REVISTA MEXICANA DE INGENIERIA QUIMICA Vol.

4 (2005) 201-212
AMIDIQ
AN ANALYTICAL STUDY OF THE LOGARITHMIC MEAN TEMPERATURE
DIFFERENCE

ESTUDIO ANALITICO DE LA MEDIA LOGARITMICA DE DIFERENCIA DE

TEMPERATURAS

A. Zavala-Río*, R. Femat and R. Santiesteban-Cos

Instituto Potosino de Investigación Científica y Tecnológica, Apdo. Postal 2-66, C. P. 78216, San Luis Potosí, S. L. P.,
México
Received May 19 2005; Accepted August 29 2005

Abstract

The logarithmic mean temperature difference (LMTD) has caused inconveniences in several applications like equation-
oriented flow sheeting programs. Such inconveniences have arisen from its indeterminate form. This is a consequence of
the incomplete model derivation generally developed in the textbooks. Heat exchanger dynamic analysis and control
synthesis through lumped-parameter models using the LMTD as driving force (fluid mean temperature difference) may
suffer from such inconsistencies too. This paper is devoted to give a solution to such inconveniences by providing a
formal mathematical treatment of the LMTD. First, a complete derivation is restated resulting in a complete well-defined
expression. Then, several interesting analytical properties of the resulting expression, like continuous differentiability on
its domain, are proved. The usefulness of the results is highlighted throughout the text.

Keywords: logarithmic mean temperature difference, heat exchangers.

Resumen

La media logarítmica de diferencia de temperaturas (LMTD, por sus siglas en inglés) ha causado inconveniencias en
diversas aplicaciones tales como ciertos programas de simulación de procesos. Su forma indeterminada es una de las
principales causas de tales inconveniencias. Tal indeterminación es una consecuencia del procedimiento incompleto que
generalmente se desarrolla para su obtención en los libros de texto. El análisis dinámico y el diseño de control de
intercambiadores de calor a través de modelos de parámetros agrupados que usan la LMTD como fuerza conductora (de
intercambio de calor, i.e. diferencia promedio de temperatura entre los fluidos) pueden también ser víctimas de tales
inconsistencias. Este trabajo está dedicado a dar una solución a tales inconveniencias a través de un análisis matemático
formal de la LMTD. Primero, un procedimiento completo para su obtención es desarrollado, dando como resultado una
expresión completa bien definida. Posteriormente, diversas propiedades analíticas de la expresión resultante, tales como
la continuidad y la diferenciabilidad en todo su dominio, son probadas. La utilidad de los resultados es comentada a lo
largo del texto.

Palabras clave: media logarítmica de diferencia de temperatura, intercambiadores de calor.

1. Introduction involved in their design (Reimann, 1986;

Incropera and DeWitt, 1990); performance
The logarithmic mean temperature
calculation (Incropera and DeWitt, 1990;
difference (LMTD)
Holman, 1997); steady-state analysis (Kern,
1950; Mathisen, 1994); dynamic modelling
ΔT2 − ΔT1
ΔTA = (1) (Reimann, 1986; Steiner, 1989; Steiner, 1987),
ΔT simulation (Papastratos, et al.; Zeghal et al.,
ln 2
ΔT1 1991), and characterization (Zavala-Río et al.,
2003, Zavala-Río and Santiesteban-Cos,
plays an important role in theoretical and
2004); (closed-loop) stability-limit analysis
practical aspects of heat exchangers. It is
(Khambanonda et al., 1991; Khambanonda et
*Corresponding author: E-mail: azavala@ipicyt.edu.mx
Phone: (44) 48342000, Fax: (44) 48342010

Publicado por la Academia Mexicana de Investigación y Docencia en Ingeniería Química, A. C. 201

 A. Zavala-Río et al. / Revista Mexicana de Ingeniería Química Vol. 4 (2005) 201-212

al., 1990); and control synthesis (Alsop and
Edgar, 1989; Malleswararao and Chidambaram,
1992). For the sake of simplicity, the use of less
involved expressions, like the arithmetic model
ΔT1 + ΔT2
ΔTa =
2 values such that ΔT1 = ΔT2 is undefined due to
or the geometric one
ΔTg = ΔT1ΔT2
is sometimes preferred to approximate the mean
temperature difference along the exchanger.
However, among these expressions, it is just the
LMTD that takes into account the exponential
distribution of the fluid temperatures over the
tube length in stationary conditions (Incropera
and DeWitt, 1990). Consequently, more
appropriate steady-state values are computed
using (1) than (2) or (3). Nevertheless, the
LMTD may cause inconvenience if it is taken
simply as shown in (1). Paterson (1984), for
instance, points out that the iterative equation-
solving procedures performed by equation-
oriented flow sheeting programs, commonly
take starting values that involve the equality of
stream temperatures, and hence zero
temperature differences where (1) reduces to an
indeterminate form; moreover, taking
ΔT1 = ΔT2 , the derivatives of ΔTA with respect to
ΔT1 and ΔT2, needed in the Newton iterative
solution of the equations, are undefined. On the
other hand, heat exchanger dynamic analysis
and control design may suffer from such
inconveniences too. In (Alsop and Edgar, 1989;
Malleswararao and Chidambaram, 1992), for
instance, the authors take the following 2nd-
order lumped-parameter model
dTco 2 ⎡ UA ⎤ (Paterson, 1984; Underwood, 1970; Chen,
= ⎢ Fc (Tci - Tco ) + ΔT ⎥
dt Mc ⎢⎣ C pc ⎥⎦
dTho 2 ⎡ UA ⎤ ΔTa + 2ΔTg
= ⎢ Fh (Thi - Tho ) + ΔT ⎥
dt Mh ⎢⎣ C ph ⎥⎦
and define ΔT as in (1) to design control
schemes for countercurrent heat exchangers.
Using Fc as control input, they propose an
algorithm to stabilize Tho at a desired value.
Nevertheless, under such a modelling, existence
of solutions going through (or starting at) state
(2)
the undeterminate form of (1). Moreover, the
proposed control schemes make use of the
(3) derivatives of ΔTA with respect to ΔT1 and ΔT2
which, as above mentioned, are undefined when
ΔT1 = ΔT2 (and it is not clear in such works how
to handle such an inconsistency). Thus, well
posedness of the LMTD turns out to be
important.
In order to deal with the above
mentioned inconveniences, some authors state
that ΔTA → ΔTa as ΔT2 - ΔT1 → 0 (see e.g.
(Kern, 1959; Gardner and Taborek, 1977)), or
simply suggest to replace ΔTA by ΔTa when
ΔT1 = ΔT2 (see e.g. (Faires, 1957)), and others
∂ΔTA 1
claim further that → , i = 1, 2, as
∂ΔTi 2
ΔT2 - ΔT1 → 0 (see e.g. (Paterson, 1984)).
However, a mathematical proof of such
assertions still seems to be lacking. Some others
suggest to replace (1) by an infinite power series
expansion that yields ΔTa when ΔT1 = ΔT2 (see
e.g. (Steiner, 1989)). Unfortunately, performing
calculations through infinite power series imply
the consideration of an infinite number of
arithmetical operations, while through a
truncated version of the series, accuracy is lost.
Other authors have proposed well-defined
replacement expressions that approximate ΔTA
over an acceptable range of ΔT1 and ΔT2
1987). Paterson (1984), for example, proposes
(4)
ΔTA P = ≈ ΔTA (5)
3

