Nils Ayral, Enno Gent, Gabriele Tomaschun Chemical Process Development Carl von Ossietzky Universitt Oldenburg June 2, 2015 Modelling and Optimization of an Extractive Dividing Wall Column Overview Basics and motivation Simulation V ISUAL B ASIC programming Scans Solver Conclusions Details on scans and solver 2 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Basics separation problems 1 Basics and motivation Here we are! separation of closely boiling mixtures and azeotropes without additive: 1 with additive: 1 1 J. Rarey, slides collection: Physical Properties for Process Simulation, N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC separation problem of cyclohexane and benzene mixture small differences in volatilities: solution add auxiliary agent to the C 6 H 12 /C 6 H 6 mixture! entrainer of choice: N-Formylmorpholine (NFM) Problem definition 1 Basics and motivation bp = Cbp = C 4 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC T-x,y diagrams (calculated with UNIFAC) 1 Basics and motivation 5 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Extractive rectification 1 Basics and motivation 2 T. Diehl, B. Kolbe, H. Gehrke, ERTC Petrochemical Conference, Prague, Fig. 1: Possible column arrangements. 2 6 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC The Morphylane process 1 Basics and motivation 2 T. Diehl, B. Kolbe, H. Gehrke, ERTC Petrochemical Conference, Prague, Fig. 2: Comparison of concepts for extractive distillation. 2 7 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Direct sequence extractive rectification 2 Simulation Fig. 3: Schematic. 2 Fig. 4: Corresponding A SPEN flowsheet. 8 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC 2 T. Diehl, B. Kolbe, H. Gehrke, ERTC Petrochemical Conference, Prague, B : C6H6 C : C6H12 Single-column extractive rectification 2 Simulation B : C6H6 C : C6H12 Fig. 6: Corresponding A SPEN flowsheet. Fig. 5: Schematic. 2 9 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC 2 T. Diehl, B. Kolbe, H. Gehrke, ERTC Petrochemical Conference, Prague, 2005. reflux ratio COLLEFT reflux ratio COLRIGHT VLEFT/VRIGHT on the fly Final flowsheet and operating parameters 2 Simulation nstages stage NFMIN stage FEED flowrate NFM nstages 10 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Key challenges 3 V ISUAL B ASIC Programming interface between A SPEN and E XCEL E XCEL -To-A SPEN A SPEN -To-E XCEL fully automated scripts that allow scans the use of the Evolutionary Algorithm (EA) built into E XCEL Fig. 7: Key challenges. operating parameter costs, duties, purities, errors, etc. 11 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Flowchart: scans 3 V ISUAL B ASIC Programming Scans Type in scan parameter Open Simulation n y Loop over all parameters Next parameter ExcelToAspen Reinit Run CheckSimStat PrintInput PrintOutput Run Scan Stop Details Simulation loaded? N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC12 Results 3 V ISUAL B ASIC Programming Scans Fig. 8: Chosen scan parameter. data points ca. 7 days of calculation for comparison (scan vs. EA) get an idea of the simulations behavior percentage of errors cost efficient points 13 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Results errors Fig. 9: Percentage of errors. 3 V ISUAL B ASIC Programming Scans data points ca. 7 days of calculation for comparison (scan vs. EA) get an idea of the simulations behavior percentage of errors cost efficient points 14 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Results errors Fig. 9: Percentage of errors. 3 V ISUAL B ASIC Programming Scans data points ca. 7 days of calculation for comparison (scan vs. EA) get an idea of the simulations behavior percentage of errors cost efficient points 14 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC 3 V ISUAL B ASIC Programming Scans 15 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Operating parameter Scan nstages COLLEFT 52 stage NFMIN 7 stage FEED 25 flowrate NFMIN [kmol/h] 460 nstages COLRIGHT 7 nstages STRIP 8 reflux ratio COLLEFT 8.00 reflux ratio COLRIGHT 7.57 VLEFT/VRIGHT 0.45 energy costs [US$/year] Results histogram Results example: 2D plot Fig. 10: 2D cut of the scan results. optimal flowrate for a given configuration 3 V ISUAL B ASIC Programming Scans 16 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Flowchart: solver 3 V ISUAL B ASIC Programming Solver 17 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Type in boun- dary conditions Solve For # of iterations Next iteration Write new parameter Type in parameter Start solver (EA) Reset solver Set solver options Set boundary conditions Cost function Check for changes Function cost(a) declarations [...] Call ReadPara, para2, para3, para4, para5, para6) Call ExcelToAspen(para1, para2, para3, para4, para5, para6) APsim.reinit APsim.run Call CheckSimStat(errstat) If errstat = 0 Then Set ihn = ihn.Elements("Results Summary") Set ihn = ihn.Elements("Utility-Sum") Set ihn = ihn.Elements("Output") Set ihn = ihn.Elements("NET_COST") cost = ihn.Value * 24 * ElseIf errstat = 1 Then cost = # End If End Function ExcelToAspen Reinit Run CheckSimStat SetFunctionValu e 3 V ISUAL B ASIC Programming Solver penalty function! 