Power-shaping of reaction systems: the CSTR

case study
A.Favache, D.Dochain 1
IMAP, Universit´e catholique de Louvain, 2 place Sainte Barbe,
B-1348 Louvain-la-Neuve, Belgium
CESAME, Universit´e catholique de Louvain, 4-6 avenue G. Lemaitre,
B-1348 Louvain-la-Neuve, Belgium

Abstract
The non-isothermal continuous stirred tank reactor (CSTR) is a classical yet complex case study of nonlinear dynamical systems. Power-shaping control is a recent
approach for the control of nonlinear systems based on the physics of the dynamical
system. In this paper we apply the power-shaping control approach to the nonisothermal CSTR study case. A global Lyapunov function is derived for the openloop exothermic CSTR. This Lyapunov function is then reshaped by the means of
a controller in order to stabilize the process at a desired temperature. Some considerations on the local and global convergence to the desired state are presented.
Key words: non-isothermal CSTR, power-shaping control

1

Introduction

Thermodynamic systems, and among them chemical reaction systems, are
usually nonlinear dynamical systems. They can therefore have a complex behaviour and be difficult to analyze and to control. Stability analysis of nonlinear systems requires the use of abstract mathematical tools such as the two
Lyapunov methods or the passivity theory. Over the past years, several works
have combined those abstract concepts with the underlying physical phenomena giving rise to the dynamical behaviour of the system. These works include
for instance the study of port-Hamiltonian systems [4][7], energy-balancing
Email address: audrey.favache@uclouvain.be,
denis.dochain@uclouvain.be (A.Favache, D.Dochain).
1 Corresponding author.

Preprint submitted to Automatica

13 October 2009

passivity based control (PBC) [13][19] or the introduction of the contact
formalism for expressing the dynamics of systems in which irreversible phenomena arise [5][6][11]. The non-isothermal continuous stirred tank reactor
(CSTR) is a classical study case of nonlinear systems. Indeed the dynamical behaviour exhibits complex features, such as multiple equilibrium points.
Up to now no precise physical interpretation of the complex behaviour of the
non-isothermal reactor has been found [10].
Power-shaping control [18] has been developed in the past years as an extension of energy-balancing passivity-based control [13][19]. In energy-balancing
passivity based control, the controller reshapes the energy function of the system so that it has a minimum at the desired equilibrium point. The controller
provides the system a finite amount of energy so as to drive the system to
the desired state. This concept has been applied to electro-mechanical systems [17][20] and also to thermodynamic systems where the storage function
is the entropy instead of the energy [1][21]. Nevertheless energy-balancing
passivity-based control cannot be applied to some systems (namely systems
with pervasive dissipation). To overcome this difficulty the concept of powershaping control has been introduced, firstly for the stabilization of nonlinear
RLC circuits [18]. Contrary to energy-balancing passivity-based control, the
storage function used for the control is related to the power and not to the
energy. Power-shaping control has subsequently been applied to the control of
mechanical and electromechanical systems [12].
This paper is dedicated to the analysis and control of the CSTR study case
by the means of the power-shaping approach. First we shall briefly introduce
the concepts of the power-shaping approach in Section 2. Then in Section 3
we shall explain the general methodology for writing the Brayton-Moser form
on the particular case of a CSTR. This methodology shall then be applied to
obtain the Brayton-Moser formulation of the dynamics of a CSTR with an
irreversible reaction (Section 4). This formulation shall be used to design a
control system for this CSTR using the power-shaping approach. Finally we
shall give some considerations about the generalization of this work to more
general cases. In Section 5 we shall extend the results obtained for the CSTR
with an irreversible reaction to more complex reactions and reaction kinetics.
Section 6.1 presents some primary results for the Brayton-Moser formulation
for a general CSTR.

2

Power-shaping control

In this section we briefly explain the principles of the power-shaping approach.
The statements are given without any proof. For more details, the reader can
refer to [12][14][15][18].
2

1 The Brayton-Moser formulation Let us consider a dynamic system of dimension N with m inputs. P (x) : RN → R is a scalar function of the state and G (x) : RN → RN × Rm .2 Power-shaping control Let us assume that the system dynamics can be expressed by using the Brayton-Moser equations presented before. The power-shaping control theory is based on the Brayton-Moser formulation of the system dynamics [3]. in the latter it is related to the Rayleigh dissipation function [14]. The state of the system is given by the vector x ∈ RN and the input is given by vector uc ∈ Rm . In the first one it is related to the so-called content and co-content of the resistances [18][15]. The desired equilibrium state is 2 This condition is equivalent to the existence of the potential function P (x). The system (3) can be written in the form (1) if there exists a non-singular matrix Q (x) fulfilling (2) and that solves following partial differential equation 2 : ∇ (Q (x) f (x)) = ∇t (Q (x) f (x)) (4) P (x) is the solution of the following partial differential equation system: ∇P (x) = Q (x) f (x) (5) and the function G (x) is given by G (x) = Q (x) g (x). Additionally the symmetric part of the matrix Q (x) is negative semi-definite. In electrical and mechanical systems. 3 . In this formulation the system dynamics are of following form: Q (x) dx = ∇P (x) + G (x) uc dt (1) where Q (x) : RN → RN × RN is a non-singular square matrix. 2.e. the potential function has the units of power and is related to the dissipated power in the system. i.: Q (x) + Qt (x)  0 (2) The function P (x) is called the potential function. Let us now assume the system dynamics is given by the following relation: dx = f (x) + g (x) uc dt (3) where f (x) : RN → RN and g (x) : RN → RN × Rm .2.