202
 A. Zavala-Río et al. / Revista Mexicana de Ingeniería Química Vol. 4 (2005) 201-212
while
ΔT11/ 3 + ΔT21/ 3
ΔTA1/U3 = ≈ ΔTA1/ 3 (6)
2
is proposed by Underwood (1970), and
ΔTA C1 = ΔTa1/ 3 ΔTg2 / 3 y ≈ ΔTA (7)
and
ΔT10.3275 + ΔT20.3275
ΔTA0.3275
C2 = ≈ ΔTA0.3725
2
(the latter being a refined modification of
Underwood’s approximation) by Chen (1987)
(a comparison study of these four
approximations is presented in (Chen, 1987)).
This work is devoted to provide a formal
mathematical treatment of the LMTD, the
results of which are intended to give a solution
of the above mentioned inconveniences. First,
we show that the indeterminate form of (1)
is a consequence of the incomplete derivation of
the LMTD generally presented in the textbooks
(see for instance (Incropera and DeWitt, 1990;
Holman, 1997; Kern, 1950; Faires, 1957;
McAdams, 1954; Walas, 1991)). By restating a
complete derivation, an expression is gotten
being equivalent to (1) when ΔT1 ≠ ΔT2 and
well-defined when ΔT1 = ΔT2 . Consequently,
approximating the LMTD through replacement
expressions may henceforth be avoided. Second,
analytical properties of the resulting complete
expression, such as continuity and
differentiability, are proved for every physically
reasonable combination of values of ΔT1 and
ΔT2, including those where ΔT1 = ΔT2 . Heat
Fig.1. Counterflow (left) and parallel flow (right) heat exchangers.
exchanger lumped-parameter dynamic models
(like (4)) using the LMTD, and control schemes
using its derivative with respect to ΔT1 and ΔT2,
will now be mathematically coherent through
the use of such a complete expression.
The work is organized as follows:
Section 2 states the notation adopted in the
present study. In Section 3, the complete
derivation of the LMTD is developed.
Analytical properties of the resulting model are
stated in Section 4. Finally, conclusions are
(8)
given in Section 5.
2. Nomenclature and notation
The following nomenclature is defined
for its use throughout this work:
F mass flow rate
Cp specific heat
M total mass inside the tube
U overall heat transfer coefficient
A heat transfer surface area
T temperature
ΔT temperature difference
Q rate of heat transfer
t time
R set of real numbers
R2 set of 2-tuples (xi)i=1,2 with xi∈R
R+ set of positive real numbers
R+2 set of 2-tuples (xi)i=1,2 with xi∈R
Subscripts:
c cold
h hot
i inlet
o outlet
203
 A. Zavala-Río et al. / Revista Mexicana de Ingeniería Química Vol. 4 (2005) 201-212

In particular, Qh, Qc, and Q represent 1950; Incropera and DeWitt, 1990).
the rates of heat convected through the hot and Furthermore, the following developments take
cold fluid tubes and that transferred from the into account both flow configurations of heat
hot to the cold fluid respectively. The exchangers simultaneously through an auxiliary
variables Th and Tc generically denote the parameter α. Any expression where α is not
temperature of the hot and cold fluids, present is valid for both flow configurations. In
respectively, at any point along the exchanger, those where α appears, such parameter
and ΔT = Th - Tc. Furthermore, let ΔT1 and determines the configuration they are valid for
ΔT2 represent the temperature difference at in the following way
each terminal side of the heat exchanger, i.e.
(see Fig. 1) ⎧1 if counterflow
α =⎨ if parallel flow
(11)
⎩−1
⎧T − Tco if counterflow
ΔT1 = ⎨ hi (9)
⎩ Thi − Tci if parallel flow At any point along the exchanger, the heat
transfer equations are
and
dQh = FhC ph dTh (12)
⎧ T − Tci if counterflow dQc = Fc C pc dTc (13)
ΔT2 = ⎨ ho (10)
⎩Tho − Tco if parallel flow dQh = α dQc = − dQ (14)
taken in this work as conventional definitions. dQ = U ΔTdA (15)
Notice that under such conventional definitions,
ΔT1 and ΔT2 shall be considered positive. To From the constant physical property assumption,
differentiate from ΔTA in (1), we denote ΔTL the and considering a mean temperature difference,
ΔTL, throughout the exchanger, we have, after
well-posed logarithmic mean temperature integration over the tube length,
difference to be derived in the following section.
ΔTa still stands (as in the previous section) for Qh = FhC ph (Tho − Thi ) (16)
the arithmetic model. Consider the sets B, C,
and D with B ⊂ C , and a function f : C → D . Qc = Fc C pc (Tci − Tco ) (17)
We denote f |B the restriction of f to B, i.e. Qh = α Qc = −Q (18)

f |B : B → D : x 6 f |B ( x ) = f ( x ), ∀ x ∈ B . Q = UAΔTL (19)

The boundary of a subset, say B, is represented

as ∂B . The next developments follow

3. Complete derivation of the LMTD d ΔT = dTh − dTc

We recall that the present derivation dQh dQc
= − (from (12) and (13))
assumes that Cpc, Cph, and U, are considered Fh C ph Fc C pc
flow and temperature invariant, that is, their ⎛ 1 1 ⎞
value is considered to be constant throughout = dQ ⎜ − ⎟⎟ (from (14))
⎜αF C
the exchanger; see for instance (Incropera and ⎝ c pc FhC ph ⎠
DeWitt, 1990; Holman, 1997; Kern, 1950;
Faires, 1957; McAdams, 1954; Wallas, 1991); and, from (15), we get
for a longer (plus exhaustive) list of the
assumptions involved, see for example (Kern,

