1997 IEEE INTERNATIONAL CONFERENCE ON SYSTEMS, MAN, AND CYBERNETICS HYATTORLANDO, ORLANDO, FLORIDA,USA OCTOBER12-15, 1997 COMPUTATIONAL CYBERNETICS AND SIMULATION
ORGANIZING COMMITTEE General Chair: James M. Tien Technical Program Chair: Charles J. Malmborg Technical Arrangements Chair: Julia Pet-Edwards Functional Arrangements Chair: Mansooreh Mollaghasemi Promotional Programs Chair: Mark J. Embrechts Special Tracks Chair: Michael H. Smith Student Programs Chair: Julie C. Adams Conference Treasurer: Jules C. Jacquin CONFERENCE STAFF Lana Leon Ingrid Cedo Mary Ellen Fullum Anthony C. Brozowski Mary S. Wagner
Robert Armacost Ileana Costea Frank DiCesare T. Govindaraj William Gruver Keith Hipel
S. B. Ahmed/USA A. S. Alfa/Canada R. Barton/USA D. Berg/USA P. Borne/France J. Buckley/USA D.Buede/USA B. Clegg/England E. Cox/USA G. Cybenko/USA C. Dagli/USA F. Duoud/Canada G.Dauphin-Tanguy/
France W. DavidUSA
N. DeClaridUSA P. DeshayedFrance W. DresdUSA
MEMBERS Timothy Kotnour James Lunhoj Linda Malone Pamela McCauley Bell Christine Mitchell Michael Mullens
James Pullin James Ragusa Ralph Rogers Kay Stanney David 172urman Mengchu Zhou
PROGRAM COMMITTEE K-C. FadChina G. Klir/USA P. Kokol/Slovenia L. FanglCanada P. Fishwick/USA R. K o m / N e w Zealand T. Fukuda/Japan R. KrajVUSA K. K. Kumar/USA Q. Gao/Canada H. W. LewidJapan M . Geirgiopoulos/USA R. Gordon/USA J. Liu/Hong Kong J. Graham/USA F. LootsmdNetherlands D. Luo/China W.A. Gruver/USA R. Hamalainen/Finland B. Malakooti/USA K. HirotaIJapan M. McGinnis/USA L. Horvath/Hungury M. MenglCanada C. HsdUSA M. Obaidat/USA Y-P. Huang/China S. OmatdJapan M. Jafari/USA G. Ortega/Holland M. Kam/USA K. R. Pattipati/USA K . Kawamura/USA M. Proctor/USA S. Kercel/USA T. L. Saaty/USA
R. Saeks/USA A. P. Sage/USA M. G. Singh/England G. StonelUSA C-Y. Su/Canada S. Sugiyama/Japan S. Sung/Singupore H. TakagiIJapan L. R. Talluru/USA H. vanl;andingham/USA J. Wang/China C. C. White/USA K. P. White/USA C. Yang/USA Y. Yu/China A. ZahediIAustralia H. Zha/Japan
Copyright and Reprint Permission: Abstracting is permitted with credit to the source. Libraries are permitted to photocopy beyond the limit of U.S. copyright law for private use of patrons those articles in this volume that carry a code at the bottom of the first page, provided the per-copy fee indicated in the code is paid through Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923. For other copying, reprint or republication permission, write to IEEE Copyrights Manager, IEEE Service Center, 445 Hoes Lane, P.O.Box 1331, Piscataway, NJ 08855-1331. All rights reserved. Copyright O1997 by the Institute of Electrical and Electronics Engineers, Inc. IEEE Catalog Number 97CH36088-5 ISBN 0-7803-4053-1 (softbound) ISBN 0-7803-4054-X (casebound) ISBN 0-7803-4055-8 (microfiche) ISSN 1062-922X
MA14: Optimization, Heuristics and Search Methods I Room: Orange D Time:8:45 - 10:30 am
6. A Test Problem Generation Methodology for Nonlinear Goal Programming. Hunter T. Albright, Peter Beling, University of Virginia, USA. .............................................. 448
Chair: B. de Andrks-Toro, Universidad Computense de Madrid, Spain. . Co-chair: Masami Shishibori, University of Tokushima, Japan.