.1 The dynamical model The general CSTR Let us consider a liquid-phase CSTR with constant volume V containing Nc species and in which Nr reactions take place. n is a column vector with ni in the ith position. As it has been shown in [10] the dynamics of such a system are given as follows (i = 1. Γil is the stoichiometric coefficient of i 4 .e. F the volume flow rate. The function Pd (x) can be used as a Lyapunov function for the closed-loop system.1 3.1.. g ⊥ (x) g (x) = 0 with rank g ⊥ (x) = N − m). . The principle of power-shaping control is to choose the input uc (x) such that in closed loop the system dynamics are given by the following relation: dx = ∇Pd (x) (6) dt where Pd (x) : RN → R is the re-shaped potential function. The reactor is cooled/heated by a surrounding jacket. n) dt V l=1  dU F  in h V − H + Q˙ = dt V (8a) (8b) with ni the quantity of species i.denoted by x∗ . U the internal energy of the mixture. Nc ): Nr  X F  in dni = Ci V − ni + Γil rl (T. Ciin is the concentration of i in the inlet flow. The function Pd (x) cannot be chosen arbitrarily since the following relation must be fulfilled: g ⊥ (x) Q−1 (x) ∇Pa (x) = 0 (7) where Pa (x) = Pd (x) − P (x) and g ⊥ (x) : RN → RN −m × RN   is a full-rank left annihilator of g (x) (i. Under these conditions.. the control input u (x) that re-shapes P (x) into Pd (x) is the following one:  uc (x) = g t (x) Qt (x) Q (x) g (x) 3 −1 g t (x) Qt (x) ∇Pa (x) The Brayton-Moser PDE for a general CSTR 3. The desired equilibrium point x∗ must be a local minimum of the potential function Pd (x) in order to be locally asymptotically stable.

3.g. T = Tˆ (n.2 The simplified CSTR In order to simplify the model. 3 5 . H the enthalpy of the mixture and Q˙ the heat transferred into the reacting mixture. • the reaction enthalpy (−∆r H)l of reaction l is independent of the temperature.1. we consider the following additional assumptions: • the specific heat Cp of the mixture are constant. Following usual thermodynamic notations (−∆r H)l > 0 for exothermic reactions and (−∆r H)l < 0 for endothermal reactions. As explained in the previous section. an ideal liquid mixture): ˆ (n. (8b) can be replaced by the following relation: Nr X dT in ˙ (−∆r H)l (T ) rl (T. hin is the volumetric enthalpy of the inlet flow. The aim is to control the reactor temperature by acting on the heat transfer. rl is the reaction rate of reaction l. In order to keep the resolution as simple as possible we shall first work with a simplified version of the CSTR model. H and T are functions of U and n depending on the thermodynamic model of the liquid (e.in reaction l. U ) H=H Using the temperature function as a state transformation. T the temperature. U ) . Using (8) (4) is written as follows: ∂ ∂xj ∂ = ∂xi Nc X Nr  X F  in qik (x) C i V − ni + Γil V k=1 l=1 Nc X " #! #! Nr  X F  in qjk (x) Cj V − nj + Γjl V k=1 l=1 " (10) The solution of these equations turns out to be quite intricate and we do not even know if a solution exists. n) = Q + Fh + Cp V dt l=1 (9) where (−∆r H)l (T ) is the molar reaction heat 3 of reaction l and Cp V = PNc i=1 Cpi ni is the total heat capacity of the mixture (with Cpi the volumetric heat capacity of species i). the change from the usual form of the system dynamics (3) into its Brayton-Moser formulation (1) requires the solution of the partial differential equation system (4).

e. · · · . 0.• the heat exchange between the reactor and the jacket is proportional to the temperature difference between them. Tw − T Integrability condition: the Brayton-Moser partial differential equation The first step to find the Brayton-Moser form of the dynamics is the solution of a partial differential equation. . . 0. l) if 1 ≤ i ≤ Nc and that contains γl in the lth column if i = Nc + 1. this relation is time independent. we have: f (x) = 3. Changing the cooling fluid flow rate influences the system dynamics via the heat transfer coefficient θ which is dependent on the cooling fluid flow rate. However implementing this control law in practice requires that the relation between the cooling fluid flow rate and the heat transfer coefficient is know. . an adiabatic reactor.. and more precisely by acting on the cooling fluid flow rate. This additional assumptions allows us to rewrite (8) as follows [2][23]:              Nr  X F  in dni Γil rl (T. Applying the notations of Section 2 to (12).. i. Tw − T ]t dt V (12) where x = [nA . The open-loop case corresponds to θ = 0. with θ the the heat transfer coefficient.e.2  F  in x − x + Γr (x) V  and t g (x) = 0. . The aim is to control the reactor temperature by acting on the heat transfer. nNc . find a matrix Q (x) that fulfills (2) and 6 . n) = Ci V − ni + dt V l=1 (11) Nr  X dT F  in = γl rl (T. The time evolution equations of n and T have the same form and the dynamics of the system can be written as follows:  F  in dx = x − x + Γr (x) + uc [0. .. T ]t . n) + uc (Tw − T ) T −T + dt V l=1 with Tw the (constant) cooling fluid temperature and T in is the inlet temperature. Γ ∈ RNc +1 × RNr is a matrix that contains the stoichiometric coefficient of species i in the reaction l in position (i. Therefore we shall consider in the sequel that the control input uc is the quantity CpθV . If the jacket dynamics are fast enough compared to the reactor dynamics. i.

