they are met by the entire class of adequately designed solution homopolymer reactors (1),
regardless of the particular model employed, and (ii) are
in agreement with well-known facts and arguments in polymer kinetics [25], reaction engineering [3,26], and material
balance control [13,21]. Moreover, the quantitative assessment of the control solvability conditions in the light of

4.1. Control model


From previous polymer reactor nonlinear estimation
and control studies [12,27,28], we know that the estimation
model and its structure are design degrees of freedom that
can be exploited to attain maximum estimator robustness
in the light of a specic estimation and objective. Following
these ideas, recall the reactor subsystem (1) and the subsystem form (12), replace the nonlinear function pair bT  bj
(13) by its image value bT  bj, introduce the map bv = bv
(21c) in the volume dynamics (1a), regard the derivatives
of the jacket temperature and initiator setpoints as dynamical states (bj and bi ) under assumption (20j) (which is
standard in signal processing to draw derivative estimates
[13]), and obtain the control model
v_ q bv ;

T_ aT T j bT ;

y v v;

yT T
20ab

T_ j aj qj bj ;

yj T j

m_ kq m qm qm  r;

s_ kq s qs qs

bj ; bi

I_ ki I wi ;

20c

T_ j ; I_ 

20fg

i_ ai I  bi kr i ci fi
p_ kp p cp fp ;

b_ v ; b_ T ; b_ j ; b_ j ; b_ i

20de

20h
0

20ij

where the function set {aT, aj, ai, bi, cp, ki, kq, kp} has been
dened in (13), and
^  a ;
^ =C
aT U
T
0

^  a
aj ^cj T je  T j =C
j
j
0

b bv ; bT ; bi bv ; bT ; bj : b
bv qm qs  em r=qm : bv r; qm ; qs ;

21a
21b

c i ai I kr i t
21cd


 =CCT


r fC j bj CbT U
j cj qj T je  T je  T j  T j 
cm qm cs qs T e  T g=D : br y; b; m; s; u; d

21e

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

468

aT (or aj) is an approximation of the steady-state value of


the function aT (or aj). Knowing that the passive controller
(17) does not depend on the polydispersity state (p), this
state has been introduced for monitoring and model calibration purposes.
From the control model (20) the auxiliary estimator
follows

Since the detectability property is not invariant under


feedback control, the estimator convergence will be
assessed in closed-loop regime. For this aim, let is introduce the estimator Lyapunov function V^ , with a passive
structure that is compatible with the one (14) of the nonlinear passive controller (17) [~ ^   denotes the estimation error of ()]:

bv q y_ v ;

22ab

V^ V^ v V^ m V^ T V^ t V^ o

22cd

V^ v ~b2v =2;

bT y_ T  aT y j


bj y_ j  aj qj ;

b :

bj ; bi

m_ qm=v qm qm  r;

T_ j ; I_  : y 

s_ qs=v qs qs

22ef

r br y; b; m; s; u; d

22g

I_ q=v Ey T I wi
i_ fi v; T ; r; m; s; I; i

22h
22i

p_ fp v; T ; r; m; s; I; i; p

22j

This dierential-algebraic system characterizes the control


model detectability structure [27,28]: (i) the unknown input
(b, b*) is instantaneously observable [29] because it is timewise determined by y; y_ ; y  , and (ii) the states (m, s, I, i, p)
can be reconstructed by integrating subsystem (22ej) driven by (y, b) [12], provided the subsystem is closed-loop
IS stable. Accordingly, the reaction rate pair rU (2a, c)
can be on-line reconstructed without needing the reaction
rate-heat exchange function pair (fr, fU), and this is model
independency feature agrees with the calorimetric estimation-based control approach for exothermic reactors
[12,30].

24a

~ 2 =2
V^ m m

24bc

V^ T ~b2T ~b2j ~b2


j =2

24d

V^i ~i2 ~I 2 ~b2


i =2;

~2 =2
V^ o ~s2 p

24ef

The derivation of V^ followed by the substitution of the


control model-estimator pair (22)(23) yields the dissipation rate
V^_ V^_ v V^_ m V^_ T V^_ i V^_ o
V^_ v xv ~b2v  b_ v ~bv ;

25a

~ wm  ~r
~ 2 m~
V^_ m kq m

25bc

_ ~
_~
_  ~
V^_ T xT ~b2T  xj ~b2j  xj ~b2
j  bT bT bj bj bj bj
V^_ k ~i2  k ~I 2  x ~b2  ^i~i~k b_  ~b

25d

~s p
~~cp  p
^~kp
~2 ~sw
V^_ o kq~s2  kp p

25f

wm qm qm ; ws qs qs ; b_ T T  aT T_ j ;
_
b_ j T j b_ i I 
b_ v v q;

25e

b_ j T j  aj q_ j ;

4.2. State estimator


Recall the auxiliary estimator (22), replace its algebraicdierential subsystem (22ad) by the battery (23ae) of
linear reduced-order single-input observers [30,31], and
obtain the state estimator for the reactor-control model
pair (1,22):
^
bv vv x v y v
bT vt x T y T
v_ T xT vT  xT xT y T aj y j ; ^
^
v_ j xj vj  xj xj y aj q ; bj vj xj y
v_ v xv vv  xv xv y v  q;


v_ j x vj  x2
j Tj ;

v_ i

x vi


x2
i I ;

23a
23b
23c

^
bj vj xj T j
^
b v  x  I 

^ v  ^r;
^_ qm qm  qm=y
m

23d
23e

^s_ q^s=y v qs qs

23fg

~r br y r ; ^
^ ^s; u; d
b; m;
^I_ q=y v Ey T ^I wi ;

23h
^i_ fi y v ; y T ; ^r; m;
^ ^s; ^I; ^t
23ij

^_ fp y v ; y T ; ^r; m;
^ ^s; ^I;^i; p
^
p
where xv ; xT ; xj ; xj and xi are
setpoint value T j (or I*) acts
(23d) [or (23e)].