204
 A. Zavala-Río et al. / Revista Mexicana de Ingeniería Química Vol. 4 (2005) 201-212

⎛ 1 1 ⎞ Tho − Thi α (Tci − Tco )

d ΔT = U ΔT ⎜ − ⎟⎟ dA (20) = − (from (18))
⎜αF C F C Q Q
⎝ c pc h ph ⎠
ΔT − ΔT1
= 2 (from (9), (10), and (11))
At this point, we consider two possible Q
situations:
which can be substituted into (22) to get
The general case. We begin by assuming that
ΔT2 ΔT − ΔT1
ln = UA 2
⎛ 1 1 ⎞ (21) ΔT1 Q
ΔT ⎜ − ⎟⎟ ≠ 0
⎜αF C F C
⎝ c pc h ph ⎠

Let us note that this is the situation that is most From this expression, we get
often found in physical heat exchangers1. Then,
Δ T2 − Δ T1
from (20), we have Q = UA
ΔT
ln 2
⎛ 1 ⎞ Δ T1
d ΔT 1
=U ⎜ − ⎟⎟ dA
ΔT ⎜αF C
⎝ c pc FhC ph ⎠ which, compared to (19) and taking into account
(23), shows that
yielding ΔT2 − ΔT1
ΔT L = = ΔTA if ΔT1 ≠ ΔT2 (24)
ΔT
ln 2
ΔT2 ⎛ 1 1 ⎞ (22)
ΔT1
ln = UA ⎜ − ⎟⎟
ΔT1 ⎜αF C
⎝ c pc FhC ph ⎠ The special case. Let us now suppose that
after integration over the tube length. Notice that ⎛ ⎞
1 1
since the right-hand side of (22) is a non-zero ΔT ⎜ − ⎟⎟ = 0 (25)
⎜αF C F C
scalar, ΔT1 and ΔT2 (in the left-hand side) must ⎝ c pc h ph ⎠
be such that ln(ΔT2 / ΔT1 ) be a non-zero scalar
Although this situation is seldom found in
too, which is satisfied if and only if ΔT1 ≠ ΔT2 .
practice2, it is considered here since its
Then, one sees from these developments that development is needed for the well-posedness of
the LMTD. From (20), we have
(21) ⇔ ΔT1 ≠ ΔT2 (23)
d ΔT = 0 (26)
Now, notice that
1 1 T −T T −T implying that the temperature difference
− = ci co − ho hi
αFC FC Qc Qh ΔT = Th − Tc has a constant value throughout the
c pc h ph

(from (16) and (17))
1
This may explain why it is in fact the case that is
generally developed in the textbooks (for a specific
flow configuration), implicitly taking (21) as a fact
(which is actually the origin of the inconsistency
problem of (1)).
exchanger. Hence, after integration of (26) and
(15) over the tube length, we have
ΔT2 = ΔT1 = ΔT (equivalent to (25)) and
2
This may explain why this case is apparently never
taken into account; it thus constitutes the part of the
derivation that is generally lacking in the textbooks.

205
 A. Zavala-Río et al. / Revista Mexicana de Ingeniería Química Vol. 4 (2005) 201-212

Q = UAΔT . From these expressions, comparing 1) ΔT1 ≠ ΔT2 . From Formula 4.1.27 in3
the latter to (19), we have (Abramowitz and Stegun, 1972), we have

ΔTL = ΔT if ΔT2 = ΔT1 = ΔT (27) 2i +1

ΔT ∞
1 ⎛ ΔT2 − ΔT1 ⎞
ln 2 = 2∑ ⎜ ⎟
ΔT1 i =0 2i + 1 ⎝ ΔT2 + ΔT1 ⎠
The well-posed LMTD. Finally, from the results
gotten in both, the general and special cases, i.e.
from (24) and (27), we get ∀( ΔT1 , ΔT2 ) ∈ R+2 .
Then, for all ( ΔT1 , ΔT2 ) ∈ R+2
such that ΔT1 ≠ ΔT2 , we get
⎧ ΔT2 − ΔT1
⎪⎪ ΔT2 if ΔT1 ≠ ΔT2 ΔT2 −ΔT1 ΔT2 −ΔT1
ΔTL = ⎨ ln (28) =
ΔT1 ΔT ∞
1 ⎛ ΔT2 −ΔT1 ⎞
2i+1
⎪ ln 2 2∑
⎪Δ ΔT1 ⎜ ⎟
⎩ T if ΔT2 = ΔT1 = ΔT i=0 2i +1⎝ ΔT2 +ΔT1 ⎠

4. Analytical properties ΔT2 −ΔT1

=
⎛ ΔT2 −ΔT1 ⎞ ⎡ ∞ 1 ⎛ ΔT2 −ΔT1 ⎞ ⎤
2i
In this section we consider the mean
temperature difference a bivariable function, 2⎜ ⎟ ⎢1+ ∑ ⎜ ⎟ ⎥
⎝ ΔT2 +ΔT1 ⎠ ⎢⎣ i=1 2i +1⎝ ΔT2 +ΔT1 ⎠ ⎥⎦
whether we refer to the logarithmic model in
(28), ΔTL(ΔT1, ΔT2), or the arithmetic one in (2),
ΔTa(ΔT1,ΔT2). Furthermore, just the cases where ΔT1 + ΔT2
heat is transferred between fluids are considered = 2
of practical interest in the present section. ∞
1 ⎛ ΔT2 − ΔT1 ⎞
2i

Hence, for analysis purposes, we choose to 1+ ∑ ⎜ ⎟

i =1 2i + 1 ⎝ ΔT2 + ΔT1 ⎠
disregard the no-heat-transfer trivial case
ΔT2 = ΔT1 = 0 . Under such perspective, R+2 is
Δ Ta ( Δ T1 , Δ T2 ) (from (2) and (29))
considered to constitute the domain of ΔTL. We =
begin by stating a useful equivalent expression. L ( Δ T1 , Δ T2 )

Lemma 1. Let 2) ΔT1 = ΔT2 . Notice from (29) and (2) that
2i
∞
1 ⎛ ΔT2 − ΔT1 ⎞ L(ΔT , ΔT ) = 1 and ΔTa (ΔT , ΔT ) = ΔT ,
L(ΔT1 , ΔT2 )  1 + ∑ ⎜ ⎟ (29)
i =1 2i + 1 ⎝ ΔT2 + ΔT1 ⎠ ∀ΔT ≠ 0 . Then, for ΔT1 = ΔT2 = ΔT > 0 , we
ΔTa (ΔT , ΔT )
have = ΔT = ΔTL (ΔT , ΔT ) .
for all (∆T1, ∆T2 ) such that ∆T1 + ∆T2 ≠ 0. L(ΔT , ΔT )
Then
ΔT (ΔT1 , ΔT2 ) (30) Remark 1. Lemma 1 is of great help in the
ΔTL (ΔT1 , ΔT2 ) ≡ a
L(ΔT1 , ΔT2 ) analysis of ΔTL. Indeed, since L(ΔT , ΔT ) ≥ 1

∀( ΔT1 , ΔT2 ) ∈ R+2 .