So we shall now restrict ourselves to a particular set of possible matrices Q(x) to find a solution to (4). i 6= j : ∂qik =0 ∂xj (14) (13) becomes then as follows: NX Nr c +1 X F ∂rl qik Γkl − qij + V ∂xj l=1 k=1 # NX Nr c +1 X F ∂rl = − qji + qjk Γkl V ∂xi k=1 l=1 # " " (15) The above set of algebraic equations is still complex to solve due to the general form of the kinetics. . Nc + 1): (Q (x) f (x))i = NX c +1  qik k=1   F  in xk − xk + Γkl rl (x) V By differentiating the above expression the condition (4) can be rewritten as follows for all i. This set is such that the partial differential equation system is transformed into an algebraic one. . kl (T ) is the kinetic coefficient of the reaction l. we shall also assume that the kinetics of reaction l can be written as a product as follows: rl (x) = kl (T ) Nc Y φ(l) m (nm ) m=1 (l) The functions φ(l) m (nm ) are smooth functions of their argument and φm (0) = 0 if the species nm intervenes in the kinetics of reaction l. j. . k = 1. . (13) is transformed into an algebraic equation by setting: ∀i. By using (12) we can write the following relation (for i = 1. . j = 1. Therefore.(4). . . . . kl (T ) is such 7 . . . Nc + 1: NX Nr c +1 X F ∂rl − qij + qik Γkl V ∂xj k=1 l=1 " + NX c +1 k=1 ∂qik ∂xj    F  in xk − xk + Γkl rl (x) V NX Nr c +1 X ∂rl F Γkl = − qji + qjk V ∂xi l=1 k=1 " + NX c +1 k=1 ∂qjk ∂xi  # # (13)   F  in xk − xk + Γkl rl (x) V This partial differential equation is difficult to solve in general. Nc + 1. otherwise φ(l) m (nm ) = 1. .

.j k=1 Nr X  (17a) (l) F ∂φi kl (T ) φ(l) (nj ) = − qji (nj ) + j V ∂ni l=1 × NX c +1 qjk (nj ) Γkl k=1 ! N Yc   φ(l) m (xm ) m6=i. some characteristics of the solution (if it exists) can be derived. we find that the following relation 4 These assumptions on k (T ) are fulfilled by the commonly used Arrhenius law. yet. However the considered ones still cover a large number of practical cases. dT lim kl (T ) = kl∞ > 0 T →0 T →∞ dkl = 0. T →0 dT dkl =0 T →∞ dT lim (16) lim Remark 1 This assumption restricts the set of reaction kinetics that are considered since it excludes for example the Michaelis-Menten kinetics for enzymatic reactions. Indeed (17a) has to be satisfied for any (n. dkl > 0. and among them the reactions with mass action law kinetics. Despite of this. By taking the limit of (17a) for T tending to 0. 8 . lim kl (T ) = 0. .m6=i  (17b) (l) Nr X F ∂φi kl (T ) = − qji (T ) + V ∂ni l=1 × NX c +1 ! qjk (T ) Γkl k=1 Nc Y   φ(l) m (nm ) m=1. (15) is written as follows (for i. . T ) and in particular for T tending to 0. j = 1.that the reaction rate rl (x) fulfills the following conditions 4 : kl (T ) > 0.m6=i A general solution of (17a) and (17b) has not been found.j • for j = Nc + 1: Nr X dkl F (l) φi (ni ) − qij (ni ) + V dT l=1 " × NX c +1 ! qik (ni ) Γkl k=1 Nc Y   φ(l) m (nm ) m=1. Nc + 1 with i 6= j): • for i. . j 6= Nc + 1:  (l) Nr X ∂φj F kl (T ) φ(l) (ni ) − qij (ni ) + i V ∂nj l=1 × NX c +1 qik (ni ) Γkl ! N Yc   φ(l) m (nm ) m6=i.

It can be shown that this system can have several equilibρcp V rium points depending on the numerical values of the parameters. . namely the case of an irreversible reaction with one reactant. whereas in the exothermic case an odd number of equilibria can occur. In this last case. we shall the particular case of an irreversible reaction A → . Thus the only elements that can depend on the composition in the matrices Q (x) that solve (17a) and (17b) are the diagonal terms. In this work we do not consider the cases where limit cycles can occur. 4 The non-isothermal CSTR with first order irreversible reaction As already mentioned. Furthermore to this stage we do not even know if a solution exists.1 The dynamical model In order to simplify the system dynamics to reduce its dimension as much as possible. All non-diagonal elements (except for the last line and the last column) are constant. Therefore we shall in a first instance work with the most simple case. a general solution of (17a) and (17b) has not been found. yet. 4. k (T ) fulfills (16). From (11) the dynamical behaviour of the CSTR can be represented by a two-dimensional system as follows [2][23]:  F  in dnA = CA V − nA − k (T ) nA dt V   F  in dT   = T − T + γk (T ) nA + uc (Tw − T ) dt V     (18) (−∆r H) . there is an alternance of asymptotically stable and unstable equilibrium points with γ = 9 . In the endothermic case there can be only one asymptotically stable equilibrium.has to be fulfilled for i. T ) V = k (T ) nA where k (T ) is the kinetic coefficient. . with reaction kinetics that follow the mass action law. j 6= Nc + 1 and i 6= j: F F − qij (ni ) = − qji (nj ) V V This means that qij = qji and qij and qji have to be constants. As a consequence the reaction kinetics can be written as follows: r (nA .

