2nd Mercosur Congress on Chemical Engineering

4th Mercosur Congress on Process Systems Engineering

BATCH REACTOR SCALE-UP ASSISTED BY STATISTICAL

EXPERIMENTAL DESIGN

O. Prieto*, G. Prieto, C. R. Gay

Departamento de Ingeniería Química – Facultad de Ciencias Exactas y Tecnología – Universidad Nacional de
Tucumán - República Argentina

Abstract. It is well known, that in general reactors scale-up is quite impossible to afford by using only
theoretical methods. Laboratory and/or pilot plant data of the reactions are necessary. The chemical reactions
are indirectly affected by the reactor type and scale in a way very difficult, if not impossible to quantify,
because changes in the reaction global velocity and/or changes in the products distributions may evolve. The
chemical reactions are influenced by physical processes such as mass transference and heat transference, in
general controlled by the reactor shape and dimensions.
Statistical experimental designs methods represent a powerful tool due to they minimize the number of
experimental runs, consequently costs are diminished, and additionally they reduce the levels o risks. The
scale-up methodology was centered in two different procedures, the gradual reactor size increment and the
use of a big reactor increasing the volume of the reaction system. The reaction selected was the catalytic
oxidation of sucrose in a glass Batch reactor. The behavior of the system could be easily visualized by
registering only the temperature evolution.

Keywords: Scale Up, Statistical Experimental Design, Chemical Reactors.

1. Introduction
The scale-up or change of scale in a chemical reactor involves multidisciplinary scientific-
technical strategies searching for practical solutions to the problems that usually appear.
Then, the use of methods based on principles of similitude, mathematical modeling or a
combination of both, as well of statistical methods, must be the result of a careful analysis to
minimize the costs of the scaling up study and to carry out an efficient investigation to get
information highly reliable.
The chemical reaction constitute one of the more complex process in the field of the
chemical engineering because it is usually associated with mass and heat transfer
phenomenon where the scale has a special signification.
In the field of the chemical reaction engineering and the scaling up of the reactor size, the
starting point is represented by the lab-scale basic research, following with the bench-scale,
and finally the pilot plant-scale or eventually the industrial-scale for low volume production
processes (e.g. multi-purpose reactors like for the pharmaceutical industry).

*
To whom all correspondence should be addressed.
Address: Departamento de Ingeniería Química. Facultad de Ciencias Exactas y Tecnología. Av Independencia 1800.
(4000) San Miguel de Tucumán. República Argentina.
E-mail: [email protected]
1
 2nd Mercosur Congress on Chemical Engineering
4th Mercosur Congress on Process Systems Engineering

In general, the methodologies employed for the scaling up do not follow common preset
lineaments despite the starting point is similar for all of them, (Bisio and Kabel, 1985),
(Trambouze, 1979), (Wei, 1990). Although the resources and the results for the scaling up are
dissimilar, the most used techniques can be classified into five groups: 1) direct experimental
approach, by which the effects of all the variables are tested over the full range of interest; 2)
mathematical models, by which differential equations can be formulated for the combined and
simultaneous processes of chemical reaction, diffusion, heat transfer and pressure drop; 3)
dimensional similitude applied to chemical reactions; 4) statistical methods, by which a
sequential program is established, based on the use of the results of prior scale as a starting
point for the following scale and 5) combined methods. (Walas, 1959), (Perry, 1992).
To employ statistical methods and in particular the statistical experimental design methods
as a procedure for scaling up, results of particular interest due to their inherent properties of
reduction of the number of experiments, reduction of costs of the changing scale process and
reduction of the levels of risks. (Box et al., 1978), (Montgomery, 1991).
The selected process for this study is the Catalytic Oxidation of Sucrose to produce Oxalic
Acid. This selection was done taking into account the perspective of the industrial
implementation in view of the regional programs of sucrose industry diversification.
Oxidation of carbohydrates, e.g. glucose, sucrose, starch, dextrine and cellulose, with nitric
acid to form oxalic acid, is an old and well known process, but suffers disadvantages of low
yields, difficulties in the reaction control and high cost of recovery of the nitrogen oxides
formed in the reaction. The oxidation of sucrose is carried out in the presence of sulfuric acid
and a vanadium salt as catalyst, between about 65.5°C and 71°C while maintaining the
carbohydrate in excess of the nitric acid during the oxidation process. The reaction of
carbohydrates with nitric acid is highly exothermic, particularly in the earlier stages of the
reaction, and the heat generated must be dissipated to maintain the reactive mass within
desired temperature limits, because sudden uncontrollable and violent temperature rises with
concomitant gas evolution. Since oxidation of lower nitrogen oxides to higher oxides of
nitrogen is also an exothermic reaction, conversion of lower oxides of nitrogen directly within
the reactive mass, places an additional burden on the reaction temperature control system,
making the effective temperature control even more difficult and greatly increases the
potential danger of the reactor (Fuchs and Watson, 1970).

