Академический Документы
Профессиональный Документы
Культура Документы
Johan Grievink
Process Systems Engineering Group, Delft UniVersity of Technology, Julianalaan 136,
2628 BL Delft, The Netherlands
The interactions between economic performance, thermodynamic efficiency, and responsiveness in reactive
distillation process design are explored in this contribution. This motivation is derived from taking a sustainable
life span perspective, where economics and avoidance of potential losses of resources (i.e., mass, energy, and
exergy) in process operation over the process life span are taken into consideration. The approach to reactive
distillation column design involves defining a generic lumped reactive distillation volume element and the
development of rigorous dynamic models for the behavior of the element. As an extension of conventional
existing models, this model, which is used as a standard building block, includes entropy and exergy
computations. Three objective functions are formulated, which account for the process performance regarding
economy, exergy loss, and responsiveness. A fundamental understanding of the strengths and shortcomings
of this approach is developed by addressing various case studies: (i) steady-state simulation of a classical
MTBE design based on economics only, useful as a reference case; the multiobjective optimization of a
MTBE reactive distillation column with respect to (ii) economics and thermodynamic efficiency; and (iii)
economics, thermodynamic efficiency, and responsiveness. Structural differences and dynamic responses are
identified between the optimized and classical designs to stress the importance of considering exergy and
responsiveness criteria. Thus, the classical (economics-driven) design is compared to a green (economics-
and exergy-driven bioptimized) design. The green design produces less entropy than the classical one, while
being marginally less attractive from an economic standpoint. The bioptimized design has an improved closed-
loop performance compared to the classical design, keeping the same control structure and settings. It is
concluded that incorporating both economic- and exergy-related objectives in the unit design results in a
process with better closed-loop performance and reduced exergy loss.
Generic Lumped Reactive Distillation Volume Element where Q(R) is assumed to only occur in the liquid phase (i.e.,
Q(R) ) QL).
The starting point for a fundamental understanding of the The Kronecker delta function defined as
strengths and shortcomings of the proposed multiobjective
design approach is the definition of a generic lumped RD
volume element GLRDVE. This element (or compartment) is
chosen as a building block in the design approach and selected
δ(R) :) { 0 (if (R) * L)
1 (if (R) ) L)
(3)
to represent the governing phenomena occurring in RD process- The energy flux associated with the interphase mass and heat
ing. This behavioral model accounts for the following: (i) mass, transfer Ψe is composed of a measurable heat flux and the partial
energy, and momentum balances, together with entropy and molar enthalpies brought by the component fluxes. Namely,
exergy computations; (ii) thermodynamic equations of state for i)nc
∑
the specific heat and specific density and for the chemical
potential per species; (iii) thermodynamic phase equilibrium Ψe ) - Ψi × h(R)
i +q
(R′-R)
(4)
i)1
conditions at the interphase; (iv) the constitutive rate equations
for chemical reactions; (v) the rate equations for interphase The entropy balance in the transient state at a macroscopic
transport of species mass and energy; (vi) the rate equation for level is given by the net amount of entropy transferred into the
heat transfer with an external solid surface; (vii) the hydrody- volume element per unit time and the amount of entropy used/
namics-based rate equations for convective mass and energy produced in the control volume per unit time. Thus, the entropy
flows with the external world and the pressure drop; and (viii) balance for a given phase (R) can be stated as,
the equations for the interphase contact surfaces and for the
void fraction of the vapor phase in the liquid phase. j)nst i)nc
∂S(R)
The following attributes are introduced for the system
proposed in Figure 1: (i) two internally (well-mixed) homo- ∂t
) ∑ ∑ Fj,(R)
j)1 i)1
i × Sj,(R)
i + Σ(R)
V (5)
Alternatively, the entropy balance can be expressed in terms Table 1. Expressions To Estimate Driving Forces in the GLRDVE
of the entropy density s(R) and the fluxes of entropy density,32,33 of Figure 1a
parameter driving force
∂s(R) Xm,i ) -(1/T)∆µi,T ) R × ln(yin
) -∇J(R) i × P /yi × P)
s + σV
(R) mass transfer in
(6)
∂t interphase energy transfer Xe ) ∆(1/TV) ) 1/2((1/Tout,V) - (1/Tin,V))
chemical reaction Xrx,j ) (ALj /T)
For stream j, the flux of entropy J js is related to the total flux external heat exchange XHX ) (1/T) - (2/(Tu,in + Tu,out))
of energy J je and mass fluxes J m,i
j
according to the following a i ∈ Znc; j ∈ Znrx .