3
Formula 4.1.27 in (Abramowitz and Stegun, 1972)
states the following well-known (infinite) series
Proof. We divide the proof in two parts: expansion of the logarithmic function:
∞ 2i + 1
1 ⎛ z − 1⎞
ln z = 2 ∑
i=0
⎜ ⎟
2i + 1 ⎝ z + 1 ⎠
, ∀z : ℜz ≥ 0, z ≠ 0.

206
 A. Zavala-Río et al. / Revista Mexicana de Ingeniería Química Vol. 4 (2005) 201-212

for all (ΔT1 , ΔT2 ) such that ΔT1 + ΔT2 ≠ 0 (see may be approximated neglecting the third and
ΔTa (ΔT1 , ΔT2 ) upper terms of the series, i.e.
(29)), one sees readily that is ΔTa ( ΔT1 , ΔT2 ) 1 ( ΔT1 − ΔT2 ) 2
L(ΔT1 , ΔT2 ) ≈ ΔTa ( ΔT1 , ΔT2 ) − . For
L ( ΔT1 , ΔT2 ) 6 ΔT1 + ΔT2
well-defined at every point on R+2
. Actually, it is
instance, ΔT1 and ΔT2 resulting in a sufficiently
easy to see that small value of ( Δ T1 − Δ T 2 ) 2 , say
(ΔT1, ΔT2 ) →(ΔT, ΔT) ⇒L(ΔT1, ΔT2 ) →L(ΔT, ΔT) =1,
∀ΔT ≠ 0 . Consequently, one sees from (30) that ( ΔT1 − ΔT2 ) 2 ≈ 0 (observe that
ΔT1 − ΔT2 → 0 ⇒ ΔTL (ΔT1 , ΔT2 ) → ΔTa (ΔT1 , ΔT2 ) ( ΔT1 − ΔT2 )2 ≈ 0 ⇔ ΔT1 ≈ ΔT2 ), may be
explaining why such an assertion is suggested considered to apply for such an approximation.
by several authors, like in (Kern, 1950; Gardner Thus:
and Taborek, 1977; Faires, 1957). Moreover, ( ΔT1 − ΔT2 )2 ≈ 0 ⇔ 2 ΔT1ΔT2 ≈ ΔT1 + ΔT2 ⇔
from (30), we have
Δ T1 + 2 Δ T1Δ T2 + Δ T2 ≈ 2( Δ T1 + Δ T2 ) ⇔
L ∂L
− ΔTa ( Δ T1 + Δ T2 ) 2
∂ΔTL 2 ∂ΔTi (31)
( Δ T1 + Δ T2 ) 2 ≈ 2( Δ T1 + Δ T2 ) ⇔
= 2( Δ T1 + Δ T2 )
∂ΔTi L2
( Δ T1 + Δ T2 )2 ( Δ T1 + Δ T2 )2
≈ 1 ⇔
3 2 ( Δ T1 + Δ T 2 )
i=1,2, where the arguments have been dropped ( Δ T1 + Δ T2 )2 1 ( Δ T1 + Δ T 2 ) 2
− ≈ − ⇔ − ≈
for the sake of simplicity, and from (29), one 3 6 Δ T1 + Δ T 2
∂ΔTL ∞
2i 2i −1 ∂S 2 ⎛ Δ T1 + Δ T 2 ⎞ Δ T1 + Δ T 2
sees that =∑ S , with −
3⎝
⎜
2
− Δ T1 Δ T 2 ⎟ ⇔
⎠ 2
−
∂ΔTi i =1 2i + 1 ∂ΔTi
1 ( Δ T1 + Δ T 2 ) 2 Δ Ta + 2 Δ Tg , i.e.
≈
ΔT − ΔT1 ∂S (−1) 2ΔT3−i i
6 Δ T1 + Δ T 2 3
S= 2 and = .
ΔT2 + ΔT1 ∂ΔTi (ΔT2 + ΔT1 ) 2 ΔTa ( ΔT1 , ΔT2 )
≈ ΔTAP , (see (5)). Alternatively, for
Hence, L( ΔT1 , ΔT2 )
⎛ ∂L ⎞ ∂ΔTL 1 ΔT1 and ΔT2 resulting in
ΔT1 − ΔT2 → 0 ⇒ ⎜
⎝ ∂ΔTi
, L ⎟ → (0,1) ⇒
⎠ ∂ΔTi
→ ,
2 (
2/3 1/ 3 1/ 2
a1 = 5 Δ T1 + 11Δ T1 Δ T 2 + 5 Δ T 2 2/3
× )
i=1,2, as claimed by authors like in [16].
Furthermore, synthetically dividing
( ΔT 1
1/ 3
− ΔT 2
1/ 3 2
) ≈0 (observe that a1 ≈ 0 ⇒

( ΔT −ΔT21/3 ) ≈ 0 ⇔ ΔT1 ≈ ΔT2 ),

1/3 2
ΔTa(ΔT1,ΔT2) by L(ΔT1,ΔT2), one gets 1 we have
( Δ T1 − Δ T2 )
2
Δ T1 + Δ T 2 1
ΔTa(ΔT1,ΔT2) 1 (ΔT1 −ΔT2)2 2 (ΔT1 −ΔT2)4 a1 ≈ 0 ⇔ −
=ΔTa(ΔT1,ΔT2) − − −... (32) 2 6 Δ T1 + Δ T 2
L(ΔT1,ΔT2) 6 ΔT1 +ΔT2 45(ΔT1 +ΔT2)3 3
⎛ Δ T 1/ 3 + Δ T21/ 3 ⎞
≈⎜ 1 ⎟ (see Appendix A), i.e.
coinciding with the power series expansion ⎝ 2 ⎠
suggested in (Steiner, 1989) as a replacement ΔTa ( ΔT1 , ΔT2 )
≈ ΔTAU (see (6)). Furthermore, for
for the LMTD. The proofs of some other L( ΔT1 , ΔT2 )
analytical features of ΔTL stated in the present ΔT1 and ΔT2 such that
section are simplified through the use of (30). ⎡ ⎛ ΔT + ΔT ⎞ 2 /3 ⎤ ⎡ ⎛ Δ T1 + Δ T 2 ⎞
2/3
a2 = ⎢2⎜ 1 2
⎟ + (Δ T 1 Δ T 2 )1 / 3 ⎥ ⋅ ⎢2⎜ ⎟
⎢⎣ ⎝ 2 ⎠ ⎥⎦ ⎣ ⎝ 2 ⎠
Remark 2. Expression (32) is helpful to see −2(ΔT ΔT ) ⎤⎦ ≈ 0 1/ 3
(observe that
1 2
the link of the replacement formulas (5)–(8)
with the LMTD. Indeed, notice that for close ⎡ ⎛ Δ T1 + Δ T2 ⎞ 2/3
⎤
a2 ≈ 0 ⇒ ⎢ 2 ⎜ ⎟ − 2( Δ T1 Δ T2 )1 / 3 ⎥ ≈ 0 ⇔
enough values of ΔT1 and ΔT2, the LMTD ⎢⎣ ⎝ 2 ⎠ ⎥⎦