23k
adjustable gains, and the
as a measurement in

4.3. Measurement-driven (MD) controller


Regard the components of the control (14) and estimator (24) LF, set the next four LFs, one for each regulated
output:
V v V pv V^ v ;

V T V pT V^ T

26ab

V i V pi V^ i

26cd

V m V pm V^ m ;

and proceed to execute a component wise control construction.


Volume controller. Take the time-derivative of the LF
(26a), substitute (20a) and (25b), and enforce the control
expression (27a) to obtain the dissipation rate (27b):
^bv  q k v y  v
v
_ u
_
V_ v k v e2  xv ~b2 sv e; ~xe ; d; d;
v

27a
27b

sv b_ v ev ~bv
xe bv ; bT ; bj ; bj ; bi ; m; s; I; i; p
 dp p  pc
d d0d ; d0pc 0 ; dd d  d;
c
c

~xe ^xe  xe ;

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

Temperature controller. Take the time-derivative of the


LF (26b), substitute (22bc) and (25d), and enforce the primary (28a) and secondary (28b) control expressions to
obtain the dissipation rate (28c):
aT T j ^
bT k T y T  T
bj aT y T  T ^
b  k j T j  T 
aj qj ^

28a

28b
2
2
2
2

2
_
~
~
~
V_ T k T eT  xT bT  k j ej  xj bj  xj bj sT e; ~xe ; d; d
~
sT b_ T eT ~
bT  b_ j ej ~
bj  b_   ej b
28c
j

Monomer controller. Take the time-derivative of the LF


(26c), substitute (20d) and (25c), enforce the control expression (29a) to obtain the dissipation rate (29b) (-m is an
adjustable control gain):
^ v qm qm  ^r k m m
^  m;

 qm=y

k m -m q=y v
29a

_
~ -m sm e; ~xe ; d; d
V_ m kq hem ; m;

29b

~  em ~
wm  ~r
sm m
2

hx; y; - : -x -  1xy y > 0


p
- 2 - ; - ; - 3  2 2

29c

MW controller. Take the time-derivative of the LF


(26d), substitute (20f, h) and (25e), and enforce the primary
(30a) and secondary (30b) control expressions to get the
dissipation rate (30c):
^ ^sI   bi y v ; y T ; ^r; m;
^ ^s;^i k i ^i  i
ai y v ; y T ; m;
^
^ ^s^i  i
 q=y v Ey T I wi ai y v ; y T ; m;

30a

^
bi  k i ^I  I 
_
V_ i kr hei ;~i; -i  ki hei ; ~I; -i si e; ~xe ; d; d;

30b
30c

-i k i =kr ; -i k i =ki
si ki ei ^
ai ei ~I kr ei  ^
ai ei ~i ei  ~ii ~i~
kr
b_ i ei ~
bi ~i  ei ~ci
Finally, solve Eqs. (27a), (28ab), (29a), and (30ab) for
the controls q; T j  qj ; qm and I*  wi, respectively, incorporate the estimator (23), and obtain the measurement-driven (MD) controller:
 Volume controller
 vv x v y v
q k v y v  v

31a

v_ v xv vv  xv xv y v  q
 Temperature controller
T  ^
bT k T y T  T =aT

31b

^
bj vj xj y j ;

k m -m q=y v ;

-m 2 - ; -

 Molecular weight controller


^I_ q=y Ey ^I w
v

31d

^i_ fi y v ; y T ; ^r; m;
^ ^s; ^I;^i

^ ^s;^i  k i ^i  i=ai y v ; y T ; m;
^ ^s
I bi y v ; y T ; ^r; m;
^ ^s
k i -i kr y v ; ^r; m;
^b v x I  ; v_  x v  x2 I  ; -i ; -i 2 - ; -
i

wi ^bi  k i ^I  I  q=y v Ey T ^I


^ ^s^i  i; k i -i ki y v ; y T ; q
 ai y v ; y T ; m;
 Polydispersity estimator
^_ fp y ; y ; ^r; m;
^ ^s; ^I;^i; p
^
p

31e

Comparing with the detailed model-based nonlinear geometric (11) and passive (17) FF-SF control implementations with (say geometric [3,79]) nonlinear estimators,
the above MD controller is considerably less model-dependent. Specically: the volume controller (31a) is model
independent, the temperature controller (31b) needs two
static parameter approximations (aT and aj) (21), the
monomer controller needs calorimetric parameters (pc)
(i.e., densities and specic heats), and the MW controller
(31d) and polydispersity estimator (31e) need the initiation-transfer kinetics parameters. In addition, the propagation-termination kinetic (fr) and heat transfer (fU) models
are not needed. The MD control interaction, decentralization, and model dependency characteristics are presented in
Fig. 2.
In classical PI form, the volume controller (31a) is given
by


Z t
q jv ev s1
e
sds
; ev y v  v
32
v
v
0
1

ja aa xa k a ;

1
sa x1
a k a ; a v;T ; j

and the primary (33ab) and secondary (33c) temperature


controllers can be written as follows
^b_  x ^b x T 
33a
j
j
j
j j


Z t
T j jT eT s1
eT s ds ; eT y T  T
33b
T
0


Z t
qj ^bj  jj ej s1
e
s
ds
; ej y j  T j
33c
j
j
0

^
bT vT xT y T ; v_ T xT vT  xT xT y T aj y j
^ v x T  ; v_ x v  x2 T 
b
j
j
j j
j j
j
j