expression, 32
and heat diffusion in the liquid-vapor interphase, chemical The driving force Xe is estimated as suggested by Koeijer
reaction in the bulk of the liquid, and heat transfer from external and Rivero.28 The measurable heat flux through the interphase
heat sources/sinks and the surrounding environment to the liquid associated with Xq is derived from the energy balance in the
phase only; and (iii) the effects of electric field, external forces, vapor side,37
and diffusive momentum fluxes are neglected. The following j)nst
∑
expression for the integral entropy production term can be then
derived,32,35,36 Jq × aΨ ) total × hi
Fj,V j,V
(13)
j)1
i)nrx
1
∫ ∫ ∫V σ (*)
dV ) Σ(*)
V )
L
∑ ALi × Li × VLRX + JLq × In this expression, the enthalpy reference state is calculated at
the average temperature at which components are transferred
T i)1
i)nc through the interphase. The expression suggested by Koeijer et
a Ψ × Xq + ∑
i)1
Jm,i × aΨ × hi × Xe + al.39 and Koeijer37 has been used to estimate the XHX average
driving force.
i)nc In the development of the GLRDVE, a reduced form of the
(- ∑
i)1
Jm,i × aΨ × Xm,i) + QL × XHX (8) momentum balance is considered. We assume a pseudo steady
state and negligible viscous contribution. Moreover, the gradient
of pressure is assumed to be much larger than the rate of
The entropy contribution due to chemical reaction contains momentum gain by convention of the fluid. These simplifica-
the affinity of chemical reaction and its degree of advancement, tions applied to the momentum balance expression given
which are defined respectively as, elsewhere,34 result in the expression,
i)nc
∇P(R) ) F(R) ‚g(R)
∑
(14)
ALi ) - νLi,j × µLi j ∈ Znrx (9)
i)1 Availability and Availability Function. Two additional
j)nrx variables of key importance for this thermodynamic-based
rLi ) ∑
j)1
νLi,j × Lj i ∈ Z nc (10) approach are the availability A and the availability function B.
In this context, availability is regarded as a measure of the
maximum amount of useful energy that can be extracted when
To estimate the driving forces and fluxes occurring in the the system is brought to equilibrium with the dead-state
volume element, additional assumptions are introduced: (i) the surroundings.35 Thus, the availability of the system in phase
GLRDVE is composed of incoming and outgoing streams of (R) and state (T1,P1) is given by the expression,
liquid and vapor phases; (ii) the vapor phase behaves ideally;
(iii) the flux of mass and heat through the interphase is positive A(R) ) (h(R) - T0 × S(R))T1,P1 - (h(R) - T0 × S(R))T0,P0 (15)
from the liquid side to the vapor side; (iv) dynamic changes in
gas phase hold-ups of species mass and energy are ignored;
where the dead-state conditions are those at which T0 ) 298.15
and (v) Gibbs free energies at liquid and vapor sides are equal.
K and P0 ) 1 atm.
The underlying reasoning behind the inclusion of these
The availability function B at state given by (T,P) is computed
assumptions is a more transparent derivation of the driving
according to the following equality,
forces and flux expressions, but without compromising the
validity of the approach. The expressions used to estimate the
averaged driving forces are summarized in Table 1.
(R)
BT,P ) (h(R) - T0 × S(R))T,P (16)
54 Ind. Eng. Chem. Res., Vol. 47, No. 1, 2008
Contrary to the availability, the availability function of the investment and operational costs of the process,
system at a given state (T1,P1) can be referenced to any other
state (T2,P2), j)nsti)nc
∆B (R)
) (h (R)
- T0 × S )T1,P1 - (h
(R) (R)
- T0 × S )T2,P2 (17)
(R)
fecon ≡ TAC ) ∑ ∑
j)1 i)1
Fj,(R)
i ci + Fcuccu + Fhuchu +
∂B(R) × M(R)
j)nst
where Wideal (Wideal < 0) is defined as by Seider et al.35 and
) ∑ Fj,(R)
total ×B j,(R)
+ fixed for a given process.37
( )
∂t j)1 From a sustainability perspective, the desired goal is the
T0 enhancement of ηII. According to eq 20 the maximization of
1- Q(R) - T0 × Σ(R)
V (18) fexergy is equivalent to the minimization of ΣV.37,40 In this
TQ contribution, we adopt this second approach and compute T0ΣV
for a fixed T0 value.
