CSTR- Example where viscosity of the feed and product are not the same.

Note: This is a fairly complex problem with detailed description. Re-read the problem

description, if necessary.

Consider the liquid phase reaction 10 M  P, i.e an approximation of polymerization. The

reaction is believed to be first order. We would like to conduct this in an isothermal CSTR

(assumed to be ideal), but first we want to know the kinetics. Specifically, the reaction order is

known to be 1, but the rate constant is not given. We conduct the reaction in an isothermal batch

reactor, at the same temperature as we would in the CSTR. We monitor the power required as a

function of time, and from that, we need to estimate the kinetics. While the viscosity changes

with conversion, the density does not.

The following data is given to us. It is known that mixing is good if the agitator is maintained at

180 rpm, independent of the vessel or agitator size, and fluid properties. The batch reactor dia =1

m, agitator dia = 0.5 m, liquid level in the tank is 1 m high, fluid density is 900 kg/m3, viscosity

of the monomer is 10-3 Pa-s and that of completely converted polymer is 104 Pa-s. The molecular

P
weight of monomer is 86 g/gmol. For this setup, the power number and Reynolds
 N 3 D5

D2 N  100
number are related by N p  for NRe < 200, and roughly 0.5 for NRe > 100. We
 N Re

monitor the power required to maintain stirring at 180 rpm, and the results (power vs. time data)

are given below.

 t (min) Power (W)
0 380
10 1129
20 2268
30 3416
40 4573
50 5738
60 6913

We have also sampled the product (mixture containing monomer and polymer) at 10 min interval

and the product at 30 min appears to be the best choice. We want to produce this product in a

CSTR at a rate of 100 kg/min. (i) What should be the volume of CSTR? (ii) How much power

will it consume, for agitation?

Solution:

We make an assumption that the viscosity of a mixture can be expressed as a simple function of
mole fraction. This is not correct, but …estimating the viscosity of mixtures is challenging; that of
polymer mixture is even more difficult. For binary mixtures of relatively simple fluids, one can
1 1 1
use Gambill method  mix3  xa a 3  xb b 3 , (proposed in 1959).A more recent method (2000) is by
Refutas described below.. We can calculate viscosity blending number (VBN) VBNi
=14.534×ln(ln(νi+0.8))+10.975, and the VBN of the liquid mixture is then calculated as follows:

VBNmixture =∑xi×VBNi

The kinematic viscosity of the mixture can then be estimated using the viscosity blending number
of the mixture using the equation below.

VBNmix=14.534×ln(ln(νmix+0.8))+10.975

We will ignore these methods and use a simple (even if incorrect) formula:
mix  MFa a  MFb b where MFa and MFb are mole fractions. The idea is this. There is a
relationship between conversion and the rheological properties of the fluid (the fluid may not be
even Newtonian), and it is complex. Since our focus is on reactor design, we will use a simple
relationship. If an accurate and proper relationship is given, if we solve this in a programming
environment, we can follow the same methodology, but use the proper equation.

From the diameter, liquid level height, density and molecular weight of monomer, we can
calculate and show that
Volume of the tank =785.4 lit.
Mass of fluid = 707 kg
Initial number of moles = 8.22 kgmol

At any conversion, the viscosity is going to be linearly related to the mole fraction. First we will
get the relationship between conversion and mole fraction.

Since 10 moles of M are needed to get 1 mole of P, the number of moles of M and P are related
1 1
by N P  N Pinit   N Minit  N M    N Minit  N M 
10 10

Total number of moles = NM + NP =

1 N 9  9 
N M   N Minit  N M   Minit  N M  N Minit 1  x 
10 10 10  10 

Mole fraction of M = NM/(NM+NP) 

1  x 
 9 
1  x 
 10 

Viscosity on the other hand will be

MFM 103  1  MFP  10 

1  x  10 3

 0.1 x  10 
4 103  103 x
 9   9   9 
1  x   1  x  1  x 
 10   10   10 

D2 N  0.52  3  900 675

Reynolds number is  1  0.9 x   1  0.9 x  .
 3
10  10 x 3
0.001  1000 x

Therefore, conversion data can be extracted from the power measurements. (Note that at < 3 %
conversion, NRe becomes equal to 200).

t (min) Power Power NRe Mu Mole x

(W) Number fraction
0 380 0.5 ≥ 200 ≤ 20 ≥0.999 ≤ 0.02
10 1129 1.49 67.2 10 0.9989 0.01
20 2268 2.99 33.5 20.16 0.9980 0.0198
30 3416 4.5 22.23 30.3631 0.997 0.0296
40 4573 6.0 16.6 40.6459 0.9959 0.0392
50 5738 7.56 13.2 51.0106 0.9949 0.0488
60 6913 9.10 11 61.4575 0.9939 0.0582
The first result is not usable (why?); we can use only the latter results.

t (min) x 1-x ln(1-x)

10 0.01 0.99 -0.01005
20 0.0198 0.9802 -0.02
30 0.0296 0.9704 -0.03005
40 0.0392 0.9608 -0.03999
50 0.0488 0.9512 -0.05003
60 0.0582 0.9418 -0.05996

Fit the equation ln 1  x   kt and we get k as 0.001 min-1.

ln(1-x)
0
0 20 40 60 80
-0.01

-0.02

-0.03 ln(1-x)

-0.04 Linear (ln(1-x))

-0.05
y = -0.001x
-0.06
R² = 1
-0.07

At 30 mins, the conversion is about 3 % in the batch reactor. We need 3% conversion in the
CSTR too.

CSTR Design equation

QCAin  VkCAin 1  x   QCAin 1  x 
Mass flow rate = 100 kg/min, density = 900 kg/m3, so volumetric flow rate = 111 lit/min.
x  0.03 
V Q  111    3432 lit
k 1  x   0.001 0.97 

We can go with CSTRs in series (to make the total volume smaller, and each reactor volume
small), or we can use multi-stage impellers to get good mixing. Let us say that we go with a
single stage impeller and single tank, just a scaled up version of the batch reactor, but operated in
CSTR mode. Usually the ratio of “impeller dia to tank dia” is 0.5. We keep the ratio of liquid
height to tank dia as approximately 1. Therefore, we get tank dia is 1.6 m, height is 1.7 m.
Impeller dia is 0.8 m. We need to maintain N = 3 rps (180 rpm).

At this conversion, the viscosity is 30.36 Pa-s. Reynolds number is 133.4. Power number is
0.7497. Power required is P  N p  N 3 D5  0.7497  900  33  0.85  18kW . Note that the design
equation does not have viscosity term, but the power requirement equation has conversion term.