^
^
q b  k j y  T  bj  aT y  T =aj
j

31c
^s_ q^s=y v qs qs ;

^r br y; ^b; m;
^ ^s; u; d
^ v  k m m
^  m=q

qm ^r qm=y
m

where h is a parameterized ellipsoidal form


2

 Monomer controller
^_ qm=y
^
qm q  ^r;
m

469

v_ j xj vj  xj xj y j aj qj

where (33b) is a primary-to-secondary feedforward lag element that performs the setpoint dierentiation.
Thus, from an industrial perspective: (i) the monomer
(31c) and MW (31d) SISO components are inventory
controllers driven by information generated in the temperature controller (31b), (ii) the monomer controller has a

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

470

Fig. 2. Dependency diagram of the proposed measurement-driven controller (31).

ratio-type feedforward component ^r=qm that sets a feedow qm contribution, (iii) the MW controller has a cascade
structure, with a primary (or secondary) element that sets
the initiator setpoint (or initiator feedrate), including a
ratio-type correction by chemical reaction ^r=^
p, and (iv)
the functioning of the volume (31a), temperature (31b),
and monomer (31c) controllers is independent of the one
of the MW controller.
4.4. Closed-loop dynamics and tuning
Having as point of departure the LFs (14) of the passive
nonlinear controller, the open-loop estimator (24), and the
closed-loop component-wise design (26), the application of
Lyapunovs direct method within the IS stability framework yields the closed-loop robust stability criteria stated
in the next proposition.
Proposition 3 (Proof in Appendix B). Consider the polymer
reactor (1) with the MD controller (31). The closed-loop
reactor is IS-stable if: (i) the control gains of the unmeasured
outputs (m and I) are chosen as follows:
p
-m ; -i ; -i 2 - ; - - 3  2 2
34a
and (ii) the gains of the primary (kv, kT) and secondary (kj)
control and estimator xv ; xT ; xj ; xj ; xi components are
tuned so that the following dynamic separation conditions
are met:

k
p < k p < k p cp k j ;

kj < k
j cj x;

x < x
34bd

where k p mink v ; k T ;

minxv ; xT ; xj ; xj ; xi

Conditions (34a) ensure the stability of the state (m, i, I)


subsystem associated with the unmeasured-regulated outputs. In fact, the value that maximizes the area of the stabilizing ellipsoidal function h (29c) is -max 3, and
therefore -m ; -i ; -i should be chosen between 1 (without
cross terms in h) and 3. In other words, the unmeasured
output gains should at most be three times faster than
the related dilution rate (q/v or r/p). In Conditions (34b
d): (i) x+ is an upper limit imposed by the high frequency
unmodeled dynamics, (ii) k 
p is a lower primary gain limit
set by the open-loop reactor instability, and (iii) c
j (or

c
)
is
an
isotonic
function
that
sets
an
upper
limit
k
j (or
p

k p ) for the secondary (or primary) gain, depending on


the faster estimation (x) [or secondary (kj)] gain. Thus,
the choice of gains aects and is aected by the sizes
of the prescribed (or to be compromised) initial state, exogenous input, model parameter disturbances, and this is in
agreement with a practical stability framework [17,32].
From the preceding stability assessment (34) in conjunction with conventional tuning rules for linear PI controllers
[33] and lters [34] the next tuning guidelines follow. (i) Set:
the unmeasured output gains at their nominal values of
their associated dilution rates -m -i -i 1, the control gains at the nominal inverse residence time k v k T
k j q=v, and the estimator gains about three times faster
x xv xT xj xi xj 3q=v. (ii) Increase x up
to its ultimate value xu (with oscillatory response), and
back o so that the behavior is satisfactory (x 6 xu/3).
(iii) Increase the volume gain kv up to its ultimate value
k uv , and back o  k v 6 k uv =3 for adequate satisfactory
response. Repeat the procedure for the secondary temperature gain k j  k j 6 k uj =3. (iv) Increase the temperature

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

gain kT up to its ultimate value k uT , and back o until an


adequate response is attained. (v) Apply the same
increase-back o procedure to the unmeasured output
gains -m ; -i ; -i , in the understanding that they cannot
be larger than two to four times the related dilution rates
(-a 6 2-to-4, a = m, i, i). (vi) If necessary, adjust the estimator gains xv ; xT ; xj ; xj ; xi .
4.5. Concluding remarks
Comparing with a previous PI-inventory design with
similar structure [13], here the temperature (or MW) secondary controller (31b) [or (31d)] has an additional term
(aTeT) [or (aie)] that compensates the modeling error associated with the primary controller design within a conventional cascade framework: the intermediate state is at its
setpoint, meaning T j T j (or I = I*). This feature implies
the relaxation of the dynamic gain separation condition of
the standard cascade designs, or equivalently, a better
behavior with an improved performance-robustness tradeo. Moreover, the use of the molecular weight inverse state
led to a simpler MW controller and facilitated the stability
analysis, and with the incorporation of the calorimetricbased polydispersity estimates the information contained
in the temperature and volume measurements can be further exploited for monitoring and MW control model calibration aims.
5. Application example
To subject the controller to a severe test, let us consider
an extreme case of an industrial situation: the operation of
a reactor at high-solid fraction with the potentially destabilizing gel-eect at play, about a nominal steady-state which
is open-loop unstable. The monomer is methyl methacrylate, with ethyl acetate (solvent) and AIBN (initiator).
The residence time is v=
q 220 minutes and nominal volume v  2000 L. The model functions and parameter val-

