Sie sind auf Seite 1von 14

Cornput. them. Engng, Vol. II, No. 6, pp.

723-736,
Printed in Great Britain. All rights reserved

1987
Copyright

0098~1354/87
1987 Pergamon

$3.00 +O.OO
Journals Ltd

A REVIEW OF SPREADSHEET USAGE IN


CHEMICAL ENGINEERING
CALCULATIONS
E. M. ROSEN
Monsanto Company FZWK, 800 N. Lindbergh, St Louis, MO 63167, U.S.A.

R. N. ADAMS
Retired Consultant, Monsanto Company, Godfrey, Illinois, U.S.A.
(Received 30 January 1987; final revision received 26 May 1987; received for publication 23 June 1987)
Abstract-The
use of the electronic spreadsheet to carry out a variety of routine chemical engineering
calculations in process design and project engineering is reviewed. Some possible formats for carrying out
these calculations are demonstrated. The limitations and advantages of the spreadsheet are discussed with
respect to a variety of its potential uses.
Scope-The electronic spreadsheets capabilities to solve algebraic and differential equations as well as
interpolation and integration are demonstrated on a series of small and generally low dimensional example
problems taken from thermodynamics, reaction equilibrium, flowsheeting and mass and heat transfer. Its
role in preparing lists, templates for routine calculations as well as economic analyses is also explored.
The interface of spreadsheets to data bases and graphical systems as a means of intergrating calculations
is indicated.
Conclusions and Significance-The electronic spreadsheet is shown to provide an attractive alternate to
other means of computation in a variety of computational instances. Its general availability, cost and
accessibility as well as ease of use will probably make it an attractive alternate for chemical engineering
calculations in both education and industry.

1.

INTRODUCTION

The first half of the decade of the 1980s can now be


seen as a revolutionary one in terms of new computing tools and computing environments for the
chemical engineer. Hardware and software have become available to enable the engineer to communicate in two of his most natural languages: drawings
and 2-D tables.
Just as graphical systems have come to support the
engineers use of drawings and sketches so have
spreadsheet programs come to respond to the engineers need to work with 2-D tables.
The broad use of both graphics and spreadsheets
have been spawned generally via the personal computer (microcomputer). In roughly 5 yr since the
introduction of VisiCalc, dozens of spreadsheet programs have arisen. For example, the Fall issue of
Microcomputer Market Place [l] lists over 200
spreadsheet programs. The popularity of Lotus l-2-3
and of Excel (on the Macintosh), attest to the attractive features of those programs. Programs using the
spreadsheet format which have ever more features
and run on minicomputers and mainframes as well as
the personal computer continue to appear [2]. Here,
however, we shall refer to a generic spreadsheet which
has features common to most programs. All of the
applications mentioned below, in fact, were carried
out with Lotus l-2-3 or Excel.
Though spreadsheets have been used most broadly

in the financial arena, they have found their way


into many aspects of process and project engineering [3-l 11. In a number of cases it provides
not only an alternative means of computation but
may indeed be the right tool for the job. However, in
his review of the book VisiCalc for Science and
Engineering Owens [12] makes the point that he
found it easier to write and understand a BASIC
program than using a spreadsheet. Hsiao [13] further
compares the programming method with the spreadsheet implementation of Gauss quadrature. Clearly
the issues of ease of use, availability, cost, system
capabilities and accessibility all determine what tool
to use for what job.
It is the purpose of this review to show that many
of the routine calculations that are undertaken by
the chemical engineer can be carried out well on a
spreadsheet. Its capability to show the details of a
calculation can make it an attractive method of
following the numerical course of a calculation. Its
use in project engineering in preparing lists, economic
evaluations and bids and its interfaces to other
utilities make the spreadsheet a broadly useful tool.

723

2. THE SPREADSHEET:
ITS FORMAT AND CONCEPTUAL

BASIS

The spreadsheet is a 2-D array of columns and


rows. The intersection of a column and row is called
a cell. The width of a cell is arbitrary.

E. M. ROSEN and R. N. An~hrls

124

Cells can contain labels or arithmetic formulae.


The cells can be executed in column-wise or row-wise
order. The labels and the results of the computations
(rather than the formulae) are shown on the spreadsheet. The cell (Lotus l-2-3) is indicated by its column
and row designation: i.e. Cl2 is column C row 12.
The spreadsheet user can follow iterative calculations by setting the recalculation
mode to
manual. Each time a pass of the worksheet is
desired, a function key (Lotus l-2-3) is pressed. This
mode is recommended for most calculations. If the
automatic mode is selected, the spreadsheet is
automatically recalculated until no changes take
place in the cells. (Note that Excel has a maximum
change as well as a maximum number of iterations
which is user controlled).
The COPY command (Lotus l-2-3) is very powerful. Cell locations can be copied either absolutely
(indicated by a $) or relatively (if no $ is used).
The very simplicity of the spreadsheets format
makes it easy to learn. Some of the advantages of the
spreadsheet are:
s
(1) complete user-defined I/O layout in a tabular
format,
(2) non-procedural,
(3) input and results are collected in one level and
the arithmetic operations are hidden in another level,
(4) easy methods are available to move around the
spreadsheet,
(5) the user completely defines the operations.
The spreadsheet does have a number of limitations
for chemical engineering computations. These are as
follows.
1. The entire spreadsheet is calculated through
entirely each time it does a recalculation. There are
no GO TO and return capabilities (without macros).
This then precludes the use of a subroutine or loop
within loop type of calculation.
2. Matrix operations of dimension greater than two
are not easily implemented.
3. Though macros are often available, their usage
is such that BASIC or FORTRAN programming may
provide a more attractive alternative.
From a practical point of view, limitation 1 means
that the entire spreadsheet must be viewed as a large
set of algebraic equations of the form
f(x) - x = 0,

(1)

where the x vector is sought. Often direct iteration


will be adequate to force convergence though such
methods as Bounded Wegstein can be used to accelerate and/or force convergence. These issues will be
discussed in the following sections.

3.SPREADSHEET
IMPLEMENTATION
FOR
NONLINEAR
AND DIFFERENTIAL EQUATIONS

It is often useful to implement spreadsheet calcu-

lations in a standard format. This is especially true in


the case of solving nonlinear and differential equations (as well as optimization algorithms) since the
form of solution for a wide class of methods is:
x(n + 1) = x(n) + tS(n)

(2)

Here x(n) and S(n) are vectors. Some initial


guess to the solution is required to start the SO~Ution. The method of solution supplies 6. The new
estimated solution (or next step) is then given by
equation (2). The scalar t is often used to permit only
a fraction of the calculated 6 vector to be used to
enhance the robustness of the particular method.
With this in mind, the following spreadsheet format is convenient to initiate the initial guess of x(n)
and then run through the iterations keeping count of
the number of iterations:
1.
2.
3.
4.