T ) "   2 F 1   in = q12 − γ CA V − nA + T in − T V 2γ F + 4α V  2 × Z T γCAin V k (τ ) +  F V γk (τ )2 T0  + k (τ ) (T in − τ ) (20)  dτ  where T0 > 0 is an arbitrary reference temperature. It is sufficient to check that the matrix Q (x) given in (19) is nonsingular and fulfills (2) and (17b) 5 . T ) is non-zero:   4α 2 F 2 det Q (nA . T ) (the dependence in T of k (T ) has not been written for the sake of shortness of the notations):   tr Q + Qt = 2q12 k 2 (γ 2 + 1) + 4  2 F V γk 2 α + 4α VF k 6= 0 2 F 2 q12 α (1 − α) ≥ 0 V k2 As a consequence the condition (17b) is fulfilled [8]. Proposition 2 The dynamics of the CSTR can be expressed in the form of the Brayton-Moser equations with:  Q (nA .[2][8][23]. Indeed the determinant of Q (nA . 10 . PROOF. The most common case which can be observed for instance with Arhenius type kinetics is the case of three equilibrium points where the temperature of the unstable equilibrium is intermediate to the two other (asymptotically stable) equilibrium temperatures. 5 (17a) does not exist because Nc = 1. The first term is related to the convection whereas the other terms are linked to the reaction kinetics. T ) = 2 q12 >0 k V Concerning (2) it is sufficient to check the sign of the trace and the determinant of the symmetric part of Q (nA . T ) =    γ 1   q12 2 F F 2 k(T ) k(T ) +4( V ) α+4α V F α k(T )+4 V k(T ) γk(T ) (19) 2 where 0 < α ≤ 1 and q12 are two real coefficients such that q12 γ < 0.   det Q + Qt = 16   The potential function P (nA . T ) is the solution of (5): P (nA .

T ) > 0 is a Lyapunov function of the open-loop system. T¯ = ∇P n ¯ A . First let us determine the nature of the critical points of P (nA . T¯ = 0 Hence the equilibrium points of the system are the critical points of P (nA . P (nA . More precisely this value is equal to one of the stable equilibrium temperatures. whereas it indefinite at the unstable ones. 11 . PROOF. T¯ = 0. T ) such that the closed loop system has an asymptotically stable equilibrium at the desired state (n∗A . T ∗ ). 4. Let us apply the methodology of Section 2.2. T ) > 0 for all (nA ≥ 0. With this particular choice of T0 . As a consequence:       Q n ¯ A .    The equilibrium points of the system n ¯ A . T ) and the unstable equilibria are saddle points. Proposition 3 The asymptotically stable equilibria are local minima of the function P (nA . T > 0). T ).2 Controller design The aim is to design a control law uc = uc (nA . H (P) is positive definite at the stable equilibria. T¯ f n ¯ A . As a consequence. T¯ are the solutions of f n ¯ A . As shown in [8] there exist a particular choice of the reference temperature T0 such that P (nA . T ) by computing the eigenvalues of the Hessian matrix H (P) at the equilibrium points:  2   q12 F F 1 + γ 2 + 4α tr (H (P)) = −  V 2 k T¯ V γk T¯ " det (H (P)) = 4α 2 q12  3 F V  2 k T¯ F dk k T¯ + − γ n ¯A V dT     F dk + k T¯ + −γ n ¯A dT V !# ! Besides the eigenvalues of the linearized system around the equilibria λ1 and λ2 are given as follows: λ1 = − F V and λ2 = γ n ¯A   dk F − k T¯ − dT V At an unstable equilibrium λ2 > 0 whereas at the stable ones λ2 < 0 [8].

T ) = Pa (nA . T ) if and only if Pa (nA . T ∗ ) is an equilibrium of the closed loop system. Lemma 4 The desired set-point (n∗A . Lemma 5 With Q (nA . if and only if:  dnA F  in = CA V − n∗A − k (T ∗ ) n∗A = 0 dt V (21) can be deduced directly from this equation. T ) is of the following form: Pa (nA . 1. Level curves of P (nA .80 nA [mol] unstable equilibrium point 70 60 stable equilibrium point 50 40 30 20 10 0 380 separatrix 400 420 440 460 480 500 520 540 T [K] Fig. γk(T )2 . T ) for the case with three equilibrium points. T ∗ ) can be an equilibrium point of the closed-loop system if and only if: n∗A = F V F V CAin V + k (T ∗ ) (21) PROOF. The low temperature equilibrium lies outside the frame. T ) given in Proposition 2 there exists a control input uc = uc (nA . T ) = w nA + | Z T ! v (τ ) dτ (22) T0 {z z } 2 F k(T )2 +4( V ) α+4α VF k(T ) where v (T ) = differentiable function. (n∗A . w (z) : R → R is some continuously 12 . T ) such that the closed-loop dynamics can be written in the form (6) with the re-shaped potential function Pd (nA . T ) + P (nA .

T ) uc (nA . If the conditions of Lemmas 4 and 5 are fulfilled and if: .PROOF. T ) = Q (nA . T ) g (nA .T ) is the solution of the differential equation (7). In our case g ⊥ (x) = ϕ 0 with ϕ ∈ R∗ so that (7) becomes as follows: ! q12 ϕ ∂Pa ∂Pa v (T ) − =0 det (Q (x)) ∂nA ∂T The solution of this partial differential equation is of the form given in (22). Proposition 6 Let us denote by u∗c (Tw − T ∗ ) = −  F  in T − T ∗ − γk (T ∗ ) n∗A V the equilibrium value of the control input. T ) This is the case if and only if Pa(nA . The closed-loop dynamics can be written in the form (6) if and only if there exists a solution uc (nA . T ) to the equation Pa (nA .

dw .

