1. MULTIPERIOD DESIGN OF AZEOTROPIC SEPARATION SYSTEMS Kenneth H. Tyner and Arthur W. Westerberg

2. OVERVIEW Problem Description Problem Challenges Previous Work Related Research Issues Solution Approach Conclusions

3. PROBLEM DESCRIPTION Design An Optimal Separation Plant Multiple Feeds –Flowrate –Composition –Operating Time Azeotropes A B CAz F1 F3 F2

4. PROBLEM DESCRIPTION A B CAz F1 F3 F2 F A B C Az

5. PROBLEM DESCRIPTION A B CAz F1 F3 F2 F A B C

6. PROBLEM DESCRIPTION FEED 1FEED 3FEED 2

7. PROBLEM DESCRIPTION FEED 1FEED 3FEED 2

8. PROBLEM DESCRIPTION FEED 1FEED 3FEED 2

9. PROBLEM DESCRIPTION FEED 1FEED 3FEED 2

10. PROBLEM DESCRIPTION FEED 1FEED 3FEED 2

11. PROBLEM CHALLENGES Highly Combinatorial –Separation Pathways –Process Units –Task Assignment Difficult Subproblems –Large Models –Highly Nonlinear –Recycle Streams –Shared Equipment

12. MULTIPERIOD DESIGN Constraints: –Column Dimensions –Heat Exchanger Dimensions –Flooding Conditions

13. MULTIPERIOD DESIGN Collocation Models: –Number of Trays and Feed Location Variable –Variable Transformations

14. MULTIPERIOD DESIGN

15. EXTEND TO AZEOTROPIC MULTIPERIOD DESIGN? Additional Feasibility Constraints How Many Columns? Large Number of Simulations Stream Characteristics Change

16. INITIAL RESEARCH THRUSTS Synthesize Designs Evaluate Designs Optimize / Modify Designs

17. AZEOTROPIC SYNTHESIS A B CAz F F A B C

18. AZEOTROPIC SYNTHESIS A B CAz F A B C F

19. AZEOTROPIC SYNTHESIS A B CAz F F A B C

20. SIMULATION Zero Slack S S S

21. SIMULATION Solve / Optimize Initialize Modify Library

22. REVISED RESEARCH THRUSTS Collocation Error Detection Scaling Solver Design

23. SIMULATION Solve / Optimize Initialize Modify Library

24. SOLUTION APPROACH Approximation –Separation Task –Column Design and Operation Shortcut Costing Autonomous Agents

25. ECONOMICS Cost = F( Feed, Distillate, Trays, Reflux )

26. ECONOMICS Cost = F( Feed, Distillate, Trays, Reflux ) Separation Task Contribution

27. ECONOMICS Cost = F( Feed, Distillate, Trays, Reflux ) Separation Task Contribution Column Design and Operation Contributions

28. TASK APPROXIMATION Variables: –Compositions –Flowrates Relations: –Mass Balance –Lever Rule –Geometric Objects A B CAz F D / F D B

29. COLUMN APPROXIMATION Cost = F(Feed, Distillate, Trays, Reflux) Reflux = F(Trays, Feed Location)

30. COLUMN APPROXIMATION Cost = F(Feed, Distillate, Trays, Reflux) Reflux = F(Trays) Optimal Feed Location = F(Trays)

31. COLUMN APPROXIMATION Reflux = C 1 * exp(-C 2 * Trays) + C 3 Opt Feed Loc = C 4 * Trays + C 5 –Numerical Difficulties Gilliland Correlation

32. DATA COLLECTION Fix Trays and Task Find Optimal Reflux

33. DATA COLLECTION

34. A B CAz Store In Database Calculate Parameters

35. SIMULATION F A B C Az A B C F Database

36. SIMULATION F A B C Az A B C F Database

37. SIMULATION Zero Slack S S S

38. ASYNCHRONOUS TEAMS Independent Software Agents Shared Memory Trial Points Newton SolverGradient Solver

39. ASYNCHRONOUS TEAMS Independent Software Agents Shared Memory Trial Points Newton SolverGradient Solver

40. ASYNCHRONOUS TEAMS Independent Software Agents Shared Memory Trial Points Newton SolverGradient Solver

41. ASYNCHRONOUS TEAMS Independent Software Agents Shared Memory Trial Points Newton SolverGradient Solver

42. ASYNCHRONOUS TEAMS Independent Software Agents Shared Memory Advantages –Scalable –Ease of Creation / Maintenance –Cooperation

43. ASYNCHRONOUS TEAMS Applications –Train Scheduling –Travelling Salesman Problem –Building Design

44. ASYNCHRONOUS TEAMS Problem Description Approximation Data Designs Database Design Agents Approximation Agents

45. MINLP DESIGN AGENT Fixed: –Separation Pathways –Intermediate Streams Variable: –Task Assignment –Number of Columns –Column Dimensions –Operating Policy

46. MINLP DESIGN AGENT Fixed: –Separation Pathways –Intermediate Streams Variable: –Task Assignment –Number of Columns –Column Dimensions –Operating Policy

47. MINLP DESIGN AGENT Fixed: –Separation Pathways –Intermediate Streams Variable: –Task Assignment –Number of Columns –Column Dimensions –Operating Policy

48. TASK ASSIGNMENT

50. PATH SELECTION Sequential Selection Genetic Algorithm Active Constraint

51. MINLP DESIGN AGENT Fixed: –Separation Pathways –Intermediate Streams Variable: –Task Assignment –Number of Columns –Column Dimensions –Operating Policy

52. ASYNCHRONOUS TEAMS Problem Description Approximation Data Designs Database Design Agents Approximation Agents

53. GENERAL BENEFITS Alternative to Hierarchical Design Persistent Data Scenario Analysis Human Agents

54. MULTIPERIOD DESIGN OF AZEOTROPIC SEPARATION SYSTEMS Kenneth H. Tyner and Arthur W. Westerberg