A
Initiator
Counter
Initial x(n)

B
0
@IF(Bl = O,O,B2+ 1)
xxxx
@IF(B2 = O,B3,B4 + BlO)

x(n)

9. t

1.0

10. t6

YYY

The initiator cell is just that. It enables the initial


value of x(n) to be set and the calculations be checked
the first time through. The initiator cell Bl is set to 1
to start the calculation. The calculations will proceed
column-wise from cell Bl to BlO. If the spreadsheet
is in the manual recalculation mode then each iteration can be followed. The value of t acts an algorithm parameter.
The @IF statement is defined to work as follows:
@IF(logical

operation,

true, false)

If the logical operation is true, then the true


action is taken to set the cells value. If the logical
operation is false then the false operation sets the
cells value.
It was pointed out above that spreadsheets do not
allow loop within loop calculations. In other words
the familiar DO loop in FORTRAN and the FOR loop
of BASIC are not available. However, the option of
repeating calculations is there using the COPY command which automatically changes the indexing of
the cells much like that of a DO loop.
Nor is there an IF statement that can be used to
test for convergence. Instead, as indicated previously,
iterations stop automatically when there is no
change in the cells values, if the recalculation mode
is set to automatic.

4.lNTERPOLATlONANDlNTEGRATlON

Interpolation
cal engineering

and integration are common chemitasks. Classically, these calculations

A review of spreadsheet usage


have often been done graphically. However, today a
large number of numerical methods are available.
Spreadsheet software generally has a rather rudimentary table lookup capability. Given a set of x vs
Y(x) values, the spreadsheet software will return the
value of Y corresponding to the value of x which is
immediately below the x argument. This capability,
however, may be used to carry out linear or quadratic
interpolation.
Consider, for example, Stirlings divided difference
formula of degree two [14]:

725

The table of x vs x* values is shown In Fig. 1


together with the evaluation of the term under the
integral.
The table is particularly simple to generate using
the COPY command once the first row is calculated.
To carry out the integration, a four interval closed
Newton-Coates integration formula (precision 5) is
used[l6]:

f(x) dx = 2h [7f(x,) + 32f(x, ) +


+32f(x,)

For the four intervals,


gration step) is

13-(x,)

+ 7f(x,)1/45.

the value of /I (the inte-

h = (0.40 - 0.03) = 0 093


4
.(

where:

y(x,,x,)==$

(4)

Xl -XII

and

Y(-%,x1, x2) =

Yh,X2) -Y(Xo,X,)
x2 -

(5)

XII

The original x vs Y table can be modified by adding


four extra columns:
2nd
dif.

av. x

av. y

1st
dif.
0
0

l
l

av.
x=2

+x_,

(6)
(7)

For a desired value of x, the @VLOOKUP function (Lotus l-2-3) can be used to extract the values
from the table needed to evaluate equation (3). (If it
is desired to interpolate linearly at a value of x
between the first two entries, where entries for 2nd
dif., av. x and av. y are missing, twice the 1st dif. can
be placed in the 1st slot of av. y).
Henley and Seader [15] describe the graphical
evaluation of the following integral to calculate the
number of transfer units N(OG):
N(OG)

where
(1 _x)

hn

o.4 (1 - xhln

o.osu
-X)(X-XL)

J-x)-(1

-x*)
lne

Since the values of x required for the functional


evaluations are not all in the table, interpolation must
be used at x = 0.123, 0.215 and 0.308. The two end
points fall on the table values and so no interpolation
is needed. Note the four columns added to carry out
the interpolation using equation (3). The interpolated
value is given by f(x) in Fig. 1. The answer of 3.73
is in reasonable agreement with the graphical value of
3.44 given by Henley and Seader. Figure 1 also gives
a portion of the cell formulas in column E.

The general format used to solve nonlinear


algebraic equations was indicated above. The 1-D
Newton-Raphson
method is:

x(n + 1) = x(n) -f(x)/f(x)

dx

(8)

(9

= x(n) + tS(n),

(12)

for the solution of the equation:

f(x)
= 0.

Y(X,,X,) fY(%,X-,)
2

av. y =

(11)

5.ALGEBRAIC EQUATION SOLVING

The 1st dif. is given by equation (4) and the 2nd dif.
is given by equation (5). Also:
x,

(lo)

(13)

In general the value off(x) will be determined


numerically by perturbing x by a small value h.
The use of the method in solving the flash equation
at a specified temperature and pressure (where the
K-values are not a function of composition) is shown
in Fig. 2a. The flash equation is;

The example was taken from Henley and Seader [ 151.


In higher dimensions
the Newton-Raphson
method requires a solution of a set of linear
equations. In 2-D this can be done easily with
determinants but in three or higher dimensions this
becomes inconvenient.
Figure 2b gives the solution to the problem of
determining the two extents of reaction such that the
equilibrium equations will be satisfied (Henley and
Rosen [ 171):

E. M. ROSEN

726

and R. N. ADAMS

Evaluation of Nunber of Transfer Units


Data From Page 653 of Henley and Seder

l-x

X-X*

X*

l-x*

(1.x)Lm
_._..._

(l-x)Lm

1st dif

(l-x)(x-x*)

0.030
0.050
0.100
0.150
0.200
0.250
0.300
0.350
0.600

0.002
0.005
0.010
0.025
0.040
0.080
0.120
0.170
0.260

0.028
0.045
0.090
0.125
0.160
0.170
0.180
0.180
0.140

0.970
0.950
0.900
0.850
0.800
0.750
0.700
0.650
0.600

0.998
0.995
0.990
0.975
0.960
0.920
0.880
0.830
0.740

0.984
;.;;z
0:911
0.878
0.832
0.787
0.736
0.668

36.227
22.746
11.658
8.575
6.856
6.526
6.243
6.293
7.947

2nd dif avg dif

-674.141
-221.733 6462.980 -337.071
-61.661 1600.716 *110.866
-34.375 272.857
-30.831
-6.594 277.817 -17.108
-5.674
9.194
-3.297
1.017
66.916
-2.837
33.072 320.550
0.509

avg x

0.065
0.100
0.150
0.200
0.250
0.300
0.350

Integration By Newton totes Formula - b intervals


0.093

i=O

X
x0
avg dif
2nd dif

avg x
YO
f(X)
ial

X
x0
svg dif
2nd dif
avg x
YO
f(X)
i=2

X
x0
avg dif
2nd dif
avg x
YO
f(X)

0.000
0.030
0.030
0.000
0.000

i=3

X
X0
wg dif
2nd dif

avg x

0.000
36.227
36.227

vo

f(X)

YO
f(X)

6.000
0.400
0.600
0.000
0.000
0.000
7.967
7.967

Integral

3.731

i=b
X
x0
evg dif
2nd dif
avg x

1.000
0.123
0.100
-110.866
1600.716
0.100
11.658
9.976
2.000
0.215
0.200
-17.188
277.817
0.200
6.856
6.661

3.000
0.308
0.300
-2.837
64.914
0.300
6.243
6.225

E20:
E21:
E22:
E23:

3
+SAS7+*tBt18%20
8VLDOKUP(E2l,SAL7..SKSlS,O)
gVLOOKUP(E2l,SAS7..tKt15,9)
Etb: 'iNlLOOKUP(E2l,SAL)7..SKS15.8)
E25: gVLDOKUP(E21;SAL7..SKt15;16)
E26: OVLCCtKUP(E21,SAt7..SKSl5,6)
E27: +E26+(E21-E22)*E23+(E21-E22)*(E21-E25)*E2b

Fig. I. Evaluation of number of transfer units.