471

ues (listed in Appendix A) were taken and/or adapted


from Alvarez et al. [3], and the nominal inputs were
adapted from a previous solution copolymer reactor study
[28]. The reactor has three steady-states (Table 1), with two
of them corresponding to extinction and ignition openloop stable operations, and one of them being open-loop
unstable. In the spirit of the nonlocal IS stability framework (4) employed in the control design developments,
the unstable steady-state is chosen as the nominal operating point, and the resulting closed-loop system will be subjected to initial state, input (persistent model parameter
and step/sinusoidal exogenous input) disturbances, and
the kind of transient, a symptotic and combined transient-asymptotic responses will be analyzed.
The adequate fulllment of the relative degree (6) and
ZD uniqueness with IS stability (8) solvability conditions
were veried with combined analytic-numerical testing.
The application of the tuning guidelines (given in Section
4.4), yielded the following gains
1

xv xj xT xi xj 1=5 min : x;


k v k j x=8; k T 2q=v;

-m -i -i 1:5

5.1. Behavior recovery


The reactor was controlled with the proposed MD control (31) and its behavior compared with the one of its
exact model-based nonlinear counterpart (16). The model
parameters were xed at their actual values, and the
closed-loop system was subjected to: an initial state deviation about the open-loop unstable steady-state in conjunction with the reactor and jacket feed temperatures step
changes shown in Fig. 3a (at t = 500 min: Te from 315 to
320 K, and Tje from 328 to 330 K). The closed-loop behavior is presented in Fig. 4, showing that: (i) as expected,
the MD controller recovers the behavior of the nonlinear
FF-SF passive controller, (ii) the volume, temperature

Table 1
Steady-states, nominal inputs, outputs and reactor states, and jacket parameter approximations
States and outputs

m (kg)
T (K)
Tj (K)
Mn (kg/kmol)
I (kg)
s (kg)
p
c
r

Steady-states
Stable (extinction)

Unstable

Stable (ignition)

1361.089
329.72
329.53
399149.03
1.685
501.283
1.9997
0.1072
0.08066

660.082
351.62
341.23
110384.75
1.3087
500.871
1.999
0.5672
0.4269

312.756
373.88
345.61
29395.15
0.3513
498.561
1.9966
0.7954
0.5997

Nominal inputs

 i 0 = (30 L/min, 9.1 L/min, 7.34 L/min, 0.0078526 kg/min) 0 , d T e ; T je ; qs 0 =(315 K, 328 K, 2.54 L/min) 0
u qj ; 
q; 
qm ; w
Jacket parameters
aT = 4.71 102 min1, aj =  3.78 102 K/L

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

Reactor (Te) and


jacket (Tje ) feed
temperatures, (K)

Reactor (Te) and


jacket (Tje) feed
temperatures, (K)

472

330

Tje

(Mn) converges in about 2 residence times, meaning twice


faster than the open-loop MW response, (iv) the control
actions eectively cancel out the eect of the input disturbances on the regulated outputs, and (v) the control actions
occur in a smooth and coordinated manner, reasonably
away from saturation.
The preceding test was repeated with sinusoidal inputs
(shown in Fig. 3b):

Te

T e 315 K 5 sin6:2832t=110

Tje

320

Te

310
330

320
310
0

1000

500

T je 328 K 2 sin6:2832t=110;

1500

t (min)

Fig. 3. Time-varying exogenous inputs: (a) step and (b) sinusoidal.

and monomer outputs reach their set points in about one


residence time, in the understanding that these controllers
ensure the stability, the production rate level, and to a good
extent the product quality, (iii) the number-average MW

and the resulting closed-loop behavior is presented in


Fig. 5, showing that: (i) again, the MD controller recovers
the behavior of its exact SF model-based counterpart, (ii)
the volume, temperature and monomer outputs converge
to their setpoints in about one residence time, and the
(number-average) MW (Mn) reaches its set point in about
two residence times, (iii) the outputs exhibit sustained oscil-

Reactor (T)
and jacket (Tj )
temperatures (K)

1.90
1.4

1.20

1.2

1.15

1.0

1.10

Mn

1.05
0.6

0.8

0.5

2.0

30

10

wi

20

10

qm
t (min)

1000

1500

Fig. 4. Closed-loop reactor behavior under step input disturbances, with


exact model-based SF passive (16) (- - -) and MD (31) (A) controllers, and
nominal operation values (  ).

2.00

680
660
640
620
600
1.25

1.95

1.90
1.4

1.20

1.2

1.15

1.0

1.10
1.05

Mn

0.6

0.8

0.5

0.4
2.1

2.0

1.9
40
15

qj

30

10

wi

20

10

qm
0

500

1000

1500

500

Tj
340

Initiator feedrate x 10,


wi (kg/min)

15

qj

Initiator feedrate x 10,


wi (kg/min)

1.9
40

Coolant (qj), exit (q),


and monomer (qm )
flowrates, (L/min)

Polydispersity,
(-)

0.4
2.1

Monomer
mass, m (kg)

1.95

Number-average
MW x 10-5,
M n (kg/kmol)

T
350

Volume x 10-3, V (L) Initiator mass, I (kg)

2.00

680
660
640
620
600
1.25

Conversion (c),
solid fraction ( ), (-)

Tj
340

Initiator mass, I (kg)

Number-average
Conversion (c),
MW x 10-5,
solid fraction (), (-)
M n (kg/kmol)

T
350

-3

Polydispersity,
(-)

360

Volume x 10 , V (L)

Monomer
mass, m (kg)

Reactor (T)
and jacket (Tj )
temperatures (K)

360

Coolant (qj), exit (q),


and monomer (qm )
flowrates, (L/min)

t P 500 min

t (min)

Fig. 5. Closed-loop reactor behavior under sinusoidal input disturbances,


with exact model-based SF passive (16) (- - -) and MD (31) (A) controllers,
and nominal operation values (  ).