207
 A. Zavala-Río et al. / Revista Mexicana de Ingeniería Química Vol. 4 (2005) 201-212

⎛ Δ T1 + Δ T 2 ⎞
2/3
⎛ Δ T1 + Δ T 2 ⎞
2
1976). Therefore Δ T a ( Δ T1 , Δ T 2 ) as
⎜ ⎟ ≈ ( Δ T1 Δ T 2 ) 1 / 3 ⇔ ⎜ ⎟ → 0
⎝ 2 ⎠ ⎝ 2 ⎠ L ( Δ T1 , Δ T 2 )
≈ ΔT1ΔT2 ⇔ ΔT12 + 2ΔT1ΔT2 + ΔT22 ≈ 4ΔT1ΔT2 ( ΔT1 , ΔT2 ) → ∂R +2
which, from Lemma 1, implies
⇔ (ΔT1 − ΔT2 ) 2 ≈ 0 ⇔ ΔT1 ≈ ΔT2 ), we have that ΔTL (ΔT1 , ΔT2 ) → 0 as ( ΔT1 , ΔT2 )

ΔT1 + ΔT2 1 ( ΔT1 − ΔT2 )

2 approaches ∂R+2 (from the interior of R+2 ). Then,
a2 ≈ 0 ⇔ − ≈ ΔTa1/ 3ΔTg2 / 3 zero can be considered the value that the LMTD
2 6 ΔT1 + ΔT2
(as a bivariable function) would take at any
ΔTa ( ΔT1 , ΔT2 )
(see Appendix B), i.e. ≈ ΔTAC1 (see point on ∂R+2 . This has been useful in the
L( ΔT1 , ΔT2 )
analyses developed in (Zavala-Río et al., 2003;
(7)). From these developments we see that as
Zavala-Río, 2004) to prove that the solutions of
long as close enough values of ΔT1 and ΔT2 are
system (4) remain within a physically coherent
considered, relatively good approximations of
(according to the assumptions made therein)
the LMTD are gotten through the replacement
domain where the outlet temperatures cannot
expressions (5)–(8) (recall that (8) is just a
become either higher than Thi or lower than Tci
refinement of (6)). On the contrary, such
(see Fig. 1).
approximations deteriorate as ΔT1 and ΔT2 are
taken far from each other.
Lemma 2. The LMTD model in (28) is
Remark 3. Let us point out that the quotient continuously differentiable and positive on R+2 .
function Δ Ta
(Δ T , Δ T ) =
Δ T (Δ T , Δ T ) is a 1 2

L
1 2
L ( Δ T1 , Δ T 2 ) Proof. Since L( ΔT1 , ΔT2 ) ≥ 1, ∀( ΔT1 , ΔT2 ) ∈ R+2
defined on a subset wider than the domain of (see (29)), one sees from (30) (and (2)) that
ΔTL. Actually, ΔTL exists and is continuous on R+2 (also
: {( Δ T1 , Δ T 2 ) ∈ R 2 Δ T1 + Δ T 2 ≠ 0 }→ R .
ΔTa
Then verifiable from (32), according to Lemma 1).
L
ΔTa ∂ΔTL
is an extension of ΔTL (actually Lemma 1 Moreover, from (31), one observes that ,
L ∂ΔTi
Δ Ta i = 1,2, exist and are continuous on R+2 ,
can be synthesized as ΔT L ≡ ). Therefore,
L R+2 proving continuous differentiability. On the
ΔTa other hand, notice (from (2)) that
may be used to extrapolate ΔTL to points ΔTa ( ΔT1 , ΔT2 ) > 0, ∀( ΔT1 , ΔT2 ) ∈ R +2 (the average
L
on R2 where the latter is not defined. This may of two positive numbers is positive).
not make physical sense but could be helpful for Consequently,
analysis purposes. For example, one sees from ΔTa ( ΔT1 , ΔT2 )
0< ≤ ΔTa ( ΔT1 , ΔT2 ), ∀( ΔT1 , ΔT2 ) ∈ R+2 .
L( ΔT1 , ΔT2 )
(29) that L ∞
1
∞
1 , and since
= ∂ R +2
= ∑j=0
2 j+1 ∑ i =1
2i − 1 Then, from Lemma 1, positivity of ΔTL
1 1 ∞
1 follows too.
> , ∀ i ≥ 1 , then ∑ is divergent
2i − 1 2i i =1 2i − 1
according to theorems 3.28 and 3.25 in4 (Rudin, Lemma 2 is specially interesting when
ΔT1 = ΔT2 . It proves to be essential when the
4
In (Rudin, 1976), Theorem 3.28 states that LMTD is involved in dynamical analysis and
Σ ∞n =01/ n p converges if p > 1 and diverges if p ≤ 1, control synthesis of heat exchangers. In (Zavala-
while point (b) of Theorem 3.25 states that if Río and Santiesteban-Cos, 2004), for example, it
an ≥ d n ≥ 0 for n ≥ N 0 (for some N 0 ), and if is shown that thanks to such continuous
differentiability property, existence and
Σ ∞n =0 d n diverges, then Σ ∞n =0 an diverges.