CH4 + H,O = CO + 3H,,


CO+H,O=CO,+H,.

(19

To solve this problem, a damped NewtonRaphson approach is used which requires the calculation of the Jacobian .I (partial derivative matrix of
the functions) using numerical perturbations.
The stoichiometric coefficients, the K,-values, the
pressure and the initial values are all given in the data
section at the top of the spreadsheet. The total moles
at any x(n) and y(n) (the extents) is given by:
total moles =

i ui, x(n )
( i-1
>
+

$,Q
y(n)+ i di, (16)
i-1
(
>
where au is the stoichiometric coefficient of the ith
component in the jth reaction and di is the initial
moles of component i.
The spreadsheet @SUM function makes this
evaluation particularly easy.

Once the Jacobian is evaluated at x(n), r(n) the S


vector is the solution to:
J(x)6 = -f(x).

(17)

The value of t chosen is 0.5. The final solution is


obtained after 20 iterations and is the one given by
Henley and Rosen.
Note that the same spreadsheet may be used to
study the effects of different initial compositions,
different &-values or different pressures and indeed
different reactions. Caution must be taken, here
however to start with extents which lead to nonnegative compositions: e.g. (0,O).
Broydens method for solving nonlinear equations
[ 171 can be implemented on a spreadsheet as well.
The calculations require only inner products and
vector additions and therefore can be done in a
straightforward manner though it is somewhat more
tedious to implement.
Wegsteins method [17] is particularly simple to
implement on a spreadsheet. Although developed for

Flesh Using Newtons Nethod


' Henley and Seeder p 279
Conponevt
K Value
Propane
n-lutane
n-Pentane
n-Hexane

t Value

4.2
1.75
0.74
0.34

initiator
counter
initial an
an

0.;
0.121003

1: r*(l-K)/(l*a*(K-1))
2: r*(l-K)/tl+a*tK-1))
3: r*(l-K)/(l+a*(K-1))
4: z*(l-K)/(l+a*(K-1))
SUll

-0.23021
-0.13743
0.000552
0.207094
-0.OOOoo

h
an+h
1: t*(l-K)/(l+a*(K-1))
2: z*(l-K)/(l+a*(K-1))
3: r*(l-K)/(l*a*(K-1))
4: r*(l-KI/(l+a*(K-1))
sun

slope

delta
t
t*delta

727

review of spreadsheet usage

0.1
0.2
0.3
0.4
BlO:
811:
912:
813:

1
GIIF(Bl0=O,O,B1l+l)
0.5
g1F(Bll=O,Bl2,Bl3+B32)

815: CS"~1-g5~/~1+BS13i~B5-l~~
816:
817:
BlO:
-Bl9:

+C6*(1-B6)/(1+0S13*(86-1))
+C7*(l-B7)/(1+BSl3'(B7-1))
+C0*(1-B0)/(1+Btl3*(B0-1))
1SUM(g15..810)

0.001
0.122003
-0.22960
-0.13734
0.000574
0.207301
0.000050

'821:
B22:
823:
B24:
B25:
826:
B27:

0.001
+Bl3*821

0.050974
0.000000
1
0.000000

B29:
830:
831:
832:

+C7'il-07i;il+0ltt'iB7-l;j
+C0*(1-80)/(l+Bt22*(B0-1))
1SUH(B23..B26)
(027-B19)/B21
-819/829
1
-Bl9/029

Fig. 2a. Solution of the flash equation.

a single equation it has been applied to higher


dimensional sets of equations when bounding is
applied. The method is applied to equations in the
form:

- @b4OW[xU)
- x@)l),

t = l/( l-slope).

(19)
(20)

= x(n) + r&(n).

~.DIF~REN~AL

EQUATION

SOLVING

As indicated, the general format for solving


differential equations on a spreadsheet is similar to
that for solving nonlinear equations. The value of 6
for the dependent variable is set by the method used.
The value of t in this case is 1.
In one dimension the equation to be solved is:

x(n) =.m(n), x(n)1

The algorithm is:


x(n + 1) =x(n) + t(4[x(n)]

the solution is found after 17 iterations.


bounding on the slope the solution would
The spreadsheet is an excellent method for
this behavior.

(18)

9(x) = x

To get started an initial estimate is made. The value


of the second estimate is set to a(x). After that a
slope and a value of t is calculated. To give the
method robustness the value of the slope is often
bounded, say between 0.8 and -9:
slope = {@IxU)l

Note
Without
diverge.
studying

-x(n)}
(21)

Here t is calculated and not set as a parameter.


Note that if t = 1 on the initial pass, the value of x( 1)
is simply equal to 4[x(O)].
The spreadsheet implementation for the solution of
volume in one form of the van der Waalss equation
(taken from Myers and Seider [18]) is shown in Fig.
2c. In order to avoid division by zero at the solution
the denominator in the expression for the slope is
modified to test for zero. The expression for the slope
is shown in cell B26.

where the value of x(n) is given at t(n), i.e. the initial


values.
The fourth order Runge-Kutta method [16] is an
explicit method that directly calculates the value of 6.
For a single equation:
6 = [k(l) + 2k(2) + ik(3) + k(4)]/6

(23)

where k(l), k(2), k(3) and k(4) are functions of h (the


step size) and the value of the function at various
points between t(n) and t(n) + h.
Explicit methods are the simplest to implement
though implicit methods involving a predictor and
corrector can be implemented by repeating the caku-

E. M. ROSENand R. N.

128

hAMS

Solution of Equil ibriun Equations Henley and Rosen p 376


Stoichiometric Coefficients
reaction 1
reaction 2
initial moles
carponcots
*1.000000
0.000000
6.000000
CH4
-1.000000
-1.000000
5.000000
HZ0
-1.000000
o.oooOOD
co
1.000000
1.000000
0.000000
3.000000
HZ
0.000000
1.000000
0.000000
co2
equilibrium K (rtm)
pressure (atm)

