This page initially listed six papers submitted/published in February-March
2008. It then includes some of the later publications.
¡¡
AISTech 2008
Metallurgical, Modeling and Software
Engineering Issues in the Further Development of the Steel Mill Level 2 Models
(Presentation Slides)
Bingji (Benjamin) Li, Ph.D.
President & CEO
Metal Data LLC
www.metalpass.com
Pittsburgh, PA, USA
John Nauman, Ph.D.
Vice President of Operation
Metal Data LLC
www.metalpass.com
Pittsburgh, PA, USA
Key words: Level 2 model, force, flow stress,
metallurgical, modeling, software engineering, learning, retained strain
Abstract
This paper introduces selected metallurgical,
modeling and software engineering issues involved in the further development of
steel mill Level 2 models. Limitation of the adaptive learning has been
identified and the Guided Two-Parameter Learning is considered the quick fix for
existing systems. Metallurgical issues involve the retained strain and the
rolling in the two-phase region, etc. The modeling issues include rolling
process models, learning logics and the intelligent learning. There are also
software engineering issues such as system design with mill process models and
the web-based Level 2 system. Finally, a concept on developing next-generation
Level 2 system was outlined.
* Paper completion deadline (2/15/08); published.
Presented in May 2008.
AISTech 2008
Level 2 Model Improvements at Evraz Oregon
Steel Mills
(Presentation Slides)
Bingji (Benjamin) Li
www.bli1.com
Metal Data LLC
www.metalpass.com
Pittsburgh, PA, USA
David Cyr
Level 2 Engineer
Department of Process Automation
Evraz Oregon Steel Mills
Portland, OR, USA
Petrus Bothma
Manager
Department of Process Automation
Evraz Oregon Steel Mills
Portland, OR, USA
Key Words: Level 2 model, metallurgical, roll
force, steckle mill, adaptive learning, flow stress, resuming passes, draft
schedule
Abstract
Level 2 force model was improved for OSM plate
steckle mill. Learning logics and metallurgical effects were identified as the
primary sources of error. Limitation of the adaptive learning was discussed.
Concept of guided two-parameter learning was proposed to resolve the issues and
over 6000 sets of the flow stress coefficients were designed. In addition, the
problems in resuming passes and the passes with large or small strain were
solved. Even with troubled grades, the testing still indicated a high accuracy
with an average absolute error of 3.4%. It was intended to make minimal code
change for the existing system.
* Paper was accepted by AISTech 2008. However,
due to certain delay, this paper is to be
published in 2009 (accepted for AISTech 2009).
Flat-Rolled Steel Processes: Advanced
Technologies. By V. Ginzburg, etc. CRC Press.
Metallurgical, Modeling and Software
Engineering Issues in the Further Development of the Steel Mill Level 2 Models
(Book Chapter 26)
Bingji (Benjamin) Li, Ph.D. (Lead Contributor)
President & CEO
Metal Data LLC
www.metalpass.com
Pittsburgh, PA, USA
John Nauman, Ph.D.
Vice President of Operation
Metal Data LLC
www.metalpass.com
Pittsburgh, PA, USA
Content
- Level 2 Model
- Metallurgical Issues in Level 2
- Retained Strain
- Rolling in the Two-Phase Region
- Metallurgical Aspect of the Flow Stress
- Others
- Modeling Issues in Level 2
- Limitation of the Adaptive Learning
- The Guided Two-Parameter Learning (GFIT2)
- Flow Stress Valid Range
- Temperature-Dependent Properties
- Intelligent Learning
- Software Engineering Issues in Level 2
- System Architecture based on Interactive
Relationship of Mill Process Models
- Web-based Level 2 System
- Others
- Next-Generation Level 2 System
- Next-generation Level 2 system
- Next-generation Level 2 model
* Paper completion deadline (2/29/08). Expanded
and modified from a similar writing. Accepted
for publishing. Book on printing.