208
 A. Zavala-Río et al. / Revista Mexicana de Ingeniería Química Vol. 4 (2005) 201-212

uniqueness of solutions of heat exchanger such ( ΔT1 , ΔT2 ) that

lumped-parameter dynamical models as in (4), ΔT2 − ΔT1 ΔT ΔT2 − ΔT1
< ln 2 < ⇔
using the well-posed LMTD as driving force, ΔT2 ΔT1 ΔT1
are guaranteed. Furthermore, as mentioned in ⎡ ΔT ΔT − ΔT1 ⎤
Section 1, the partial derivatives of the ( −1) i ⎢ln 2 − 2 ⎥>0, i=1,2, proving
⎣ ΔT1 ΔTi ⎦
LMTD (with respect to its arguments) are
used in the control laws synthesized in (Alsop the lemma.
and Edgar, 1989; Malleswararao and
Chidambaram, 1992). Provided that the well- Lemma 3 has been helpful in (Zavala-
posed LMTD in (28) is used, their closed-loop Río and Santiesteban-Cos, 2004) to prove the
analyses are correct (according to the respective existence of a unique equilibrium solution of
assumptions made in each of those works). On system (4) to which every trajectory converges.
the contrary, such control strategies could cause Furthermore, it has been helpful in (Zavala-Río
inconvenience if ΔTL were not continuously et al., 2003) to characterize heat exchangers as
cooperative systems (under the assumptions
differentiable on R+2 , or ΔTA in (1) were used
made therein). This means that a temperature
instead. This is true for any kind of (numerical) raise at any of the outlets entails a temperature
algorithm using the partial derivatives of the increase at the other outlet. The dynamics of
LMTD, like the equation-oriented flowsheeting heat exchangers may then be analyzed under
program application mentioned by Paterson the framework of cooperative systems, which
(1984) (see Section 1 above). constitutes a complementary way to
comprehend their behaviour.
Lemma 3. The LMTD model in (28) is
strictly increasing in its arguments, i e. Discussion and conclusions
∂ΔTL
> 0 , i=1,2, ∀( ΔT1 , ΔT2 ) ∈ R+2 . Some insights on the usefulness of the
∂ΔTi
results presented above are in order. For
instance, a well-known calculation problem is
Proof. From (28) (for ΔT2 ≠ ΔT1) and (31) (for that of finding a suitable ΔT2 satisfying an
ΔT2 = ΔT1; recall Remark 1), we have LMTD relation to a given ΔT1 (Paterson, 1984).
An exact solution to such a problem may be
⎧ i ⎡ ΔT2 ΔT2 −ΔT1 ⎤ gotten considering the complete LMTD
⎪(−1) ⎢ln −
⎪ ⎣ ΔT1 ΔTi ⎥⎦ expression in (28). Indeed, departing from (16)-
if ΔT1 ≠ ΔT2 (33) ⎛ α 1 ⎞
∂ΔTL ⎪⎪ ⎡ ΔT2 ⎤
2
UA⎜ − ⎟
⎜ Fc C pc Fh C ph ⎟
=⎨ ln (19), (28), one gets ΔT2 = e ΔT1 ,
⎝ ⎠
∂ΔTi ⎪ ⎢ ΔT ⎥
⎣ 1⎦
⎪ which is a simple linear expression. Numerical
⎪1 if ΔT2 = ΔT1 algorithms (whose convergence is not even
⎪⎩2 guaranteed) or nonlinear complex
i=1,2, existing and being continuous on R+2 approximation expressions (Paterson, 1984;
according to Lemma 2. Notice from (33) that Chen, 1987), are not any longer needed to
the proof of the lemma amounts to perform such a calculation. Furthermore, the
demonstrate positivity of results developed in Sections 3 and 4 have been
⎡ ΔT ΔT −ΔT1 ⎤ very helpful to characterize dynamic properties
(−1)i ⎢ln 2 − 2 ⎥ , i=1,2, for all of heat exchangers through simple lumped-
⎣ ΔT1 ΔTi ⎦ parameter dynamic models (Zavala-Río and
such that ΔT1 ≠ ΔT2 . Then,
∀( ΔT1 , ΔT2 ) ∈ R+2 Santiesteban-Cos, 2004; Zavala-Río et al.,
from Formula 4.1.33 in [23], we have, for all 2003). In (Zavala-Río and Santiesteban-Cos,

209
 A. Zavala-Río et al. / Revista Mexicana de Ingeniería Química Vol. 4 (2005) 201-212

2004), for instance, dynamic properties of avoiding such a problem. Other numerical
model (4), using (28), have been analyzed and problems arising from the choice of the system
proved to be equivalent to those of the parameter values may still take place if
distributed-parameter model where (4) comes simulations are performed through a simple
from. The complete expression of (28) and its model as (4), like numerical rigidity, for
analytical features have played a key role for instance. Nevertheless, such problems may be
such results in three main directions: 1) avoided if more than one bi-compartmental cell
existence and uniqueness of solutions of (4) is considered in the lumped-parameter
have been demonstrated for any physically modelling, as suggested in (Papastratos, et al.;
possible initial state conditions (such is not a Zeghal et al., 1991; Reimann, 1986). Finally,
property of (4) if (1) is used); 2) (4) was proved the results proposed in this work find potential
to have a unique equilibrium point, To* = applications in control design of heat
(Tco*,Tho*)T, linearly related to the inlet exchangers too. For instance, based on (4),
temperatures, Ti = (Tci,Thi)T, i.e. To* = ATi, where taking Fc as input and Tho as output, input-
the complete expressions of the elements of output (partial) linearization control algorithms
A∈R2×2 have been obtained in terms of the have been proposed in (Alsop and Edgar, 1989;
system properties, whatever value these ones Malleswararao and Chidambaram, 1992). In a
take; 3) such a unique equilibrium point was natural way, such design methods lead the
proved to be exponentially stable, and to keep proposed control laws to involve the partial
its asymptotical stability character globally on derivatives of ΔT1 and ΔT2. The authors use (1)
the system state-space domain. On the other but do not treat its inconsistency either. Again,
hand, based on the results in Section 3, other numerical results avoiding such inconsistencies
interesting dynamic properties of system (4) are presented. The results developed here give
have been proved in (Zavala-Río et al., 2003). sense to their results if (28), and (33), are
Basically, under certain standard assumptions, considered in their analysis and their proposed
model (4) has been characterized as a positive, algorithms.
compartmental, cooperative system. These Thus, an analytical study of the LMTD
features provide a new framework for the turns out to be essential in view of the important
comprehension of the dynamic behavior of heat role that the LMTD plays in several theoretical
exchangers. Furthermore, the use of the LMTD and practical aspects of heat exchangers. The
has found an important application in dynamic results developed here are intended to support
simulation too (Papastratos et al.; Zeghal et al., all those works where the LMTD is involved
1991; Reimann, 1986). However, none of the and the consideration of its analytical properties
previous works on the subject treat the is important, like dynamical modelling,
inconsistency of (1). Results avoiding such an simulation, characterization, and control of heat
inconsistency are generally shown. The exchangers.
complete expression in (28) may now be used