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

lations with rather small amplitudes (0.01, 0.12, 0.1, 0.25)%


for (v: volume, T: temperature, m: monomer, Mn) with respect to industrial reactor operations.
These tests verify the IS stability property of the closedloop reactor system with strongly nonlinear FF-SF and the
mildly nonlinear MD control: after combined initial-state
step-input disturbances, the closed-loop system undergoes
a linear-like vanishing transient response, with asymptotic
convergence to the prescribed steady-state. In the case of
persistent sinusoidal input disturbances, the same transient
behavior is exhibited, and the system converges to a sustained oscillatory regime within an acceptable-size compact
set about the nominal steady-state.
5.2. Behavior with parameter errors
Finally, the MD control was tested with the following
typical parameter errors: 14% error in the chain transfer
parameters (am, as) (Appendix A), and 5% in the initiator

eciency factor (fd), in the understanding that the highly


nonlinear and uncertain polymerization rate-heat exchange
function pair fr  fU is not needed by the MD control
scheme. I the closed-loop system was subjected to the
above stated initial state deviation with step (Fig. 3a) and
sinusoidal (Fig. 3b) exogenous input disturbances, and
the behavior results are presented in Fig. 6 (with step
inputs) Fig. 7 (with sinusoidal inputs). Comparing with
the corresponding errorless test (Fig. 4), in the case of initial state and step input disturbances, the parameter errors
do not appreciably aect the control behavior (Fig. 6): (i)
the volume, temperature and monomer responses are
imperceptibly aected, and (ii) the MW weight response
exhibits 3% asymptotic oset (i.e., smaller than experimental measurement uncertainty). Should this oset be
unacceptably larger, the initiator and/or transfer constants
should be occasionally recalibrated on the basis of freemonomer, solid fraction and Mn measurements which are
routinely taken for operation monitoring and product

Reactor (T)
and jacket (Tj)
temperatures (K)

1.95

1.15

1.2
1.0

1.10

0.6

0.5

0.8

Conversion (c),
solid fraction (), (-)

1.05

Mn

2.0

30

wi

20

10

10

5
500

1000

1500

t (min)

Fig. 6. Closed-loop reactor behavior under step input disturbances and


model parameter errors, with MD (A) controllers (31), and nominal
operation values (  ).

1.95

1.90
1.4

1.20

1.15

1.2
1.0

1.10
1.05

Mn

0.6

0.5

0.8

0.4
2.1

2.0

1.9
40

qj

30

15

wi

20

10

10

qm
0

500

1000

1500

qm

Initiator feedrate x 10 ,
wi (kg/min)

15

qj

Initiator feedrate x 10 ,
wi (kg/min)

1.9
40

Coolant (qj), exit (q),


and monomer (qm)
flowrates, (L/min)

Polydispersity,
(-)

0.4
2.1

2.00

680
660
640
620
600
1.25

Initiator mass, I (kg)

1.20

Tj
340

-3

1.90
1.4

Monomer
mass, m (kg)

Number-average
MW x 10 -5,
M n (kg/kmol)

2.00

680
660
640
620
600
1.25

T
350

Volume x 10 , V (L)

Number-average
MW x 10 -5,
M n (kg/kmol)

Tj
340

Initiator mass, I (kg)

Conversion (c),
solid fraction (), (-)

350

-3

Polydispersity,
(-)

360

Volume x 10 , V (L)

Monomer
mass, m (kg)

Reactor (T)
and jacket (Tj)
temperatures (K)

360

Coolant (qj), exit (q),


and monomer (qm)
flowrates, (L/min)

473

t (min)

Fig. 7. Closed-loop reactor behavior under sinusoidal input disturbances


and model parameter errors, with MD (A) controllers (31), and nominal
operation values (  ).

474

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

quality assessment purposes. The behavior with sinusoidal


input disturbances (Fig. 3b) is shown in Fig. 7: (i) the transient responses are similar to the ones of the case without
parameter errors (Fig. 4 with step inputs, and Fig. 5 with
sinusoidal inputs), and to the case with parameter errors
and step inputs (Fig. 6), and (ii) the asymptotic response
exhibits an oscillatory behavior with a amplitudes of
(0.01, 0.12, 0.1, and 0.25)% in (v, T, m, Mn) about imperceptible osets in (v, m, T) and an  3% oset in MW.
5.3. Concluding remarks
The responses to initial state deviations, exogenous
input disturbances, and model parameter errors verify: (i)
the (nonlinear passive control) behavior recovery as well
as the IS robust stability property of the MD controller,
and (ii) the simplicity and eectiveness of the tuning guidelines drawn from the closed-loop stability assessment. The
closed-loop testing illustrated and made quantitative the
assessment of IS stability features, like transient overshoot,
settling time, and asymptotic (constant or oscillatory)
behavior. From the comparison of responses with and
without parameter errors, it follows that the closed-loop
outputs exhibit a nearly linear (i.e., superposition-like)
behavior, in accordance with the nearly closed-loop output
behavior associated with the MD control design.
6. Conclusions
A constructive robust MD control design methodology
for continuous solution homopolymerization reactors with
ow and temperature measurements has been presented.
Structural (relative degree, ZD and detectability) solvability conditions that bear physical meaning were identied
and employed to set a simplied model for control design.
The application of a Lyapunov interlaced estimator-control design yield led to a MD control scheme with: (i) maximum linearity, decentralization, and model independency
features, (ii) linear-decentralized PI volume and control
components, (iii) material balance monomer and MW
controllers which exploit the information contained in the
integral actions of the PI controllers, (iv) a systematic
construction procedure, and (v) a closed-loop nonlinearnonlocal stability criteria coupled with simple tuning guidelines. The proposed controller: is considerably simpler and
less model dependent than previous nonlinear geometric
controllers, recovers the behavior of its exact model-based
FF-SF counterpart with optimality-based robustness characteristic, is equivalent to an unconstrained MPC with innite receding horizon, and amounts to a set of coordinated
decentralized and cascade components that resemble the
ones employed in industrial polymer reactors.
The polymerization of MMA in an open-loop unstable
industrial size reactor was considered as representative case
example with numerical simulations. The closed-loop testing veried the control behavior robustness and recovery