0.540000
1.000000

initiator
counter
initial xn
initial yn
xn
yn
sun moles
p CH4
p HZ0
P CD
P 142
p co2
fl(xn,yn)

fZ(xn,vn)
h
xmh
sun moles xn+h,yn
p CH4
p HZ0
P CD
P H2
p co2
fltxn+h,yn)
fZtxmh,yn)

sux n!aLesxn,yn+h
D CH4
p HZ0
P CD
P R2
p co2
flCxn,yn+h)
fZ(xn,wh)
J(l,l)
J(2,l)
dct J
dct J : xn
dct J : )n
delta xn
delta yn
t
Vdelta xn
t*deLta yn

J(1,2)
J(2,Z)

2.490000

1.000000
20.000000
3.000000
1.000000
2.421094
0.842178
15.843787
0.225837
0.109S65
0.099706
0.511737
0.053155
0.000001
0.000001

813:
814:
815:
816:
617:
Bl8:
819:
820:
821:
BZZ:
B23:
024:
825:
026:

0.001000
2.422894

B28:
B29:

15.845787
0.225745
0.109488
O.OW756
0.511862
0.053148
0.002347
0.000314

831:
832:
B33:
834:
835:
836:
837:
838:

0.843178

640:

15.043787
0.225837
0.109502
0.099643
0.511801
0.053218
0.000314
0.002520

B42:
043:
844:
845:
046:
047:
048:
049:

1
aIFtBi3=0,0,8i4+i)
3
1
a1F(814=0,815,817+859)
a1FtB14=0,816,s18+860)
aSUM~LOU..~Dt8)+aSUn(SaS4..SBS8)*B17+aSVn(tCS4..SCS8)'818
lSRS11*WJ4+SB4'SBS17+SC4*S8Sl8)/SaSl9
+tBS11*(SD5+Ss~iiES17+SC5*SBS18)/SaS19
+SBS11'(SD6+Ss6*SBS17+SC6'SsSl8)/S8Sl9
+Sss1l*(SD7+SB7'SBS17+SC7*S8Sl8)/SsSl9

2.345948
0.313344
5.812484
-0.000003
-0.000001
-o.oooodi
-0.000000
0.500000
-0.000000
-0.000000

0.312989
2.519475

C51: (837.B25)/~28
CSZ: (838-826)/028

051: (648.B2S)IstE
Et: (849.B26)/828

853: +Csl*D~2-C52'DSl
854: -825*D52+826*D51
855: -C51*B26+C52'825
856: +B54/853
I?;;:;B;5/053
859: +856'850
860: +857*B58

Fig. 2b. Solution to the equilibrium equations.

lation a fixed number of times. This indeed is similar


to what is often done in larger codes.
Simple predictor-corrector
methods such as the
trapezoidal corrector [19] are not difficult to implement, i.e.
predictor:
x(n + 1) = x(n - 1) + Zhx(n),

To get started, the Euler method is used:


x(n + 1) = x(n) +/2x(n).
Figure 3a gives the solution to the integration of
the equation describing the differential distillation of
a binary mixture using the trapezoidal predictorcorrector. The problem (see Henley and Seader[ 151,
p. 362) is to find the composition in the pot as a
function of time:

(24)
dx,/dr = -[D/W(t)](y

corrector:
x(n +

1) = x(n) + [x(n + 1) + x(n)]h/2.

(25)

-x,).

(26)

W(r) is the amount of the mixture remaining in the


pot at any time and is given by:

A review of spreadsheet usage


Bounded Wegstein
Myers and Seider p 454
van der Uea\s' equation version 2
v - -?*v**S/s
+171*v**Zfa +b*P*v**Z/a
upper bound on s\ope
lower bound on rlope

0.8

86: Q.8

-9

87:

.9

89: 1351000
BjO: 38.64
911: 50
612: 82.06
813: 173.15

1351000

;
P'atm
R cm3 otm/ mote K
1 .K

38.66
50
82.06
173.15
1
17
600
222.4454
222.4454

815:
816:
917:
818:
619:

phi(xl)
x2
phi(x2)

222.4454
222.4454
222.4454
222.4454

B21: @1~(~16=0,0,823)
922: @1~(816=0,0,824)
823: +B18
024: 4B19

slope
t
t*de\te

-0.17872
0.840377
3.1E-13

026: @M1N(86,@MAX(B7,~824-922)/~lF~~ABS~B23~B21~~0,~823-~~~~,1~~~
827: l/(1-826)
828: l827"(024-823)

inttlstor
counter
initial xn
zltxn)
xl

729

1
alF~e15=0,O,B16+1)
600
~lF(B16=0,817,818rBZ8)
~8114818~E18~818/89~~l2*Bl3'~l~~8l8/8P,BlO~~ll*8l~~Bl~/~9'~lO

Fig. 2c. Wegstein solution to van der Waal's equation.

H'(t) = -Dt

+ W(0).

(27)
The value of y (vapor composition) is given as a
function of the relative volatility, (I, which is set to the
value of 2.41. X, is the pot composition:
y = a&,/[1 + x,(a - 111.

(28)

D = 10 kg h-r (overhead distillate) and W(0) = 100 kg

initially. The solution is given at t = 5 h. It compares


well with the analytical solution given by Henley and
Seader. The solution where the equilibrium relationship is given graphically easily follows using the
interpolation method discussed previously.
The use of spreadsheets for the solution of partial
differential equations using finite difference methods

Binary Differential Distillation


TrapezoidaL Predictor-Corrector
Henley and Scader p 362
initiator
counter
initial xg
initial t

1.0000000
50.0000000
0.5000000
0.0000000

86:
87:
08:
BP:

10.0woooo
2.4100000
100.0000000

611:
612:
613:

x(n-1)

0.3560276

815:

xn
tn
h

0.3517684
5.0000000
0.1000000

017:
El&
819:

50.0000000
-q.O42984g

021:
622:

0.3474308

024:

5.1000000
49.0000000
-0.0437892
-0.0043387

826:
B27:
828:
BZO:

D-dist rata kg/hr


alpha
initial liq in still

U(w)
f(xn,tn)=x'(n)
predictor: r(rt+lY
tn+h
U(tn+h)
x'Ul+l)
corrector: delta

1
gJFfB6=O,O,B7+1)
0.5
0

Fig. 3a. Binary differential distillation.

