Accepted Manuscript

Logarithmic mean: Chen’s approximation or explicit solution?

J.J.J. Chen

PII: S0098-1354(18)30979-7
DOI: https://doi.org/10.1016/j.compchemeng.2018.10.002
Reference: CACE 6245

To appear in: Computers and Chemical Engineering

Please cite this article as: J.J.J. Chen , Logarithmic mean: Chen’s approxima-
tion or explicit solution?, Computers and Chemical Engineering (2018), doi:
https://doi.org/10.1016/j.compchemeng.2018.10.002

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and
all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT
Highlight

 The Chen‟s approximation has been widely used in research on the

optimization of heat exchanger and mass exchanger networks because it
allows easy handling of the numerical problems associated with the
logarithmic term.
 This work shows how a long forgotten function, the Lambert W-function
(Euler 1779, Lambert 1758, revived recently by Corless et al. 1996), can
be employed to provide an explicit solution involving the logarithmic
mean.

T
IP
CR
US
AN
M
ED
PT
CE
AC

Page 1 of 10
 ACCEPTED MANUSCRIPT
Logarithmic mean:
Chen‟s approximation or explicit solution?

J.J.J. Chen
Chemical and Materials Engineering Department,
The University of Auckland, PB92019, Auckland 1142, New Zealand
Email: [email protected]

Keywords: Logarithmic mean, approximation, HEN, Lambert W-function

T
Abstract

IP
An explicit solution has been obtained for the logarithmic mean temperature

CR
difference method of heat exchanger calculation by making use of the Lambert

US
W-function. The results might be of use where an explicit solution involving the

logarithmic mean is required.

AN
Chen’s Approximation

An approximation to the logarithmic mean (LMTD) between two numbers 1 and

M

2 attributed to Chen (1987) is given in Eq. (1).

ED

{ ( )} Eq. 1
( )
PT

Eq. (1) has been widely used in optimization models for heat exchanger network
CE

and other engineering equipment (see, for example, Yee et al. 1990, Yee &
AC

Grossman, 1990, Amarger et al, 1992, Lewin, 1998, Adjiman et al, 2000,

Jackson & Grossmann, 2001, Davis & Sandall, 2003, Ponce-Ortega et al, 2008,

Gabriel, et al., 2016, Bongartz & Mitsos, 2017, Pavao et al., 2017). The Chen

approximation slightly overestimates the area requirement, but it avoids

numerical problems associated with the logarithmic term, and it also has the

important advantage that when either 1 or 2 equals zero the driving force will

Page 2 of 10
 ACCEPTED MANUSCRIPT
be approximated to be zero (Yee et al. 1990, Yee & Grossmann, 1990, Lewin,

1998, Floudas et al., 1999, Adjiman et al., 2000, Davis & Sandall, 2003).

Lambert W-function

The earliest mention of the problem stated in Eq. 2 is attributed to Euler (1779),

but Euler himself credited Lambert (1758) with it (see also Corless et al. 1996,

Hayes 2005). Corless et al. (1996) gave a detailed analysis of the Lambert W-

T
function while Barry et al. (2000) gave analytical approximations for it. Hayes

IP
(2005) and Stewart (2005) gave simplified accounts of the properties of the

CR
Lambert W-function. Others, including Keady (1998) who applied the Lambert

W-function to the Colebrook-White equation, Valluri et al. (2000) who discussed

US
its possible applications in physics, and Golicnik (2012) applied the function to
AN
enzyme-catalysed biochemical reactions, while Disney & Warburton (2012)

applied it to certain economic problems. Hayes (2005) suggests that scientific

M

calculators should have a built in Lambert key, and on the American Scientist

Website, Hayes (2005) asked: “Should Lambert W be added to the canon of

ED

standard textbook functions?” Stewart (2005) also suggested that the Lambert
PT

W function should be included in the mathematical curricula.

CE

In Eq. 2, it is easy to calculate y when the value of x is given.

Eq. 2
AC

However, in the inverse case, i.e. given y, the value of x is not readily obtained.

This is where the Lambert W-function comes in because it is the inverse of Eq. 2,

i.e. if y is the input, the Lambert W-function gives x = W(y) such that Eq. 2 is

satisfied.

Page 3 of 10
 ACCEPTED MANUSCRIPT
For example, with reference to Fig. 1, for y=e, W(e) gives x=W(y)=1 because e

= 1*e1, and this is plotted as (e, 1); and for y=( ), W( ) gives x=W(y)=-1

- -
because -1*e 1=-e 1, and this is plotted as ( , -1).

The definition of W(y) may be written as

( ) ( ) Eq. 3

T
The inverse W function given by Eq. 3 has two values of W when y is between

IP
and 0. When y is greater than 0, W is single valued (Hayes, 2005).

CR
US
AN
M
ED
PT
CE
AC

Fig. 1. The Lambert W-function.