Flat-Rolled Steel Processes: Advanced
Technologies. By V. Ginzburg, etc. CRC Press.
The State-of-the-art of Infrared, Laser and
Microwave based sensors and systems
(Book Chapter
22)
Francois Reizine
Lead Contributor, President of American Sensors Corp.
Pittsburgh, PA, USA
Xiaoqing Zhang
Operations Manger of American Sensors Corp.
Pittsburgh, PA, USA
Bingji (Benjamin) Li
President of Metal Data LLC
Pittsburgh, PA, USA
John Nauman
Lead Contributor, Vice President of Metal Data LLC
Pittsburgh, PA, USA
Content
- Current Sensor Technologies
- Principles of Selected Applications
- Continuous Caster Optimization of Cut
- Width Measurement of Slab
- Strip Centering/Camber and Width Measurement
- Sensor Systems
- Systems Developments
- Systems Techniques
- System Examples in Slab Casting
- System Examples in Hot Rolling
- System Examples in Finishing
* Paper completion deadline (2/29/08). Accepted
for publishing. Book on printing.
Materials Science & Technology 2008
(MS&T'08)
Significance and Development of a
Next-Generation Level 2 Model as a Metallurgical System
(Presentation Slides)
Abstract
Level 2 model improvement projects have revealed
various metallurgical issues that negatively affect the current Level 2 model.
The considerable retained strains due to uncompleted recrystallization, and the
metallurgical phenomena during hold and two-phase region rolling, etc., cause
significant model errors which cannot be removed by adaptive learning. Wide
application of metallurgical processes in today’s steel rolling calls for a
Level 2 model to fully consider metallurgical principles. The next-generation
Level 2 model should include a hybrid system by combining a full-range of
metallurgical models with intelligent learning such as neural network, together
with an expert system to guide the learning. The new model would also improve
the pass schedule in controlled rolling principle and provide assistance for the
Level 3 scheduling. The revealed metallurgical issues, the general concepts of
the next-generation Level 2 system and the related metallurgical models, etc.,
will be introduced.
- Level 2 model as a metallurgical system
- Incomplete recrystallization and retained strain
- Softening during the hold
- Two-phase region
- Metallurgical nature of the flow stress
- Property variations
- Benefits of metallurgical Level 2
- High Accuracy of the Force Prediction
- Improved pass schedule and slab selection
- Development of Next-Generation Level 2 Model as a
Metallurgical System
- Level 2 System
- Rolling mill Level 2 model
- Reheating furnace Level 2 model
- Controlled cooling Level 2 model
Submitted by
Bingji (Benjamin) Li, Ph.D.
www.bli1.com
Metal Data LLC
www.metalpass.com
Pittsburgh, PA, USA
John Nauman, Ph.D.
Metal Data LLC
Pittsburgh, PA, USA
www.metalpass.com/jnauman
* Paper abstract submitted (3/4/08); accepted for
publishing. To be presented on Oct., 2008.
Materials Science & Technology 2008
(MS&T'08)
Career Development to be a Multi-National and
Multi-Disciplinary Engineer
(Presentation
Slides)
Abstract
Experiences are shared on how to perform
self-training to become one of the most dynamical engineers, for integrating
German engineering, US IT and Chinese market. With over 30 years of training,
the author has gained three countries' working experiences, four languages and
skills on material engineering, mechanical engineering, software engineering and
industry automation. After receiving Ph.D., working on rolling process modeling
and publishing a book, the author spent recent 10 years to be a mill-automation
software engineer and to do mill application development. To be a highly
qualified software engineer, the author completed 30 computer classes. Critical
factors for success are to plan ahead and to brew interest in the things to be
done. The paper also outlines author's results on the mill process models,
web-based applications, general design on the next-generation Level 2 systems
and a book in writing on steel mill process modeling and computer application,
etc.