Design Decision Variables. According to a degree of
The responsiveness index fresponse is approximated to the
freedom analysis performed in the GLRDVE the following
settling time defined by Luyben et al.,31
design variables are to be specified for a given operating
conditions and feed specifications:
∆B(x)
(a) reaction volume V(R) fresponse ≡ τ ) max (21)
{
RX
x T0∆ΣV(x)
V(R)
RX )
V(R) (for liquid phase homogeneously
catalyzed reaction(s) ) At steady-state conditions B(x) is at its minimum value and
-1 for heterogeneously catalyzed
mcat × Fm,cat ( reaction(s) ) increases when the system is perturbed by a disturbance. Since
a similar situation holds for ΣV(x), the response time τ measures
(b) external and heating/cooling flow rate Q(R), and how fast a small variation ∆B(x) would dissipate due to an
(c) the interfacial area for heat and mass transfer aΨ. increased ∆ΣV(x).31
Connecting streams and associated flow rates Fj,(R) are In the previous definition of fresponse, Luyben et al.31 assumed
i
assumed to be given. Operational variables such as reflux ratio that both the availability and entropy production rate are
and pressure are of key relevance in RD and should be chosen exclusively functions of the vector of state variables x. Although
with thoroughness. In our approach, these variables are specified this assumption is fundamentally correct, it somehow disguises
at a higher decision making layer (cf. Volume Elements to a the way of estimating the effect of a small disturbance or a
Column Structure, Definition of Optimization Criteria, and change in an input on the system. Therefore, we propose a subtle
Optimization Approach), aiming at operating windows where modification of the arguments of B and ΣV, by which a second
steady-state multiplicity and reactive azeotropy phenomena are argument is included, representing the systemic view on a
not likely to occur. dynamic system. Thus, the availability and entropy production
The perspective adopted for selecting the design decision rate are regarded as functions of an internal state vector x and
variables is task-oriented rather than the conventionally used a vector of disturbances or manipulated external excitations u,
equipment-oriented design approach. First, the task-based
concept would allow us to derive the phase volume V(R) from exergy: B(x,u) (22a)
the reaction time (i.e., residence time in element) to meet a entropy production rate: ΣV(x,u) (22b)
specified degree of advancement of reaction; second, the amount
of heat to be added or removed allows one to compute the heat Combining the definitions of eqs 22 with eq 21 leads to a
exchange surface and coolant flows; and third, the interfacial more generalized expression for fresponse,
areas for species mass and heat transfer may be translated into
the mechanical design of a tray and its prevailing hydrodynamic
∆B(x,u)
regime such that the required areas can be met within constraints. fresponse ≡ τ ) (23)
The main benefit of this functional, task-oriented perspective T0∆ΣV(x,u)
involves the process design unconstrained by equipment ge-
ometry. Thus, task-oriented design prevails against equipment- The previous expression could be further elaborated by
oriented design. defining the values of state variables x and disturbances u at
Economics/Exergy Efficiency and Responsiveness Objec- each state. Hence,
tive Functions. A set of three performance criteria indicators
is defined for the GLRDVE: the economic performance index ∆B(x,u) ) B(x,u) - B(x0,u0) (24)
fecon, the sustainability performance index fexergy, and the
responsiveness performance index fresponse. ∆ΣV(x,u) ) ΣV(x,u) - ΣV(x0,u0) (25)
The economic performance indicator fecon represents the total
annualized cost TAC of the GLRDVE, including capital with
Ind. Eng. Chem. Res., Vol. 47, No. 1, 2008 55
be found in the research work of, for example, Ismail et al.41 (29a)
and Papalexandri and Pistikopoulos.42 total
The rules of connectivity used in the RD superstructure state (Wideal)
fexergy ) (29b)
the following: (i) the outgoing liquid at a given compartment (Wideal)total - T0(ΣV)total
enters the compartment located below; (ii) the outgoing gas at
a given compartment enters the compartment located above; (∆B(x,u))total
(iii) the liquid stream leaving the condenser is partially recycled fresponse ) (29c)
back to the top compartment and withdrawn from the unit as
T0(∆Σ V(x,u))total
top product; (iv) the liquid leaving the bottom compartment
enters the reboiler; (v) the streams leaving the reboiler are a The validity of an additivity property (i.e., the objective
vapor stream (entering the bottom compartment) and a liquid functions per compartment can be added up to a corresponding
stream (withdrawn from the unit as a bottom product); and (vi) objective function over the entire structure) is implicitly assumed
side feed streams and side withdrawal streams may occur. in these definitions.