E. M. ROSEN and R. N. ADAMS

730

Carnahen, et. al p 486


Steady State Heat Conduction in a Plate
initiator
counter
100
100
too
100
100
100
too
100
100

48.252
66.1045
73.0543
74.9291
73.0543
66.1045
48.2586
0

26.930:
43.1052
51.1835
53.6079
51.1835
4?.1052
26.9301
0

16.356;
28.2028
34.9668
37.1353
34.9668
28.2028
16.3567
0

10.294:
la.3823
23.3455
24.9999
23.3455
18.3823
10.2941
0

0
6.43TJ2
ii.6866
15.0331
16.1753
15.0331
tt.686B
6.43732
0

0
3.76832
6.89467
8.92666
9.62732
8.92666
6.89468
3.76832
0

0
1.74130
3.19687
4.15153
4.48259
4.15153
3.19686
1.74130
0

c7: 0
ca: arF(scss=o,o,(~a+c7+c9+oa)/4)
c9: a~wcss=o,o,cs9+ca+cto+o9v4~
cto: aIF(sCss=o,O,~BlD+C9+Cll+olov4)
Fig. 3b. Steady stateheat conduction in plate.

Unsteady-State Heat Conduction in a Slab


Parabolic Differential Equation
Backward Difference Formulation
Carnahan et al. Page 446
Initiator
Counter
dx
dt
la&da

1.00000
52.00000
0.10000
0.01250
1.25000
__________..~____.~.~~___~~----______--__,

<----~~~~~~____~---__..._----_._.__~

time
0.00000
0.01250
0.02500
0.03750
0.05000
0.06250
0.07500
0.08750
0.10000
0.11250
0.12500
0.13750
0.15000
0.16250
0.17500
0.18750
0.20000
0.21250
0.22500
0.23750
0.25000
0.26250
0.27500
0.28750
0.30000

0.00000

0.10000 0.20000 0.30000 0.40000 0.50000 0.60000 0.7OOOiI 0.80000 0.90000 1.00000

0.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000

0.00000
0.42054
0.59326
0.68069
0.73394
0.77129
0.80014
0.82381
0.84392
0.86138
0.87671
0.89026
0.90227
0.91295
0.92246
0.93092
0.93845
0.94516
0.95114
0.95647
0.96122
0.96544
0.96921
0.97257
0.97556

ct4:
C15:
C16:
C17:

0.00000
0.17751
0.32471
0.43131
0.51047
0.57247
0.62336
0.66654
0.70394
0.73673
0.76568
0.79135
0.81416
0.83445
0.85252
0.86860
0.88293
0.89569
0.90707
0.91720
0.92623
0.93427
0.94144
0.94783
0.95352

0.00000
0.07650
0.17391
0.26722
0.35033
0.42324
0.48730
0.54382
0.59387
0.63831
0.67782
0.71299
0.74430
0.77218
0.79702
0.81915
0.83887
0.85644
0.87209
0.88604
0.89846
0.90953
0.91940
0.92819
0.93602

0.00000 0.00000
0.03668 0.02620
0.10104 0.07965
0.17779 0.14975
~IJ.256670.22612
0.33235 0.30200
0.40248 0.37377
0.46631 0.43987
0.52385 0.49986
0.57544 0.55385
0.62157 0.60222
0.66275 0.64546
0.69948 0.68404
0.73222 0.71846
0.76141 0.74913
0.78741 0.77648
0.81059 0.80084
0.83124 0.82255
0.84963 0.84190
0.86603 0.85913
0.88063 0.87449
0.89365 0.88817
0.90524 0.90037
0.91558 0.91123
0.92479 0.92091

0.00000
0.03668
0.10104
0.17779
0.25667
0.33235
0.40248
0.46631
0.52385
0.57544
0.62157
0.66275
0.69948
0.73222
0.76141
0.78741
0.81059
0.83124
0.84963
0.86603
0.88063
0.89365
0.90524
0.91557
0.92478

0.00000
0.07650
0.17391
0.26722
0.35033
0.42324
0.48730
0.54382
0.59387
0.63831
0.67782
0.71299
0.74430
0.77218
0.79702
0.81915
0.83887
0.85644
0.87209
0.88603
0.89846
0.90953
0.91939
0.92818
0.93601

0
~lF(SCS5=0,0.5,(Cl4+SCS9*Bt~+SCS9*Ol~~/~t+2~SCS9~~
~IF(SCS5=0,0.5,(C15+SCS9*E16+SCS9*016~/~1~2*SCS9~~
@IF(SCS5=0,0.5,(C16+fCfP*B17+fCtP*Dt7)/(t~2*SCS9~~

Fig. 3c. Unsteady-state

heat conduction in a slab.

0.00000
0.17751
0.32471
0.43131
0.51047
0.57247
0.62336
0.66654
0.70394
0.73673
0.76568
0.79135
0.81416
0.83445
0.85252
0.86860
0.88293
0.89569
0.90707
0.91720
0.92623
0.93427
0.94144
0.94782
0.95351

0.00000
0.42054
0.59326
0.68069
0.73394
0.77129
0.80014
0.82381
0.84392
0.86138
0.87671
0.89026
0.90227
0.91295
0.92246
0.93092
0.93845
0.94516
0.95114
0.95647
0.96121
0.96544
0.96921
0.97257
0.97556

0.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
1.00000
t.00000
1.00000
1.00000
1.00000

review of spreadsheet usage

731

The Uavc Equation


Burden and Faircs p 597
Lankxia
Initiator
PI

;
3.14159

Time lncrnnent

<_.._____x ,~~~t$__.__..._.......-_;...........--._--.~....~....~~_~.....____....,

0.1

0.2

0 0.30901 0.58778
0 0.29389 0.55Wl
0 0.26999 O.b7SS2
0 0.18163 0.3bSb9
0 0.09Sb9
0.18163
0 0.00000 0.00000
0 -0.09Sb -0.1816
0 -0.1816 -0.3454
0 SO.2499
-0.b7SS

0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.b
O.bS

0 -0.2938

-0.5590

0.3

0.b

0.5

0.80901 0.95105
0.76942 0.9ObSO
0.65450 0.76942
O.bT552 0.55901
0.25000'0.29389
o.oOOoO 0.00000
-0.tb99 -0.2938
-0.bT55 -0.55W

1
0.9SlOS
0.80901
0.58778
0.30901
6E-17
-0.3090
-0.5877

-0.6545
-0.76%

-0.76%
*0.9ObS

Fig.