Submitted by
Bingji (Benjamin) Li, Ph.D.
www.metalpass.com/bli
Metal Data LLC
Pittsburgh, PA, USA
www.metalpass.com
¡¡
* Abstract submitted (3/4/08). Accepted;
presented in Oct. 2008.
AISTech 2009
Development of Model-Intensive Web-based Rolling Mill Applications
(Presentation
Slides)
Bingji (Benjamin) Li
www.metalpass.com/bli
Metal Data LLC.
www.metalpass.com
Pittsburgh, PA, USA
Abstract
Model-intensive and web-based steel rolling mill applications have been
developed in metalpass.com. They include pass design suites AutoForm and
FreeForm, mill force/torque prediction suite, temperature profile program with
finite-differential method for rolling and water/air cooling, and microstructure
prediction application, etc. Coupled with tension models, the FreeForm is
particularly useful for high-speed rolling blocks, and for both designing new
passes and examining existing ones. Multiple algorithms are applied to ensure
both speed and accuracy. Issues in developing each of the applications, such as
process modeling, data modeling, model verification, object-oriented
programming, and data management, etc., are discussed.
* Abstract submitted (July 2008), accepted for
AISTech 2009. Paper submitted on 2/13/09.
¡¡
AISTech 2009
Level 2 Model Improvements at Evraz Oregon
Steel
(Presentation Slides)
Bingji (Benjamin) Li
www.metalpass.com/bli (www.bli1.com)
Metal Data LLC
Pittsburgh, PA, USA
www.metalpass.com
David Cyr
Level 2 Engineer
Department of Process Automation
Evraz Oregon Steel Mills
Portland, OR, USA
Petrus Bothma
Manager
Department of Process Automation
Evraz Oregon Steel Mills
Portland, OR, USA
Key Words: Level 2 model, metallurgical, roll
force, steckle mill, adaptive learning, flow stress, resuming passes, draft
schedule
Abstract
Level 2 force model was improved for OSM plate
steckle mill. Learning logics and metallurgical effects were identified as the
primary sources of error. Limitation of the adaptive learning was discussed.
Concept of guided two-parameter learning was proposed to resolve the issues and
over 6000 sets of the flow stress coefficients were designed. In addition, the
problems in resuming passes and the passes with large or small strain were
solved. Even with troubled grades, the testing still indicated a high accuracy
with an average absolute error of 3.4%. It was intended to make minimal code
change for the existing system.
* Accepted for publication in AISTech 2009. Paper
submitted on 2/13/09.
¡¡
Materials Science & Technology 2009
(MS&T'09)
Development of Web-based Metal Property and Metal Information Databases
Abstract
A list of web-based metal property and metal information databases have been
developed and made accessible through metalpass.com. The property databases
include Flow Stress, High-Temperature Property and General Property, etc. The
information databases consist of Metal Dictionaries (both Tech Terms and
Translation), Metal Software, Metal Patents, and Metal Directory, etc. Number of
entries in each database usually ranges from several thousand to over fifty
thousand. Flow stresses in dependence of strain, strain rate and temperature,
etc. are provided in the form of both data and models, while high-temperature
properties are available in temperature dependence. As extensions to the
databases, over a thousand pages of short papers describing technical details
and dozens of applications for predicting process/product parameters are
provided. Besides data development, data management and data application, etc.,
technical significance of the data such as temperature dependence of metal
properties in hot forming modeling is also covered.
Submitted by
Bingji (Benjamin) Li, Ph.D.
www.metalpass.com/bli
Metal Data LLC
Pittsburgh, PA, USA
www.metalpass.com
* Abstract submitted (3/6/09); Accepted; Paper
completed and accepted; To be presented in Oct. 2009.
