Separation of n-hexane,

cyclo-hexane and n-heptane.

Chalce Mhlanga
u19250020

2022-06-03
Abstract

Distillation is a method of separating chemical substances based on differences in their

volatilities in a boiling liquid mixture, the main purpose of distillation is to increase the
purity of the desired substance and increase its monetary value, hence it is used for many
commercial processes.

This report explores the design of a distillation column to separate a ternary mixture
of N-Hexane, C-Hexane, and N-Heptane, to achieve a fractional recoveries of 95% N-
Hexane, 98% C-Hexane, and 98% N-Heptane. The Wilson thermodynamic model was
used to correlate the experimental data, which made it possible to calculate the total
number of stages and the optimum feed stage using DWSIM. A sortcut column(DWSIM)
was used as a preliminary design then a ChemSep column was used for the final design.
Sensitivity analysis was performed to optimize the design, followed by the calculations
of the column diameter, height and the internls. The obtained answers agree with the
theory and and the design can be improved by using advanced softwares for simulation.

Figure 1: Separation Process

ii
Contents

Abstract ii

Nomenclature iv

1 Thermodynamic data and Model 1

1.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Physical data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.3 Thermodynamic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Column Design and Calculations 2

3 Sensitivity Analysis 5

4 Column size and internals 6

5 Heat Exchanger Design 8

6 Conclusions and Recommendations 8

A Appendix A.1

iii
List of Figures

1 Separation Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii

2 TXY diagram for C-Hexane and N-Heptane. . . . . . . . . . . . . . . . . 2

A.1 Mass Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.1

A.2 Colunmn 1 results from DWSIM simulations . . . . . . . . . . . . . . . . A.2

A.3 Colunmn 2 results from DWSIM simulations . . . . . . . . . . . . . . . . A.2

List of Tables

1 Provided information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Outcomes from shortcut Method . . . . . . . . . . . . . . . . . . . . . . 4

3 Number of feed and plates obtained from different multipliers . . . . . . 5

A.1 Summary for the calculation of Eo . . . . . . . . . . . . . . . . . . . . . . A.2

Nomenclature

αi relative volatility of component i with respect to Heavy Key

B, bot Bottoms

D, dist Distillate

N Number of plates/stages

Nf eed Feed position

P∗ Vapour Pressure

xi,B Mole fraction of component i in the Bottoms

xi,D Mole fraction of component i in the Distillate

FRi,B Fractional Recovery at the Bottoms

iv
FRi,D Fractional Recovery at the Distillate

µi Viscosity of component i

ϕ liquid phase packing parameter

C-Hexane Cyclohexane

CMO Constant Molal Overflow

F Feed

HK Heavy Key

HNK Heavy Non Key

LK Light Key

LNK Light Non Key

NK’s Non Keys

v
1 Thermodynamic data and Model

1.1 Assumptions

• CMO is assumed:

– Distillation column is adiabatic. Specific heat changes are small compared to

latent changes.
– Heat of vaporization is constant, independent of concentration. Saturated
vapour and liquid lines are parallel on the H-x-y diagrams.

• All the NK’s are non-distributing .

• Constant relative volatilities.

• Well-insulated distillation columns. In the preliminary design, it is assumed that

the column is 100% efficient, no loss in energy.

• Incompressible liquid and vapour(density and viscosity remains constant)

• Atmospheric pressure is 101.325 kP a . CP values are constant(not a function of

temperature).

• Ideal mixing occurs, due to lack of appreciable forces of repulsion or attraction

amongst unlike molecules.

• The latent heat of vaporization is constant, the heat of vaporization is the function
of temperature.

1.2 Physical data

Table 1: Provided information

Key Compound FRi,D xi,D FRi,B xi,B Zf eed Boiling point P ∗ (kP a)
LNK Hexane 0.95 0.15 69 20.1
LK CycloHexane 0.98 0.70 80.7 12.9
HK Heptane 0.98 0.15 98 6.13

The vapour pressures were obtained from Fisher Scientific (2022), at a temperature of
298.15K.