2
 2nd Mercosur Congress on Chemical Engineering
4th Mercosur Congress on Process Systems Engineering

The main goal of this research is the application of the statistical experimental design
techniques to find strategies and working methodologies to implement the changes of scale in
a batch reactor for a process with a low volume of production with the aim to infer about the
final scale represented by almost a pilot plant.

2. Experimental
2.1. Experimental Setup
Three different round-bottom batch reactors, immersed into an open jacket with room
temperature water as a refrigerating bath in a close circuit of circulation were used. Each
reactor was equipped with a magnetic stirrer or a laboratory mixer. The process was
completed with three gas absorption bottles, manometers, thermometers and a plunger pump
with variable flow to feed the nitric acid. All the chemicals used were laboratory grade.
Synthetic liquor of the following composition was prepared: sulfuric acid (100%), 50% of
the total liquor; oxalic acid (on anhydrous basis), 5% of the total liquor; catalyst iron (as
Fe2(S04)3), 0,05%; V205, 0,005%. Vanadiun was taken as sodium vanadate. A 60% sugar
solution was mixed with the above mentioned liquor. The temperature of the reaction mixture
was adjusted as desired, and nitric acid was slowly added. The reaction mixture was stirred
vigorously. After the end of the reaction, or when reaction slowed down sufficiently, the
reaction mixture was cooled and kept overnight. It was finally cooled into the refrigerator.
The crystals of oxalic acid were separated, filtered, and weighed. The percentage yield of
oxalic acid obtained at various initial temperatures, with different concentrations on nitric
acid, and the ratio of mother liquor to the 60% sugar solution is tabulated, (Deshpande, Vyas,
1979).

2.2. Statistical Experimental Design Setup

The complete factorial experiment techniques were selected.
- Design 1: For the selection of variables, a first design was programmed to make the
screening of the variables according to their sensibility in the process. In this design, with a
reactor of 500 gr of capacity the studied variables and their respective range were: To (initial
temperature) at the levels of 30-45-60 in centigrade degrees; R (sucrose/mother liquor/nitric
acid) at the levels 3.5-5.5-7.5 and Q (flow rate) at the levels of 0.4-1.2-2 in mL/min.

3
 2nd Mercosur Congress on Chemical Engineering
4th Mercosur Congress on Process Systems Engineering

- Design 2: The reactor volume was of 4000 gr (eight times greater than in Design 1). The
selected variables and their respective range were: To (initial temperature) at the levels of 30-
45-60 in degrees centigrade; R (sucrose/mother liquor/nitric acid) at the levels 3.5-5.5-7.5 and
Q (flow rate) was kept constant fixing only the total time of nitric acid fed in two hours.
- Design 3: The reactor volume (V) was selected as a variable at the levels of 3000-9000-
15000 gr. The other selected variables and their respective range were: To (initial
temperature) at the levels of 30-45-60 in centigrade degrees and R (sucrose/mother
liquor/nitric acid) at the levels 3.5-5.5-7.5. The flow rate (Q) was kept constant fixing only
the total time of nitric acid fed in two hours.

3. Results and Discussion

The selected output variable was the process yield defined as:
Yield = Oxalic Acid filtered after reaction/(Oxalic Acid in the feed + stoichiometric
Oxalic Acid to be obtained)
Table I shows the experimental results obtained for the Design 3 as an example of all the
experimental runs carried out.

Table 1. Experimental results for Design 3

Initial Ratio Max

Volume Yield
Run Temperatura (sucrose/mother Temperature
(gr) (%)
(ºC) liquor/nitric acid) (ºC)
1 30 3.5 3000 35.24 45.5
2 60 3.5 3000 32.22 71.5
3 30 7.5 3000 42.97 43.0
4 60 7.5 3000 43.39 65.8
5 30 3.5 15000 34.21 96.5
6 60 3.5 15000 26.63 97.4
7 30 7.5 15000 47.91 64.0
8 60 7.5 15000 52.57 94.0
9 45 5.5 9000 60.90 74.5
10 45 5.5 9000 45.24 77.4
11 45 5.5 9000 43.36 78.4
12 45 5.5 9000 57.35 71.8

From the analysis of the experimental results, when the selected output variable (considered
the variable of interest) is the yield, the input variable temperature has a marked positive
effect. This means that when the initial temperature increases, an increment of the yield
should be expected. Nevertheless, due to the characteristics of the reaction, an increment of
the temperature may trigger the reaction temperature to an incontrollable state. In

4
 2nd Mercosur Congress on Chemical Engineering
4th Mercosur Congress on Process Systems Engineering

consequence, the maximum temperature reached, the time needed to reach this maximum, as
well as the maximum temperature gap are followed and registered during all the runs. Figure
1 shows the temperature profiles for one of the experimental runs.