features, and illustrated and made quantitative the assessment of the closed-loop IS stability property.
The proposed polymer reactor control design methodology suggests the pursuit of a general-purpose design framework that can fruitfully blend theoretical nonlinear and
applied chemical process control techniques.
Acknowledgements
The authors gratefully acknowledge the support from
the Mexican National Council for Research and Technology (CONACyT Scholarship 118632) for P. Gonzalez.
Appendix A. Polymerization reactor model
A.1. Kinetics [13,20]
ri ET I : fri T ; I
r k p v;m; T ; I;sk0 v;m; T ; I;sm : fr v; m;T ;I; s
r0 cd fri T ; I it T ;m; sfr v;m; T ; I;s : f0 v; m;T ; I; s
cd 2f d =mwi
ET ead bd =T ; k po T eap bp =T ; k to T eat bt =T
k p v; m; T ;I; s k po T =1 k po T hp T Ef v;m; T ; sk0 v; m;T ; I; s
k0 v; m;T ; I; s k0 v;T ; I k2g v;m; T ; I;s1=2 kg v; m;T ; I; s
k0 v; T ;I 2f d fri T ; I=vmwi k to T 
kg v; m;T ; I; s fd ht v;T ; IEf v; m;T ; sfri T ; I=vmwi
Ef v; m; T ;s e2:3m=vqm s=vqs =fAT BT m=vqm s=vqs g
hp T eahp bhp =T ; ht v; T ; I eaht bht I=vmwi d ht =T
AT aA  bA 1  T =T g  cA 1  T =T g 2 ;
BT aB  bB 1  T =T g
it T ; m;s jm T s=mjs T ; jm T 1=mwm eam bm =T ;
js T 1=mws eas bs =T

A.2. Heat capacity and transfer [2,3]


cm qm cpm ; cs qs cps ; cj qj cpj ;
C mcpm scps fp v;m;scpp : fC V ;m; s
q qm 1  em m=vqm  es s=vqs =1  em : fq v; m; s
p Vfq v; m; s  m  s : fp v; m;s
U Afh T ;T j ; v;m;s : fU T ;T j ; v;m;s
b

h ah k=DLN q=l h cl=k h l=lw h : fh T ;T j ; v; m;s


c

l al T  273:15bl Ef l T ;v;m;s : fl T ; v;m;s; lw fl T j ; v;m;s

A.3. Numerical values [3,20]


ad ; bd ; fd 35:811324; 14896:127; 0:52
ap ; bp ; at ; bt 18:39228; 2609:199; 22:49482; 352:758
am ; bm ; as ; bs 8:396; 6472:83131; 5:664; 4570:25734
aA ; bA ; cA ; aB ; bB ; T g 0:1678; 0; 1:23517; 0:03; 0; 387:15
ahp ; bhp ; aht ; bht ; cht ; d ht
35:11094; 13964; 47:03; 48:85; 637:19; 17956
k; D; L; N ; D 0:2768; 13; 7; 250; 134:76

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

lgq x; q; qm ; qs ; q_ s ; T e mm  qm mv 1  em oxwi fr

al ; bl ; cl 2:484 1029 ; 9:6973; 1:99


ah ; bh ; ch ; d h 0:74; 2=3; 1=3; 0:14

#  qm q_ s =qm  m=v
lgqm x; q; qm ; qs ; q_ s ; T e fmm  m=vvv

mw ; mws ; mwi 100:11; 88:1; 164:21


qm ; qs ; qp ; em ; es 950; 901; 1170; 0:188; 0:23

1  em m=qm voxwi fr #  m=vq_ s g=qm  m=v


g
lwi x; q; qm ; qs ; T e q=v ET  mi  hoxwi /i ; fwi i=oI fi
lgqj x; q; qm ; qs ; T e ; T j e; q_ s ; T_ e mT  /j  hoxqj /j ; fqj i

Appendix B. Solvability assessment


Proof of Proposition 1 (relative degree existence). Recall
the reactor system (3), incorporate the dynamic extension
(5a), and obtain the augmented system
x_ a fa xa ; d; u; xa x0 ; x0a 0 ; xa q; qm 0
_ q_ m ; qj ; wi
ua q;

 hodwi /j ; q_ m ; q_ s ; T_ e 0 i=fU oTj fT


B4
0

where f wi fT ;fv ;fm ;fs ; xwi T ;v; m;s


0

fqj fw0 i ;fI ; xqj x0wi ;I

# f1  em =qm m=vfr qm  v=mqm  m=vqs gq=v;

This system has relative degree j (5b) i the the maps / and
u of the coordinate change
v v; fv ; m; fm ; T ; fT ; i; fi 0 : /xa ; d;

oua / 0

B1a
0
0_
_
m oxa fv ; fm ; fT ; fi od fv ; fm ; fT ; fi d : uxa ; ua ; d; d
/ is xa -invertible

