1. 1 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 New results for divided-wall columns Deeptanshu Dwivedi (PhD Candidate, NTNU) Ivar Halvorsen (Senior Scientist, SINTEF) Sigurd Skogestad * (Professor, Department of Chemical Engineering, NTNU, Trondheim)

2. 2 Trondheim Oslo UK NORWAY DENMARK GERMANY North Sea SWEDEN Arctic circle

3. 3 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Introduction: Divided wall columns for 3- and 4- product separations Structures “V min diagrams” Experiments: 4- Product Kaibel Column – Experimental Setup – Control Structure – Experimental Runs- Steady state profiles – Experimental data- model fitting – Experimental Runs- Vapor Split Experiment Conclusions Outline

4. 4 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Introduction Distillation is energy intensive process. Exergetic analysis requires minimizing irreversibilities: -mixing effect -large ΔT across column Complex distillation arrangement: -Kaibel Arrangement -Petlyuk Arrangement -intermediate Reboilers & Coolers etc -HIDIC distillation Potential Energy Savings up to ~30 % in Kaibel Arrangement and Petlyuk Arrangements

5. 5 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 3-product separation: Conventional “direct split” ABC A/B A BC B/C B C

6. 6 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Improvement: Prefractionator (Easy split first) ABC A/C AB BC C B/C B A/B A B ABC A/C AB BC C A/B B/C A B Simplification: JOIN

7. 7 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Simplification: Direct coupling (“Petlyuk”) A/B B/C ABC AB C A A/C BC B C ABC A B A/C AB BC + single shell (divided wall column) Petlyuk column

8. 8 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 V min diagram for three components V min | Petlyuk = max (V AB, V BC ) = V BC V Prefractionator = V AC

9. 9 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 4-product separation: Extended Petlyuk A/B B/C C/D ABCD ABC BCD B D S1 S2 A/C B/D A/D AB CD BC ABCD D B S1 S2 ABC BCD AB CD BC 4-product extended Petlyuk column up to ~50 % energy savings

10. 10 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 4-product separation: Simplified (“Kaibel column”) D ABCD CD A B C AB ABCD AB CD D A B C B/C A/B B/C C/D 4-product extended Kaibel column up to ~30 % energy savings

11. 11 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 V min diagram for four components

12. 12 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Experimental Set up 4 products Packed Column Magnetic funnel-liquid split & Product valves Number of theoretical stages (experimentally determined): –Prefractionator: 13 –Main column : 21 Feed ABCD A (Methanol) B (Ethanol) C (Propanol) D (Butanol) 8m

13. 13 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Experimental Set up (Labview Interface)…

14. 14 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Control Structure (As used in experiments) Boilup V=constant 4 control degrees of freedom: Liquid split ratio R L1, Reflux ratio R L2 (top) Reflux ratio R L3 (middle) Reflux ratio R L4 (bottom) Decentralized Control with 4 PI Temperature Controllers: T 2s is adjusted to get large temperature change in the prefractionator T 3s, T 5s, T 7s is adjusted to get the temperature of product stages close to the boiling points of their main components 1 4 5 6 7 3 F D S1 S2 B T5T5 T3T3 T7T7 TC R l2 R l1 R l3 R l4 TC T 3S TC T 5S TC T 7S T2T2 2 T 2S V

15. 15 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Start-up T 2s is adjusted to get large temperature change in the prefractionator T 3s, T 5s, T 7s is adjusted to get the temperature of product stages close to the boiling points of their main components Temperatures Reflux ratios

16. 16 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Start-up

17. 17 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Steady Profiles with 4 temperature loops TEMPERATURESReflux ratios 1 4 5 6 7 3 F D S1 S2 B T5T5 T3T3 T7T7 TC R l2 R l1 R l3 R l4 TC T 3S TC T 5S TC T 7S T2T2 2 T 2S V

18. 18 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Liquid Split Loop -2 C Steady Profiles with 4 temperature loops.. TEMPERATURESReflux ratios 1 4 5 6 7 3 F D S1 S2 B T5T5 T3T3 T7T7 TC R l2 R l1 R l3 R l4 TC T 3S TC T 5S TC T 7S T2T2 2 T 2S V