MA16: Panel on Computational Cybernetics and Simulation: Issues, Contentions, and Perspectives Room: Osceola B Time:8:45 - 10:30 am Invited Session Chair: Paul Werbos, National Science Foundation, USA Co-chair: Michael H. Smith, University of California, Berkeley, USA
1.Application of Genetic Algorithms and Simulations for the Optimization of Batch Fermentation Control. B. de AndrksToro, Jose M. Giron-Sierra, J.A. L6pez-Orozco, C. Fernhndez-Conde, Universidad Complutense de Madrid, Spain. ................................................................................... 382 2. Minimum Turns / Shortest Path Problems: A FramedSubspace Approach. Robert J. Szczerba, Lockheed Martin Federal Systems, USA. Danny Z. Chen, Kevin S. Klenk, University of Notre Dame, USA. ......................................... 398
3. A n Algorithm f o r The Special Two-Dimensional Cutting Problem. Zhiping Fan, Jian Ma, Peng Tian, City University of Hong Kong, Hong Kong. ................................................. 404 4. A Stochastic Tabu Search Strategy and Its Global Convergence. Peng Tian, Jian Ma, Zhiping Fan, City University of Hong Kong, Hong Kong. ............................... 410 5. An Efficient Compression Method f o r Patricia Tries.
Masami Shishibori, Masafumi Okuno, Kazuaki Ando, Junichi Aoe, University of Tokushima, Japan. ......................... 415 MA15 Optimization Applications and Methodology Room: Osceola A Time:8:45 - 10:30 am Invited Session Chair: K. Preston White, Jr., University of Virginia, USA. Co-chair: Peter Beling, University of Virginia, USA.
1. Darwin Meets Computers: New Approach to Multiple Depot Capacited Vehicle Routing Problem. * Minea Filipec, Davor Skrlec, Slavko Krajcar, University of Zagreb, Croatia. ................................................................................ 421
2. A Branch-And-Bound Method f o r Finding Independently Distributed Probability Models that Satisj’j Probability Order Constraints. Bon K. Sy, Xiao Ying Han, Queens CollegeKity University of New York, USA. ........................................... 427 3. Optimization of Integrated Circuit Design with Respect to Yield. Robert Athay, University of Virginia, USA. .......... 433 4. A Simulation-Optimization Methodology for Sensor Placement. Donald E. Brown, University of Virgina, USA. Jeffrey B. Schamburg, United States Military Academy, USA. .................................................................................... 439 5. Simulation Study of An X-Ray Lithography Cell: Background and Objectives. K. Preston White, Jr., University of Virginia, USA. Walter J. Trybula, SEMATECH, USA. .................................................................................... 444
William Gruver, Simon Fraser University, Canada. (No Paper) Paul Werbos, National Science Foundation, USA. (No Paper) Toshio Fukuda, Nagoya University, Japan. (No Paper) I.B. Turksen, University of Toronto, Canada. (No Paper) Lotfi Zadeh, University of California, Berkeley, USA. (No Paper) Michael H. Smith, University of California, Berkeley, USA. (No Paper) MA17: Biocybernetics Room: Osceola C Time:8:45 - 10:30 am Chair: Ken Ohta, Bio-Mimetic Control Research Center, Japan. Co-chair: John R. Alexander, Towson State University, USA.