1
1.3 Thermodynamic Model

The Wilson thermodynamic model presented the best fit against experimental data. The
mixture being separated contained organic compounds which range form C6 C7. Since
the C of the mixture lies between C4 C18, it is best to use the Wilson model(Suppes,
2002). To verify this information DWSIM was used to plot (T)xy diagrams of C-
hexane and N-heptane. Peng-Robinson (PR), Soave-Redlich-Kwong (SRK) models were
added for comparison. Then VLE experimental data, for C-Hexane and N-Heptane, was
obtained from Jan et al. (1994) and plotted with the data generated by DWSIM. The
Wilson model gives the best (T)xy fit, it can be seen in Figure 2 that the graph generated
by the Wilson model(black in colour) is almost indistinguishable form the experimental
data points’ path.

Figure 2: TXY diagram for C-Hexane and N-Heptane.

2 Column Design and Calculations

Short Cut Method

On the first distillation column C-Hexane was chosen as the Light key and N-Heptane
was chosen as the heavy key, this is because when comparing the change in the boiling
point temperatures, N-Hexane and C-Hexane have the smallest change in temperature,
∆T = 11.7 K, compared to C-hexane and N-Heptane, ∆T = 17.2 K. As a result its
more likely for C-Hexane to go to the distillate than the bottoms. N-Hexane is our Light
Non-Key. On the second distillation column Hexane was LK and C-Hexane was the HK.

The Fenske equation was used to estimate Nmin (1) at total reflux, followed by calculation
of ϕ(2), with the condition that αHK < ϕ < αLK , Vmin (3), D(4), Lmin (5) and Rmin (6)

2
using the Underwood equation then NActual was determined from the Gilliland Correlation
P∗
followed by NF eed . The relative volatilities were calculated using αi−ref = ∗i (Wankat,
Pref
2012). For the LK a fractional recovery of 99% on the distilate was chosen on the first
column. The results and mass balance are summarised in table 2 and figure A.1.

 
(F RA )dist ·(F RB )bot
ln [1−(F RA )dist ][(F RB )bot ]
Nmin = (1)
ln αAB
C
X αi−ref · F · zi
∆Vf eed = Vmin − V min = F (1 − q) = (2)
i=1
αi−ref − ϕ
C
X αi−ref · (D · xi )dist
Vmin = (3)
i=1
αi−ref − ϕ
X
D= (D · xi )dist (4)
Lmin = Vmin − D (5)
Lmin
Rmin = (6)
D

 
L
RActual = was determined by multiplying Rmin by 1.5(Wankat, 2012), followed
D Actual
by the calculation of the abscissa(7). The Gilliland Correlation(Wankat, 2012) and
equation 8 was used to determine NActual . The minimum feed state was calculated using
equation 9 and actual feed stage from equation 10. The obtained answers are reflected in
table 2. The calculated answers were compared with the results obtained from DWSIM
simulations, figure A.2 and figure A.3 and there was quite a noticeable difference in the
second column and little differences in column 1. These differences may be due to the
chosen multiplier. For the rest of the report the calculated answers were chosen over
DWSIM one’s because they have less trays, therefore less costs.

RActual − Rmim
abscissa = (7)
RActual + 1
ygraph + Nmin
NActual = (8)
1 − ygraph
h   i
ln xxHKLK
/ zzHK
LK

dist
NF,min = (9)
ln αLK−HK
NActual · NF,min
NF,actual = (10)
Nmin

3
 Table 2: Outcomes from shortcut Method

Culumn 1 Column 2

Minimum Reflux Ratio 0.923 4.39

ϕ 1.09 3.07
Minimum number of stages 10.7 20.0
Actual Number of stages 20.9 33.4
Optimum Feed Stage 10.3 10.2
Distillate Flowrate( kmol
h
) 97.4 17.2
kmol
Bottoms Flowrate( h ) 17.7 80.2

Efficiency

Efficiency was calculated using equation 11, which is found on Wankat (2012):

Nequil
Eo = (11)
Nactual

However Eo is given by equation 12, equation 13 was used to calculate the viscosity of
the mixture in the feed, where xi is the mole composition of component i in the feed.
The viscosity of pure components was found at the temperature Tave , which is given by
equation 14. Tdistillate and Tbottoms were taken from DWSIM simulation. The results are
summarised in table A.1

Eo = 0.52782 − 0.27511log10 (αµ) + 0.044923 [log10 (αµ)]2 (12)

ln µmix = x1 ln µ1 + x2 ln µ2 + x3 ln µ3 (13)
Tdistillate + Tbottoms
Taverage = (14)
2

Using the O’Connell correlation(Wankat, 2012), which is described by equation 12, the
efficiency was determined to be ≈ 69% for column 1 and ≈ 63% for column 2. Nequil =
14.4 ≈ 15 and Nequil = 21.0 respectively.