The connectivity rules, however, impose some configuration An indication of the implications of process responsiveness
limitations: (i) heat integration over volume elements is not for design is revealed by the definition of fresponse. According to
considered; (ii) side reactors are not present; and (iii) no parallel eq 30, a short responsiveness time is desired and can be attained
structures are considered. by a large change in the entropy production rate for a given
During the development of the RD unit, the following set of change in stored exergy. Using an ideal heat exchanger as a
simplifications are made: (i) the liquid and vapor streams case study, for instance, Almeida-Rivera45 and Meeuse46
leaving a tray are in thermodynamic equilibrium; (ii) both liquid demonstrated that better process responsiveness is attained at
and vapor hold-ups are fully considered; (iii) the liquid and smaller driving forces and at small hold-up values. Moreover,
vapor on each tray are perfectly well mixed (i.e., no radial Almeida-Rivera45 suggested that as a heuristic for heat-transfer
gradient in temperature and composition and the liquid on the processes the exchange surface per unit mass should be
tray has a concentration equal to that of the liquid leaving the maximized, while driving force is kept small at constant transfer
tray); (iv) the reactions are taking place in the liquid phase and duty.
56 Ind. Eng. Chem. Res., Vol. 47, No. 1, 2008
Table 2. Utopia Point in Optimization Problems with More Than Table 3. Wilson Interaction Parametersa for the System
One Objective Function iC4-MeOH-MTBE-nC4 at 11 × 105 Pa5,74
Utopia point U in a RΦ-space is given by the Φ-objective function values iC4 MeOH MTBE nC4
that solve independently the Φ scalar optimization problems.62 For
aij
instance, in an optimization problem involving two objective functions
iC4 0.00000 -0.74200 0.24130 0.00000
the mathematical formulation of the utopia point U is
MeOH 0.74200 0.00000 0.98330 0.81492
min f1(x) ) flow f2(x*) ) fup MTBE -0.2413 -0.98330 0.0000 0.00000
1 (x*) and 2 (x*) (43)
x∈Γ nC4 0.00000 -0.81492 0.00000 0.00000
min f2(x) ) flow
2 (x*) and f1(x*) ) fup
1 (x*) (44) bij (K)
x∈Γ
iC4 0.0000 -85.5447 30.2477 0.0000
where f low
i (x*)and f up
i (x*) are the lower and upper limits of
the MeOH -1296.719 0.0000 -746.3971 -1149.280
objective function fi, respectively, and the set of constraints
defines the MTBE -136.6574 204.5029 0.0000 0.0000
space of design variables x. The coordinates of the utopia point are then nC4 0.0000 -192.4019 0.0000 0.0000
given by (f low low
1 , f2 )
W,i ) ∑j)1 xj × Λij; ln(γi) ) 1 - ln(SW,i) -
a ln(Λ ) ) a + b /T; S j)nc
ij ij ij
Approach To Design Optimization and References Cases. ∑k)1
k)nc
(xk × Λki/SW,k).
Two approaches are adopted to solve the optimization problem Table 4. Equilibrium Constant and Kinetic Constant for the
involving more than one objective function, Synthesis of MTBE58
( ) ()
(a) Each objective function is used independently, disregard-
1 1 T
ing the other two; this approach localizes the coordinates of ln Keq ) ln Keq,0 + Rk - + βk × ln + γk × (T - T0) + δk ×
T T0 T0
the utopia point (cf. Table 2) and, subsequently, reduces the
(T2 - T02) + k × (T3 - T03) + φk × (T4 - T04) (45)
computational effort by limiting the search space,
(b) The interactions and trade-offs among the objectives are parameters Rk ) -1.49277 × 103; βk ) -7.74002 × 101;
γk ) 5.07563 × 10-1; δ ) -9.12739 × 10-4;
accounted; in this approach, a multiobjective problem is
k ) 1.10649 × 10 ; φk ) -6.27996 × 10-10;
-6
formulated by using a weighted p-norm for the vector of T0 ) 298.15 K; Keq,0 ) 284
objective functions, as given by Clark and Westerberg,47 validity heterogeneous catalyst (Amberlyst 15)
min F ≡ [
D∈Γ
∑i (wifi)p]1/p p ∈ [1, ∞) (30) [ (
kf(T) ) kf(T*) × exp -
Ea 1
-
R T T*
1
)] (46)
( )
aiC4 The GLRDVEs are integrated to a column structure in such
aMTBE
grx(x,T) ) - (32) a way that the superstructure is comparable to the unit studied
aMeOH Keq × aMeOH2 elsewhere4,8,59 and specified in Table 5. In this column structure,
the catalyst load is evenly distributed along the reactive trays4,8,59
In this study, we use the γ - φ thermodynamic formulation.52 and all the sections of the column have a common value. The
The vapor phase is assumed to behave ideally, whereas the liquid structure to be further considered during the course of this
phase is modeled using the Wilson activity coefficient equa- contribution is depicted schematically in Figure 2.
tion.53 The interaction parameters for the MTBE system at a To prepare for the optimization task and to reduce the problem
given operation condition are listed in Table 3. complexity, we identify the dominant contributions to entropy
The reaction equilibrium and kinetics of MTBE synthesis production rate over the MTBE column. The expressions to
have been the subject of intensive research during the past compute all contributions in the GLRDVE are the terms of eq
Ind. Eng. Chem. Res., Vol. 47, No. 1, 2008 57
s.t. gx(‚) ) 0
hx(‚) e 0
hf i - hf min E; w ) [1 0 ]T
i
f′i ) (35)
hf max
i - hf min
i
and
to w ) [0.3 0.7]T.