1. Dynamic Control of Bipeds Using Postural Adjustment Strategy. Philippe Gorce, Laboratoire de Genie Mecanique Productique et Biomecanique, France. ................................ 453
2. Human Perceptual-Motor Coordination in Unkown Dynamical Environments. Ken Ohta, Zhi Wei Luo, Masami Ito, Institute of Physical and Chemical Research, Japan. ... 459 3. A Three-Neuron Controller (TNC) MI. John R. Alexander, Towson State University, USA. Jacob P. Cox, Oxford Molecular Group, USA. ..................................................... 463 4. An Inter-Segment Allocation Strategy for Postural Control in Human Reach Motions Revealed by Differential Inverse Kinematics and Optimization.* Xudong Zhang, Don B. Chaffin, University of Michigan, USA. ............................. 469 5. Biologically-Motivated Learning in Adaptive Mobile Robots. T.W. Scutt, University of Nottingham, United Kingdom. R. I. Damper, University of Southampton, United Kingdom. .... 475
6. Fuzzy Sets and Semiotics. I. Burhan Turksen, University of Toronto, Canada. ................................................................. 481
A Plenary: Information Granulation and Its Centrality in Human and Machine Intelligence Dr. Lotfi A. Zadeh, University of California, Berkeley, USA ....................................... 486 Room: Cypress Ballroom Time: 11:OOam - Noon
APPLICATION OF GENETIC ALGORITHMS AND SIMULATIONS FOR THE OPTIMEATION OF BATCH FERMENTATION CONTROL
B. Andres-Toro, J. M. Giron-Sierra, J. A. Lopez-Orozco, C. Fernandez-Conde Departamento de Informatica y Automatica. Facultad de Ciencias Fisicas Universidad Complutense de Madrid 28040 Madrid, Spain e-mail:
[email protected]
ABSTRACT
2. THE OPTIMIZATION PROBLEM
Batch fermentations are dynamic processes that must be guided along convenient paths to obtain the desired results. Our research deals with the application of computers for advanced control of such processes. We selected beer fermentation, and started to investigate whether it is possible to optimize the process, taking as reference to be improved a real industrial fermentation. A good mathematical model is needed for that, and, as we refer to realistic industrial conditions, we had to develop a new one. Then we started optimization studies, exploring the adaptation of Genetic Algorithms for our problem. Good results are obtained, fumishing a promising ground for additional improvements. In this paper we describe the process, the new model, the optimization problem, and the solution by Genetic Algorithms.
To begin our work, we established contacts with a brewery, to get knowledge from experts. The industrial batch fermentation begins after filling big tanks (after meticulous cleaning and disinfecting) with wort and yeast: a temperature profile is applied, so the fermentation process follows a dynamic trajectory, along an interval of time. Wort degrades rapidly, so it is stored at low temperatures before being driven to the tanks. The yeast is added to the wort in the conduits to the tanks. During the fermentation no stirring is applied, and IK) substances are added. The biochemical process is exothermic: consequently, cooling devices are needed to achieve the desired temperature profile. The initial substrate (sugars) is employed by the yeast for energetic and growing purposes, being mostly converted to ethanol. Some undesirable byproducts appear, that degrade beer quality.
1. INTRODUCTION Economical results will be better if we can save time to reach the desired ethanol concentration. increasing the temperature will mean an acceleration of the process, but perhaps attaining a lower ethanol concentration, or yielding byproducts' concentrations that surpass established limits. In addition, brewing experts say that for each centigrade degree over 15 "C,the risk of spoilage (for instance, due to Lactobacillus Plantarum) doubles.
We are interested in applied control problems, requiring the use of advanced computer-based methods. Batch fermentations are good candidates for two reasons: the complexity of the biological phenomena taking place, and the dynamic nature of the process. Fermentations are the basis of many important industrial activities [l]. Both for modelling and testing purposes, we selected the conventional beer fermentation, as a representative example that can be experimentally studied at laboratory scale with a moderate equipment effort. The results obtained in this way, can be useful for many other fermentation processes.
Any chemical intervention on the process is forbidden. Temperature is the only variable we can control, to drive the fermentation process along convenient paths.