19. 19 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Distillate Loop ±1 C Steady Profiles with 4 temperature loops.. TEMPERATURESReflux ratios 1 4 5 6 7 3 F D S1 S2 B T5T5 T3T3 T7T7 TC R l2 R l1 R l3 R l4 TC T 3S TC T 5S TC T 7S T2T2 2 T 2S V

20. 20 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 S1 Loop ± 1 C Steady Profiles with 4 temperature loops.. TEMPERATURESReflux ratios 1 4 5 6 7 3 F D S1 S2 B T5T5 T3T3 T7T7 TC R l2 R l1 R l3 R l4 TC T 3S TC T 5S TC T 7S T2T2 2 T 2S V

21. 21 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Steady Profiles with 4 temperature loops.. S2 Loop ± 1 C TEMPERATURESReflux ratios 1 4 5 6 7 3 F D S1 S2 B T5T5 T3T3 T7T7 TC R l2 R l1 R l3 R l4 TC T 3S TC T 5S TC T 7S T2T2 2 T 2S V

22. 22 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Steady Profiles with 4 temperature loops.. All Loops ± 1 C TEMPERATURESReflux ratios 1 4 5 6 7 3 F D S1 S2 B T5T5 T3T3 T7T7 TC R l2 R l1 R l3 R l4 TC T 3S TC T 5S TC T 7S T2T2 2 T 2S V

23. 23 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Model (lines) and experiments (points) fit well DS1S2B SimulationExperimentSimulationExperimentSimulationExperimentSimulationExperiment Methanol92.6% 15.4%17.2%0.21%000 Ethanol7.3% 51.5% 4.52%5.38%00 Propanol0032.9%31.2%89.6% 3.14%6.68% Butanol00005.67%5.02%96.86%93.32%

24. 24 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Vapor Split So far: Vapor split (Rv) kept constant But: Energy usage depends on Rv. Implement adjustable Rv But: Difficult to set Rv at desired value –Solution: Use Rv for temperature control (feedback) –The more precise liquid split (Rl) can be preset V/F vs R V for Kaibel column

25. 25 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Vapor Split Experiment.. From top left: Valve in fully open position Top right: Rack and pinion arrangement Schematic of the vapor split valve

26. 26 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Vapor Split Experimental run (Total Reflux, two component)

27. 27 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Vapor Split Experimental run (Kaibel Column)

28. 28 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Conclusions Four-Product Kaibel column –Experimentally demonstrated 4-point temperature control for stabilizing and startup operation –Experimentally demonstrated active vapor split control –Experimental data fits well with the model

29. 29 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Introduction 4- Product Kaibel Column – Four-product Kaibel column – Control Structure – Experimental Setup – Experimental Runs- Steady state profiles – Experimental Runs- Vapor Split Experiment 3- Product Petlyuk Column – Three-product Petlyuk column – The “V min diagrams” – Control Structures – Close Loop Results Conclusions Outline

30. 30 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Three-product Petlyuk column A/B B/C ABC AB C A A/C BC B ABC A C B A/C AB BC

31. 31 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Control Structure 1 Feed ABC B D C22 C1 C21 CC xCxC xAxA xBxB xCxC xBxB S Five degrees of freedom including vapor split Control key impurities using “close-by” parings Side product has two side impurities In CS1, S is paired with heavy key (x C )

32. 32 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Closed-loop result from CS1 Fails for feed composition disturbance z f =[53 13 33] from nominal equimolar feed

33. 33 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Why CS1 failed ?? For nominal equimolar feed, B/C is the most difficult split For the new feed A/B is more difficult feed and CS1 can not provide sufficient vapor in top section of main column

34. 34 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Control Structure 2 Feed ABC B D C22 C1 C21 CC xCxC xAxA xBxB xCxC xBxB S > xAxA Same as CS1, but boilup now has a maximum select controller with: light key, x A at S or, light key, x B at reboiler

35. 35 EFCE Working Party on Fluid Separations, Bergen, 23-24 May 2012 Closed loop results from CS2 Works for all feed composition disturbance from nominal equimolar feed The purity of bottom product may be over purified for some disturbances