3d. The wave

0";:
0.8
0.85
0.9
0.95
1

can be particularly attractive in terms of ease of use


and overall speed of solution.
The solution of the Laplace equation using finite
differences is discussed by Camahan et al. [16] on
p. 483. The equation for the steady state temperature
distribution in a rectangular plate is:
a=T

a=T

(29)

2+7=0.

ay

In the example problem one face is at 100F and


the other three sides are kept at 0F. The finite
difference approximation is given by (Ax = Ay)
I.1 -

2wi,j

wi+

I.1

(Ax)=
wi.j-

I -

2wi,j

wi,j+ I

(Ax)=

= 0

(30)

(31)

or
wj,j =

wi-l,j+

0.7

0.8

0.9

0.95105 0.80901 0.58778 0.30901


0.9Ob50 0.76912 0.55901 0.29369

0.769b2

0.65650

0.55901
0.29389
~0.0000
-0.2938
-0.5590

0.57552
0.2b999
-0.0000
-0.2500
-0.b7SS

-0.76%

-0.6565

0.61552
0.3bSb9

0.16163
-0.0000
-0.1816
-0.3bSb

-0.b75S

0.21999

0.18163
0.09529
-0.0000
-0.09Sb
-0.1816

-0.2699

-0.9510 -0.9ObS -0.769b -0.5590 -0.2938


0 -0.3090 -0.5877 -0.8090 -0.9510
-1 -0.9510 -0.8090 -o.su'TT -0.3090
0 -0.2938 -0.5590 -0.76% -0.9045 -0.9510 -0.9ObS -0.769b -0.5590 -0.2938
0 -0.2bW
-0.4755
-0.65bS
-0.769b
-0.0090
-0.76% -0.6545 -0.b7SS -0.2694
0 -0.1816 -0.3454
-0.47SS
-0.5590 -0.5877 -O.SSW -0.b7SS -0.3654
-0.1816
0 -0.09%
-0.lt316
-0.2bW -0.2956 -0.3090 -0.2938 -0.2500 -0.1816 -0.09sb
0 0.00000 0.00000 0.00000 0.00000 -6E-17 -0.0000 -0.0000 -0.0000 -0.0000
0 0.09549
0.18163 0.25000 0.29389 0.30901 0.29389 0.26999 O-18163 0.09559
0 0.18163 0.34549
0.41552
0.55901 0.58778 O.SfWl 0.67552 O-34549 0.18163
0 0.24999 0.47552
0.65450
0.769b2
0.80901 0.769b2
0.6SbsO
O.b7S52
0.2b999
0 0.29389 0.55901 0.76942
0.90450 0.95105 0.90450 O-76962 0.5SWl O-29389
0 0.30901 0.58778 0.80901 0.9$105
1 0.9SQ)S 0.80901 0.5m-a 0.30901

0%
0.6
0.65

wi-

-0.8090

0.6

wi+l.,+
4

wi.j-I

wi,j+l

Figure 3b gives the spreadsheet solution to the


problem after 88 iterations. It compares almost exactly with the solution given by Carnahan et al. [16]
using a FORTRAN program after 88 iterations starting
with a field that is zero everywhere except at the
100F boundary. The spreadsheet solution required

equation.

but a single algebraic expression, equation (31),


which can be copied to all the other cells. Clearly the
effort for the spreadsheet calculation is an order of
magnitude simpler than the FORTRAN solution shown
by Camahan.
Hirschel[6] discusses the solution of the Laplace
equation as applied to the temperature distribution in
a rectangular oven with a rectangular open interior.
He points out the spreadsheet solution for the explicit
method used is equivalent to solving a set of linear
equations by relaxation, the terms being applied
to what is mathematically a modification of the
Gauss-Seidel iterative method [20]. It should he
noted how easily an irregular boundary shape can
be handled with the spreadsheet.
Wiggins [3] also discusses the Laplace equation
applied to a heat transfer problem. He argues that the
spreadsheet solution is more efficient than the use of
TK! Solver on the same problem.
Holland and Liapis [19] and Camahan et al. [16]
discuss the solution to the simple parabolic differential equation:
aT
-=-.

at

a=T

ax2

(32)

The finite difference form can be rearranged into


the backward-difference
form (i =x index, n = t
index).

E. M. ROSENand

732
~wi-,,+wi,.-,+~wi+,,.
Wi,n

1+2L

(33)

where
At
= (Ax)*
This is unconditionally stable.
Consider the initial and boundary
t=O:
t >O:

T=O

T = 1

at

conditions:

for

O<x<l,

x=0

and

R. N. ADAMS

Camahan et al. (p. 446) used the implicit method


to solve the problem. However, the solution given in
Fig. 3c using the backward-difference method gives
almost identical answers. Comparing
the given
FORTRAN program, the spreadsheet solution is again
an order of magnitude easier.
Consider the hyperbolic problem (the wave
equation):

(35)
x= 1,

(36)

(X, t) - 4 2

(x, t) = 0,

o<x< 1, 0 < t, (37)

and let 1 = 1.25.

Nitric Acid Flowsheet

- Rcklaitis p 367

Initiator

1.000

Counter

60.000
110.957
0.290

Burner Extent 1
Burner Extent 2

carponants

N2
HZ
NH3
02
HMO3

H20
NO

Stoichiometric Coefficient8
Converter
Burner-2
Burmr-1
-1.000
0.000
0.00000
-3.000
0.000
o.oOQoo
2.000
-1.000
-2.owoo
0.000
-2.000
-2.50000
0.000
1.000
o.ooooo0.000
1.000
3.ooow
0.000
0.000
2.ooow

SUn
N2
HZ
NH3

02
HNo3
HZ0
NO

sun
Converter Extent

Initial
400.000
400.000
0.009

0.000
0.000
0.000
0.000
800.000

N2
HZ
NH3
02
HMO3
HZ0
NO

Air
1343.000
0.000
0.000
357.000
0.000
0.000
0.000

sull

1700.000

N2
HZ
NH3
02
HMO3
HZ0
NO

SUM

Fctd

SO6
0.000
0.000
127.536
0.000
0.000
0.000
0.000

SW
1343.000
0.000
-0.000
114.362
,110.957
135.826
16.580

so2
94.010
0.000
0.000
0.000
0.000

127.536

1720.n5

94.010

Reac In
669.6%
765.217
0.000
0.000
0.000
0.000
0.000

so7
30.242
28.6%
0.000
0.000

1433.823
63.768

58.930

SO4
604.838
573.913
127.536

0.000
0.009
0.000

Sll
0.000
0.000
0.000
o.ow
110.957
135.826

0.000

0.000

0.000

220.000

1396.287

1119.g15

1473.942

Uater
0.000
0.000
0.000
0.000
0.000
llO.OW
o.ow

SOS
604.838
m.913
0.000

URNER IN
1343.000
0.000
127.536
357.000
0.000
0.000
0.099

s13
1248.999
0.000
0.000
114.362
0.000
0.000
16.580

110.000

1178.751

1827.536

1379.932

0.000
0.000
0.000
0.000

X:E

0.000

XZ
8:Z

Fig. 4.

