Combination of Equations (18) and (20) gives

For convenience, Figure 12 plots

(1 + g;)
as a function of B.
Equation (21) reduces to E* ED"as /3 + O and E* + nK7' as
-+

p + which are proper limiting forms.

T h e dashed line on Figure 10 is thc prediction o f 1q:quatioii
(1 5) for the initial conditions and temperature variation of
Run 3 3 .

Acknowledgment
The early portion of this research was sponsored by Texas Butadiene
and Chemical Corporation and the final portion by National Science
Foundation Grant GP 376. The computations were done at the M.1.T.
Computation Center.
References
Nomenclature
( 1 ) Howninn. W. 13.. Sc.11. Thrsi.;. Chcm. En$?.. h1.I.T. ( l95J ).
A particlv radius, f t .
= ( 2 I Weisz,P. H., and Prater, C.1).. Advancc.s i n Catalpi.;. 6 ( 1954 I .
( 0 ) Hagrrhanrnrr, W'. A.. and L w . H., Triins. ASME. 69, 779 ( 1947 ).
C, bulk oxygen concentration
= ( 4 I Haldc~miin,R . 11.. and Hiitty, h1.C.. J . Phy.;. Clicwi., 63, 489 ( 19.501.
9 = effective clilfusivity, ft.*/hr. ( 5 ) Johnson, hl. E'. L., and !vl21ayl;nid, I I . C., Irid. Ens. CIi<m.. 47, 127
(l9:55).
E, = actiLVation cnergv, B.t.ii./Ib. niolc ( 6 ) Pasing, W. F., A.1.CIi.E. Jorirn;il, 2, 71 (19.56).
E* = apparent activation energy
e = retardation factor, diiiiensionlcss * * *

Experimental Verification of the Mathematical

Model for a Continuous Stirred.Tank Reactor'
RlCHARD I. KERMODE' and WlLLlAM F . S T E V E N S :

This paper reports the dynamic behavior of a particular Lr comportement dynamiqur d'un r&acteurcontinu avec
continuous stirred-tank reactor system, which has been agitation, suite i I'dtude diiaillir tant expiriment;rle qut-
investigated in detail both experimentally and o n an analog mathbmatique i I'aide d'nne caIeuIatrice analcbgique, est
computer. The t w o sets of data were taken under compar- prCaenti.. Ida roniparaison e s t possible car leu t1onnht.s sont
able conditions to facilitate comparison. The reaction prises sous des conditions similaires. On a choisi la ri.aetiun
between formaldehyde and ammonia was chosen for study, entre la formaldihyde et I'ammoniar i r a u w d r la fnrte
because of its high heat of reaction and its signifirant chaleur de ri.action ct do taux d r ri.artion suli~t:inticlB la
reaction rate at room temperature. Agrerment between the tenipbrature anibiante. Ides rCsultats thdoriques 4.1 esl&ri-
experimental data and the computer results was quite good, mentuux sont en acrord et 1;i majorit6 drs d b v h t i o n * s'rx-
with most deviations being explainable as arising from the pliqurnt par I'utilisation d'one rxprrwsion approximiitivt.
use of an approximate form for the kinetic rate exprrssion. du taux de ri.action.

\\ere i i i o r c easily solved. I Io\\.cvcr, such P procedure inti-oduccs

D tiring the past few years, research in process dynamics
has emphasized the dc\&pmcnt and use of mathcmatical
models for many types of chemical engineering cq~iil~iiiciit",'~:l).
dout)ts a s to the applical)ilit!. of the solution ot)taincd from the
linear prol)lciii a s a solution to the origin;il iioii-Iiiicar diffcrcntial
Most of these niodcls have involved non-linear differential equation, cspcciall!. in the rcgion f.u. rcniovcd from tlic oi-iginal
equations, which ofrcn arc very dificult to solve. In sonic cases, stcadv state.
the original equations \vcre linearized t)y expanding each of the l'hc best general incthoti for solving non-lincar oi-din;ir\,
non-linear ternis in a 'Taylor's sctics and truncating after t\vo differential equations o f the t!.pc that arise in process (i! n.inii&
terms, after which the resulting linear differential equations studies involves the use o l an xialog coiiipiitcr. \ \ itli s t a n h d
............................................................................... non-lincar components (iiiultipiicrs, tiinction gciicr;itors. ere.).
'Manuscript rcwioed May 5; acceptcsd lunc 22, 1964. a niodcrn analog computer is capt)lc ot' solving almost mi\.
Tarnegie Institute of Technology. Pittsburgh, Pa., U.S.A. non-lincar differential equation s>'stciii \\it11 good accurac\'.
3Northwcstcrn Uniccrsity, Eoanston, I l l . , U.S.A.
Contribution from Northwestern Unioersity. Eoamton, I l l . , U.S.A. Hence, it should be possible to prc'iict the d!.naiiiic t)chavi;r

