ISA NEW DIR BIOPROC MDL CNTRL-2007 New Directions in Bioprocess Modeling and Control - Maximizing Process Analytical Technology Benefits.pdf

1、New Directionsin Bioprocess Modelingand Control:Maximizing Process AnalyticalTechnology BenefitsBioProcess2006.book Page i Thursday, August 17, 2006 3:58 PMBioProcess2006.book Page ii Thursday, August 17, 2006 3:58 PMNew Directionsin Bioprocess Modelingand Control:Maximizing Process AnalyticalTechno

2、logy BenefitsByMichael A. BoudreauGregory K. McMillanBioProcess2006.book Page iii Thursday, August 17, 2006 3:58 PMNoticeThe information presented in this publication is for the general education of the reader. Because nei-ther the author nor the publisher have any control over the use of the inform

3、ation by the reader, both the author and the publisher disclaim any and all liability of any kind arising out of such use. The reader is expected to exercise sound professional judgment in using any of the information presented in a particu-lar application.Additionally, neither the author nor the pu

4、blisher have investigated or considered the affect of any patents on the ability of the reader to use any of the information in a particular application. The reader is responsible for reviewing any possible patents that may affect any particular use of the information pre-sented.Any references to co

5、mmercial products in the work are cited as examples only. Neither the author nor the publisher endorse any referenced commercial product. Any trademarks or tradenames referenced belong to the respective owner of the mark or name. Neither the author nor the publisher make any repre-sentation regardin

6、g the availability of any referenced commercial product at any time. The manufac-turers instructions on use of any commercial product must be followed at all times, even if in conflict with the information in this publication.Copyright 2007 ISAThe Instrumentation, Systems, and Automation SocietyAll

7、rights reserved. Printed in the United States of America. 109876543 2ISBN-13: 978-1-55617-905-1 ISBN-10: 1-55617-905-7 No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form orby any means, electronic, mechanical, photocopying, recording or otherwise, withou

8、t the prior writ-ten permission of the publisher.ISA67 Alexander Drive, P.O. Box 12277Research Triangle Park, NC 27709www.isa.orgLibrary of Congress Cataloging-in-Publication Data in process.BioProcess2006.book Page iv Thursday, August 17, 2006 3:58 PMvTABLE OF CONTENTSAcknowledgments . . . . . . .

9、. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiAbout the Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10、 . . . . . . . . . . . . . . . . . . . . . . . . xiChapter 1 Opportunities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31-1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31-2. Analysis of Variabili

11、ty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61-3. Transfer of Variability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161-4. Online Indication of Performance . . . . . . . . . . . . . . . . . . . . . . . . . . 241-5. Optimizing Performance

12、 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271-6. Process Analytical Technology (PAT) . . . . . . . . . . . . . . . . . . . . . . . 28References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Chapter 2 Process Dynamics . . . .

13、. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352-1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352-2. Performance Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362-3. Self-Regulat

14、ing Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472-4. Integrating Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54Chapter 3

15、Basic Feedback Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573-1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573-2. PID Modes, Structure, and Form . . . . . . . . . . . . . . . . . . . . . . . . . . . 603-3. P

16、ID Tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 713-4. Adaptive Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 873-5. Set-Point Response Optimization. . . . . . . . . . . . . . . . . . . . . . . . . . . 91Re

17、ferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96Chapter 4 Model Predictive Control. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 994-1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

18、. . . . . . 994-2. Capabilities and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1004-3. Multiple Manipulated Variables . . . . . . . . . . . . . . . . . . . . . . . . . . 1094-4. Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

19、. . . 116References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127Chapter 5 Virtual Plant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1315-1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

20、 . . . . . . . . . . . . . . . 1315-2. Key Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1325-3. Spectrum of Uses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1385-4. Implementation. . . . . . . . . . . . . . . . . . .

21、 . . . . . . . . . . . . . . . . . . . . . . 141References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147Chapter 6 First-Principle Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1516-1. Introduction. . . . . . . . . . . . . .

22、 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1516-2. Our Location on the Model Landscape . . . . . . . . . . . . . . . . . . . . . 1526-3. Mass, Energy, and Component Balances . . . . . . . . . . . . . . . . . . . 1536-4. Heat of Reaction . . . . . . . . . . . . . . . . . . . . . . .

23、 . . . . . . . . . . . . . . . . . 1586-5. Charge Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1596-6. Parameters and Their Engineering Units . . . . . . . . . . . . . . . . . . . 162BioProcess2006.book Page v Thursday, August 17, 2006 3:58 PMvi Table of

24、Contents6-7. Kinetics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1676-8. Mass Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1806-9. Simulated Batch Profiles . . . . . . . . . . . . . . . . . . . . . .

25、. . . . . . . . . . . 185References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188Chapter 7 Neural Network Industrial Process Applications. . . . . . . . . . . 1937-1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

26、 . . . . . . . 1937-2. Types of Networks and Uses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1987-3. Training a Neural Network. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2007-4. Timing Is Everything . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

27、. . 2037-5. Network Generalization: More Isnt Always Better . . . . . . . . . . 2067-6. Network Development: Just How Do You Go about Developing a Network? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2087-7. Neural Network Example One . . . . . . . . . . . . . . . . . . . . . . . . . . .