4
3 Sensitivity Analysis

The optimization of Reflux Ratio is of vital importance profit-wise, at minimum reflux

ratio the column requires an infinite number of trays, and consequently the capital cost
becomes infinite, but the energy cost will be the least(Abolpour and Mohebbi, 2014). As
R increases the number of trays decrease until they reach a minimum and so does the
cost. To find the optimum R, without changing the FR, a multipliers that ranges from
1.5 − 4 were used and N, Nf eed were calculated. The R value that gave the minimum N
and Nf eed was taken as the optimum R, Minimum value of Nf eed means less piping and
cost. The R optimum for both columns is:

3 + 3.25 3.5 + 3.75 + 4

M ultiplieroptimum1 = M ultiplieroptimum2 =
2 3
= 3.13 = 3.75

therfore R:

Roptimum1 = 3.13 × Rmin Roptimum2 = 3.75 × Rmin

= 2.66 = 16.4

Table 3: Number of feed and plates obtained from different multipliers

Column 1 Column 2
Multiplier N Nf eed N Nf eed

1.5 20.9 10.3 33.4 10.2

1.75 18.8 9.21 30.1 9.18
2.00 17.4 8.53 28.2 8.59
2.25 16.4 8.06 26.9 8.19
2.50 15.7 7.71 25.9 7.91
2.75 15.1 7.44 25.2 7.69
3.00 14.7 7.22 24.7 7.52
3.25 14.4 7.04 24.2 7.38
3.50 14.1 6.90 23.8 7.27
3.75 13.8 6.78 23.5 7.18
4.00 13.6 6.67 23.3 7.10

5
4 Column size and internals

Column dimensions

To calculate the diameter, flooding velocity must be taken into account. Flooding velocity
is determined by using Fair’s flooding correlation, which takes into account mass flows,
densities and surface tension.(Wankat, 2012). The column diameter and tray hydraulic
calculations were done using imperial sytem because all equations were developed using
imperial units. For the calculations, tray spacing of 9 in was first assumed in order to
calculate the correlation factor using equation 16.

The flooding velocity was calculated using the equation 15, where Csb,f is the capacity
factor and it is calculated by equation 16, with a = 1.1622, b = 0.56014 and c = 0.18168.

 α 0.2 r ρ − ρ
L V
uf lood = Csb,f (f t/s) (15)
20 ρV
log10 Csb,f = −a − b log10 Flv − c(log1 0Flv )2 (16)

At the top of the column, Flv is given by equation 17. ρV and ρL were found to be
WL
2.96 kg/m3 , 698 kg/m3 respectively, from DWSIM. M MV = M ML therefore =
WV
Lf low
Surface tension was also found in DWSIM to be 15.5 dynes/cm. The calculations
Vf low
yielded... Flv = 0.0500, Csb,f = 0.182 and uf lood = 2.11 f t/s. The diameter was then
calculated from equation 18, with f raction and η chosen to be 0.75 and 0.8 respectively, as
advised by Wankat (2012). The diameter of the column was found to be Dia = 5.41 f t.
The same procedure was followed and the diameter at the bottom was found to be
Dia = 4.99 f t

r
WL ρV
Flv = (17)
WV ρL
s
4V M W V
Dia = (18)
πη(f raction)uf lood 3600

6
Packing and Trays

Trays were chosen over packing on the design because the diameters of the column are
greater than 3 f t, since the liquids that will be separated are non-corrosive, the usage
of packing is of no importance. Because both diameter of the column are greater than
2.5 f t, staged columns will be used.