The Cnrtadinn Journnl of Cherriicnl Engineering, A p r i l , 1965

 a .-

0
i!
I
i
oo-\ 0 --!

0 O . !
0
L L-
JW 1200 uoo
.

RECIPROCAL TEMFZRATLUE PI( XlO'l

3403
10 30 40 so
I
TEMPERATURE CC)
Figure 1-Experimental valuer of reaction rate conrtent.
Figure %Heat of readon data-formaldehyde plm am-
monia to five hexaxnine.

of any piece of process equipment, if a good mathematical model simultaneous reactions, some of which may be reversible, rather
is available for analysis. than the single, irreversible reaction assumed in correlating the
The present paper summarizes the results of an experimental data. Therefore, it was decided that statistically fitted straight
investigation of the dynamic behavior of a particular continuous lines would be used to represent both sets of data, as indicated
stirred-tank reactor, along with an analog computer solution on the two plots.
of the non-linear dynamic model of the reactor system('). The The second portion of the experimental work consisted of
two sets of data were taken under comparable conditions to an examination of the transient behavior of the reactor. The
facilitate comparison and evaluation of the accuracy of the following four independent variables were stepped, both simul-
mathematical model. The reaction between formaldehyde and taneously and one at a time: inlet reactant temperature, cooling
ammonia to give hexamine was chosen for study, because of water temperature, cooling water flow rate, and total reactant
its high heat of reaction and its significant reaction rate at room flow. Experimental data resulting from these studies are plotted
temperature. This reaction follows the stoichiometric relation, on Figures 3, 4, 5 and 6.
4NHs + 6HCHO + N,(CHz)s +
6Hz0. A major part of In obtaining the experimental data for the phase plane plot
the investigation involved the measurement of kinetics data and of Figure 3, the reactor was first filled with formaldehyde
heats of reaction for the reaction being studied, after which the solution of the desired temperature and composition. At time
reactor system was operated in the unsteady state to give a t = 0, the recorder was started and all variables were stepped
phase plane plot. Subsequent analog computer analysis of the to their desired values. If each independent variable was always
mathematical model then gave results which could be compared stepped to the same value, then the final steady-state composition
to the experimental data. and temperature in the reactor should always have been the same.
As can be seen from the plot of the data, this came very close to
Experimental Work being true. It should be noted that some of the lines on Figure
The reaction system studied in the laboratory consisted of 3 appear to cross each other. This is quite as expected, since
a small stainless steel reactor, equipped with an internal heat a three-dimensional phase plot is necessary to describe completely
exchanger('). Since it was assumed that perfect mixing took the system under study because of its three dependent variables-
place in the reactor, several precautions were taken to insure this. temperature and two compositions.
Four evenly spaced baffles helped the mixing. The impellers Figures 4,s and 6 present the results of conventional transient
on the mixer shaft were twisted in a manner to prevent circulation analysis, with one variable being stepped at a time. The data
in one direction. In addition, the reactant inlets and outlet were were obtained by first bringing the system to the steady state
positioned to prevent channeling. The speed of the impeller mrres nding to the center point on Figure 3, and then stepping
was about 1800 r.p.m. and the mixer power was very high one op"the variables to a new value. The composition and tem
with respect to the reactor volume. Reactants were fed con-
tinuously, through metering orifices, and the product was
ature history of the reactor are presented for the period rom F-
the instant of the step to the attainment of a new steady state.
removed by displacement. The reactor temperature was
recorded continuously, and the reactor concentration was
followed by withdrawing periodic samples from the outlet Development of Mathematical Model
stream. These were neutralized in a known excess of acetic In the original derivation of the equations, several simplifying
acid and analyzed for formaldehyde by a srandard procedure(Q. assumptions were made('). These include: perfect mixing in
The first part of the experimental work was the determination the reactor, no heat losses, all physical properties the same as
of kinetics data and heats of reaction. These were obtained water, a thirdader, irreversible reaction, and reaction velocity
while operating the reactor at steady state. Although hexamine constant given by the Arrhenius equation. Equations (1) and
has been produced from formaldehyde and ammonia for many ( 2 ) are the two mass balances.
years, little basic data were available in the literature, except
for some reaction rate studies carried out at much lower re-
actant concentration~(~.7.8).
The kinetics and heat of reaction data obtained in the present
study, over a range of inlet compositions, are presented in
Figures 1 and 2. Reaction rate measurements at temperatures
below 25°C. gave reaction velocity constants which seem to
be somewhat erratic. In addition, the heat of reaction appears
to change by about 50% over the temperature range examined.
These results are probably due to the existence of several

