Level 2 Model Improvement
Case Study: Oregon Steel
5. Second Improvement
Temperature Regions Based on Metallurgical Feature
The trials in March 2007, though
demonstrated the significant improvement of the Level 2 model, revealed several
new sources of error. After the tests, further improvements could be achieved by
optimizing temperature ranges for each temperature region. The temperature range
for the third region (lower-temperature region) was narrowed, so that there
would not be too many passes in this region. The temperatures with metallurgical
meaning, such as the hot/cold forming boundaries, were used as reference in
determining the dividing points for temperature regions. By technical
definition, the hot forming is the forming conducted over the recrystallization
temperature range, while the cold forming is the one below the recrystallization
temperature. The flow stress modeling in the hot forming is different from that
in the cold forming. Traditional rolling in the hot mills was conducted in the
hot forming temperature range. However, today the rolling in hot mills may be
carried out below the recrystallization temperature, especially for finishing
passes and those passes in the second stage of controlled rolling. Technically,
it is cold rolling (it is often referred to as warm rolling to avoid confusion).
The philosophy behind the EOS Level 2
models, which was installed over a dozens of years ago, was primarily for the
hot rolling. Due to the invalidity of the flow stress model in the passes below
the hot forming range, usually in the finishing passes when this occurs,
improvements can be made using a narrowed range in the third temperature region.
In this aspect, the Guided Two-Parameter Learning (FIT2G) fits well a narrowed
temperature range because it only has two learning parameters (degrees of
freedom) instead of three or four, and thus necessitates fewer passes to perform
a regression.
Expansion of the Valid Range for the Flow stress
Formula
The flow stress model was improved for
passes with drafts 10% and lower, and for drafts 30% and higher. Those passes
were often identified to be with high force errors. Technically, the flow stress
formula described in Equation (1) is not valid for the strain below 0.1 [6]. The
strain value 0.1 corresponds to a draft of 10%. Unfortunately, many finishing
passes were rolled with a draft at or below 10%, and the resultant force error
more easily caused shape problems in the finishing passes than other passes
because the thickness is smaller. In addition, this same formula is with a
relatively narrow valid range of strain, so for a relatively high draft, it
introduces errors, too.
In order not to make significant change to
the Level 2 source code, the improvement of the flow stress valid range problem
was tackled by scaling up the strain for the passes where the strain was below
0.15, and scaling down the strain for the passes where strain was over 0.35. The
improvement in this way only required a modification to the function that
defines the strain.
Force Modification for the Resume Pass
The controlled rolling technique is applied at EOS when
rolling X-grade and some HSLA grade products. The rolling process is divided
into two or three stages by one or two delays. A delay (or hold) begins when the
piece reaches a PDI defined intermediate thickness. During a delay, the piece is
held until its surface temperature drops to a specified value. A hold involves
complicated metallurgical evolution processes. The Level 2 models in EOS, like
most others, ignores the metallurgical effects during a hold. This frequently
caused errors 10% to 30% in the resume passes (the passes after holds). Some
passes often had errors of -10% to 40%. Fig. 4 shows the flow stress
errors in the resume passes.
The modification was done with the results of error
pattern analysis with empirical formula, so the flow stress for the resume pass
was scaled down (or up) by multiplying a factor:
S = a * T + b
(3)
In which T is the entry temperature of the resume pass, S
is the scale factor, and a and b are constants for a given grade. Constants a
and b were calculated from historical data for each model grade with holds. The
value of a and b could be stored in the grade file; it may be recommended that a
database (file or table) be created to store the constants a and b for all those
model grades with a hold.
Fig. 4:
Resume Pass Flow Stress Error
Rolling in the Two-Phase Region
Entering into the austenite/ferrite
two-phase region was not intended; rather, it was due to inaccurate scheduling.
Primarily the metallurgist, through Level 3 production scheduling system, should
perform improvement for the problem. Because this problem, as long as it occurs,
would cause significant error as showed in the Fig. 2, a quick fix was
proposed for the Level 2. First, the system could be set up to allow the
temperature coefficient C2 to be negative, and secondly, narrowing
the range of the lower-temperature region. The Guided Two-Parameter Learning is
recommended in this case since it requires a very narrow temperature range (or
number of passes). For a complete solution of this Two-Phase Region problem,
metallurgical modeling for the flow stress should be performed, during which the
volume fraction of each phase should be predicted and the flow stress for each
material (phase) should be determined.
<To
Be Continued>
<Previous>
|
<Next>
|
Part1
|
Part2
|
Part3
|
Part4
|
Part5
|
Part6
|
Part7
|
Part8
|
Part9
|
|