¡¡
AISTech 2010
Mill Level 2 Model in Improvement of Product Quality and Productivity
Bingji (Benjamin) Li, Ph.D.
www.metalpass.com/bli
Metal Data LLC
Pittsburgh, PA, USA
www.metalpass.com
Key Words: Level 2 Model, draft schedule, product property, shape defect,
productivity, rolling mill models
Abstract
Mill Level 2 model, with major manufacture logics and mill intelligence in
it, is responsible for creating draft schedule and stage plan. Level 2 model
errors would cause equal deformation target, metallurgical temperature target
and maximal productivity target, etc. to be missed. This paper first discusses
association of Level 2 model quality with product property and shape and mill
productivity. Then it covers application of a full range of rolling mill models
in Level 2, and integration of models on product defects (camber, center/edge
wave, etc.) with Level 2 and AGC. Examples are from author’s past and ongoing
consulting projects.
* Paper presented by Douglas Stalheim, a mill
consultant associated with Metal Data (Benjamin Li was on onsite in China on
mill Level 2 project).
¡¡
AISTech 2010
Developments on a Web-based Metal Technology and Metal Information Network
Key Words: Metal Technology, Metal Information, Web-Based, Tech Resources,
Tech Directories, Categorization, Data Population
Abstract
Developments on metal technology and metal information network metalpass.com
are summarized and experiences shared. The website includes metal properties
(flow stress and high-temperature properties, etc.), tech resources (e.g. Metal
Dictionaries for Tech Terms and Translation, Metal Patents, Software Database
and Tech Papers), and business showcase applications (Metal Directory, Product
Profile and Tech Profile, etc.). Development processes such as data collection,
data categorization and data population are discussed, together with
implementations for automatic upload of technical paper, expandable submission
of listing in every page of the showcase applications, and web-based mill
software suites, etc.
Submitted by
Bingji (Benjamin) Li, Ph.D.
www.metalpass.com/bli
Metal Data LLC
Pittsburgh, PA, USA
www.metalpass.com
* Paper presented by Douglas Stalheim, a mill
consultant associated with Metal Data (Benjamin Li was on onsite in China on
mill Level 2 project).
¡¡
10th International Conference on Steel Rolling
Plate/Coil Mill Level 2 Model Upgrade with Metallurgical Modeling and
Advanced Learning
(Presentation Slides)
Bingji (Benjamin) Li, Ph.D.
Metal Data LLC
www.metalpass.com
Pittsburgh, PA, USA
(001) 412-621-3836
Abstract
Mill Level 2 model, with major manufacture logics and mill
intelligence in it, is responsible for creating draft schedule
and achieving parameter targets for production optimization.
Level 2 model errors cause equal deformation targets,
metallurgical temperature targets and maximal productivity
targets, etc. to be missed. This would lead to product shape
problems (e.g. center/edge waves) and low mechanical
properties. A large AGC movement due to initial roll gap error
caused by inaccurate force prediction, for example, could
result in plate head-end geometry problem. Traditional Level 2
model does not contain metallurgical principle such as those
for thermomechanical rolling and microalloying strengthening.
Due to the recovery, recrystallization and grain growth, etc.,
flow stress is very dynamic. The retained strain from
incomplete recrystallization, and the microstructural changes
during intermediate hold, etc., cause significant parameter
prediction errors. In addition, there are problems in learning
logics, such as scatter of adapted coefficients due to their
potential dependence on each other. Blind learning only
reaches limited accuracy.
This paper first provides various fixes on metallurgical
issues and learning limitations mentioned above, and discusses
application of a full set of rolling mill models in Level 2.
Then it primarily introduces a simple but very effective
learning mechanism, a so-called guided two-parameter fitting
(FIT2G), as the solution to all those problems. The FIT2G uses
carefully designed strain coefficients and strain rate
coefficients, and performs adaptation by adjusting temperature
coefficients and material coefficients. It can not only remove
limitations of adaptive learning, but also include the
metallurgical effects into the Level 2 model. The large number
of flow stress coefficients, usually about 6,000 to 12,000
sets for a mill, is the integration of all the solutions for
the learning logics and metallurgical effects. In addition, it
only requires very limited modifications to the Level 2 source
code and needs a very small temperature range to perform
regression. Therefore, it is the right choice in improving
existing Level 2 model, with very little concern on
introducing potential bugs to the existing system. Past
results in USA and current project warrantees in China all
assure a low force prediction error below 5% for the
plate/coil steckle mills. Economic value of the upgrade is
usually millions of US dollars per year for each of the mills.