Function: min
Ki,Bi,τc,i,SPi
F) ∫0t [(xb - xjb)2 + (xd - xjd)2] dt
f
Jk ≈ R × Xk (39)
Figure 8. Driving forces as a function of the MeOH feed flow rate. MeOH feed flow rate varies from the nominal value (168 mol s-1) to 150 mol s-1; the
optimal design variables and sizing parameters corresponds to a fecon - fexergy design, with w ) [0.3 0.7]T.
Figure 9. Response time as a function of the MeOH feed flow rate. MeOH feed flow rate varies from the nominal value (168 mol s-1) to 150 mol s-1; the
optimal design variables and sizing parameters corresponds to a fecon - fexergy design, with w ) [0.3 0.7]T.
fresponse by Luyben et al.,31 this is a very specific case. Their not optimized over the feasible domain of states for normal
(Luyben’s et al.) definition demands an optimization of the states operation. Our approach might be locally optimized for the given
x to find the worst case situation over a domain of states. operational window, but might have a low degree of validity to
Therefore, there is a latent risk that the presented results are other cases.
62 Ind. Eng. Chem. Res., Vol. 47, No. 1, 2008
w ) [0 1 ]T
{
decision variables was task-oriented rather than the convention-
FMeOH )
FSS
MeOH (t ∈ [4 × 104 s, 6 × 104 s])
if t ∈ [0 s, 1 × 104 s] or
(42)
ally used equipment-oriented design approach. The chosen
design decision variables included the reaction volume, the
amount of heat to be added or removed, and the interfacial areas
1.1 × FSS
MeOH (if t ∈ [1 × 10 s, 4 × 10 s])
4 4
for species mass and heat transfer. Analyzing the steady-state
entropy production profiles in a MTBE RD column showed that,
The dynamic response of the classic and green designs are in terms of entropy production, the separation function domi-
depicted in Figure 10. In this figure, the MTBE composition at nates over the reaction function. A biobjective function was then
the bottom stream is presented as a function of time in the studied regarding economic and exergy performance. It is
presence of the MeOH feed disturbance. The bioptimized design suggested that designs based exclusively on economic or exergy
(i.e., green) has an improved closed-loop performance compared performances differ significantly in their structure, showing an
to the classical design. The deviations from xMTBE set point are existing trade-off. The Pareto curve provides sensitivity infor-
less pronounced in the green design, which smoothly returns mation on the cost effects of improving the exergy efficiency.
back to the initial steady state when the disturbance disappears. In addition to economics and exergy objectives, this optimiza-
The set point of the green design is slightly lower (0.05%) than tion-based approach covers responsiveness aspects. To define
the one of the classic design. This negligible variation in set a responsiveness criterion, the availability and entropy produc-
points does not affect the quality of the analysis results. It is tion rate were regarded as functions of an internal state vector
suggested from our results that incorporating both economic x and a vector of disturbances or manipulated external excita-
and exergy-related objectives in the unit design results in a tions u. In our approach, the maximization approach suggested
process with better closed-loop performance, provided that by Luyben et al.31 was not considered due to its inherent need
control structure and tuning are kept invariant.A well-tuned for high computational effort. However, a rather formal defini-
control can compensate for any change in the process operability tion of fresponse was adopted together with a single input
feature. excitation. By not searching over all possible state trajectories,
Ind. Eng. Chem. Res., Vol. 47, No. 1, 2008 63
the generality of the results might be questionable and the Xm,j ) conjugated driving forces for interfacial mass diffusion
external excitation is a fairly conservative one. By analyzing of component j
the responsiveness criteria of the bioptimized design, it is XHX ) conjugated driving forces for heat exchange with the
reconfirmed that small driving forces are desirable for reactive external world
sections and large driving forces are desirable for separation Xrx,j ) conjugated driving forces for the chemical reaction of
(nonreactive) sections.31,46 The first statement suggests that, in reaction j
the case of reactive sections, driving forces should be manipu- hf ) scaled objective function
lated to control the unit. The second statement suggests that in xj ) product specification