.

.

= − |q12 | u∗c (Tw − T ∗ ) dz .

z∗ with (23) 2 k ∗ VF Ψ∗ ω 1 d2 w .

.

=− .

>  2 γ γq12 dz 2 .

z∗ 4α VF k ∗ + VF − k ∗2 Ψ∗ .

where (24) .

F dk .

.

Ψ =− k + + .

γn∗ V dT .

T ) must be positive definite at x∗ = (n∗A . T ∗ ) = 0. Lemmas 4 and 5 combined with (23) ensure that (n∗A . T ∗ ) is a local minimum of Pd (nA . T ∗ ) is such that ∇Pd (n∗A .  H (Pd ) (x∗ ) =   −γ VF −ω − VF − ωv (T ∗ ) 13 − VF ∗  − ωv (T )   q12 H22 .  ∗  ∗ If F 2 F 2 k∗ + − k ∗ Ψ∗ < 0 4α V V then the closed-loop system has no asymptotically stable equilibrium. T ∗ ) if and only if (n∗A . To have a minimum the Hessian matrix H (Pd ) of the function Pd (nA . T ∗ ) is an asymptotically stable equilibrium of the closed-loop dynamics. T ∗ ). The closed-loop system (6) has an asymptotically stable equilibrium at the desired set-point (n∗A . T ).T ∗ A then (n∗A .   (25) PROOF.

Note that the possible values are constrained by (24) and (25). Global convergence is ensured if and only if the shaped potential function Pd (nA . This means that the actual control input is the transferred heat. 4. i. if and only if x is a global minimum of the function Pd (nA .e. T ) acts in (18) via the term   0   g (nA .3 Considerations on local and global convergence The controller designed in the previous section only guarantees a local convergence to the desired equilibrium point.F   with H22 = − γkV∗2 k ∗ − 4α VF Ψ∗ − ωv (T ∗ )2 . T ) (Tw − T ) = ω n∗A − nA − Z T T∗ ! v (τ ) dτ + u∗ (Tw − T ∗ ) (26) The control law has two parameters to be chosen: α and ω. T ). By analogy with (21) let us define the function neq A (T ) by the following relation: F in C V eq V A nA (T ) = F + k (T ) V 14 . i. 2 The simplest function w (z) that fulfills the above proposition is the following one: Pa (nA . In the other cases ω must fulfill the condition (24) so that the Hessian matrix H (Pd ) is definite positive [8]. If (25) is fulfilled. T ) = w (z) h i = |q12 | ω (z − z ∗ )2 − u∗c (Tw − T ∗ ) (z − z ∗ ) The final expression of the control input is finally deduced from (8): uc (nA .e. T ) uc (nA . T ) =     R T ∗ ∗ ∗ ω nA − nA − T ∗ v (τ ) dτ + u (Tw − T ) which does not depend on the difference (Tw − T ) anymore. T ) does not have other local minima. then the Hessian matrix H (Pd ) cannot be positive definite [8]. Finally let us mention that the control action u (nA . the system will converge to desired point if and only if the initial conditions are close enough to it.

the right-hand term is equal to   # ∆ T . T   # By replacing nA by the expression of neq in the first equation of (28). let us consider a state n# A. Hence we have:              dnA =0= dt dT =0= dt             F V F V  # CAin V − n# n# A −k T A   T in − T # + γk T # n# A +ω  n∗A n# A −  T #      − Z T# T∗ ! (28) v (τ ) dτ + u∗c (Tw − T ∗ ) ∗ The first equation can be rewritten as n∗A = neq A (T ). When replacing this expression into the second equation of (28).T )−z }  # Lemma 7 A state n# is an equilibrium state of the closed loop system A.Let us also consider the following function: ∆ (T ) = γ  F F  in CA V − neq (Tin − T ) A (T ) + V V  ∗ + u∗c (Tw − T ∗ ) − ω neq A (T ) − nA + ZT  v (τ ) dτ  (27) T∗ |  {z ∗ z (neq A (T ). it A T is obvious that dynamics: dnA dt   # = 0 if n# ∈ Ω. T of the closed loop system. let n# be an equilibrium state A. T (18) if and only if it is contained in the following set: n    eq # Ω = ∆ T # = 0. n# A = nA T  o  # PROOF.   # that is contained in Ω. T      F  in # = T T − T # + γk T # neq A V ! +ω n∗A − neq A  T + u∗c (Tw − T ∗ ) 15 #  − Z T# T∗ v (τ ) dτ . To prove the ”if” part. Let us now consider the temperature A. To prove the ”only-if” part. T    dT F  in = T − T # + γk T # n# A dt V    # + uc n# Tw − T # A.