18 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Flowchart: solver N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Type in boun- dary conditions Solve Initialization Done y n ny Evolutionary Algorithm Evaluation Mutation Maximum # of iterations Exceeds a given fitness level Recombination 3 V ISUAL B ASIC Programming Solver 19 Solver solutions 3 V ISUAL B ASIC Programming Solver 20 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Comment : blank test, moderate boundary conditions, moderate population size N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Solver solutions 3 V ISUAL B ASIC Programming Solver Comment : bad initialization, broad boundary conditions, moderate population size 20 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Solver solutions 3 V ISUAL B ASIC Programming Solver Comment : good initialization, narrow boundary conditions, moderate population size 20 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Solver solutions 3 V ISUAL B ASIC Programming Solver Comment : good initialization, broad boundary conditions, little population size, best solver result 20 Solver solutions N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC 3 V ISUAL B ASIC Programming Solver Comment : good initialization, same boundary conditions as the scan, little population size 20 Comparison 3 V ISUAL B ASIC Programming Fig. 11: Final simulation flowsheet. Tab. 1: Comparison of best configurations. N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Operating parameter Scan Solver nstages COLLEFT stage NFMIN 7 8 stage FEED 25 flowrate NFMIN [kmol/h] nstages COLRIGHT 7 5 nstages STRIP 8 6 reflux ratio COLLEFT 8.00 reflux ratio COLRIGHT 7.57 VLEFT/VRIGHT 0.45 energy costs [US$/year] Summary 4 Conclusions successful simulations of an extractive rectification with two- column and single-column (EDWC) arrangements well-working interface between E XCEL and A SPEN scans and EA found good sets of parameters, which display comparable low energy costs The solver provides better results within a shorter time! 22 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Experience and outlook 4 Conclusions 23 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC complex and multifaceted assignment building-blocks : VBA, E XCEL, A SPEN proof of concept study further research possible: determination of TACs! further optimization with linear algorithm chemical problems of real relevance parallelization? combination with neural networks? Thank you for attention. Any questions? 24 N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Stop Type in scan parameter Simulation loaded? Open Simulation n y Loop over all parameters Next parameter ExcelToAspen Reinit Run CheckSimStat PrintInput PrintOutput Run Scan Flowchart: scans N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC 5 Appendix Scans 12 Stop Type in scan parameter Open Simulation n y Loop over all parameters Next parameter ExcelToAspen Reinit Run CheckSimStat PrintInput PrintOutput Run Scan Public Sub RunScan() Declarations [...] If APsim Is Nothing Then Call OpenSimulation End If [...] Flowchart: scans Simulation loaded? N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC 5 Appendix Scans 12 Type in scan parameter Open Simulation n y Loop over all parameters Next parameter ExcelToAspen Reinit Run CheckSimStat PrintInput PrintOutput Run Scan Sub OpenSimulation() Dim path As String Dim file As String Set ws = ActiveSheet path = ws.Cells(1, 2) file = ws.Cells(2, 2) Set APsim = GetObject(path + file) End Sub Stop Flowchart: scans Simulation loaded? N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC 5 Appendix Scans 12 Loop over all parameters Next parameter Run Scan Type in scan parameter Open Simulation n y Stop Flowchart: scans ExcelToAspen Reinit Run CheckSimStat PrintInput PrintOutput ExcelToAspen Reinit Run CheckSimStat PrintInput PrintOutput Simulation loaded? N. Ayral, E. Gent, G. Tomaschun Modelling and Optimization of an EDWC Public Sub RunScan() declarations [...] start(0) = ws.Cells(4, 4) Do While start(0)