nonreactive sections flux should be manipulated to control the E ) economic-based optimal design
unit. Finally, the economic-driven design (classical design) was Q ) optimal design based equally on economic performance
compared to the new bioptimized design (green design) for the and exergy efficiency
case of the MTBE RD column. The green design produced less U ) Utopia point
entropy than the classical one, while being marginally less X ) exergy-based optimal design
attractive from an economic standpoint. The bioptimized design A Li ) affinity of chemical reaction i in the liquid phase
had an improved closed-loop performance compared to the D ) set of design or decision variables
classical design, keeping the same control structure and settings. R ) universal gas constant
It is concluded that incorporating both economic and exergy- A ) availability
related objectives in the unit design results in a process with B ) availability function
better closed-loop performance and reduced exergy loss. B(x,u)total ) availability over all the superstructure
In this contribution, we have focused on the performance of aΨ ) mass and heat transfer area
RD process designs, paying less explicit attention to the ai ) activity of component i
industrial conventional reaction-separation processing route. aij, b ij ) Wilson model parameters
Therefore, a comparison between RD and non-RD processes c ) investment cost of unit and catalyst; cost of component or
with regard to entropy production and the design implications utility per molar unit
for the principle of equipartition of forces are recommended dx(‚) ) set of disturbances
topics to be further researched. f ) objective function
f′ ) normalized objective function
Nomenclature fecon ) economic performance indicator
fexergy ) sustainability performance index
List of Symbols
fresponse ) responsiveness performance index
A ) area fx(‚) ) set of differential expressions
B ) controller’s bias gj,(R) ) external force exerted on component i per molar unit in
i
Dcol ) column diameter stream j and phase (R)
total ) total flow of stream j in the vapor phase; Ftotal
F j,V j,V )
gx(‚) ) set of algebraic expressions
i)nc
Σ i)1 Fij,V grx(‚) ) chemical reaction driving force based on activities
F j,(R)
i ) convective flow of component i in phase R in the hj,(R)
i ) molar enthalpy of component i in stream j and phase
incoming or outgoing stream (R)
i ) convective flow of component i in compartment k in the
F j,k hx(‚) ) set of inequalities and path constraints
incoming or outgoing ix(‚) ) set of initial conditions
stream j for the exchange with the external world m ) mass
e ) total energy flux in phase (R)
J (R) nc ) number of components
J Lq ) measurable heat flux into the liquid phase no ) number of objective functions
np ) number of phases
s ) entropy flux in phase (R)
J (R)
nt ) number of compartments in the integrated structure
(R)
J m,i ) mass flux of component i in phase (R)
nrx ) number of chemical reactions
K ) controller’s gain nst ) number of incoming and outgoing streams
Keq ) temperature-dependent chemical equilibrium constant p ) norm
M(R) ) molar holdup in phase (R) q(R′ - R) ) heat flux from phase (R′) to phase (R)
P ) absolute pressure ri(R) ) net change of the amount of component i in phase (R)
Q(R) ) external heat exchange rate in phase (R) due to all reactions
RR ) reflux ratio s(R) ) entropy density in phase (R)
SW ) variable in Wilson model t ) time
Se ) entropy produced by interfacial energy transfer u ) internal energy density of the volume element
S j,(R)
i ) molar entropy of component i in stream j and phase w ) vector of weighting factors
(R) x ji ) liquid molar fraction of component i in stream j
Sm ) entropy produced by interfacial mass diffusion y ji ) vapor molar fraction of component i in stream j
Srx ) entropy produced by chemical reaction R ) nominal design
T(R) ) absolute temperature in phase (R) U(R) ) internal energy of phase (R) or compartment
V ) volume of compartment
V LRX ) reaction volume in the liquid phase Acronyms
Wideal ) minimum work required by the process nC4 ) n-butane
Xq ) conjugated driving forces for heat exchange between iC4 ) isobutene
phases GLRDVE ) generic lumped reactive distillation volume element
Xe ) conjugated driving forces for interfacial heat diffusion MeOH ) methanol
64 Ind. Eng. Chem. Res., Vol. 47, No. 1, 2008
(29) Takeshita, K.; Ishida, M. Optimum Design of Multi-stage Isotope (53) Kooijman, H.; Taylor, R. The ChemSep Book; Library books on
Separation Process by Exergy Analysis. Energy 2006, 31, 3097. demand: Norderstedt, 2000.
