| Home | My Account | Post Article | Upload Showcase | Site Shortcuts | Feedback | About us |

  Home > Consulting > Level 2 Model

Seven Tech
Pat. Warn
Pat. Level2
Prj cases
Model Sys
New L2
Ind. Softw.
Key Prjs
Smart Equip
Model Prjs
Li Batt Ind.
Steel Ind.


Defect Warn

Equip Intelli

Domain Sale


- Class Time
- Classes
- Resources
- Lrn Method
- Status Need
- Metaverse
- Li Batt Tech
- Equip Softw.
- Intelli System

Save Money with Better Tech

Product Defects and Level 2 Model Error  (Ask a question)

Rolling mill Level 2 model creates pass schedule (draft distribution and stage plan) based on a long list of predicted parameters such as
force and temperature. When the Level 2 model goes wrong, it causes various problems that usually leads to product defects.

Product Defects due to Missing Production Targets

When the mill Level 2 model goes wrong, there would be all kinds of quality and operation problems, repeatable or unrepeatable. The draft schedule would be inappropriate and the rolling operation would be shifted away from the planned rolling conditions (temperature, draft force, torque, and number of passes, etc.). Consequently, targeted properties would be missed, product shape would be poor. In worst case, there could be equipment damage.

Examples of the problems caused by a poor Level 2 Model:

  1. Center buckle
  2. Edge wave
  3. Kinks, knuckles, missed slots, etc.
  4. Other types of bad finish shape
  5. Too few passes (with the risk for equipment damage and product shape defect)
  6. Too many passes (with low efficiency and poor temperature profile)
  7. Rolled product properties lower than usual
  8. Geometry variation (over thickness, width or length)

If the Level 2 log (files or database tables) is examined, it would show that the troubled passes are usually accompanied with parameter prediction error (in force, temperature, and roll deformation, etc.). Level 2 model creates pass schedule based on predicted parameters for the purpose of meeting various targets. If one or more targets are missed, it indicates problems with product defects. If the parameter prediction is wrong, the targets cannot be met unless there is a very lucky situation that two or more wrong factors make a right one. In the following sections the commonly pursued targets are discussed.

Equal Deformation Target

The Level 2 model combines roll ground crown, roll wear, roll thermal crown, roll deflection, mill stretch and roll flattening, etc., to form an environment that the draft across the stock width is equal. This is very critical in the finishing passes when the strip/plate is thin. Unequal deformation across width would cause either center buckle (if the deformation in the width center is too high) or edge wave (with too large deformation in the edges). The roll deformation is very sensitive to roll separating force. Error in the force prediction or the roll deformation modeling would easily let this target be missed. In addition, because the temperatures in the head end and tail end are usually lower, the higher roll force leads to larger deformation for the roll and stand (mill stretch) and consequently, a smaller roll gap is needed for the head end and tail end, in order to achieve equal deformation among the head end, tail end and body. Inaccurate prediction or inappropriate handling such issues may cause various shape defects (kinks, knuckles, missed slots, etc.). One of the worst cases is that the rolling process in the finish pass may be dragged into the two-phase region due to temperature prediction error, etc., during which a huge force error, sometimes over 40%, causes significant roll deformation errors and very poor geometry shape to the thin strip/plate.

Metallurgical Temperature Target

The temperature of the stock during rolling is affected by two factors contrary to each other: heat loss through heat transfer and heat gain from deformation energy. Any variation of the draft and rolling speed, etc. would lead to temperature variation. When the temperature error is high, the system would miss the metallurgical targets such as those for controlled rolling to improve rolled steel properties. Consequently, the product properties would be poorer than usual. In certain situation, as mentioned above, the finishing stage rolling may be actually conducted in the austenite/ferrite two-phase region due to the temperature error, though the rolling was initially scheduled above the two-phase temperature region. This could lead to severe shape defect, as discussed above. Most Level 2 systems do not consider metallurgical effects. For the rolling in the two phase region, with lower temperature, the measured flow stress may also go lower because of a higher fraction of ferrite, which is much softer than austenite.

Assume the temperature 1000C is pursed based on the metallurgical principles. If the Level 2 model has an error of 50C, then both 1050C and 950C could be regarded as 1000C by the Level 2. The consequence for this is that, the production may not be in the optimal condition (1000C). Suppose two production lines A and B. A has Level 2 potential error of 80C while B with possible error of 10C.The optimal temperature is 1000C. After rolling, products from the line B would have better properties than A, because metallurgical strengthening and other optimization measures are better achieved in the product line B. The actual temperature in the A line is 920-1080C, in B line 990-1010C.

Mill Capacity and Productivity Target

As long as the Metallurgical Temperature Target and the Equal Deformation Target are satisfied, the Level 2 model would schedule as few passes as possible within the mill capacity, in order to achieve high productivity. In a Level 2 system, if the number of passes is too many, the reason is likely that the predicted force is too high, or the Metallurgical Temperature Target or the Equal Deformation Target are hard to reach, among others. Usually, a small draft makes it easier to achieve the Equal Deformation Target; this is at least a partial reason that the finishing passes are usually with smaller draft than rest passes. To be noted is that for some alloy steels, amount of draft should also be within the formability limit to avoid possible crack.

For further discussion on the product defect vs. Level 2 model error, please review:

Common Sources of Level 2 Model Error

Some examples are listed below:

  1. Poor tuning, especially in the initial stage of the system installation. In this stage, learning is not yet fully established due to insufficient history data and thus the accuracy is not yet high enough. A poorly designed Level 2 model needs longer learning establishment period than usual.
  2. Poorly applied properties and other input parameters. This would lower down the accuracy. Most property data should be used as temperature-dependent.
  3. Weak system met with tough production environment (hard grade, thin gauge, low temperature, etc.). In those tough conditions, poorly designed Level 2 models may fail.
  4. Learning logics. Blind adaptive learning has certain limitation. In force learning, the strain and strain rate are actually not totally independent; this reduces the quality of the adaptive learning. With added logics or improved input data, this limitation can be removed. See related discussion in this paper.
  5. Model error or inaccuracy. Inaccurate models for e.g. roll deflection; flow stress model with a narrow valid range, which often has significant error in the finish pass.
  6. Lack of consideration for slab variation (temperature, size, etc.) and strip variation (body/head, body/tail).
  7. Various technical issues, which would cause significant error if not handled correctly.

To be mentioned is that, many error sources of Level 2 model interactively affect each other. When the force error is big, the temperature, often calculated based on force prediction, may also be inaccurate. An unreasonable draft schedule would further downgrade the prediction quality of various parameters, e.g. temperature and roll deformation.

Related Resources

Metal Data Work List on Level 2 and Mill Modeling

Metal Data Recent Publications

Technical papers published in the February and March of 2008. Those publications are primarily on Level 2, Level 2 model and process automation. 

Metal Data has dozens of research reports and some software applications available at no charge for the consulting clients (only for the projects in the name of Metal Data and approved by Dr. Benjamin Li).


  | Private Policy | Terms & Conditions | About Us | AdvertisePartnerInvestorSponsorlistings |  
Copyright: 2022 Metal Pass LLC. All right reserved