The curve may be divided into three regions (Golicnik, 2012, Barry et al., 2000)

and the branches are:

Page 4 of 10
 ACCEPTED MANUSCRIPT
1. Region 1: y>0, the W0+ branch;

-
2. Region 2: -e-1 < y < 0, and 0 > W > -1, the Wo branch;

3. Region 3: -e-1 < y < 0, W < -1, the W-1 branch.

Application of the Lambert W-function to the Logarithmic Mean

Writing the heat exchanger equation as

T
Eq. 4

IP
( )

CR
Dividing by 2 and re-arranging

US Eq. 5
AN
Consider 2 as known and 1 the unknown to be solved,

Let Eq. 6
M

Let Eq. 7
ED

Thus, ( ) Eq. 8
PT
CE

( )
Giving Eq. 9
AC

Eq. 9 may be re-cast as

Eq. 10

Multiplying both sides by

Eq. 11

Page 5 of 10
 ACCEPTED MANUSCRIPT
In Eq. 11, 2 is known, while  the unknown to be solved appears on the left-

hand-side of the equation only.

This is where the Lambert W function comes in. Taking the W of both sides of

Eq. 11, the LHS is of the form xex, where x  according to Eq. 3.

Thus, Eq. 11 becomes

T
 ( ) Eq. 12

IP
All terms on the RHS of Eq. 12 that are within the W function are known and

CR
hence ( ) can be evaluated. In fact, by inspection with reference to Eq.

3, it should give a numerical value of

US . However, from Fig. 1, W of a
AN
negative number for -e-1 < y < 0 will yield two roots, and the value of is

expected to be one of them. The two roots are obtained as one each from the
M

-
Wo branch and the W-1 branch. Thus  can be evaluated.
ED

Example 1 (for the general case when 1  2)

PT

To illustrate the application of Eq. 12, using the same problem considered in

Patterson (1984):
CE

2 = (50-30)=20oC; 1 = (125-t1)oC;  = 1 /2

AC

Q = 7500kW; UA = 175 kW/oC

K = Q/UA = 7500/175 = 42.857oC

2 /K = 20/42.857 = 0.4667

Thus, if  is solved, 1 and t1 can be readily evaluated.

Page 6 of 10
 ACCEPTED MANUSCRIPT
Applying Eq. 12, the LHS gives -0.4667, and the RHS gives W(-0.4667e-0.4667),

-
i.e. W(-0.2927). From Figure 1, W(-0.2927) has two roots, one from the Wo

branch and one from the W-1 branch.

The exact values of the two roots may be obtained using a standard

mathematical software package such as Wolfram MathWorld, Matlab, Macsyma,

Maple, Mathematica etc. (Hayes 2005). Note that the Lambert W-function is

T
also referred to as the Product Log function in some software packages. Barry et

IP
al. (2000) and Golicnik (2012), among others, have provided equations for the

CR
evaluation of W, and tabulated values are also given in Disney & Warburton

(2012).

-
US
W(-0.2927) gives -0.46 from the Wo branch (which is = -0.4667 as
AN

discussed earlier) and -1.84 from the W-1 branch. Thus

M

 
ED

The first solution is „trivial‟ as this occurs when =1, but 1  2 , and hence this

root can be discarded. The second solution gives =3.94, resulting in 1 = 78.9
PT

and t1 = 46.1oC. As expected, this solution is exactly the same as that obtained
CE

by iteration using the actual LMTD.

Example 2 (for the case when 1 = 2)

AC

Basically the Paterson (1984) case but setting 2 to be equal to K.

2 = (72.86-30)= 42.86oC; 1 = (125-t1)oC;  = 1 /2

Q = 7500kW; UA = 175 kW/oC

K = Q/UA = 7500/175 = 42.857oC

Page 7 of 10
 ACCEPTED MANUSCRIPT
2 /K = 42.86/42.857 1.

Applying Eq. 12, the LHS gives  , and the RHS gives W(-e-1). From Figure 1,

-
this occurs at the junction of Wo and W-1, i.e. there is only one root, and

W(-e-1) = -1, i.e. =1 giving 1 = 2.

Discussion

T
The two examples shown above are based on the case considered by

IP
Paterson(1984) and Chen(1987) using the transformed equation given in Eq.

CR
12. Eq. 12 is equivalent to Eq. 4, but it allows for the exact evaluation 2

without a trial-and-error process which is necessary when using Eq. 4 because

US
of the appearance of 2 in two places on the right-hand-side of the equation,

inside and outside of the logarithmic term. In the case where Q (or A) is the
AN

only unknown, Q may be evaluated from an energy balance, and Eq. 4 will
M

provide an explicit solution.

The above examples considered 2 as the unknown. If 1 is the unknown instead

ED

and is required to be solved, it is simply a matter of interchanging 1 and 2 in

PT

Eq. 4, 5, and 6, and the solution will follow immediately.

CE

It is also noted that in the case where 1 = 2 as illustrated in Example 2, Eq. 4

will become .
( )
AC

Conclusion

An explicit solution has been shown for heat exchanger calculations using the

logarithmic mean temperature method by employing the Lambert W-function

due originally to Lambert (1758) and Euler (1779). It remains to be seen if it

Page 8 of 10
 ACCEPTED MANUSCRIPT
will replace the use of approximations for the logarithmic term in the research

on optimization such as those referenced in this note.