Proposition 8 If the parameter α fulfills the following condition: ∀T ∈ R+ : 1 ≥ α >  Ψ (T ) k 2 (T )  4 VF k (T ) + F V (29) 2  dk where Ψ (T ) = − k (T ) + VF + dT γneq A (T ). As it can be seen in Figure 2 if the value of ω is sufficiently high. but increasing on one determined interval (see also [8]). As a consequence for ωγ > 0. then there exists a lower bound ω on γ such that ∆ (T ) has only one zero. then ∆ (T ) has only one zero: the desired closed-loop equilibrium. The derivative of ∆ (T ) with respect to d∆ d   ω γ =− ZT T∗ ω γ  is given as follows: −k 2 (τ ) Ψ (τ ) + 4α VF k (τ ) + k 2 (τ )  F V F V 2  + k (τ ) dτ If the numerator is positive.  The right-hand term is equal to ∆ T # and consequently. we have dT dt = 0. 16 . PROOF. (29) and (30) imply (25) and (24). respectively. then ∆ω (T ) is of the sign of − (T − T ∗ ). Using Proposition 6 the desired equilibrium is locally stable. i. T > T ∗ ). we have ∆ (T ) > ∆0 (T ) (resp. depending on the term ω∆ω (T ). ∆ (T ) < ∆0 (T )) for T < T ∗ (resp. if (29) is fulfilled. Proposition 9 If (29) is fulfilled and ∀T ∈ R+ : F Ψ (T ) k 2 ω V   > γ −k 2 Ψ (T ) + 4α VF k + VF (30) then the controller (26) achieves global stability of the desired equilibrium point. PROOF. ∆ (T ) can have several zeros. Therefore it can be written as follows: ∆ (T ) = ∆0 (T ) + ω∆ω (T ) As it can be deduced from the work in [10] the existence of the three open-loop equilibria implies that ∆0 (T ) is decreasing for low and high temperatures. The function ∆ (T ) is linear in the parameter ω.e. As a consequence.

T ) has two local minima and consequently the convergence to the desired state is only local.g. It would be interesting to discuss the size of the attraction region with the proposed controller in case of input constraints compared to other controller proposed in the literature (e. Remark 10 The conditions for the global convergence (29) and (30) are very similar to the ones for the local convergence (25) and (24). global convergence can be lost if the input is bounded. the parameters α and ω were chosen such that local convergence is achieved (i.∆(T ) [J/min] increasing ω γ 0 ∆0 (T ) desired set point T [K] Fig. The function Pd (nA . ∆ (T ) is monotonConditions (29) and (30) imply that d∆ dT ically increasing and hence has a unique zero. (24) is fulfilled but not (30)). In the second case the parameters α and ω were chosen such that global convergence is achieved (i. but instead of being satisfied only at the desired set point temperature T ∗ .e. namely the desired closed-loop equilibrium. Local convergence will however be kept. (30) is fulfilled).e. they have to be satisfied at any T . T ) has a single minimum and the convergence to the desired state is global. T ) is shown in Figures 3 and 4. The shape of the function Pd (nA . 2. the derivative of ∆ (T ) is given as follows: F Ψ (T ) k (T )2 d∆ V  − = dT k (T )2 k (T ) + VF ω γ   −k (T )2 Ψ (T ) + 4α VF k +  k (T )2 k (T ) + F V F V   < 0 for all T > 0. in [24]). respectively. Influence of ω γ on ∆(T ) Moreover. Remark 11 Even if the conditions of Proposition 9 are satisfied. In the first case. The function Pd (nA . 17 .

T ∗ ) 440 460 480 500 520 540 560 T [K] Fig. It is therefore of major interest to see if this approach can be extended to more complex systems. T ) 60 50 40 locus of possible closed-loop equilibria 30 20 desired set point (n∗A . T ) (global convergence) 5 Extension to more complex systems The power-shaping approach has given interesting results on the case study of a CSTR with an irreversible reaction. and more particularly to systems with more than one reacting chemical species and/or with multiple reactions. Level curves of Pd (nA . 3. 4. Level curves of Pd (nA . We shall now look at some particular cases of more complex reactions to get 18 . T ) (local convergence) 80 nA [mol] 70 60 locus of possible closed-loop equilibria 50 40 30 20 10 0 420 desired set point (n∗A .nA [mol] other convergence point 70 # (n# A. T ∗ ) 10 0 390 400 410 420 430 440 450 460 470 480 490 500 T [K] Fig.

This difficulty can be overcome by also considering the dilution coefficient VF as an input of the system. In the other cases a general solution has been found. 6 6.1. given a matrix Q (x) that fulfills (4) other solutions to (4) can be built departing from the first one. As shown in [12]. But if the Brayton-Moser form exists. In the three cases it has been assumed that the kinetic functions φ(l) m (nm ) can be expressed as powers of nm .a first intuition of the existence of non-singular and negative semi-definite solutions of (17a) and (17b).: (l) ζm φ(l) m (nm ) = (nm ) It has been shown that in some cases (17a) and (17b) have no solution.1. One of the difficulties is due to the term in VF .1 Some considerations on the general CSTR dynamics General solution As it has been explained in Section 3.2.e. Let us consider 19 . However. they did not manage to extend their result to more general cases. Contrary to [22] who used the simplified form of the dynamics (12) with mass action law dynamics we shall use the more general form (8). as the authors mention. then the relation (14) is not fulfilled and the partial differential equation (13) has to be solved. Therefore the corresponding potential functions cannot be used in general as a Lyapunov functions for the open-loop system because they may be non-decreasing along the system trajectories. The Brayton-Moser formulation of the CSTR has been found with the same assumptions as in Section 3.1 the solution to (10) is quite intricate. but the obtained matrix may have a non-negative symmetric part. i. This does not mean that the dynamics cannot be put into the Brayton-Moser form. Applying this methodology on the matrices that have been found could lead to another matrix Q (x) that would be negative-definite and hence give a Lyapunov function for the open-loop system. consecutive reactions and one reaction with two reactants is given in [8][9]. This corresponds to the system studied in [22]. The solution of (17a) and (17b) in the case of parallel reactions.