Therefore, we have into view a multi-objective optimization problem, that combines time and efficiency criteria,
along with some constraints. To solve this problem we have designed an iterative strategy, computing a sequence of shorter and shorter optimal processes. Our first task h a s been to formulate the objective function to be optimized, in mathematical terms. For this purpose, we defined the following functions, that include weighting factors to make their values comparable:
The objective of our work is to optimize the beer production, taking advantage of the flexibility provided by computers in control tasks. We take as reference for our research, the actual industrial conditions and practices.
0-7803-4053-1/97/$10.00 1997 IEEE @
392
better; and that the crossover operator is more efficient if applied to mid-sections (twocrossover points).
In addition, we have to continue with industrial implementation aspects, considering a more general optimization problem: economic factors, limitations of the temperature control devices, and energy management.
c
'i.
Acknowledgment The authors would like to thank the support of this research work by the Spanish CICYT Committee, Project TAP94-0832-CO2-0 1, and by the Universidad Complutense de Madnd, Project PR188/92-4088, and the collaboration of CNzcampo's brewery .
profile for industry
0
50
Time, Hours
100
7. REFERENCES
150
[l] A. Johnson, "The Contrcil of Fed-batch Fermentation Processes A Survey," Automatica, vol. 23, no. 6, pp. 691-705, 1987.
-
fig. 8: Optimal 130 hours profile adapted for industq
[2] D. A. Gee, and W. Fred Ramirez, "Optimal Temperature Control for Batch Beer Fermentation,'' Biorechnol. d Bioeng., no. 3 I , pp. 224-234, 1988.
smoother profile. Figure 7 shows the result for the 130 hours optimal profile (J/= 558.6).
[3] D.A. Gee, Modeling. Optimal Control. State Estimation and Parameter Identijication Applied to a Batch Fermentation Process. Doctoral Thesis, University of Colorado at Boulder, 1990.
The industrial procedures determine that the temperature of wort, at the beginning o f fermentation, is 10 "C. From this point, the heat generated by the reactions is employed to let the temperature rise up to a desired higher level. That means that, to offer a better temperature profile to industry, we should retain a fixed initial trajectory, and include this constraint in the optimization problem. Fortunately, the evolutionary approach is open to maintain a fixed part along the evoIution. So we created a first part of the chromosomes as follows:
141 D. A. Gee, and W. Fred Ramirez, "A Flavour Model
for Beer Fermentation," J. Inst. Brew., no. 100, pp. 321-329, 1994.
Andres-Toro, J.M. Gron-Sierra, C. FemandezConde, J.M. Peinaclo, F. Garcia-Ochoa, "A Kynetic Model for Beer Production: Simulation under Industrial Operational Conditions", hoc. IMACSIIFAC Int. Symp. on Mathematical Modelling and Simulation in Agricultural and Bio- Industries, (MMSAB1'97), Budapest, pp. 203-208,1997
Chromosome= [ l o 10 10 10 10 10 10 10 10 I 1 11 1 1 12 12 13 13 14 14 14 1 5 15 # # # ...I Starting again our optimization procedure, we obtain a 130 hours smoothed profile (figure 8), more adequate for industrial implementation, with JJ= 552.51.
E. Goldberg, Generic Algorithms in Search, Optimization, and Machine Leaming. Addison Wesley, 1989.
6. CONCLUSIONS
171 L. Davis, Handbook of Genetic Algorithms. Van Nostrand, 1991.
Our paper shows a positive experience with genetic algorithms and simulations, for the optimization of batch fermentation processes. Good results are obtained with moderate development efforts and computational results, using MATLAB (a well known tool). Thanks to the velocity of the method, we were able to study fine grained discretizations, paving the way for smooth adequate sohtions.
[S]
D. B. Fogel, Evolurioiaary Computation: Toward a New Philosophy ofMachine Intelligence. IEEE Press, 1995.
191 Z. Michalewicz, Genetic Algorithms + Data Structures = Evolution PmAvams. Springer Verlag, 1996.
We are now refining the method, to establish better conditions for a fast evolution. For example, our experiments indicate in this moment that less initial individuals is
397