The Canadian Journal of Chemical Enginseriw, April, 1965 69

 130-

120- ...... COMPUTER RESULTS

+AV a 0 0 EXPERIMENTAL RESULTS

110-

IOO-

-
U
g 090-
.
A

:0 8 0 - ---__
0
L, 06
-
0 70-
TOYE (MINUTES)
0
0
* 060- Figure &Experimental results-step of cooling water flow
rate.
z
e
050-
0
x
=: 0 4 0 -
0 30-
T,, S I E R STEP
El
€2
10 9
25P I
.
€3 31 8 A
020- c* 365.
€5 *50 D
I,. .
C !
0 10-

-
4 4 0.

420t
I
.- €3 !
0 OOL
E2

2 360
-- --- __ ~ --,
Figure 3-Phase plane plot-xperimental and computer 3401

results. $ 3201
3oot
0- I
i
2
L~
3 4
A _ _
5 6
---I I - * - % ~

TlYE (YINUTLSI

If a log mean A 7 is used, the cooling water outlet temper-

ature appears, which is not readily available. T,, was eliminated Figure %Experimental results-step of inlet formaldehyde
by the manipulation shown in Ikpations (4) to (8). temperature.

\Vhen all of the numbers representing the physical quantities

in Kquations ( I ) , ( 2 ) and ( 3 ) are substituted, lquations ( I Z ) ,
( I 3 ) and (14) result. These equations,

10 d CA = +2.49 - 1.226C~- 200r.. . . . .

~
(12)
dr

- Tci
T__ dCp
- (,.\I . .............. , . . . (6 ) 10 - = +.1.87 - 1.226Cl. - 300r. . . . . . , . .(13)
T Tc. ~
dr
d 7'
r" - 1 ~~ ~~ = $35.90 -- 0.1225T - 20r(0.1210T - 18.87) ~-
( Tco - T<,)= ~

If
( f - 7'ct). . . . . . . . . . . .( 7 ) dr
0.1051(T - 293.2) . . . , . . . . . . (14)
along with the rare expression, were programmed on the analog
computer(4). T h e equations were speeded up bv a factor of
20, to keep the total computer time low. T h e computer results
The heat of reaction has been approximated by the straight arc presented in Figures 3, 7, 8 and 9 . In all cases, the same
dotted line shown in Figure 2 . The equation for this line is runs were made as were carried out experimentally.
shown below :
-AHA = 16.61 X lOJ + 0.1210(T - 293.2) X l o 3 . .. . ( 9 ) Comparison of Results
expressed in calories per gram mole of ammonia reacted. The In trying to compare the two sets of data, some difficulty
expression for the reaction velocity constant is given by lqua- was experienced in defining "good agreement". The computer
tion (10). results had an internal consistency which was lacking in the
3090
~
system data. &is an example, the computer always started at
k = 1.42 X 10%- exactly the same composition and temperature. T h e actual
data, however, had a certain amount of scatter in the steady
The reaction rate expression is Fquation ( 1 1). state composition and temperature before a variable was stepped,
I = kcAcF*. . . . . . . . . . . . . . . . . . .(11) even when all conditions appeared to be the same from run to run.

