VIII 
Process Modeling  Empirical 





No 
Name 
Description 
Client / Employer 
63 
Morgan?s roll force calculation procedure examination to identify weakness 
Requested by Morgan?s top management to examine the force prediction procedure. Attention was also focused to examine the algorithm and material data used for Morgan?s roll pass program CAPE. It was found that the procedures used at the time was only valid for the strain rate below 100/s, while the actual rolling process had a strain rate up to 3000/s. 
Morgan 
64 
Analysis of Morgan?s lab mill data acquired from fiveyear rolling tests 
Data was collected in a fiveyear lab mill test, during which about 1000 samples were rolled. For each rolled piece, over 70 parameters, including free radius, cross section area, forward slip, etc. were measured or calculated. It was one of the primary data sources to develop rolling process models. However, due to certain weaknesses in the experiment design and limited understanding of rolling process, the prior engineers who designed the tests had difficulty deriving meaningful models from such largescale rolling tests. In the new work the missing parts were added from other data sources (e.g., those from 15year rolling tests from the institute in Germany, in a fourstand highspeed continuous mill), to make the data fully useful. 
Morgan 
65 
Development of spread model, particularly for high speed rolling 
The existing Shinokura formula used in Morgan?s computer aided design was expanded to add more influence factors. Various spread prediction procedure, such as Hensel, Wusatowski, etc. were studied and verified/modified based on the large amount of experiment data. Spread procedures were later also verified by the field data collected from the NoTwist mills (e.g. ASW mill), etc. 
Morgan 
66 
Development of accurate flow stress models for both low and high speed mills 
Traditional flow stress models were verified and improved with the flow stress data I collected worldwide, and with mill results and especially field data. To fit high speed mill, traditional models (valid up to the strain rate 100/s) were extended to including the temperature/strain rate interactive factor to allow it to be valid up to the strain rate 500/s and can be roughly used for the highspeed rolling up to the strain rate 3000/s (100m/s). This is the nearest approach to the highest rolling. Neither formula is available nor the flow stress measurable in the range of 3000/s with constant strain rate values. Flow stress models for major grade groups and key grades were determined based on the new formula. 
Morgan 
67 
Development of empirical models for spread estimation, for RDOV and OVRD passes 
This was the direct modeling of the spread based on Morgan?s lab mill data. Model can be used for the first assessment of the roll pass because it does not need so much geometry calculation as Shinokura formula does. Some colleagues did roll pass design with ΔB = k ΔH to estimate width spread from a height reduction, with k ~ 0.33 for OVRD and k ~ 0.6 for RDOV. Therefore, I developed a simple formula to determine value of the k by considering several key parameters. 
Morgan 
68 
Development on steel rolling mill force model, with study on mean flow stress, contact area and shape factor. 
Roll separating force during rolling actually depends on three primary factors: the mean flow stress, the projective contact area and the shape factor. Mean flow stress depends on material only (including deformation history), and the shape factor is affected by the strain state and stress state in the deformation zone. Accurate prediction of the rolling force requires that all those three factors be accurately modeled, which was one of my primary areas of the rolling process modeling. I also studied various force prediction procedures for flat rolling, which lead to only different shape factor formulas. 
Morgan 
69 
Development of contact area prediction model, particularly for the shape rolling, for accurate force prediction 
The shapes of the contact areas for various rolling processes are complicated and are different in various pass sequence. Mathematical description of the actual shape and the projective contact area is critical for accurate force prediction. Contact area models for various pass sequences, such as RDOV, OVRD, SQOV, OVSQ, etc., were developed with the formulation of the actual shapes. 
Morgan 
70 
Development of shape factor prediction model for accurate force prediction 
Actual modeling for the shape factor needs to consider friction condition and groove geometry. My modeling for the shape factor started from an universal formula that covers all the pass sequences, created by some German colleagues based on the tens years of measurement and modeling. With rich experience results and mill field data at hand, I did verification, modification and simplification, so I established accurate models for RDOV, OVRD, RDRD (For Morgan Sizing Mills), SQOV, OVSQ, etc. The models fit both the low speed rolling and high speed rolling. 