B1b

/ is ua -invertible

are xa and ua-invertible (inv) [14], respectively, meaning


that (B1a) [or (B1b)] has a unique solution for xa (or ua).
Since the inverse (v, m, T, i) 0 = (v1, v3, v5, v7) 0 is trivially given, / (or u) is xa or (or xu)-invertible i Condition (B2a)
[or (B2b)] is met:
0

fv ; fm is q; qm -inv () 6a
fT is T j -inv () 6c; f i is I-inv () 6d

B2a

_ q_ m -inv () 6a
uv ; um 0 is q;
B2b

Consequently, (/,u) is (xa,ua)-invertible i the conditions


of Proposition 1 are met. h
Geometric controller derivation (11). Recall the stateinput pair v  m (B1), write the Brunovskys controllability
form [14] associated with the relative degree j (5b)
v_ 3 v4 ; v_ 4 m2 ;

v_ 5 v6 ; v_ 6 m1

Proof of Proposition 2 (ZD stability). In nominal steadystate regime the ZD (7) become
 2 mwi  2;
p

s
sro s=v w

B5ab
0

 i Condiimplying that there is a unique steady-state s; p


tion (8a) 2 is met. Recall the ZD LF (9a), take its derivative, and obtain the dissipation rate (9b) with
ao es ; ep ; do loq es ; dos lop es ; dop
o

b es ; d

bos es ; dos

B6a

bop es ; dop

B6b


loq es ; dos q=v; q 1  em r=qm qs =1  m=p
o
r fr v; m; T ; lI s; s
^s ~
qs
lop es ; dop r=p; bos es ; dos es f
ws 1  loq es ; dos =
q q

~s ges

qs q

~
bop es ; dop ~cop  p
kop ; dos d0z ; dqs ; dpr 0 ; dop d0z ; dpr 0 ; dz z z
dqs q  
qs ; dpr pr  pr

Write the steady-state solution (B5) in perturbed form (i.e.,


loq bos ; lop bop ), and recall condition (8a), to establish the
existence of K-class functions (cos and cop ) that bound the
perturbed steady-state solutions, this is,
loq es ; dos es bos es ; do ) jes j 6 cos jdos j

B7a

lop es ; dp ep

Compare these equations with the ones of the LNPA output error dynamics (10), obtain the controller (in v  m attening coordinates)
c c
ma xc2
a ea  fa xa fa ;

d wi qm ;qs ;T e : 

p fp v; m; s

uT is qj -inv () 6bc; ui is wi -inv ) 6ad

v_ 1 v2 ; v_ 2 m1 ;
v_ 7 v8 ; v_ 8 m2

475

a v; m; T ; i

B3

substitute these equations into (B1a), obtain the equation


0

_ K p C z x  z  K d fv ; fm ; fT ; fi xa ; d
uxa ; ua ; d; d

) jep j 6

bop es ; dop ) jep j 6 cp jes j; jdop j


cp jes j; jdop j 6 cp cos jdos j; jdop j :

B7b
cop kdo k

This leads us to conclude the existence of the asymptotic


gain functions (cos and cop ) that make non-positive the associated dissipation rate (9b), or equivalently, the ZD (7) are
IS stable. h
Proof of Proposition 3 (Closed-loop stability with MD
controller). Introduce the LF V,

c2
c2
c2
K p diagxc2
v ; xm ; xT ; xi
K d diagfcv xcv ; fcm xcm ; fcT xcT ; fci xci ;

_
V V p V^ ) V_ ae; ~xe ; d e; ~xe ; d; d
ae; ~xe ; d P 0; s0; 0; 0; 0 0

recall the ua-invertibility of u (B1b), and solve to last equation to obtain the dynamic nonlinear FF-SF controller (11)
with the following maps [mi is dened in (B3)]:

where Vp (or V^ ) is the LF (14) [or (24)] of the nonlinear


passive FF-SF controller (17) [or estimator (23)]. Take
the derivative of V along the motion of the reactor-MD

B8ab

J. Alvarez, P. Gonzalez / Journal of Process Control 17 (2007) 463476

476

control pair (1)(31), substitute the dissipation rates V^_ (25),


V_ v (27b), V_ T (28c), V_ m (29b), and V_ i (30c), recall Vo (9a),
determine its dissipation rate component V_ o , and obtain
the closed-loop dissipation rate (B8b) with
a k v e2v xv ~
b2v k T e2T xT ~
b2T k j e2j
2
~2
xj ~
s2 e2p p
b2j xj ~
b2
j kq es ~
~ -m kr hei ;~i; -i ki hei ; ~I; -i
kq hem ; m;
d dd ; d_ d ; dp
e

_ u
_ si e; ~xe ; d; d
_
_ sT e; ~xe ; d; d
s sv e; ~xe ; d; d;
~~
ws ~cp  ~
k
pep ~cp  p p
kp ~
p
es ~s~
where {kq, kr, ki, kp} (13) is the dilution rate set, h is the
ellipsoidal function (29c), and the functions sv (27b), sT
(28c), and si (30c) have been already dened. Set the Eq.
(B9a), recall the closed-loop IS stability (18) with the passive controller (17), conclude the existence of a local
asymptotic gain c (B9b), draw the dissipation rate inequality (B9c):
_
ae; ~xe ; d se; ~x; d; d
) je0 ; ~x0 0 j ckd0 ; d_ 0 0 k