Sieve trays were used for the design because they are easy to manufacture, cost effective,
easy to clean, have low entrainment and reduced fouling(Wankat, 2012). Since the designs
are not proprietary, anyone can build a sieve tray column, the efficiency is good at design
conditions. The tray will have a cross flow pattern, these systems contain downcomers
that allow the liquid to flow from plate to plate and weirs that retain the liquid on the
plate, giving it enough time to be heated and evaporate before moving to the next plate.
The material chosen to build the tray is carbon steel as it is most commonly used and
available. The picture of a tray layout is available on Wankat (2012) page 400, figure
10-20 and the tray support on page 385, figure 10-7.

Downcomers and Weirs

A segmented vertical downcomer with a notched weir will be used For the downcomer of
the sieve, in order for the trays to handle the different liquid flow rates. One advantage
is that it is cheap to build and easy to install(Wankat, 2012). Notched weirs have the
least weeping compared to other weirs.

Entrainment and Weeping

Entrainment was calculated using equation 19. Flv was calculated above and Ψ was read
from the graph, figure 10.7 in Wankat (2012), and it’s numerical value is 0.070. This
yielded an entraiment of 38.4 lbmol/h

ΨL
e= (19)
1−Ψ

The whole operation must be above the weeping point, to prevent the occurrence of
weeping. Liquid will not drain through the holes if the sum of heads due to surface
tension and gas flow are greater than the function 20, x = hweir + hcrest + hgrad . As long
hδp + hsigma ≥ the given correlation, weeping won’t be an issue.

7
 0.10392 + 0.25119x − 0.021675x2 (20)

5 Heat Exchanger Design

Shell and tube heat exchanger was chosen for the condenser because it is the most versatile
heat exchanger, it can handle a wide range of pressures and temperatures. Kettle reboiler
was used because it is cost friendly and suitable for fluids with low viscosity . It was
assumed that the temperature of the liquid that cools the reboiler decreases by 10◦ C and
the steam that enters the reboiler fully condenses, i.e saturated liquid comes out. To
estimate the required heat transfer area, equation 21 will be used, where: Q is the rate
heat to be transferred, H is the overall heat transfer coefficient and LM T D is the Long
Mean Temperature difference.

Q = A × H × LM T D (21)

Q was taken from DWSIM simulation and...

6 Conclusions and Recommendations

The answers obtained from the calculations seem convincing because most of the values
lies within acceptable values, and most of them meet the theoretical requirements. There
is always a room for improvement, and if the design can be done again better results
can be obtained, this time the sensitivity analysis must be done using a trustworthy
software. This will decrease human error, increasing the reliability of the results. Better
softwares, like ASPEN, are recommended for simulating the design because DWSIM have
inconsistent answers and the answers generated by DWSIM differ a lot from the answers
found by using the shortcut method in the second column.

8
References
Abolpour, B and A Mohebbi (2014). “Optimization of the reflux ratio of benzene-toluene
stage distillation columns by the Cuckoo algorithm”. In: Petroleum Science 11.3, pp. 446–
453.

Fisher Scientific (2022). url: https://www.fishersci.com/us/en/home.html.

Jan, DS, HY Shiau, and FN Tsai (1994). “Vapor-liquid equilibria of n-hexane+ cyclo-
hexane+ n-heptane and the three constituent binary systems at 101.0 kPa”. In: Journal
of Chemical and Engineering Data 39.3, pp. 438–440.

Suppes, GJ (2002). “Selecting thermodynamic models for process simulation of organic

VLE and LLE systems”. In: ChapterONE Online (2002).

Wankat, PC (2012). Separation process engineering. Pearson Education. Chap. 2,7, pp. 75,
285, 390. isbn: 978-0-13-138227-5.

A Appendix

Figure A.1: Mass Balance

A.1
 Table A.1: Summary for the calculation of Eo

Column 1 Column 2
N-Hex C-Hex N-Hept N-Hex C-Hex N-Hept
Tave (◦ C) 88.1 75
µ (cP ) 0.178 0.320 0.325 0.190 0.480 0.238
µmix (cP ) 0.294 0.406

Figure A.2: Colunmn 1 results from DWSIM simulations

Figure A.3: Colunmn 2 results from DWSIM simulations

A.2