A. Developments for Underwood’s approximation

( 5ΔT + 11ΔT11/3ΔT21/ 3 + 5ΔT22/3 )( ΔT11/3 − ΔT21/3 ) ≈ 0 ⇔ [5(ΔT12 / 3 + ΔT11/ 3ΔT21/ 3 + 5ΔT22 / 3 ) + 6ΔT11/ 3ΔT21/ 3 ]
2/ 3 2
1

( Δ T 1
1 / 3
− Δ T 2
1 / 3
) 2
≈ 0 ⇔ [5 ( Δ T 1 − Δ T 2 ) + 6 ( Δ T 1
2 / 3
Δ T 2
1 / 3

− Δ T 11 / 3 Δ T 2
2 / 3
) ] ( Δ T 11 / 3 − Δ T 2
1 / 3
) ≈ 0 ⇔ 5 Δ T14 / 3
+ Δ T1Δ T 2
1 / 3

− 1 2 Δ T12 / 3
Δ T 2
2 / 3
+ Δ T 11 / 3
Δ T 2 + 5 Δ T 2
4 / 3
≈ 0 ⇔ 4 Δ T14 / 3
+
4 Δ T 1
2 / 3
Δ T 2
2 / 3
+ 4 Δ T 2
4 / 3
+ 8 Δ T 1 Δ T 2
1 / 3
+ 8 Δ T 1
2 / 3
Δ T 2
2 / 3

210
 A. Zavala-Río et al. / Revista Mexicana de Ingeniería Química Vol. 4 (2005) 201-212

+ 8 Δ T 11 / 3 Δ T 2 ≈ 9 Δ T14 / 3
+ 9 Δ T1Δ T 2
1 / 3
+ 9 Δ T 11 / 3 Δ T 2 + 9 Δ T 2
4 / 3

⇔ 4 ( Δ T12 / 3
+ Δ T 11 / 3 Δ T 2
1 / 3
+ Δ T 2
2 / 3
) 2
≈ 9 (Δ T1 + Δ T 2 ) ( Δ T 11 / 3
4 ( Δ T12 / 3
+ Δ T 11 / 3
Δ T 1 / 3
+ Δ T 2 / 3
) 2

+ Δ T 2
1 / 3
) ⇔ 2 2
≈ Δ T 11 / 3
+ Δ T 2
1 / 3

9 ( Δ T1 + Δ T 2
)

⇔ −
3
(Δ T 1 / 3
+ Δ T 1 / 3
) 2
⋅
4 Δ T ( 1
2 / 3
+ Δ T 11 / 3
Δ T 21 / 3 + Δ T 2
2 / 3
) 2

1 2
8 9 (Δ T 1 + Δ T 2 )
3 1 ( Δ T1 − Δ T 2 ) 2

≈ − ( Δ T 11 / 3 − Δ T 2
1 / 3
) 2
( Δ T 11 / 3 + Δ T 2
1 / 3
) ⇔ −
8 6 Δ T1 + Δ T 2
ΔT11/ 3 + ΔT21/ 3 ⎡ 1 Δ T11 / 3 + Δ T21 / 3 ⎡ 1
≈ ⎢⎣ 4 (−3ΔT1 + 6ΔT1 ΔT2 −3ΔT2 ) ⎤⎦ = + 2 Δ T11 / 3 Δ T21 / 3 + Δ T22 / 3 )
2/3 1/ 3 1/ 3 2/3
⎢⎣ 4 ( Δ T1
2/3

2 2

]=
1 / 3 1 / 3
2 / 3 1 / 3 1 / 3 2 / 3 Δ T 1 + Δ T 2
− Δ T 1 + Δ T 1 Δ T 2 − Δ T 2 ×
2
3
⎡⎛ 2 ⎤ ⎛ Δ T 1 1 / 3 + Δ T 21 / 3 ⎞
⎞
( ) Δ T1 + Δ T 2
1/ 3 1/ 3
⎢ ⎜ Δ T1 + Δ T 2 ⎟ − Δ T12 / 3 − Δ T11 / 3 Δ T 21 / 3 + Δ T 22 / 3 ⎥ = ⎜ ⎟ − ⇔
⎢ ⎜⎝ 2 ⎟
⎠ ⎥ ⎝ 2 ⎠ 2
⎣ ⎦
(Δ )
2 3
Δ T 1 + Δ T 2
−
1 T 1 − Δ T 2
≈
⎛ Δ T 1
1 / 3
+ Δ T 2
1 / 3
⎞ .
⎜ ⎟
2 6 Δ T 1 + Δ T 2 ⎝ 2 ⎠

B. Developments for Chen’s approximation

⎡ ⎛ ΔT1 + ΔT2 ⎞2/ 3 ⎤ ⎡ ⎛ ΔT1 + ΔT2 ⎞2/ 3 ⎡ ⎛ Δ T1 + Δ T2 ⎞ 2 / 3

⎟ − 2( Δ T1 Δ T2 ) ⎦⎤ ≈ 0 ⇔ ⎟ −
1/ 3
2
⎢ ⎜ ⎟ + ( ΔT1ΔT2 )1/ 3
⎥ ⎢2 ⎜ ⎢2 ⎜
⎢⎣ ⎝ 2 ⎠ ⎥⎦ ⎢⎣ ⎝ 2 ⎠ ⎢⎣ ⎝ 2 ⎠
(ΔT ΔT )1/3 ⎤ 9 ⎛ ΔT + ΔT2 ⎞
4/3
( Δ T1 + Δ T2 ) 2 / 3
− 1 2 ⎥ − (ΔT1ΔT2 )2/3 ≈ 0 ⇔ 4⎜ 1 ⎟ − 2( Δ T1Δ T2 )1/ 3 − 2( Δ T1Δ T2 ) 2 / 3
2 ⎦ 4 ⎝ 2 ⎠ 2
4/3
⎡⎛ ΔT + ΔT2 ⎞4/3 1/3 (ΔT1 + ΔT2 )
2/3
⎛ Δ T1 + Δ T 2 ⎞
≈ 0 ⇔ 2 ⎢⎜ 1 ⎟ + (ΔT1ΔT2 ) + ( Δ T 1 Δ T 2 ) 2/3
⎤
⎦ ≈ 6 ⎜ ⎟ =
⎢⎣⎝ 2 ⎠ 2 ⎝ 2 ⎠

