Designing a Separations Process Without VLE Data

1
Designing a Separations Process Without VLE Data

• by Thomas Schafer - Koch Modular Process
  Systems, LLC
• This presentation utilizes as its example a
  problem presented to KMPS by a pharmaceutical
  client who was incinerating a valuable solvent
  stream.
• Components to be separated - Toluene from
  Acetic Anhydride

2
Problem Definition - Customer Objectives
Physical Properties
Table 1

• Table 2

3
Approaches That Should Be Considered When VLE
Data is Not Available

• Engineers may choose from the following
  alternatives
• 1. Assume the compounds form an ideal solution,
  which means vapor pressure
• vs. temperature data can be used to predict VLE.
  This is generally acceptable
• when the compounds are closely related, such as
  members of a homologous series.
• Examples include linear alcohols, paraffinic
  hydrocarbons, aliphatic substituted
• benzene (benzene, toluene, xylene), polymeric
  glycols.
• 2. Find VLE data for an analogous system, one
  that contains one of the compounds of
• the pair. The 2nd compound should be closely
  related to the other compound of the
• pair, containing the same or similar structure
  and functional groups. An example of
• this technique would be to use liquid activity
  coefficients of benzene and ethanol to
• predict VLE for benzene and propanol.
• 3. Develop VLE data for key pairs of components.
  Set up a VLE apparatus to test
• each component pair. Data developed this way can
  be regressed to provide
• interaction coefficients which can then be used
  in a process simulator to
• explore a range of design alternatives.

4
Analogous Systems DataThe literature was then
searched for VLE data for an analogous system.
Datafor benzene-acetic anhydride (Figure 1A) and
cyclohexane-acetic anhydride(Figure 1B) were
found in Dechema. These data clearly indicate
non-ideality.

• Figure 1B

Figure 1A
5
Conclusions Drawn From Analogous System
DataAfter analyzing the analogous system data,
an engineer should expect that the
toluene-acetic anhydride system will exhibit
similar but more severe non-ideality because
tolueneboils closer to acetic anhydride than
either cyclohexane or benzene.Two initial
predictions of the VLE curve for toluene-acetic
anhydride were made by usingthe benzene-acetic
anhydride and cyclohexane-acetic anhydride NRTL
coefficients, with toluene vapor pressure data.
The predicted curves are plotted in Figure 2.
The data indicate that azeotropic behavior is
probable.
Figure 2

• Since the plotted data showed significant
  non-ideality it was decided that generating VLE
  data was
• the preferred alternative.

6
VLE Apparatus To generate the VLE data, a simple,
inexpensive apparatus was constructed of glass
and polytetrafluoroethylene components, similar
to the design shown in Figure 3. It is
critical that the apparatus yield exactly one
theoretical stage.

• Figure 3

7

• Calibration of VLE Appartus with Known System
• After assembly of the apparatus shown in Figure
  3, a known system was checked to ensure
• that the apparatus will yield exactly one
  theoretical stage. The known system should boil
• in a similar temperature range to the
  experimental system. Internal condensation must
  be
• avoided as it can result in up to two theoretical
  stages in the test apparatus. Data for
• ethanol-water was then compared to literature
  data as shown in Table 3. As can be seen,
• the test system results in almost exactly one
  theoretical stage.

Table 3
8
Table 4

• Experimental Toluene-Acetic Anhydride Data
• Using the calibrated experimental setup, VLE data
• for toluene-acetic anhydride was generated over a
• range of compositions. The data is shown in
• Table 4. More data was collected at the toluene
• rich end of the curve due to predictions from the
• analogous systems that there may be an azeotrope
• or possibly an asymptote in the VLE curve.

The data was then regressed and NRTL
coefficients were derived. A smooth curve was
then developed for the system using the NRTL
coefficients as shown in Figure 4.
Figure 4
9
Figure 5

• Azeotrope Found
• A minimum boiling azeotrope was calculated at 96
  mole (95.6 wt) toluene and 4 mole
• acetic anhydride. Figure 5 is an enlarged plot
  of the toluene-acetic anhydride VLE curve in
• the range of 95-100 mole toluene.

Because the components form an azeotrope, it is
not possible using simple distillation
to separate the components into pure acetic
anhydride and pure toluene.
10

• Process Design
• Given the feed composition, a single distillation
  column is adequate to recover relatively pure
• acetic anhydride as a bottoms product and a
  mixture which approaches the azeotropic
• composition as a distillate. Approximately 93
  of the acetic anhydride was recovered in
• one pass through this distillation column. Some
  design parameters for the distillation
• column are

• Theoretical stages 19
• Packing Type Flexipac 2Y
• Packed Height 33 ft.
• Reflux Ratio 1.4
• Distillate Composition 91 wt Toluene
• Bottoms Composition 99 wt Acetic Anhydride

The toluene in the distillate was recovered by
water extraction to remove the small amount of
acetic anhydride. Figure 6 is a photo of a
modular process system that was built to perform
the separation described. The resulting process
recovers 92 of the acetic anhydride and 99 of
the toluene from a stream that was previously
incinerated.
11
Figure 6

• Modular Separation System
• for Recovery of
• Toluene Acetic Anhydride