nA . 2. ψ (U. Using the notations of (3) we have:   f (nA . U ) = −r (T. nA . nB ) and ρ (U ) are real scalar functions. U ). nB . U ) = ZZ φ (nA . nA . 3): ∂ ∂ [ai (x) rˆ (x) V ] = [aj (x) rˆ (x) V ] ∂xj ∂xi where the following notations have been used: (31) ai (U. The general solution of (31) is given as follows: a1 (x) rˆ (x) V = a2 (x) rˆ (x) V = a3 (x) rˆ (x) V = Z Z φ (U. U ) dnA dnB + + 20 Z Z ψ (nA . nB and x = (nA . U ) dnB + Z ρ (U ) dU + ω . U ) 1 V hin − H  with uc = F V  t Q˙ .a reversible reaction A νB. nA . nB ) V       νr (T. U ) and the potential function P (x) are given respectively by:  Q (x) =  νq12 (x) − a1 (x) q12 (x) q13 (x)     νq22 (x) − a2 (x) q22 (x) q23 (x)     νq32 (x) − a3 (x) q32 (x) q33 (x) P (nA . nA . nA ). nB ) . nA . nB . j = 1. nB ) dnA + ϕ (U. The corresponding matrix Q (nA . U ) dnA ϕ (nB . nB . nB . nB )   rˆ (U. nB ) = r Tˆ (U. nA . nB . nB . nB ). nA ) φ (U. nA . nB ) dnB + ψ (U. nA . ϕ (U. U ) = CAin V        − nA 0 CBin V − nB 0    ˆ (nA . The partial differential system (4) becomes (for i. nA . nB ) Z dψ dφ dnA dnB + dnA dU dU Z ∂ϕ dnB + ρ (U ) + ∂U Z Z where φ (U. nB . nB ) − qi1 (U. nA . nB ) = νqi2 (U. nB ) V      0  g (nA .

Indeed the open-loop system uc = 0 is an isolated system and the entropy variation is equal to the entropy production. 6. nB ) = 0 ρ (U ) = 0 q32 = 0 q13 = 0 q23 = 0 where Sˆ (nA . nB . The functions qij (x) can be chosen arbitrarily as far as Q (x) is non-singular and (2) is fulfilled.2 Particular solution Let us choose the following particular solution: φ (nA . the reaction evolves in the direction of decreasing affinity µA − νµB . U ) = − ∂ 2 Sˆ ∂nA ∂nB ω=0 ψ (U. The potential P (x) is the opposite of the entropy function −Sˆ (nA . nB . U ) is always decreasing along the systems trajectories. U ) is the entropy function. Finally let us mention that the solution to the partial differential equation system has been obtained without restricting to a particular subset of solutions.e. Following thermodynamics i 6 the reaction kinetics must fulfill the condition : sign [ˆ r (x) V ] = sign [µA − νµB ] Using this property it can be seen that Q (x) in (32) fulfills (2). µi is the ∂ Sˆ chemical potential of the species i with µTi = ∂n . U ) with the following matrix:  Q (x) =  µB T rˆV Λ T rˆV        4α µB 4α Λ Λ  − 2 T r˜V ν T rˆV  ν T rˆV  1 T rˆV 0 0 0 β 4(1−α)T rˆV Λ       (32) where 0 ≤ α < 1 and β ≥ 1 are two real constants and Λ = νµB −µA . This is in accordance with the second principle of thermodynamics which claims that the entropy production is always positive. 21 . 6 i.where ω is some real constant. contrary to what has been done in the previous sections. nA ) = 0 ϕ (U. nB . The dynamics can be written as follows: dx = −Q−1 (x) ∇Sˆ dt (33) Using (2) it is possible to see that −Sˆ (nA . nB .

n ¯ B . Remark 12 Note that the availability function proposed by Alonso. −Sˆ (x) as a Lyapunov function candidate and (33). There are an infinity of points x¯ contained in this set. nB . The open-loop equilibria are locally stable but not asymptotically stable. Contrary to what has been shown in Section 4 the gradient of the potential function is not zero at the open-loop equilibria.  The open-loop equilibria x¯ = n ¯A. This approach requires to write the system dynamics in the Brayton-Moser form. U 0 ). Indeed the matrix Q (x) is not defined at the open-loop equilibria and Q−1 (x) is singular at these points. We have given the form of this partial differential system in the case of a CSTR with general kinetics and we have proposed a methodology for finding some of its solutions. U¯ are the set of points such that µA (¯ x) = νµB (¯ x). Ydstie and coworkers (e. 7 Conclusion In this paper we apply the power-shaping approach to the control of a CSTR. U corresponding matrix Q (x) is very similar to (32) and is also not defined at the open-loop equilibria. This methodology has been successfully applied to a CSTR with irreversible kinetics with one reactant. However the dynamics an be written in the Brayton-Moser form only if a partial differential equation system is solved. Indeed let us consider an initial point (n0A . the stability is deduced from La Salle’s theorem [16]. n0B . This particular form of the dynamics also allows to find a Lyapunov function for the open-loop system. [1]) can also be used as a potential function for the BraytonMoser form of the dynamics:    ref ref P (x) = S ref − S (x) − − TµAref − TµBref 1 T ref nref − nA   A   ref nB     − nB    U ref − U     ref ref is a reference state and S ref = S xref . The where xref = nref A . We have also 22 . From (8) it can be seen easily that along the trajectories we have: νnA + nB = νn0A + n0B and U = U 0 (34) Using (34) as the invariant set. However the power-shaping methodology cannot be applied using the above proposed matrix Q (x) and potential −Sˆ (x). The Lyapunov function for the open-loop CSTR that has been found using the Brayton-Moser formulation of the dynamics has been related to the physical phenomena.g.