* Presented in Sept. 14-17, 2010, Beijing,
China.
¡¡
AISTech 2011
NISCO Plate/Coil Mill Level 2 Force Model improvements
Abstract
Level 2 force prediction serves as basis for draft schedule generation and
initial roll gap setup before AGC adjustment. This paper covers Level 2 force
model improvements in NISCO plate/coil steckle mill, especially for continuously
increasing steel grades, and for automatic design of flow stress coefficients
based on chemical composition, slab and product mix, and rolling process
including thermomechanical rolling. The guided two-parameter learning, by
designing flow stress coefficients with metallurgical effects integrated and
learning logic issues resolved, proves to be an easy-to-apply and very accurate
solution. Force prediction accuracy is thus significantly increased. Further
development fields are also listed.
Submitted by
Bingji (Benjamin) Li, Ph.D.
Metal Data LLC
Pittsburgh, PA, USA
www.metalpass.com
Pengju Zhu
Nanjing Iron & Steel Co. (NISCO)
Nanjing, Jiangsu, PR China
Daoyuan Wang
Nanjing Iron & Steel Co. (NISCO)
Nanjing, Jiangsu, PR China
¡¡
AISTech 2011
Development of Next-Generation Level 2 Model as Metallurgical System
Bingji (Benjamin) Li
www.metalpass.com/bli
Metal Data LLC
www.metalpass.com
Pittsburgh, PA, USA
Daoyuan Wang
Nanjing Iron & Steel Co. (NISCO)
Nanjing, Jiangsu, PR China
Pengju Zhu
Nanjing Iron & Steel Co. (NISCO)
Nanjing, Jiangsu, PR China
Key words: Level 2 Model, Metallurgical processes,
Metallurgical models, Intelligent learning, Hybrid system
Abstract
Various metallurgical issues negatively affect current
Level 2 model and cause significant model errors which cannot
be removed by adaptive learning. Wide application of
metallurgical processes (e.g. controlled rolling) calls for
metallurgical principles in Level 2 model. The next-generation
Level 2 should include a hybrid system by combining a
full-range of metallurgical models with intelligent learning.
The new model would also improve the pass schedule in
controlled rolling. The revealed metallurgical issues, the
general concepts of the next-generation Level 2 system, and
the related metallurgical models in rolling, reheating and
accelerated-cooling will be particularly discussed.
¡¡
International Symposium on the Recent Developments in Plate Steels
19-22 June 2011 ?Winter Park, Colorado
Improved Level 2 Draft Scheduling for Good Plate Shape and
Properties
Bingji (Benjamin) Li
www.metalpass.com/bli
Metal Data LLC
www.metalpass.com
Pittsburgh, PA, USA
Key words: Level 2 Model, draft scheduling, plate shape, product property
Abstract
Primary purpose for plate mill Level 2 model is to create
quality draft schedule for good plate shape and product
properties. Three key issues should be well addressed in the
Level 2 model. Firstly, the logic in creating draft schedule
based on predicted parameters should be good; secondly,
parameter prediction as basis for the draft scheduling should
be accurate; and thirdly, metallurgical issues such as
retained strain, recrystallization, etc., for model accuracy
and rolled steel properties, should be well considered. Level
2 model should achieve equal deformation target, metallurgical
temperature target and maximal productivity target. Draft
scheduling is also aimed at grain refinement for the finish
product. This paper summarizes work results in several plate
mills on Level 2 model logic improvement, Level 2 parameter
prediction improvement and metallurgical integration into
Level 2 model. Features for the next-generation Level 2 model
as metallurgical system are also introduced.