90

80
70
Temperature (ºC)

60
50

40
30
Design 1
Design 2
20
Design 3
10
0
0 20 40 60 80 100 120 140

Time (min)
Fig. 1. Temperature profiles for Designs 1, 2 and 3

Regarding the time elapsed to reach the max temperature, it depends only on the initial
temperature in a negative way, which means that when the initial temperature increases, the
time to get the peak of temperature decreases.
When the experimentation is moved to bigger reactor volumes, the initial temperature has
almost no effect on the yield. The amount of catalyst only for its highest level cause a slight
increment on the yield.
Nevertheless, the maximum temperature reached will be higher when the reactor volume
increases, as the initial temperature increases. This effect becomes less significant as the
amount of catalyst diminishes. Figure 2 shows the temperature profiles for one of the
experimental runs from Design 3.

5
 2nd Mercosur Congress on Chemical Engineering
4th Mercosur Congress on Process Systems Engineering

100
90
80
70
Temperature (ºC)

60
50
40
30 3000 gr
20 9000 gr
10 15000 gr
0
0 20 40 60 80 100 120 140
Time (min)

Fig. 2. Design 3: Temperature profiles for the different reactor volumes

100

Yield

15000 gr
min
Volume
Ratio
3000 gr
max

Fig. 3. Response Surface of the Yield as a function of Ratio (sucrose/mother liquor/nitric acid) and reactor
volumes in Design 3. To = 45 ºC

6
 2nd Mercosur Congress on Chemical Engineering
4th Mercosur Congress on Process Systems Engineering

Figure 3 is a response surface representing the effects on the yield of the Ratio
(sucrose/mother liquor/nitric acid) and Volume of the reactors assayed.. The graphic shows
no effect of the ratio and the reactor volume on the yield. To increase the Ratio means to
increase the amount of catalyst, but this has no effect on the reaction yield for the different
reactors assayed. Similar graphics are obtained for the other initial temperatures (To).
An independence of the yield for an increasing reactive mass, as the amount of catalyst
increases means that probably there are no mass transfer limitations on the conversion rate.
These results prove that the reactor agitation was good enough combined with the use of
sufficiently small particles of catalyst which offers another technique to escape from the mass
transfer-limited regime.
The influence of the temperature on mass transfer is manifested by its effects on the
physical properties involved in the dimensionless groups, but the total effect is very small
compared with the effect on the chemical reaction expressed by the Arrhenius equation,
(Walas, 1959).

100

Yield
9000 gr

3000 gr
0
500 gr
60 ºC
min
To
Ratio
30 ºC
max

Fig. 4. Response Surface of the Yield as a function of Ratio (sucrose/mother liquor/nitric acid) and Initial
Temperature for reactor volumes in Designs 1, 2 and 3.

7
 2nd Mercosur Congress on Chemical Engineering
4th Mercosur Congress on Process Systems Engineering

Figure 4 is a response surface which represents the joint effect on the yield of the Ratio and
the To (initial temperature) for the Designs 1, 2 and 3. The graphic shows that in general the
process exhibit almost the same effects for the all the different reactor vessels, but it is
difficult to assure no mass-transfer limitations for each one. Only if the enlargement of the
reactor dimension is made in the same vessel, just increasing the reactive mass, the mass
transfer limitation can be controlled and consequently avoided.

4. Conclusions
Statistical experimental designs methods represent a powerful tool due to they minimize the
number of experimental runs, consequently costs are diminished, and additionally they reduce
the levels o risks. The scale-up methodology was centered in two different procedures, the
gradual reactor size increment and the use of a big reactor increasing the volume of the
reaction system. The second procedure seems to be the better one because the mass transfer
limitations can be controlled and consequently avoided. Another advantage of the method is
that the behavior of the system can be easily visualized by registering only the temperature
evolution.

References
Bisio, A., Kabel, R. L (1985). Scale-up of Chemical Processes, Ed. John Wiley.
Box, G.P., Hunter, W.G., Hunter, J.S. (1978). Statistics for Experimenters, John Wiley and Sons, New York.
Deshpande, S.D., Vyas, S.N. (1979). Ind. Eng. Chem. Prod. Res. Dev., Vol.18, 1, pp.69-71.
Fuchs, G.H., Watson, W.E. (1970). Patent Specification, C 07 c 55/06 international classification.
Montgomery, D.C. (1991). Diseño y Análisis de Experimentos, Grupo Editorial Iberoamérica.
Perry J.H. (1992). Manual del Ingeniero Químico. Sexta Edición (Tercera Edición en español), Tomo II, Ed. McGraw-Hill,
Mexico.
Trambouze, P. (1979). Chemical Engineering, September 10, pp. 122-130.
Wei, J. (1990). Chemical Engineering Science, Vol. 45, 8, pp. 1947-1952.
Walas, S. M. (1959). Reaction Kinetics for Chemical Engineers, Ed. McGraw-Hill, New York.