GE
0:ow
0.900
0.000
0.000
0.009

steP
0.800
0.000
0.000
0.000
0.000
0.090
0.099

246.783

so3
574.5%
545.217
0.000
0.000
0.000
0.000
0.000

22i%,

EZ

2606
765:217
0.000
0.000
0.000
0.000
0.000

s12
1343.OaO
0.000
-0.000
114.362
0.090
0.000
16.580

x-z
0:Ooo
0.000
0.000

5.000
1.ooo
1.ooo
1.ooo
1.000
1 .ooo
1 .ooo

x2
668.606
765.217
0.000
0.990
0.009
0.990
0.000

da1
0.000
0.000
0.000
0.000
0.000
0.000
0.000

6L6
765:217

A review of spreadsheet usage

733

N?

Fig. 4. Nitric acid flowsheet.


~(0, t) = ~(1, t) = 0, 0 < 1

with boundary conditions


and initial conditions:

u(x, 0) = sin(I7x),

0< x< 1

and
$(x,0)=0,

O<x<l.

(38)

Burden [20], p. 597, shows that the finite difference


approximation is:
wi.n+l=2(1

-A2)wi,n

+12(Wi+I.n+Wi-I,n)-Wi,n-I

(39)

and
wi.,=(1-n21f(xi)+~f(xi+l)

g_ftxi-

I ) +

ktxh

t40)

where i = x index and II = t index.


In the above problem:
f(x) = sin(I7x)
g(x) = 0

7. FLOWSHEETING

The spreadsheet was early recognized as a way of


carrying out flowsheet material balance calculations
especially in the preliminary design phase. The
spreadsheet format giving stream compositions, temperature, pressure and specific gravity has been used
for many years by engineers on the PFD output of
process simulation systems [21].
Grulke [S] discusses the use of the spreadsheet for
material balancing in an educational environment
while Schmidt and Upadhye [7] discuss its use to do
flash calculations and to create templates for routine
industrial calculations. The most detailed examination of flowsheets and spreadsheets, however, is
given by Julian [4] who uses the spreadsheet to draw
the flowsheet with keyboard characters used as labels
in the cells. The material balance flows are placed to
fall in line with the graphical description of the
streams giving an integrated picture of the flowsheet
and material balance.
Many of the routine methods used with sequential
process simulators can be applied directly to carrying
out material balances on a spreadsheet.
Figure 4 gives the flowsheet for a nitric acid plant
taken from Reklaitis [22]. The ammonia reaction:
N2 + 3H, = 2NHS,

a=2

takes place in the converter (fractional H, converted


is 0.25) and the following two reactions take place in
the burner with all the NH, reacted:

k = 0.05 = l/20
h = 0.1 = l/10
(41)

NH3 + 2Or = HNOJ + H,O,


2NH, + 2.502 = 2N0 + 3H,O,

Figure 3d gives the solution to this problem which


compares well with the analytical solution:
14(x,t) = sin(I7x)cos(217)t.

(43)

(42)

(44)

with the yield of nitric acid equal to 87%. Five


percent of the gas stream is purged in the splitter and
93% of the nitrogen is removed in the separator. The

E. M. RCSEN and R. N.

134

extent of each of the reactions can than be determined


as a function of the input to the respective reactor.
For the converter:
extent = 0.25 (input flow of hydrogen)/3.
For the burner:
extent 1 = 0.87 (input flow of ammonia),
extent 2 = 0.13 (input flow of ammonia)/2.
To implement this flowsheet on a spreadsheet, a
sequential modular approach [17] can be used. The
inlet to the reactor is guessed (torn) and each
stream is then calculated in turn. The Bounded
Wegstein method described above is used to force
convergence. The solution is essentially achieved after
some 60 iterations.
Note that since each stream must be calculated in
order, it is necessary to calculate the spreadsheet in
a column-wise fashion. Starting with the Reac In
stream (the torn stream) the order follows with S04,
SOS, S06. . . S02. The extents of reaction are shown
preceding the unit in which the extent will be used.
It is easy of course to change the conversion of the
hydrogen in the reactor to study its effects on the
flowsheet.
In this example the entire flowsheet can be viewed
on a single page. This is a very appealing feature of
the spreadsheet.
8.

PROJECT

ENGINEERING

Spreadsheets can be useful in project engineering


for at least three functions: material balances, economic comparisons and lists.
A. Material balances

As indicated in the previous section, the spreadsheet can be an attractive means of carrying out a
material balance. A particular advantage of the
spreadsheet is that the implications of changing any
flow rate can quickly and easily be seen. However,
there are three other potential uses of an electronic
spreadsheet in material balancing that have not been
widely used or discussed. These are in trace components, in matching existing operations and in
communicating material balances to others.
After the design material balance has been completed, it is easy to add columns and rows for trace
components. More than half of the major problems
contamination,
environ(corrosion, by-products,
mental) with process units have been associated with
very minor constituents in the input stream. A few
parts per million of chlorides, fluorides, arsenic, etc.
can and do lead to major corrosion, contamination
or environmental problems. Everything that goes in
comes out, sooner or later in some form. A fraction
of a compound per day of chlorides has built up and
left a process as a colored corrosion product contaminating a distillation column stream. Less than a
pound per day of arsenic has provided an environ-

ADAMS

mental waste disposal problem and/or a nonfood


grade by-product. The ease with which a spreadsheet
material balance can be modified to account for
minor constituents can help to reduce the effort to
communicate and understand the total operation.
An extra column and added rows can be inserted
into a spreadsheet material balance (daily, batch,
weekly) when trying to use sample data and measurements to synthesize an actual material balance. Doing
this work manually is very time consuming. The extra
column has been used to provide a balance of input
minus output for all compounds and/or elements.
Since it is not possible for the measurements and
analysis to provide complete closure, the extra
column provides a synopsis of the lack of closure. If
two elements have errors, do they have a common
analysis/sample and/or quantity measurement? This
work has been done by hand to provide some insight
as the variation in the process balance, but the
spreadsheet can easily and quickly handle many
scenarios.
For study the material balance does not have to
balance! How many times have we been misled (or
misled others) unintentionally
because a simple assumption was required in order for the material
balance to balance. Review of differences and the
data used may require adjustments of some of the
compositions and flows. After a few trials the balance
can be used for development of a design balance for
say a debottlenecking project.
Design material balances may start out using research data in the form (g/100 g product) but the
spreadsheet can quickly and easily be modified or (in
some systems) be linked to provide balances for
lb/100 lb product or lb per batch. These balances are
useful for communicating with research or design
groups. Balances figured as lb/av. day (365daysfyr)
are useful for communicating with planning, utility
and service groups. The lb/av. day serves as a reference base for projects that contain batch and continuous steps. Upper management may find a balance
in lb/yr very useful. An average day or design day
balance can provide a good vehicle for communication with the waste treatment design group. Balances used for communication
with the environmental groups can also include rows for TOC (Total
Organic Carbon), COD (Chemical Oxygen Demand)
and tons sulfur dioxide/yr.
A full set of communication
balances could
include:
DESIGN, M lb/yr (that didnt change during the
project and could include sulfur dioxide in tons/yr)
AVERAGE,
lb/day (to tie together batch/
continuous steps and could include TOC and/or
COD values)
DESIGN, lb/batch (which would change as the
batch size, time cycle or on-stream time charges)
DESIGN, lb/h (which would change with changes
in on-stream time)