(30) Iwakabe, K.; Nakaiwa, M.; Huang, K.; Nakanishi, T.; Rosjorde, (54) Colombo, F.; Cori, L.; Dalioro, L.; Delogu, P. Equilibrium Constant
A.; Ohmori, T.; Endo, A.; Yamamoto, T. Energy Saving in Multicomponent for the Methyl tert-Butyl Ether Liquid Phase Synthesis by Use of UNIFAC.
Separation Using an Internally Heat-integrated Distillation Column (HIDiC). Ind. Eng. Chem. Fundam. 1983, 22, 219.
Appl. Therm. Eng. 2006, 26, 1362. (55) Rehfinger, A.; Hoffmann, U. Kinetics of Methyl Tertiary Butyl
(31) Luyben, W. L.; Tyreus, B. D.; Luyben, M. L. Plantwide Process Ether Liquid Phase Synthesis Catalyzed by Ion Exchange ResinsI. Intrinsic
Control; McGraw-Hill: New York, 1999. Rate Expression in Liquid Phase Activities. Chem. Eng. Sci. 1990, 45, 1605.
(32) Kjelstrup, S.; Bedeaux, D. Elements of IrreVersible Thermodynam- (56) Zhang, T.; Datta, R. Integral Analysis of Methyl tert-Butyl Ether
ics; International Centre for Applied Thermodynamics: Istanbul, 2001. Synthesis Kinetics. Ind. Eng. Chem. Res. 1995, 34, 730.
(33) Marquardt, W. Modeling and Analysis of Chemical Process (57) Izquierdo, J.; Cunill, F.; Vila, M.; Tejero, J.; Iborra, M. Equilibrium
Systems. Presented at Delft University of Technology, Delft, The Nether- Constants for Methyl tert-Butyl Ether Liquid-phase Synthesis. J. Chem.
lands, July 2004. Eng. Data 1992, 37, 339.
(34) Bird, B.; Stewart, W.; Lightfoot, E. Transport Phenomena; John (58) Thiel, C.; Sundmacher, K.; Hoffmann, U. Residue Curve Maps for
Wiley and Sons, Inc.: New York, 2003. Heterogeneously Catalysed Reactive Distillation of Fuel Ethers MTBE and
(35) Seider, W.; Seader, J. D.; Lewin, D. Process Design Principles; TAME. Chem. Eng. Sci. 2002, 52, 993.
John Wiley and Sons, Inc.: New York, 1999. (59) Seader, J. D.; Henley, E. J. Separation Process Principles; John
(36) Bedeaux, D.; Kjelstrup, S. Irreversible Thermodynamics. A Tool Wiley and Sons, Inc.: New York, 1998.
to Describe Phase Transitions Far from Global Equilibrium. Chem. Eng. (60) Bansal, V.; Perkins, J.; Pistikopoulos, E.; Ross, R.; van Schijndel,
Sci. 2004, 59, 109. J. Simultaneous Design and Control Optimisation under Uncertainty.
(37) Koeijer, G. Energy Efficient Operation of Distillation Columns and Comput. Chem. Eng. 2000, 24, 261.
a Reactor Applying Irreversible Thermodynamics. Ph.D. Dissertation, (61) Process Systems Enterprise. gPROMS AdVanced User Guide;
Norwegian University of Science and Technology, Trondheim, Norway, Process Systems Enterprise: London, 1999.
2002. (62) Lim, Y. I.; Floquer, P.; Joulia, X.; Kim, S. D. Multiobjective
(38) Kjelstrup, S. Norwegian University of Science and Technology at Optimization in Terms of Economics and Potential Environment Impact
Trondheim, Trondheim, Norway. Personal communication, 2005. for Process Design and Analysis in a Chemical Process Simulator. Ind.
(39) Koeijer, G.; Kjelstrup, S.; vander Kooi, H. J.; Gros, B.; Knoche, Eng. Chem. Res. 1999, 38, 4729.
K. F.; Andersen, T. R. Positioning Heat Exchangers in Binary Tray
(63) Jimenez, E.; Salamon, P.; Rivero, R.; Rendon, C.; Hoffman, K.;
Distillation Using Isoforce Operation. Energy ConVers. Manage. 2002, 43,
Schaller, M.; Andresen, B. Optimization of a Diabatic Column with
1571.
Sequential Heat Exchangers. Ind. Eng. Chem. Res. 2004, 43, 7566.
(40) Johannessen, E.; Rosjorde, A. Equipartition of Entropy Production
(64) Larsson, M.; Wang, C.; Dahl, J. Development of a Method for
as an Approximation to the State of Minimum Entropy Production in
Analysing Energy, Environmental and Economic Efficiency for an Integrated
Diabatic Distillation. Energy 2007, 32, 467.
Steel Plant. Appl. Therm. Eng. 2006, 26, 1353.
