New-Generation Level 2 System
Technology and Application Overview

Technology and Application Overview

1.   Microstructure Model, Intelligent Learning and High-end Software Engineering

In the mode of modern production using microalloyed steel, the addition of microalloyed composition makes the existing model not accurate enough, so it is necessary to add microstructure model, and adopt intelligent Learning and high-end software engineering. In addition, related technologies remove the common Learning logic defects of conventional Level 2 system.

This batch of model data is especially aimed at the situation that micro alloying elements are added to steel in recent years, which significantly improves various properties in combination with rapid cooling and large deformation under the condition of slight increase in cost, and enables the smooth production of hard, wide and thin products. The metallographic microstructure model, based on the use of a large number of microalloyed elements and the basic model required by machine learning, needs to be further modified. The Level 2 with these three characteristics can be called a new generation of Level 2.

The Level 2 is a huge large-scale software system, which was originally written by automation personnel. Therefore, insufficient consideration of the properties of materials leads to the logic problem of machine learning in terms of mechanical properties of materials; There may be more than 100000 sets of models based on chemical composition rather than use and corresponding model design.

The high-end software engineering enables the further developed computer windows module to understand the engineering problems, especially the continuous upgrading, which avoids the problem of throwing away the old Level 2 system and buying a new set every 5 to 10 years (because various changes make the old model no longer accurate)

2.  The Guided Two-Parameter Learning

The Guided Two-Parameter Learning can only use the parameters of deformation and deformation speed for Learning, and take the material parameters and temperature parameters into account through a large number of model design. The Learning of the resulting model avoids the problems that may be very accurate, such as large deformation coefficient parameters plus small deformation velocity parameters, or small deformation parameters plus large deformation velocity parameters. Japanese colleagues once praised the Learning method using deformation and deformation speed in published articles, and the Learning method of using a set of functions to describe the change trend of steel grade and temperature is very accurate; What we have adopted is not only one set, but more than 4000 sets! When the microstructure of materials is different, the customized design is far better than the general mathematical description! The model data is designed based on software and can be directly integrated into the Learning mechanism of each Level 2. Therefore, it is a low-cost and efficient industrial upgrading scheme.

3.  Field Application

The Level 2 of Oregon company is based on the version developed by NASA experts. After more than 4000 improvements in five years, it has been excellent in the optimal production of normal materials. However, when this system is used to optimize and control the wide, hard and thin (3.5m wide X80 high-strength steel 5mm thick) series products, the above problems are so prominent that there are defective products almost every day. After the optimization according to the above-mentioned new generation Level 2, the site responded that there had never been the same defective product during the return visit half a year later.

In addition to applying the optimization adopted by Evraz Oregon Steel, NISCO has made a series of new improvements, such as changing the steel grade from the previous classification according to purpose to the classification according to chemical composition, developing software, designing more than 4000 sets of models, optimizing the Two-Parameter self-learning under the guidance, and so on. For the 8000 ton equipment of the enterprise, the operators did not dare to operate when they saw the forecast on the screen to 4000 tons. After optimization, everything was normal and the utilization rate increased by 50%; The original situation that the wrong temperature and wrong steel grade can be used to make a useful production plan in the manufacturing of production procedures has also been completely improved, and the formulation of correct production procedures has been realized by using all real parameters. All this has increased the investment utilization rate from 25% to more than 90%!

This set of model-based technology has high parameter prediction accuracy, optimized production procedure logic and reasonable production optimization mechanism, which greatly improves the process technology and automation level of new variety development and significantly optimizes the production of high-end products.

Details can be found in the attached document: Level2 Model and New Product Development.

New Generation Level 2 System Development Project Series

1.  Improvement Of Level 2 Model In Medium And Heavy Plate Coil Plant Of NISCO

Summary / overview (Part I)

1.1 summary (Part II)
1.1.1 Development background
1.1.2 Technical introduction

1.2 Improvement of rolling force model of Level 2 in NISCO (Part III)
1.2.1 Problems and Countermeasures of self-learning logic in Level 2
1.2.2 Two parameter Learning under the guidance of FIT2G

1.2.3 Data collection and historical log processing (Part IV)
1.2.4 Steel grade integration
1.2.5 Model steel design

1.2.6 Steel family design (Part V)
1.2.7 Automatic design of rheological stress coefficient

1.2.8 Generation of steel grade documents (Part VI)
1.2.9 Improvement of temperature waiting pass model
1.2.10 Miscellaneous improvements
1.2.11 Source program modification

1.2.12 Solutions to the continuous increase of steel grades and the loss of steel grade documents (Part VII)
1.2.13 Onsite Testing
1.2.14 Rolling force prediction accuracy

1.3 Areas for further development (Part VIII)
   literature