2 Process intensification in integrated membrane processes
2.5 Conceptual example
2.5.3 Result of the PBS workflow
2.5.3.1 Step P3: Identification of desirable phenomena
The knowledge base is contacted to identify possibilities for PI to overcome the iden-tified limitations. As this case has an unfavorable equilibrium, the integration of the reaction task with a second reaction (of IPAc and/or H 2 O) or a separation task (to
2.5 Conceptual example 47
Table 2.3: Decision table regarding PI solutions to limitations in a process
Necessary task: Reaction Reaction Reaction Reaction
Desirable Task: Separation Heat supply Mixing Reaction Limitation Contact problems of
reactants
Product reacts further/is intermediate
Activation problems
Degradation by T
Energy management
Slow reaction
Limiting equilibrium
this particular intensifi ed option is reported to overcome this limitation;
the activation of an option through a knowledge search.
remove IPAc and/or H 2 O) using suitable phenomena are options obtained from the database ( Table 2.3 ).
Applying the method based on thermodynamic insights ( section 2.4.2.3 ) and a reaction screening, 21 different phase transition phenomena serving the identified (six are membranes) and two reactions have been identified. For example, phase transition by relative volatility has been identified by occurring differences in boiling points for the removal of H 2 O and IPAc [PT(VL)] as well as phase transition by per-vaporation identified by differences in radius of gyration to remove H 2 O [PT(PVL)].
However, not all of them are feasible because the operating window does not match the liquid phase reaction phenomenon and/or potential promising because their DF is too low. Once H 2 O by pervaporation is removed, two membranes found in the data-base, one of them is a zeolite [ 56 ], the other one is polymeric [ 57 ]. The pervaporation has been modeled using correlations for selectivity and flux of a membrane based on experimental data of a zeolite membrane from Van Hoof et al. [ 56 ] and for a polymeric membrane from Sanz and Gmehling [ 57 ].
The zeolite membrane also has higher fluxes and selectivities for small water concentrations. Hence, the polymer membrane is removed from the search space.
Table 2.4 gives an overview of the screening. Only two-phase transitions are kept in the search space: the zeolite pervaporation [PT(PVL)] and relative volatility [PT(VL)].
Subsequently, additional phenomena are in addition to the necessary and desi-rable phenomena. Those are selected from the knowledge base using a set of rules:
– mixing (liquid (L): ideal (M id ), tubular flow (M tub ), rectangular flow (M rec ) – vapor (Mv): ideal
– two-phase V-L: ideal (2 phM) – stream dividing (D)
– convective heating (H) and cooling (C)
– heterogeneous reaction (R) – ideal phase contact of V-L (PC)
– ideal phase separation of V and L (PS).
Overall, n P,tot = 13 phenomena are in the search space.
2.5.3.2 Step P4: Generate feasible operation/flowsheet options
Firstly, all 13 phenomena are interconnected to SPBs. The number of competing phe-nomena (Eq. 2.15) that cannot be present within one SPB is n P,compete = 4 because of competing: heating or cooling (-1); liquid flow mixing phenomena (-2); dividing is by definition one SPB alone (-1). Based on n P,tot and the maximum number of phenomena allowed within an SPB n P,max (Eq. 3.2), a total number of 4019 SPBs ( NSPB max , Eq. 2.16) are generated.
n P,max = n P,tot – n P,compete = 13 – 4 = 9 (2.15) NSPB max =
∑
k = 1 n P, max(
n P,tot,w/oD !/( n P,tot,w/oD − k)!k!))
+ 1 = 4019 (2.16)Table 2.4: Identified phenomena for each desirable task using the algorithm APCP and AMP with pure component properties, and result of the screening of phenomena for each desirable task using the algorithm SoP
Task Identified phenomena Method of
determination
Screened out
Separation H 2 O/Rest Pervaporation R g , V M , V VdW
Vapor permeation R g , V M , V VdW Not matching operating window with necessary task Nanofiltration R g , V M , V VdW , δ SP Suitable membrane not in
database
Separation IPac/Rest Pervaporation R g , V M , V VdW Suitable membrane not in database
Vapor permeation R g , V M , V VdW Not matching operating window with necessary task Nanofiltration R g , V M , V VdW , δ SP Suitable membrane not in
database Separation IPAc and
H 2 O/Rest
Relative volatility T B , P LV Note: Possible until the ternary azeotrope
Reaction with H 2 O Reaction phenomenon No additional reaction found Reaction with IPAc Reaction phenomenon Not desired
T M , melting point; T B , boiling point; P LV , vapor pressure; R g , radius of gyration; V M , molar volume;
V VdW , Van der Waals volume; δ SP , solubility parameter).
The remaining phenomena and tasks kept in the search space are written in bold.
2.5 Conceptual example 49
Subsequently, the generated SPBs are screened for feasibility using connectivity rules and the information of the operating window of each phenomenon (not shown here).