(41) Ismail, S.; Proios, P.; Pistikopoulos, E. Modular Synthesis Frame-
work for Combined Separation/Reaction Systems. AIChE J. 2001, 47, 629. (65) Tsatsaronis, G. Definition and Nomenclature in Exergy Analysis
(42) Papalexandri, K.; Pistikopoulos, E. Generalized Modular Repre- and Exergoeconomics. Energy 2007, 32, 249.
sentation Framework for Process Synthesis. AIChE J. 1996, 42, 1010. (66) Almeida-Rivera, C. P.; Swinkels, P. L. J.; Grievink, J. Economics
(43) Bermingham, S. A Design Procedure and Predictive Models for and Exergy Efficiency in the Conceptual Design of Reactive Distillation
Solution Crystallisation Processes. Ph.D. Dissertation, Delft University of Processes. AIChE Symp. Ser. 2004, 2004, 237.
Technology, Delft, The Netherlands, 2003. (67) Huang, K.; Iwakabe, K.; Nakaiwa, M.; Tsutsumi, A. Towards
(44) Coulson, J. M.; Richardson, J. F. Chemical Engineering Design; Further Internal Heat Integration of Reactive Distillation Columns. Part I:
Chemical Engineering, Vol. 6; Butterworth-Heinemann: Oxford, Boston, The Design Principle. Chem. Eng. Sci. 2005, 60, 4901.
1997. (68) Nummedal, L.; Kjelstrup, S. Equipartition of Forces as a Lower
(45) Almeida-Rivera, C. P. Designing Reactive Distillation Processes Bound on the Entropy Production in Heat Exchange. Int. J. Heat Mass
with Improved Efficiency. Ph.D. Dissertation, Delft University of Technol- Transfer 2001, 44, 2827.
ogy, Delft, The Netherlands, 2005. (69) Sauar, E.; Siragusa, G.; Andresen, B. Equal Thermodynamic
(46) Meeuse, M. On the Design of Chemical Processes with Improved Distance and Equipartition of Forces Principles Applied to Binary Distil-
Controllability Characteristics. Ph.D. Dissertation, Delft University of lation. J. Phys. Chem. A 2001, 105, 2312.
Technology, Delft, The Netherlands, 2003. (70) Georgiadis, M.; Schenk, M.; Pistikopoulos, E.; Gani, R. The
(47) Clark, P.; Westerberg, A. Optimization for Design Problems Having Interactions of Design, Control and Operability in Reactive Distillation
More than One Objective. Comput. Chem. Eng. 1983, 7, 259. Systems. Comput. Chem. Eng. 2002, 26, 735.
(48) Jimenez, L.; Wanhschafft, O.; Julka, V. Analysis of Residue Curve (71) Wang, S. J.; Wong, D. S. H.; Lee, E. K. Effect of Interaction
Maps of Reactive and Extractive Distillation Units. Comput. Chem. Eng. Multiplicity on Control System Design for a MTBE Reactive Distillation
2001, 25, 635. Column. J. Proc. Control 2003, 13, 503.
(49) Peters, U.; Nierlich, F.; Schulte-Körne, E.; Sakuth, M. Methyl tert- (72) Sneesby, M. G.; Tade, M. O.; Smith, T. N. Steady-state Transitions
Butyl Ether. In Ullmann’s Encyclopedia of Industrial Chemistry; John Wiley in the Reactive Distillation of MTBE. Comput. Chem. Eng. 1998, 22, 879.
and Sons, Inc.: New York, 2000. (73) Hasse, R. Thermodynamics of IrreVersible Processes; Addison-
(50) Sundmacher, K.; Hoffmann, U. Multiple Reactions in Catalytic Wesley Publishing Co.: Darmstadt, Germany, 1969.
Distillation Processes for the Production of the Fuel Oxygenates MTBE (74) Ung, S.; Doherty, M. Vapor-Liquid Phase Equilibrium in Systems
and TAME: Analysis by Rigourous Model and Experimental Validation. with Multiple Chemical Reactors. Chem. Eng. Sci. 1995, 50, 23.
Chem. Eng. Sci. 1999, 54, 2839.
(51) Schrans, S. M.; Wolf, S.; Baur, R. Dynamic Simulation of Reactive ReceiVed for reView November 28, 2006
Distillation: An MTBE Case Study. Comput. Chem. Eng. 1996, 20, S1619. ReVised manuscript receiVed September 10, 2007
(52) Malanowski, S.; Anderko, A. Modelling Phase Equilibria. Ther- Accepted September 11, 2007
modynamic Background and Practical Tools; John Wiley and Sons, Inc.:
New York, 1992. IE061521J