K. December 2005. Amundson. Brayton and J. Banga. Finally first results for applying the power-shaping approach in the case of a general CSTR with two control inputs have been presented. 7(3):121–131. On representations and integrability of mathematical structures in energy-conserving physical systems.J. Favache is funded by a grant of aspirant of Fonds National de la Recherche Scientifique (Belgium). Universit´e Claude Bernard. van der Schaft. 44th IEEE Conference on Decision and Control and European Control Conference. Alonso. In Proc. 2006. Eberard. An analysis of chemical reactor stability and control–I : The possibility of local control. B.A. Quaterly of applied mathematics.shown that this methodology needs to be extended in the case of more complex reaction kinetics.E. These results show how the power-shaping approach can be related to the entropy of the system. The work of A. Maschke. 37(1):54–91. Sevilla (Spain).J. 22:1–33. Dalsmo and A. Extension des syst`emes hamiltoniens ports aux syst`emes irr´eversibles: une approche par la g´eom´etrie de contact. initiated by the Belgian State. and A. The work presented in [12] offers promising tools to extend these results.J. B.M. [6] D. 2002. A theory of nonlinear networks I. 1964. Control. Aris and N. van der Schaft. and A. Lyon 1. Maschke. Port contact systems for irreversible thermodynamical systems. [3] R. Science Policy Office. Acknowledgements This paper presents research results of the Belgian Network DYSCO (Dynamical Systems. The scientific responsibility rests with its authors. [7] D. Eberard.R. B. PhD thesis. 1958. Chemical Engineering Science. 1998. with perfect or imperfect control mechanisms. van der Schaft. [5] D. Energy-conserving formulation of RLC-circuits with linear resistors.R. Journal of Process Control. In Proc. 12:507–517. From irreversible thermodynamics to a robust control theory for distributed process systems. Ydstie. 17th International 23 . and J. and Optimization). References [1] A. Moser. [4] M. Eberard.K. by instance for finding alternative potential functions departing from the proposed one. [2] R.M. funded by the Interuniversity Attraction Poles Programme. Journal on Control and Optimization.

M. Scherpen.A. Dynamic analysis and control of chemical and biochemical reaction networks.A. [18] R. Favache. Ortega. September 2004.J.M. A power-based description of standard mechanical systems. J. July 2007. B. International Symposium on Advanced Control of Chemical Processes (ADCHEM). Scherpen. Jeltsema.Dochain. Garcia-Canseco. Favache and D. May 2007. Jeltsema. Prentice Hall. April 2006. A. Putting energy back in control. April 2001. content and co-content.Escobar. Maschke. Energy-based Lyapunov functions for forced Hamiltonian systems with dissipation.A. Kos (Greece). [10] A. Scherpen. Interconnection and damping assignment: passivity-based control of port-controlled Hamiltonian systems. On mechanical mixed potential. G.M. Maschke. In Proc.M. Kyoto (Japan). [14] D. [15] D. 56(5):349–356. and R. 2002. IEEE Transactions on Automatic Control. 2008. 48(10):1762–1767. 40(9):1643–1646. [17] B. 24 . Thermodynamics and Process Control. Jeltsema and J. Jeltsema and J. Alonso. Power shaping: A new paradigm for stabilization of nonlinear RLC-circuits. July 2006. In Proc. Ortega. Gramado (Brazil). [19] R. D. Scherpen.J. [11] A. van der Schaft. 21(2):18–33. Nonlinear Systems. and B. Power-shaping control of an exothermic continuous stirred tank reactor (CSTR). Upper Saddle River. [9] A. 2003. october 2003. Automatica. Otero-Muras. Szederknyi. Dochain. pages 165–170. Universit´e catholique de Louvain. and A. and J. and J. [12] E. pages 101–110. Contact structures: application to interconnected thermodynamical systems. In Proc. IEEE Transactions on Automatic Control. Khalil. In Proc. [13] D. Jeltsema.A. [8] A. Scherpen. Maschke. A.M. van der Schaft.M. Mareels. and D. Thermodynamics and chemical systems stability: The CSTR case study revisited. Favache.M.Symposium on Mathematical Theory of Networks and Systems. Automatica. IEEE Control Systems Magazine. PhD thesis.A. A. 17th IFAC World Congress. European Control Conference. Systems & Control Letters. van der Schaft. [21] I. pages 73–78. 2008. 3rd edition. 2009. and G. R. Seoul (Korea).A. D. 38(4):585–596. [16] H. [20] R. 45(8):1498–1502. Ortega. 2002. Ko¸c (Turkey). Maschke.J. European Control Conference 2007. Ortega. An energy-balancing perspective of interconnection and damping assignment control of nonlinear systems. Ortega. B. Ortega. 2009. Hangos. 2000. Dochain. International Symposium on Advanced Control of Chemical Processes. M. R. Journal of Process Control. I. and K. Favache and D. Powerbased control of physical systems: two case studies.

F. 29(4):967–985. Automatica. Bastin. Ray. 33:1437–1448. April 1974. 19:405–414. Chemical Engineering Science.[22] H. 1997. Ram´ırez. On the control of non-linear processes: an IDA-PBC approach. and A. Global stabilization of exothermic chemical reactors under input constraints. Uppal. W. 2009. and R. and G. [23] A. H. 25 . Ortega. B. On the dynamic behavior of continuous stirred tank reactors. Jadot. D. Sbarbaro. Journal of Process Control. [24] F. Poore. Viel.