71 
Development of forward slip models for RDOV and OVRD, for both lowspeed and highspeed rolling 
Neutral angle models for RDOV and OVRD were established based on the lab mill data and followed by necessary modification. On the basis, the forward slips were calculated with Wusatowski procedure. Predicted result fits excellently with the field measurement from ASW NTM, etc. 
Morgan 
72 
Rolling mill friction model development to improve Morgan?s design procedure 
The model considers factors such as type of groove, temperature, speed, steel grade, etc. As effects of temperature and steel grade, the metallurgical constituent of the scale and its property were also studied. 
Morgan 
73 
Development of accurate temperature prediction model and models for heat transfer coefficients 
The model is based on the energy balance between the stock and rolls. Heat transfer coefficients were based on the past measurement in various researches. Thermal properties were all in temperature dependence. 
Morgan 
74 
ASW NTM data processing and Model verification 
Processed ASW data pass by pass; used the models I had developed to repredict the measured data and so, to verify and improve the developed models. 
Morgan 
75 
Development of the microstructure model for steel hot rolling processes to predict recrystallization, grain size, etc. 
The model was developed by combining the major models (both formulas and data) published worldwide. Quite a portion of my $10,000 research funds was spent for acquiring publications. The model consists of the submodels to determine: (1) strain or time, for start and 50% completion of the dynamic and static recrystallizations; each of those may consist of two or more ways of prediction; (2) volume fractions of, and grain sizes after, the dynamical and static recrystallizations; and (3) grain growth, equivalent grain sizes at the start and end of a pass, and the equivalent interpass time under temperature modification. Over 10 formulas and overt 30 coefficients were used to describe the model of a steel grade. Microstructural models for over 10 steel grades were collected. Models were examined with data collected from publications. 
Morgan 
76 
Development of tension correction model to modify spread, forward slip and roll force 
Tension correction to spread and forward slip is critical for roll pass study of existing NoTwist Mill (NTM) because all stands are tied (driven by a single motor) and it is very hard to avoid tension. Without this correction, any field data from the NTM would not lead to any meaningful model. In the development, German results were accepted but Japanese formula (definition of tension and tension correction formulas) were used. However, Japanese results, which were based on products of larger size, didn?t fit the NTM data and thus were filtered out. 
Morgan 
77 
Development on interstand tension determination 
The development involved the definition of tension. Many German results on tension effects were based on the definition of tension as the relative difference of the speed. The actual tension, calculated from the specially modified flow stress model was used (even the best flow stress formula available still doesn?t work for a small strain below 0.05). 
Morgan 
78 
Modeling of the freecontours of the stock after rolling 
The free contour was modeled based on the measurement in Morgan?s lab mill data. This model was used for accurate calculation of the rolled crosssection area. 
Morgan 
79 
Flow Stress Modeling Program to create flow stress models based on the data in the flow stress database 
The program accesses database to read in the flow stresses in various strains, strain rates and temperatures, and calculates flow stress coefficients for temperature, strain and strain rate. The calculation was based on various criteria. See www.meta40.com/bli/L2Net/FSModel.htm. The program determines factors m1 to m5 and A1 to A3, and Kf0, through both linear and nonlinear regression. It also allow user to select special needs for modeling, such as minmidmax, best peak strain, best Rsquare, etc. 
Metal Data 
80 
Extensive data collection on mill test results from reports, publications, etc.; data storage in database 
Mill test results on rolling and controlled cooling, collected in past years as published or unpublished reports, published papers, books, Ph.D. dissertations, etc., were processed and stored in the relational database. Great portions of the data were roll pass related measurement, force measurement, controlled cooling data, rolled product properties and a portion of the microstructure data. Further data would be added into the database. 
Metal Data 
81 
Collection of flow stress data for about 2000 steel grades 
Flow stress data and models were collected from various available sources. Currently about 2000 models are available on the web. Other data, another 2000 sets, et., would soon be processed and uploaded. 
Metal Data 
82 
A Coordinate Measuring Tool to read data from curves (e.g. flow stress curves) 
This application was initially developed to measure flow stress from flow stress curves. The scanned picture with the curves is to be measured for coordinate. Mouse clicking is made against the points on the existing curve to draw another curve, in order to make sure the clicking is not off the line. For every click, the coordinate is recorded. The coordinate is immediate calculated into the physical value based on an initial setup against the coordinate system in the picture. The value from every click is displayed on the form for doublechecking. When the Submit button is clicked, the data is sent to the database. 
Metal Data 
83 
Compilation of data list for the metal properties (about 3000) 
Compiled data list for the metal properties. Particular focus was on high temperature properties while room temperature properties were also collected. 
Metal Data 
84 
Development of heat transfer coefficient models for rolling, controlled water cooling and controlled air cooling 
For interface between rolls and hot metals during rolling, for steel cooling in the air in the function of the travel speed and environment temperature, and for the controlled water cooling depending on water flow volume, water pressure, water temperature and turbulence, etc. For example, for the water box cooling alone, several hundred pages of the fieldtesting reports were used. High quality models were derived from the rich mill data. 
Metal Data 
85 
Development of spread model for various steel grades during various type of rolling 
Some stainless steel may have twice as high as spread as the plain carbon steels, while other stainless steels may have low spread. Different metals have different spread tendency. 
Metal Data 
86 
Development of forward slip model for various steel grades during various types of rolling 
This part of the model focused on the steel grade influence on the forward slip. It was the further development based on the unpublished writing of Dr. A. Hensel. Data for over 30 steel grades were developed. Not to be published but is available in the further mill projects. 
Metal Data 
87 
Friction model further development  to create a oneforall friction model for steel rolling 
Different materials have different friction values; for the same pair of the materials, friction depends on surface condition (roughness, amount and type of the lubricants, etc.), groove type, speed, and temperature. Temperature has different effects to friction in different temperature range. Formulas were collected and further developed with available new data. 
Metal Data 
88 
Publishing a book on the computer simulation of steel hot rolling process (Germany 1996) 
Compared Experimental and Theoretical Investigations of Forming Technical Parameters in Shape Rolling with Example of the Hot Rolling of Angle Steels. TU Bergakademie Freiberg, Freiberg, Germany, 1996 (in German). ISBN 3860120298. 

89 
Wrote a book  Steel Mill: Process Modeling and Computer Application 
Broad topics, with fundamentals, technologies and industrial project examples. The book is not yet intended for publishing. Contents may be available for clients or work colleagues. See www.meta40.com/bli/home/book2.htm (under updating). 
Metal Data 
90 
Rolling mill resource collection and processing 
Collected rolling mill modeling resources for over 20,000 pages; processed over 10,000 pages of the resources 
Metal Data 