A review of spreadsheet usage

DESIGN,

lb/cwt (to reference research practice)

All these balances can readily be obtained by using


matrix multiplication
(Lotus l-2-3, version 2) or
array multiplication (excel).
B. Economic

comparisons

The original development and use of the electronic


spreadsheet was for economics. The business literature and software tutorials provide a comprehensive
description of the use and development of economic
comparisons. Indeed the vast majority of projects do
require economic justification and the electronic
spreadsheet makes it easy to document and reproduce
almost any model used by management/accounting
for comparisons. Grulke [5] gives an example of a
cash flow analysis for a methanol project. What if
questions are very easily posed and answered by the
spreadsheet.

735

Conversion of spreadsheet data to/from database


data is not always straightforward. Lotus l-2-3 does
supply a translate program to access and store data
in dBase II and dBase III. However, since probably
80% of all data used for communication between
disciplines is in form of simple lists, there is often little
motivation for converting spreadsheet data to a form
for input to a database. Lotus l-2-3 also has macros
to read external files into the spreadsheet. Though
possible, one needs to evaluate whether the convenience of interfacing with data bases and files is
worth the effort.
Spreadsheet programs do provide a means of generating ASCII files of their output which can be sent
through communication networks. If, for example, a
laser printer is not available, a PRN (print) file in
Lotus l-2-3 may be generated of the spreadsheet
output and sent to a network on which the laser
printer is available

c. Lists
4 large portion of the communication
requirements of a project consist of lists [lo]. Lists are
prepared for:

Equipment Information
Utility Requirements
Instrument Indexes
Manpower Assignments
Motors
Cable Schedules
Pipeline Schedules
Power Summaries
Control Summaries
The current spreadsheets provide simple, straightforward generation of lists with or without summation of columns. Use of these lists may possibly
eliminate the need to use and learn a separate database program.
9. INTERFACES

TO GRAPHICS

AND DATABASES

Most of the currently available spreadsheet programs do provide some direct or indirect link to a
data graphing program. Generally the data values for
the abscissa and values of the ordinate are pulled
from consecutive cells in the spreadsheet. A large
portion of the available graphs are arithmetic only,
however, i.e. they do not directly supply capabilities
for semi-log or log-log plots. (Excel and MS Chart
do have this capability, however). As a result data
analysis may be more convenient with one system
than another. For example plotting vapor pressure vs
temperature on arithmetic coordinates will not spot
inconsistent data as well as plotting it on a semi-log
scales, Likewise an arithmetic plot of vapor and
liquid compositions may not highlight the pinch
point in a distillation. Of course one could convert
the data by setting up an additional set of cells and
taking the log of the numbers before calling the
arithmetic graphing program.

10. CONCLUSIONS

Spreadsheets make a powerful tool available to the


chemical engineer to carry out a wide range of
functions. Its general format and ease of use make it
an attractive alternative to the use of a FORTRAN/
BASIC program
on the PC or a FORTRAN/BASIC
program on a mainframe. However, the point at
which it ceases to have an advantage over other
methods is clearly often a function of its availability,
accessibility and knowledge of its user. As with other
tools it can be very productive or only marginally
useful as the situation warrants.
REFERENCES

1. Retailers Microcomputer Market Place, Vol. 1, No.

2.

R. R. Bowker Co., New York (1985).


2. D. hen-Aaron, With 20/20, Access Has Success In Its
Sights. Informafion Week Jan. 13 (1987).
3. E. G. Wiggins, VisiCalc vs TK! Solver in engineering
problem solving. Computers in Education Division of
ASEE VI, 04 (1986).
4. F. M. Julian Flowsheets and Spreadsheets. AIChE Mrg,

Houston (1985).
5. E. A. Grulke, Using spreadsheets for teaching design.
Chem. Enann Education uW3) (1986).
6. S. M. Goldfarb and R. *Hirschel, Spreadsheets for
chemical engineering. Chem. Engng St(S) (1985).
I. W. P. Schmidt and R. S. Upadhye, Material balances
on a spreadsheet. Chem. Engng 91(M) (1984).
8. S. Selk, Spreadsheet software solves engineering problems. Chem. Engng 90, 51 (983).
9. E. H. Rasmussen, Virtues of VisiCalc. Gem. Engng 90,
5 (1983).
10. R. N. Adams, Impact of Microcomputers In Process
Engineering: Creativity or Confusion?. AIChE Mrg,
New Orleans (1986).
11. C. J. Sowa, Engineering calculations on a spreadsheet.
Chem. Engng 94, 67 (1987).
12. A. J. Owens, Book review: VisiCalc for Science and
Engineering (by S. R. Tost and C. Pomemacki).
ACCESS July/August (1984).
13. F. S. T. Hsiao, The Gauss Quadrature Numerical
Integration-A
comparison
of the Programming

736

14.
15.
16.
17.
18.

E. M. ROSEN and R. N.
Method and the Spread&et
Method. ACCESS
September/October (1986).
F. Shied, Numerical Analysis. Scbaums Outline Series
(1968).
E. J. Henley and J. D. Seader, Equilibrium-Stage Separation Operations in Chemical Engineering. Wiley, New
York (1981).
B. Camaban, H. A. Luther and J. 0. Wilkes, Applied
Numerical Metho&. Wiley, New York (1969).
E. J. Henley and E. M. Rosen, Material and Energy
Balance Computations. Wiley, New York (1969).
A. L. Myers and W. D. Seider, Introduction to Chemical

19.
20.
21.
22.

ADAMS

Engineering and Computer Calculations. Prentice-Hall,


Englewood Cliffs (1976).
C. D. Holland and A. I. Liapis, Computer Metho&&
Solving Dynamic Separation Problems. McGraw-Hill,
New York (1983).
R. L. Burdeh and I. D. Faires, Numerical Analysis, 3rd
Edn. Prindle. Weber & Schmidt. Boston (19851.
E. M. Rosenand A. C. Pauls, Computer AidedCbemical Process Design: Tbe FLOWTRAN System. Comput.
them. Engng 1, 11 (1976).
G. V. Reklaitis, Introduction to Material and Energy
Balances. Wiley, New York (1983).

Das könnte Ihnen auch gefallen