Bingji Li,Ph,D
Metal Pass LLC, Pittsburgh, PA, USA.
www.Meta4-0.com
Email: bli68@qq.com,
Phone (China): 13400064848
Level 2 system, usually counting for below 0.5% of the
strip/plate equipment investment, is nevertheless the brain
and nerve of the entire production line, and determines the
utilization performance of the product line. Usually, after
5-10 years of Level 2 installation, an upgrade or new
installation is needed for optimized equipment utilization. In
view of product quality, the primary purpose of mill Level 2
model is to create quality draft schedule for both shape and
property. Current Level 2 systems in the market usually have
many weaknesses such as issues in learning logics and lack of
microstructure evolution consideration in model calculation.
For example, plate rolling after hold is technically a cold
rolling since recrystallization is not completed; traditional
modeling based on hot rolling leads to high parameter
prediction error and insufficient product quality. The
author?s investigation through technical exchange with six
steel enterprises from TISCO to Sha-Steel and Baosteel
indicates that most rolling mills have issues of production
process instability, poor strip/plate shape of high grade
products, and instable product properties, etc. Those, from
technical aspect, result from insufficient quality of the
Level 2 system.
In the current steel industry with a small profit margin, it
is pursued by every steel producer to produce high grade
products at low cost in optimal process. For 8 years, Metal
Pass has developed a simple, low-cost but very effective way
of Level 2 system upgrade, to transform existing Level 2
system into the world?s top-quality new-generation Level 2
with microstructure models and intelligent learning. This way
of upgrade does not need major modification of the Level 2
source code, but rather integrate over 6,000-20,000 sets of
proven, high-quality data models into the current Level 2
system, by using Metal Pass?Guided Dual-Variable Learning.
Since most models are ready to use and no interfaces to other
systems need to be developed, the upgrade only costs 1/10 of
the new Level 2 installation. This paper summarizes work
results in several mill-Level 2 upgrades, especially Level 2
parameter prediction improvement and metallurgical integration
into Level 2 model. Features for the new-generation Level 2
model as metallurgical system are also introduced. In this way
of upgrade, offline design was performed first on the basis of
the rolling process and product models the author created in
the past 20 years during the work in Germany and USA. The
models take into account metallurgical phenomena such as
incompleted recrystallization and thus induced residual strain
due to micro-alloy application. This procedure has also
removed learning logic defects incurred in almost every Level
2 package in the market. Then the results were integrated into
the online learning of the Level 2 System. The upgraded Level
2 system achieves extremely high parameter prediction
accuracy. This makes it possible to increase operational upper
limit of rolling force.
Keywords: Level 2 Upgrade, Level 2 Accuracy, Microstructure
Model, Product Quality