70 The Canadian Journal of Chemical Eng.ineering, April, 1965

 1
,.I
1 " " " " "

::
0 66

52 ~
eC I SECOND
so 0
'
*not
;.so /
_ -0 2
J -
.
.0
84
w 42 0
BI

BZ

83

32 0
3 0 Oo
I 2 3 4 5 6 I B 9 80
TIME (MINUTES 1

Figure 7 -Computer results -step of cooling water flow

Figure &Experimental results-step of total reaetant flow. rate.
1

i
TlYE (MINUTES1

Figure &Computer results-step of inlet formaldehyde Figure 9-Computer results-step of total reactant flow.
temperatures.

It has already been noted that sonic of the phase planc thc computer and experimental results is very good. Once again,
trajectories cross in the actual system data. This is also present the major disagrcemcnt is at t = 0. -411 the final experimental
in the computer results, and can bc explained in the same manner steady state temperatures are within 1 "C.of the comparable
as beforc. From Figure 3, it can be seen that the ratc of conver- computer run, some being virtually the same, as can be seen by
sion of formaldehyde at lour temperatures (below 25'(:.) is not reference to Figures 4 and 7.
as rapid for the simulated system as for the rcal system. This Figures 5 and 8 show rhc results of stcpping the inlet
is clearly shown by comparing the maximum height of compar- formaldehyde trmperature. Both sets of results show a time
able trajectories. In all cases, the computer compositions reached lag before a change in composition takes place. T h e temperature
a higher valuc. This means that the computer predicts a lower comparison does not seem to be as good as in previous runs.
reaction ratc, which is not surprising considering the limited T h e final steady state difference between runs L<1 and E S is
amount of kinetics data. . i t teniperatures below 25"(:., thc only 8OC. This smaller total diffcrence tends to magnify the
straight line approximation used for thc computer simulation effect of scatter at t = 0.
is considerably different from the actual ratc constant data, T h c final step investigated was a step of the total reactant
1:igure I . At thc higher rcnipcratures (25°C. and above), a How, Figures 6 and 9. Each reactant was stepped by the samc
similar deviation is present, but not in all the trajectories o r to amount at the samc time, thus keeping the mixed inlet coniposi-
the same extent. tion the same. T h e diffcrence between rhesc runs cannot be
T h e steady state agreemcnt bctwecn thc cxperimental and accounted for by analysis o r computer errors. In examining
computer results is within the accuracy of the analog computer. the differences, the most plausible explanation is inadequate
T h e steady state as predicted by the computer is 37.3'C. and kinetic dam. T h c computer predicts a steady-statc composition
0.637 moles of formaldehyde per liter. If all the results for the of 0.470 moles of formaldehyde per liter for run B3, Figure 9.
actual runs are averaged, the tcmperature and composition in T h e actual data give 0.608 moles of formaldehyde per liter
the reactor are 36.1 'C. and 0.640 moles of formaldehyde per for the same run. Although the differcncc in composition is
liter, respectively. T h e ( x - y ) plotter used in this study reads significant, the difference in conversion is very small. Thus,
somewhat in error, such that, if the lowest temperature wcre one might expect that the temperature agreement would be
set exactly, then the upper temperatures would be at least better, and this is exactly what Figures 6 and 9 show.
I.O°C. high. A t 3 7 O C . , a correction of 0.4OC. was subtracted In examining the computer results, it is apparent that the
from the plotter reading before comparison, but this correction computer always predicts a greater change in concentration
has not been applied to the drawings. than the actual data do. O n e method of explaining the large
For the case of a step of cooling water flow rate, the scatter computer k is to examine the batch kinetics data. T h e normal
of the composition data is very small, and the agreement between method for utilizing batch data is to plot concentration versus