Declaration of Interest

None. This research did not receive any specific grant from funding agencies in

the public, commercial, or not-for-profit sectors.

References

T
Adjiman, C.S., Androulakis, I.P., Floudas, C.A., 2000, Global optimization of mixed-

IP
integer nonlinear problems, AIChE Jl., 46, 1769-1797.

Amarger, R.J., Biegler, L.T., Grossmann, I.E., (1992) An automated modelling and

CR
reformulation systems for design optimization, Computers and Chemical Engineering,
16, 623-636.

Barry, D.A., Parlange, J.-Y., Li, L., Prommer, H., Cunningham, C.J., Stagnitti, F. (2000),

US
Analytical approximations for real values of the Lambert W-function, Mathematics and
Computers in Simulation, 53, 95-103.
AN
Bongartz, D., Mitsos, A., 2017, Deterministic global optimization of process flowsheets
in a reduced space using McCormick relaxations, J. Global Optimization, 69, 761-796.

Chen, J. J. J., 1987, Letter to the editor: comments on improvement on a replacement

M

for the logarithmic mean. Chemical Engineering Science, 42, 2488-2489.

Corless, R.M., Gonnet, G.H., Hare, D.E.G., Jeffrey, D.J., Knuth, D.E., 1996, On the
lambert W function, Advances in Computational Mathematics, 5, 329-359.
ED

Davis, R.A., Sandall, O.C., 2003, A Simple Analysis For gas separation membrane
experiments, Chemical Engineering Education, 74-80, Winter 2003.
PT

Disney, S.M., Warburton, R.D.H., 2012, On the Lambert W function: Economic order
quantity applications and pedagogical considerations, Intl. J. Production Economics, 40,
754-764.
CE

Euler, L., 1779, De serie Lambertina plurimisque eius insignibus proprietatibus,

Leonhardi Euleri Opera Omnia, Ser. 1, Opera Mathematica 6 (1921) [orig. date 1779]
350-369.
AC

Floudas, C. A., Pardalos, P.M., Adjiman, C.S., Esposito, W.R., Gumus, Z.H., Harding,
S.T., Klepeis, J.L., Meyer, C.A., and Schweiger, C.A., 1999, Handbook of Test Problems
in Local and Global Optimization, Kluwer Academic Press.

Gabriel, K.J., El-Halwagi, M.M., Linke, P., 2016, Optimization across the Water-Energy
Nexus for Integrating Heat, Power, and Water for Industrial Processes, Coupled with
Hybrid Thermal-Membrane Desalination, Industrial & Engineering Chemistry Research,
55, 3442-3466.

Golicnik, M., 2012, On the Lambert W function and its utility in biochemical kinetics,
Biochemical Engineering Journal, 63, 116-123.

Page 9 of 10
 ACCEPTED MANUSCRIPT
Hayes, B., 2005, Why W? American Scientist, 93, 104-108. See also
https://www.americanscientist.org/article/why-w “Why W? “Should Lambert W be
added to the canon of standard textbook functions?” (Downloaded 18 September 2018)

Jackson, JR; Grossmann, IE., 2001, A disjunctive programming approach for the
optimal design of reactive distillation columns, Computers and Chemical Engineering,
25, 1661-1673.

Keady, G., 1998, Colebrook-White formula for pipe flows, J. Hydraulic Engineering,
124, 96-97.

Lambert, L.J.H., 1758, Observationes variae in mathesin puram, Acta Helvetica,

physico-mathematicoanatomico-botanico-medica 3, Basel (1758) 128-168.

Lewin, DR, 1998, A generalized method for HEN synthesis using stochastic optimization

T
- II. The synthesis of cost-optimal networks, Computers and Chemical Engineering, 22,

IP
1387-1405.

Paterson, W. R., 1984, A replacement for the logarithmic mean. Chemical Engineering

CR
Science, 39, 1635-1636.

Pavao, L.V., Borba Costa, C.B., Ravagnani, M. A. S. S., 2017, Heat Exchanger Network
Synthesis without stream splits using parallelized and simplified simulated Annealing

US
and Particle Swarm Optimization, Chemical Engineering Science, 158, 96-107.

Ponce-Ortega, J. M., Jimenez-Gutierrez, A., Grossmann, I.E., 2008, Optimal synthesis

AN
of heat exchanger networks involving isothermal process streams, Computers and
Chemical Engineering, 32, 1918-1942.

Stewart, S., 2005, A new elementary function for our curricula? Australian Senior
Mathematics Journal, 19, 8-26.
M

Yee, T.F., Grossmann, I.E., Kranja, Z., 1990, Simultaneous-optimization models for
heat integration. 1. Area and energy targeting and modelling of multi-stream
ED

exchangers, Computers and Chemical Engineering, 14, 1151-1164.

Yee, T.F., Grossmann, I.E., 1990, Simultaneous-optimization models for heat

integration. 2. Heat-exchanger networks synthesis, Computers and Chemical
PT

Engineering, 14, 1165-1184.

CE
AC

Page 10 of 10