OPTIMISATION BASED DESIGN OF MEMBRANE ASSISTED HYBRID

SEPARATION PROCESSES USING AN EVOLUTIONARY ALGORITHM

Katharina Koch, Daniel Sudhoff, Stefan Krei, Peter Kreis

Motivation Chemical system Acetone

(TB = 56.1 C)

Design of energy efficient distillation processes for the Production of acetone by dehydrogenation

0.1

0.9
separation of non-ideal multicomponent mixtures of isopropyl alcohol (IPA)

0.2

0.8
Distillation

0.3

0.7
boundary
Conventional separation methods: extractive, heteroazeotropic or Separation of ternary azeotropic
INTRODUCTION

0.4

0.6
system acetone-IPA-water

x(
pressure swing distillation

R)

AC 5
TE

ET
5

0.
WA

0.

ON
x(

E)
High capital costs Potential use of membrane:

0 .6

0.4
0 .7

0.3
Feed
Low energy efciency To cross the distillation

0 .8

0.2
Innovative process concept: membrane assisted hybrid processes boundary (Fig. 1)

0 .9

0.1
Strong synergies due to interactions between both unit operations To separate the close Water 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 IPA
(TB = 100.0 C) x (IPA) (TB = 82.3 C)
boiling binary mixtures Azeotrope
Overcoming limitations of stand-alone processes (TB = 80.3 C)
at high organic Fig. 1: Residue curve map for ternary system
Lack of general design methodology and in-depth understanding concentrations (Fig. 2) acetone-IPA-water at p = 1 atm
of complex interactions
110 105

Signicant economical potential hardly exploited in industry 100 Acetone/Water IPA/Water

Temperature [C]

Temperature [C]
100
Fig. 2: Binary VLE at p = 1 atm. 90
95
Lines: Data from Aspen Prop- 80
90
Development of a reliable methodology for the optimi- erties using NRTL-RK; 70

60 85

Symbols: experimental data

sation based design of hybrid separation processes 50
0 0.2 0.4 0.6
x (Acetone), y (Acetone) [mol/mol]
0.8 1
80
0 0.2 0.4
x(IPA), y (IPA) [mol/mol]
0.6 0.8 1

Design of hybrid processes Modelling

State of the art: Non-equilibrium stage model of distillation
Early decision on process conguration Multicomponent mass and heat transfer
Design approaches mostly based on short-cut methods and simplied Hydrodynamics of column internals considered
models, especially for membrane separation Calculation of optimal diameter based on correlation of Makowiak1)
APPROACH

Application of rigorous approaches limited to binary systems Detailed model of membrane separation
Generic design approach Transmembrane mass transfer based on solution-diffusion model
Evolutionary Process Optimisation Consideration of all relevant mass transfer resistances
Separation Optimal 1) J. Makowiak, Fluid dynamics of packed
Process- Process- Process-
Task Process columns, Springer-Verlag (2010)
alternatives simulation evaluation Evolutionary optimisation
Process superstructures for the consideration of all possible process Algorithm based on Population
Population Generation
Generation
Individual
Individual// Individual
Individual //
Parent
Parentvector
vector Trial vector
Parent vector
alternatives modied differential
Mutation
Mutation &
&

Generic process model with rigorous models for both unit operations evolution approach2) Recombination
Recombination

(Fig. 3) Fitness Selection

Selection Fitness
Simultaneous optimisation of process conguration, dimension of ap-
time

evaluation evaluation

paratus and operating conditions Economic objective

function (= tness)
Experimental determination of membrane parameters at lab-scale and
2) R. Angira et al., Chem. Eng. Sci., Fig. 3: Simplied owchart of evolutionary algorithm
experimental validation of hybrid process at pilot-scale 61 (2006), 4707-4721

Experimental parameter determination Optimisation of hybrid process Position of

Feed = Dist. 3
Pervaporation of IPA-water 2.0 Hybrid process Side = Dist. 3
E 1 1
mixture Qi = Qi,0 exp - A,i - exp ( A i w H2O ) Recycle = Dist. 4
R T T0 Packed column (Sulzer BX) Column
DDFi = Dfi = fFeed,i - fPerm,i xAcetone = 99.7 mol-%
Hydrophilic PVA/PAN membrane Pervaporation module with
D = 1.4 m Dist. 1
Permeate flux JExp [kg/(hm2)]

1.5 n = 6.3 YAcetone = 99.7 %

Ji = Qi DDFi +10%
(Sulzer Pervap 2201D) Sulzer Pervap 2201D
h1 = 8.1 m

-10% Dist. 2 Sidestream

2 .
Membrane area: 162 cm Simultaneous optimisation of h2 = 1.7 m
mSide= 7.1 t/h
1.0
RESULTS

Experimental conditions: 3 discontinuous and

Dist. 3

h3 = 1.0 m Membrane
wFeed,H2O = 2,5 - 18 wt.% 9 continuous variables (Fig. 5) Feed AMemb= 340 m
0.5 Dist. 4
.
mFeed= 2.8 t/h
TFeed = 60 - 78 C Objective function: costs per ton h4 = 2.2 m Recycle
Splitfactor = 0.7
Dist. 5
pFeed = 2 bar of puried acetone (CPT)
0.0
pPerm = 30 mbar 0.0 0.5 1.0 1.5 2.0 CPT = 106 /t xIPA = 99.5 mol-%
Permeate flux JSim [kg/(hm)] Condenser Heater YIPA = 97.1 %
4% 3%
Permeate Retentate
Model parameters determined Fig. 5: Superstructure of hybrid process
Fig. 4: Parity plot of simulated and measured Overhead Annuity
for an empirical correlation Costs (5a) and results of evolutionary optimisation
permeate ux for IPA-water using the empirical Column 17% 18%
Module incl. incl.
(Fig. 4) correlation Membrane & Packing
Labour
Costs 15%
10% Column
Feed Pump 41%
39%
40%

Future work Condenser

8% 5%
Membrane
Module Heater
Retentate
6%

21%
7%
6% Replacement

Feed
Membrane

Column Reboiler Pump

Future work will focus on the development and optimisation of more com- Fig. 6: Distribution of investment costs (left) and
Condenser Permeate
(Cooling brine)
plex superstructures, which consider all possible process congurations. operating costs (right) for optimised hybrid process

We acknowledge the nancial support of the Deutsche Forschungsgesellschaft for the project Optimisation-based framework for the synthesis of
hybrid separation processes. This project is a cooperation with the Process Systems Engineering Group of Prof. W. Marquardt, RWTH Aachen.

Department of Biochemical and Chemical Engineering Contact: Dipl.-Ing. Katharina Koch

phone: +49(0)231-755-2356
Univ.-Prof. Dr.-Ing. Andrzej Grak katharina.koch@bci.tu-dortmund.de
Laboratory of Fluid Separations www.fvt.bci.tu-dortmund.de