28、 . 2117-8. Neural Network Example Two . . . . . . . . . . . . . . . . . . . . . . . . . . . 2177-9. Designing Neural Network Control Systems. . . . . . . . . . . . . . . . 2337-10. Discussion and Future Direction . . . . . . . . . . . . . . . . . . . . . . . . . . 2357-11. Neural Network PointCounte

29、rpoint . . . . . . . . . . . . . . . . . . . . . . 239References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242Chapter 8 Multivariate Statistical Process Control . . . . . . . . . . . . . . . . . . . 2478-1. Introduction. . . . . . . . . . . . . . . . .

30、 . . . . . . . . . . . . . . . . . . . . . . . . . . . 2478-2. PCA Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2498-3. Multiway PCA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2658-4. Model-based PCA (MB-PCA) . . . .

31、. . . . . . . . . . . . . . . . . . . . . . . . . 2728-5. Fault Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282Appendix A Definition of Terms. . . .

32、 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289Appendix B Condition Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301Appendix C Unification of Controller Tuning Relationships. . . . . . . . . . . . 305Appendix D Modern Myths . . . . . . . . .

33、 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317Appendix E Enzyme Inactivity Decreased by Controlling the pH with a family of Bezier Curves 1 . . . . . . . . . . . . . . . . . 321Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

34、. . . . . . . . . . . . . . . . . . 333BioProcess2006.book Page vi Thursday, August 17, 2006 3:58 PMviiAcknowledgmentsThis book would not have been possible without the enthusiastic support and commitment of resources by Mark Nixon and Grant Wilson of Emerson Process Management. In particular, the a

35、uthors thank Grant Wilson for setting the path and establishing the perspective to achieve the full opportunities of a Process Analytical Technology initiative. The authors express their appreciation to Joseph Alford of Eli Lilly and Company for his detailed and informative answers to our questions

36、on current practice in the biopharmaceutical industry. This book tried to capture some of his extensive expertise in terms of the practical considerations in pursuing opportunities. The authors also express their appreciation to Harry Lam of Genentech, D. Grant Allen of the University of Toronto, an

37、d Nick Drakich of Invista (Canada) Company for their review of chapter 6 on first-principle modeling.The authors thank the following people from Emerson Process Management for their technical contributions: Michalle Adkins (biopharmaceutical applications), Terry Blevins (basic and advanced control),

38、 Bruce Campney (demos), Ashish Mehta (neural networks), Dirk Thiele (model predictive control), Peter Wojsznis (adaptive control), and Willy Wojsznis (model predictive control). The authors thank Robert Heider from Washington University in Saint Louis and Martha Schlicher from Renewable Agricultural

39、 Energy, Inc., for contributions to the application of neural-network modeling in ethanol production. The authors thank Thomas Edgar and Yang Zhang from the University of Texas for their investigation and survey of principal component analysis for batch fault detection and analysis.The kinetics and

40、operating conditions in chapter 6 for the first-principle modeling of the production of ethanol from cellulose were developed by Nathan Mosier and Michael Ladisch in the Laboratory of Renewable Resources Engineering at Purdue University. A portion of chapter 7 is extracted from a thesis presented by

41、 Michael N. May to the Henry Edwin Sever Graduate School of Washington University in Saint Louis in partial fulfillment of the requirements of the degree of Master of Science, 1 July, 2005.Appendix E is an excerpt from an article by Michael J. Needham titled “Enzyme Decreased by Controlling the pH w

42、ith a Family of Bezier Curves” published in the May 2006 issue of Pharmaceutical magazine.Finally, the authors express their gratitude to Jim Cahill and Brenda Forsythe for the cover graphics.BioProcess2006.book Page vii Thursday, August 17, 2006 3:58 PMBioProcess2006.book Page viii Thursday, August

43、 17, 2006 3:58 PMAbout the AuthorsMichael A. Boudreau, P.E., is a control systems engineer on the final leg of a multi-year, world-wide tour of distributed control system manufacturers; testing and developing advanced control systems on Emerson Process Managements DeltaV. During his career, Michael

44、has worked on bioreactor control systems ranging in scale from 10 liter pharmaceutical pilot fermentors to 100,000 liter bulk chemical fermentors. He helped design the first EPO and G-CSF process control systems for Amgen, managed the control group at a Miles Labs (Bayer) citric acid plant and desig

45、ned, started and maintained bacterial and cell culture fermenter control systems at Genentech.Gregory K. McMillan, CAP, is a retired Senior Fellow from Solutia and Monsanto where he worked in engineering technology on process control improvement. Greg was also an affiliate professor for Washington U

46、niversity in Saint Louis. Presently, Greg is a contract consultant for Emerson Process Management in DeltaV R rather it is the achievement of the desired material attribute. Quality decisions should be based on process understanding and the prediction and control of relevant process/product attribut

47、es.”The discovery and implementation of more optimal batch profiles, end points, and cycle times are encouraged by the PAT guidance document. The proper integration of technologies can actually speed up rather than slow down the time to market. Ideally, the optimization process is coincident with th

48、e commercialization process. Waiting until a pharmaceutical is out of patent protection increases the risk of not having the lowest cost position. The return on investment (ROI) gained by optimizing the production process for a new drug can be enough to justify the use of analyzers, models, and cont

49、rol. The goal of this book is to provide the basis for taking advantage of this opportunity.Making the optimization process coincident with the commercialization process can lead to a low-cost position, an important factor for a competitive advantage when a drug goes off of patent protection.BioProcess2006.book Page 5 Thursday, August 17, 2006 3:58 PM6 Chapter 1: OpportunitiesLearning ObjectivesA. Be able to track down the major sources of variability.B. Recognize how instruments aggravate or mitigate variability.C. Understand how to reduce variability through different levels of con