T h e Canadian Journal of Chentical Engineering, April, 1965

time at various temperatures. Evaluation of the two constants in the system. Good kinetics data are at best difficult to obtain,
in the Arrhenius equation requires that conversion data at two and represent the most important part of most reactor designs.
temperatures be available. As an example, knowing the con- However, even with limited kinetic data, the present work
versions at time t, for different temperatures, and the general has confirmed the usefulness of the conventional mathematical
rate expression, an equation can be developed that will predict model for predicting the dynamic behavior of a chemical reactor.
the conversion for any temperature at a given time. It is
important to remember that the rate expression must be correct, Nomenclature
and the reaction velocity constant must be represented by the A = area of the heat exchanger, cni.z
.4rrhenius equation. ,411 the previous data, including the kinetics C, = heat capacity of a n y stream in or out of the reactor,
data, were taken with the same retention time in the reactor. 1 cal./gm.'K.
C, = heat capacity of the cooling water, 1 cal./gm."K.
In the series of runs where the total reactant flow was stepped, EIR = energy of activation divided bv R. 3086°K.
a different retention time resulted. The reaction velocity -AH +
= heatof reaction. 16.61 X lo3 0.1210 ( T - 293.2) X
constants were all determined at one retention time. If the lo3 cal./gm. moie of ammonia'reacted '
constant based on another retention time were different, then
the computer results would not agree with the experimental data. = reaction velocity constant, 2-
C" c2
Unfortunately, no additional data are available to confirm these = frequency factor, 1.419 X lo3
conclusions; however, when the complex nature of the kinetics = inlet ammonia concentration, 4.06 moles/liter
is considered, the conclusions do not seem unlikely. = concentration of ammonia in the reactor, moles/liter
= inlet formaldehyde concentration, 6.32 moles/liter
To summarize, the transient data agree very well for all = c o n c e n t r a t i o n of f o r m a l d e h y d e i n t h e r e a c t o r ,
the cases where the conditions of the steps are such that they moles/li ter
do not alter the conditions under which the kinetic data were = inlet ammonia flow rate, 1.50 cc./sec.
= inlet formaldehyde flow rate, 1.50 cc./sec.
determined. For the two cases where this is not true, the phase = inlet cooling water flow rate, 2.65 cc./sec.
plane plot and the step of total reactant flow, the agreement = reaction rate, gm. moles of ammonia converted/liter ser.
might be improved by more complete kinetic information. The = density of all streams, 1 gm./cc.
stirred tank reactor has flexibility in determining kinetic data. = a n y temperature, OK.
= inlet cooling water temperature, 293.2"K.
In addition, heat of reaction data are obtained with no additional = outlet cooling water temperature, "K.
work. This gives an added check on the accuracy of the model = inlet ammonia temperature, 293.2"K.
for an exothermic or endothermic reaction. = inlet formaldehyde temperature, 293.2"K.
= time, seconds
= speeded-up time, t/20
Conclusion = overall heat transfer coefficient, cal./cm.* "K. sec.
= reactor volume, 490 cc.
This study has compared the results of an experimental
investigation of the unsteady behavior of a continuous stirred-
tank reactor system with the data from a companion analog References
Kermode, R . I., M.S. Thesis, Northwestern Univ., Evanston, Ill.
computer solution using a mathematical model for the reaction (1959).
system in question. The constants, such as the overall hear Kermode, R . 1. and Stevens, M'. F., Can. J. Chem. Eng., 81 (1961).
van Heerden, C., Ind. Eng. Chem., 45, 6, 1242 (1953).
transfer coefficient, heats of reaction, etc., necessary to program Kermode, R . 1.. Ph.D. Thesis, Northwestern Univ., Evanston, 111.
the mathematical model were all evaluated by operating the (1962).
Bose, S.. J. Indian Chem. Sac.. 34, 6 6 3 (1957).
system at steady state. The values of thcsc constants proved Bauer, E. and Ruetschi, W., Helv. Chim. Acta. 24, 7 5 4 (1941).
to be very satisfactory for dynamic comparison of the modcl Boyd, M. L. and Winkler, C. A,, Can. J . Research, 25B, 3 8 7 (1947).
Richmond, H. H., hlyres, G. S . and Wright, G. F., J. Am. Chem.
and system. As has been previously mentioned, the kinetics Sac., 70, 3659 (1948).
data were the most questionable part of the work. These dara
are very important, as they account for most of the non-linearitics * * *

CORRECTION

Operating Lines in Cascade Separation of Binary Mixtures

S-T.HWANG and KARL KAMMERMEYER

Equation ( 2 2 ) on page 39 of the February 1965 issue

of T h e Canadian Journal of Cheinical Engineering should
contain a plus sign rather than a minus sign, and should
read as follows:

Y"+l = yu.. . . . (22)

:(jYj)
I-1 +1
Editor.

72 The Canadian Journal of Chemical Engineering, April, 1965