For example, the operating window with respect to temperature/pressure for reaction phenomenon is restricted to have a liquid present, and by the deactivation energy of the catalyst while the maximum allowable temperature/pressure of the phase transition phenomenon by pervaporation is limited to the membrane stability of P max = 2bar and T max = 300K . In total, 58 SPBs are feasible in terms of conditions of the operating windows of the integrated phenomena (not shown here). Subsequently, the performance by using the conversion as criterion of each SPB is checked by using the extension of the Kremser method to identify the number of necessary stages and their connection. For example, when the reaction phenomenon is coupled with phase transition by pervaporation a conversion of 0.99 may be achieved within one stage. In other SPBs, such as reaction with phase transition by V–L, a conversion of 0.1 can achieved in 13 countercurrent stages. In order to allow the integration of different SPBs using different phase transition phenomena but still aiming at a simple operation, it is ideal to use a crossflow-current SPB arrangement and the use of betweenone and three connected stages. The simplified superstructure for this form is retrieved from the model library ( Figure 2.7 ).
The introduction of a dividing phenomenon into one stage of the three-stage super-structure enables up to six different recycles of the liquid while for configurations with two stages, three recycles are possible. That means that based on the 58 feasible SPBs in the search space 218,892 different process options are generated (Eq. 2.17).
NOO max = 58 3 + 58 2 + 58 1 + 6 · 58 2 + 3 · 58 1 = 218892 (2.17) Subsequently, all options are screened by logical constraints ( Table 2.5 ), fixing the
binary variables Y for feasible processes by ensuring a matching operating window of inlet/outlet streams of the SPBs and by ensuring the formation of the product. That gives 24,142 remaining options. By subsequent structural screening (Eq. 2.3; see Table 2.5 ), for example, by removing energy redundant options, 506 structural promising processes are identified (see Table 2.5 ).
2.5.3.3 Step P5: Fast screening for process constraints
The remaining 506 feasible and structural most promising options are screened by solving operational constraints and process models (Eqs. 2.4 – 2.5) which lead to 118
Stage 1 Stage 2 Stage 3
1 2 3 4
5 6
7
Figure 2.7: Superstructure of maximum three stages in crossflow. Only the liquid streams are shown
remaining options ( Table. 2.5 ). Subsequently, all options are ranked by its energy consumption as PI screening criterion (see Table 2.2 ) and 22 most promising options are selected to remain in the search space ( Table 2.5 ). All those 22 phenomena-based options are transformed into unit operations using a knowledge base and a set of rules. For example, because the material of the membrane is stiff (and not flexible as for polymers), the membrane is arranged either tubular or as plate-and-frame module. Additionally, boundaries for the unit operational level are retrieved from the knowledge base, leading to a subsequent screening at the unit operation level. In total, only two options remain in the search space that have been identified as achie-ving the lowest operational costs (represented by raw material and energy consump-tion). For example, an integrated CSTR type pervaporator system has been removed because necessary reactor sizes are too large. The two remaining options are:
– a tubular
(#1; D- MFl,tub = R = H = 2phM = Mv = PC = PT(PVL) = PS(VL)-MFl,tub = R = H = 2phM = Mv = PC = PT(PVL) = PS(VL)) – or a rectangular
(#2; D- MFl,rec = R = H = 2phM = Mv = PC = PT(PVL) = PS(VL)-MFl,rec = R = H = 2phM = Mv = PC = PT(PVL) = PS(VL))
flow reactor-pervaporator with integrated simultaneous heating (see Figure 2.8 ).
2.5.3.4 Step 6: Solve the reduced optimization problem and validate most promising
In step 6, a reduced NLP optimization problem involving the remaining two options is solved in which the diameter (and/or volume) of the tube/channel is optimized to Table 2.5: Results of the screening in steps P4 – P5
Constraint Number of options
Generated 218,892
Formally feasible operation 121,610
Logical: Product formed (L1) 102,424
Logical: Reaction before or simultaneously with phase transition (L2) 64,179
Logical: Process within operating window 24,142
Structural: Energy redundant operations are removed (S5) 12,244 Structural: Efficiency redundant stages are removed (S1, S2, S4, S9) 11,153 Structural: SPBs with highest effect on improving the yield are last (S3) 7,619
Structural: Process in 1 unit possible (S6, S7) 518
Structural: Process in 1 unit operation possible and no external recycle (S8) 506
Operational: Yield 118
Performance screening: Energy 22
Operational: Feasibility at unit operational level 2
2.5 Conceptual example 51
achieve a conversion of 0.99, a purity of 98 mol.% IPAc in the outlet and a pressure drop Δ p < 1 bar. It is assumed that 100 kg/m 3 catalysts can be put into the channel of a static mixer type of packing. Additional operating constraints are fluid conditions on the membrane (Re > 20) and that the number of parallel arrangements should not exceed 100. The kinetic model can be found in Lai et al. [ 53 ] while the experimental data for the correlations of the membrane can be found in Van Hoof et al. [ 56 ]. The results of the optimization are given in Table 2.6 . The best option is the plate-and-frame heat exchanger-reactor-pervaporator with a rectangular channel (option #2;
see Table 2.6 , Figure 2.8 ).