⎛ ΔT + ΔT2 ⎞
1/ 3
1 (ΔT1 + ΔT2 )
2 ⎡ ⎛ ΔT + ΔT2 ⎞ 4/3
( ΔT1 + ΔT2 ) 2 / 3
3( ΔT1 + ΔT2 )⎜ 1 ⎟ ⇔− × ⎢⎜ 1 ⎟ + ( ΔT1ΔT2 )1/ 3
⎝ 2 ⎠ 6 ΔT1 + ΔT2 ⎢⎣ ⎝ 2 ⎠ 2

⎛ ΔT − ΔT2 ⎞ ⎛ ΔT1 + ΔT2 ⎞

2 1/ 3
⎡ ⎛ Δ T1 + Δ T2 ⎞ ⎤ ⎛ Δ T1 + Δ T2 ⎞
2 1/ 3

+ (ΔT1ΔT2 ) 2/3
⎤⎦ ≈ − ⎜ 1 ⎟ ⎜ ⎟ = ⎢ Δ T1 Δ T2 − ⎜ ⎟ ⎥⎜ ⎟
⎝ 2 ⎠ ⎝ 2 ⎠ ⎣⎢ ⎝ 2 ⎠ ⎥⎦ ⎝ 2 ⎠
1/ 3 1/ 3
1 (ΔT1 − ΔT2 ) ⎡ ⎛ ΔT + ΔT2 ⎞ ⎤ ⎛ ΔT1 + ΔT2 ⎞ = ⎛ ΔT1 + ΔT2 ⎞ (ΔT ΔT )1 / 3 −
2 2/3
⇔− ≈ ⎢(ΔT1 ΔT2 )1 / 3 − ⎜ 1 ⎟ ⎥× ⎜ ⎟ ⎜ ⎟ 1 2
6 ΔT1 + ΔT2 ⎢⎣ ⎝ 2 ⎠ ⎥⎦ ⎝ 2 ⎠ ⎝ 2 ⎠

1 ( Δ T1 − Δ T 2 )
2 1/ 3
Δ T1 + Δ T 2 Δ T1 + Δ T 2 ⎛ ΔT + ΔT2 ⎞
( ΔT1ΔT2 )
1/ 3
⇔ − ≈⎜ 1 ⎟ .
2 2 6 Δ T1 + Δ T 2 ⎝ 2 ⎠

211
 A. Zavala-Río et al. / Revista Mexicana de Ingeniería Química Vol. 4 (2005) 201-212

References Paterson, W.R. (1984). A replacement for the

logarithmic mean. Chemical Engineering
Abramowitz, M. and Stegun, I.A. (1972).
Science 39, 1635–1636.
Handbook of mathematical functions (9th
Reimann, K.A. (1986). The consideration of
printing). Dover Publications, New York.
dynamics and control in the design of heat
Alsop, A.W. and Edgar, T.F. (1989). Nonlinear
exchanger networks. Ph.D. Thesis, Swiss
heat exchanger control through the use of
Federal Institute of Technology Zürich
partially linearized control variables.
(ETH), Switzerland.
Chemical Engineering Communications 75,
Rudin, W. (1976). Principles of mathematical
115.
analysis. (3rd ed.). McGraw-Hill, USA.
Chen, J.J.J. (1987). Comments on improvements
Steiner, M. (1987). Dynamic models of heat
on a replacement for the logarithmic mean.
exchangers. In: Chemical engineering
Chemical Engineering Science 42, 2488–
fundamentals, XVIII Congress The Use of
2489.
Computers in Chemical Engineering, pp.
Faires, V.M. (1957). Thermodynamics (3rd ed.).
809–814. Giardini Naxos, Sicily, Italy.
The MacMillan Company, New York.
Steiner, M. (1989). Low order dynamic models of
Gardner, K. and Taborek, J. (1977). Mean
heat exchangers. In: International
temperature difference: a reappraisal,
symposium on district Heat Simulations.
AIChE Journal 23, 777–786.
Reykjavik, Iceland.
Holman, J.P. (1997). Heat transfer (8th ed.).
Underwood, A.J.V. (1970). Simple formula to
McGraw-Hill, USA.
calculate mean temperature difference.
Incropera, F.P. and DeWitt, D.P. (1990).
Chemical Engineering, June 15, 192.
Fundamentals of heat and mass transfer
Walas, S.M. (1991). Modelling with differential
(3rd ed.). John Wiley & Sons, New York.
equations in chemical engineering.
Kern, D.Q. (1950). Process heat transfer.
Butterworth-Heinemann, Stoneham.
McGraw-Hill, Tokyo.
Zavala-Río, A. and Santiesteban-Cos, R. (2004).
Khambanonda, T., Palazoglu, A., and Romagnoli,
Qualitatively reliable compartmental
J.A. (1990). The stability analysis of
models for double-pipe heat exchangers. In:
nonlinear feedback systems. AIChE Journal
Proc. of the 2nd IFAC symposium on
36, 66–74.
system, structure, and control, pp. 406–411.
Khambanonda, T., Palazoglu, A. and Romagnoli,
Oaxaca, Mexico.
J.A. (1991). A transformation approach to
Zavala-Río, A., Femat, R. and Romero-Méndez,
nonlinear process control. AIChE Journal
R. (2003). Countercurrent double-pipe heat
37, 1082–1092.
exchangers are a special type of positive
Malleswararao, Y.S.N. and Chidambaram, M.
systems. In: Positive systems – proceedings
(1992). Nonlinear controllers for a heat
of the first multidisciplinary international
exchanger. Journal of Process Control 2,
symposium on positive systems: Theory and
17–21.
applications (L. Benvenuti, A. de Santis,
Mathisen, K.W. (1994). Integrated design and
and L. Farina, eds.), pp. 385–392. LNCIS
control of heateExchanger networks. Ph.D.
294, Springer, Rome, Italy.
Thesis, University of Trondheim, NIT,
Zeghal, S., Isambert, A., Laouilleau, P., Boudehen,
Norway.
A. and Depeyre, D. (1991). Dynamic
McAdams, W.H. (1954). Heat transmission (3rd
simulation: a tool for process analysis. In:
ed.). McGraw-Hill, Tokyo.
Proceedings of the Computer-oriented
Papastratos, S., Isambert, A. and Depeyre, D.
process engineering, EFChE Working
(1993). Computerized optimum design and
Party, pp. 165–170. Barcelona, Spain.
dynamic simulation of heat exchanger
networks. Computers and Chemical
Engineering 17, S329–S334.

212