B9b

0
; ~x0e j

B9c

V_ 6 08je

P ckdtk

B9a

with c(0) = 0, and conclude (4), the IS stability of the


closed-loop reactor (1) with the MD driven controller
(31), provided the gains are tuned according to Conditions
(34) of Proposition 3, to ensure that the stabilizing term (a)
dominates the potentially destabilizing one (s)[13]. h
References
[1] J.W. Hamer, T.A. Akramov, W.H. Ray, The dynamic behavior of
continuous polymerization reactors II. Nonisothermal solution
homopolymerization and copolymerization in a CSTR, Chem. Eng.
Sci. 36 (12) (1981) 1897.
[2] S. Padilla, J. Alvarez, Control of continuous copolymerization
reactors, AIChE J. 43 (2) (1997) 448.
[3] J. Alvarez, R. Suarez, A. Sanchez, Semiglobal nonlinear control
based on complete input-output linearization and its application to
the start-up of a continuous polymerization reactor, Chem. Eng. Sci.
49 (21) (1994) 3617.
[4] J.P. Congalidis, J.R. Richards, Process control of polymerization
reactors: an industrial perspective, Polym. React. Eng. 6 (2) (1998) 71.
[5] B.R. Maner, F.J. Doyle III, Polymerization reactor control using
autoregressive-plus Volterra-based MPC, AIChE J. 43 (1997) 1763.
[6] P. Daoutidis, M. Soroush, C. Kravaris, Feedforward/feedback
control of multivariable nonlinear processes, AIChE J. 36 (10)
(1990) 1471.
[7] M. Soroush, C. Kravaris, Multivariable nonlinear control of a
continuous polymerization reactor: an experimental study, AIChE J.
39 (12) (1993) 1920.
[8] J. Alvarez, Output-feedback control of nonlinear plants, AIChE J. 42
(9) (1996) 2540.
[9] J.P. Gauthier, I. Kupka, Deterministic Observation Theory and
Applications, Cambridge University Press, UK, 2001.

[10] R.K. Mutha, W.R. Cluett, A. Penlidis, On-line nonlinear modelbased estimation and control of a polymer reactor, AIChE J. 43 (11)
(1997) 3042.
[11] I. Saenz de Buruaga, P.D. Armitage, J.R. Leiza, J.M. Asua,
Nonlinear control for maximum production rate of latexes of welldened polymer composition, Ind. Eng. Chem. Res. 36 (1997)
4243.
[12] J. Alvarez, F. Zaldo, G. Oaxaca, Towards a joint process and control
design framework for batch processes: application to semibatch
polymer reactors, in: P. Seferlis, M.C. Georgiadis (Eds.), The
Integration of Process Design and Control, Elsevier, Amsterdam,
The Netherlands, 2004, pp. 604634.
[13] P. Gonzalez, J. Alvarez, Combined proportional/integral-inventory
control of solution homopolymerization reactors, Ind. Eng. Chem.
Res. 44 (2005) 7147.
[14] A. Isidori, Nonlinear Control Systems, third ed., Springer-Verlag,
New York, 1995.
[15] A. Isidori, Nonlinear Control Systems II, Springer-Verlag, London,
1999.
[16] R. Sepulchre, M. Jankovic, P. Kokotovic, Constructive Nonlinear
Control, Springer-Verlag, NY, 1997.
[17] R.A. Freeman, P.V. Kokotovic, Robust Nonlinear Control Design:
State-Space and Lyapunov Techniques, Birkhauser, Boston, 1996.
[18] M. Krstic, I. Kanellakopoulos, P. Kokotovic, Nonlinear and Adaptive Control Design, Wiley, New York, 1995.
[19] E.D. Sontag, The ISS philosophy as a unifying framework for
stability-like behavior, in: A. Isidori, F. Lamnabhi-Lagarrigue, W.
Respondek (Eds.), Nonlinear Control in the Year 2000, Lecture
Notes in Control and Information Sciences, vol. 2, Springer-Verlag,
Berlin, 2000, pp. 443468.
[20] W.Y. Chiu, G.M. Carratt, D.S. Soong, A computer model for the gel
eect in free-radical polymerization, Macromolecules 16 (3) (1983)
348.
[21] F.G. Shinskey, Process Control Systems, third ed., McGraw-Hill,
New York, 1988.
[22] H.K. Khalil, Nonlinear Systems, third ed., Prentice-Hall, New Jersey,
2002.
[23] F. Allgower, A. Zheng (Eds.), Nonlinear Model Predictive Control,
Birkhauser, Germany, 2000.
[24] R.E. Kalman, When is a linear control system optimal? Trans. ASME
Ser. D: J. Basic Eng. 86 (1964) 1.
[25] P.J. Flory, Principles of Polymer Chemistry, Cornell University Press,
New York, 1953.
[26] J.A. Biesenberger, D.H. Sebastian, Principles of Polymerization
Engineering, Wiley, New York, 1983.
[27] J. Alvarez, Nonlinear state estimation with robust convergence, J.
Proc. Cont. 10 (2000) 59.
[28] T. Lopez, J. Alvarez, On the eect of the estimation structure in the
functioning of a nonlinear copolymer reactor estimator, J. Proc.
Cont. 14 (2004) 99.
[29] R. Hermann, A.J. Krener, Nonlinear controllability and observability, IEEE Trans. Auto. Control. AC-22 5 (1977) 728.
rez, J. Alvarez, A. Morales, An adaptive cascade
[30] J.J. Alvarez-Ram
control for a class of chemical reactors, Int. J. Adapt. Cont. Signal
Process. 16 (2002) 681.
[31] R.T. Stefani, C.J. Savant Jr., B. Shahian, G.H. Hostetter, Design of
Feedback Control Systems, third ed., Saunders College Publishing,
Florida, 1994.
[32] J. LaSalle, S. Lefschetz, Stability by Lyapunovs Direct Method,
Academic, New York, 1961.
[33] W.L. Luyben, Process Modeling, second ed.Simulation and Control
for Chemical Engineers, McGraw-Hill, Singapore, 1990.
[34] J.J. Dazzo, C.H. Houpis, Linear Control System Analysis and
Design, McGraw-Hill, New York, 1981.

S-ar putea să vă placă și