Development Of Smart Manufacturing
Model System

Model Overview

The main technical advantage of our team lies in model development, and the focus of model development is basically consistent with the development trend of the industry. Sometimes, although it is a few years later, considering the development and maturity of the industry, it is also within the normal time range.

The team's model development began in 1990. At that time, the industry trend was mainly offline models, so the development of team members in their early years was also based on this. For example, the development of finite element model in DFG project in Germany, the development of model involving more than 100 sets of traditional industry material processing in Morgan in the United States in four years, the development of finite difference model in later website system, and the model data of high-temperature materials varying with temperature sold on the Internet, The development of Flow Stress model, etc. are all offline models.

Since the late 1990s, the era of online model has come, such as the Level 2 model and the model of a new generation of Level 2 unique to our team. The team's online model development is mainly combined with software development, because it involves machine learning and various software architectures.

1.  The development of DFG finite element model in Germany has created a world precedent for the factory application of finite element simulation in the thermal and mechanical simulation of complex and large deformation, and developed a set of full model and a set of simplified model suitable for today's machine learning. The simulation field involves a variety of large deformation of section steel.

2.  More than 100 sets of models developed by Morgan in the United States, including models for material deformation in all directions, force and energy demand, microstructure and mechanical properties prediction, etc.

3.  The model development of finite difference method is further optimized on the basis of the original software system of a world-class large company. With reference to the field measurement results of more than 10 books and more than 1000 pages, the temperature field at any time of more than 50 processes can be calculated in a few tenths of a second.

4.  Models for online high temperature data sales, such as a large amount of data related to Flow stress with temperature, deformation, deformation speed and material; Model data of mechanical and thermal properties varying with temperature at high temperature.

5.  Model data of mechanical and thermal properties at room temperature.

6.  In the online model, there are hundreds of models in the Level 2 of the production line, which involve all the details of the online control of the production process.

7.  The corresponding model development in a series of processing processes related to lithium battery, such as slurry stirring, coating, baking, rolling, slitting, winding, volume separation and so on.

8.   In a series of sections of PI film production process, the processing operation model is developed.

9.  Other relevant model development, design many models related to biotechnology.
 

Model Based Intelligent System Example

1.  German DFG Model Project

This is a finite element artificial intelligence simulation project, which is a national project in Germany. In order to certify that this set of simulation technology can be used in factory production in terms of calculation cost and model prediction accuracy and in predicting the deformation of large and complex section profiles.

2.  Empirical Model in Morgan (Primetal) in the United States

On the eve of graduation from Germany, our team leader was employed as a senior engineer of Morgan construction company (merged with Siemens as Primetal). The data he modeled comes from a large amount of data collected by the company from customers around the world for more than 100 years, the data collected in the five-year experiment conducted by the company itself, and a large amount of data (including more than 40 doctoral theses) of 15 years of research results of high-speed wire rod rolling brought by him from Germany, And in order to solve the problem of the company's leading high-speed wire rod mill in the world, we collected a large amount of data in various customer factories. For example, in Cleveland, the United States, morgan's engineers lived on the site for a few months and dug a hole in the high-speed wire rod mill machine to measure the data. These large amounts of data were finally integrated into his hands, so he developed three groups of more than 100 sets of models, including material deformation model series, material force energy and power requirement model series, and material microstructure and mechanical property prediction model series.

In rolling, the relative difference between the speed along the length direction when the material is pressed by the roll and the linear speed along the length direction of the roll is called forward slip. Like all other models, the forward slip model is based on the prediction of rolling process parameters. The predicted value of forward slip is compared with the measured value (see the figure below). The overall error is very small. Interestingly, for example, if the size is 0.668 (0.668 inch in diameter), the predicted value is between the two measured values, which seems to be more accurate than each measured value! In the production of high-speed wire rod, a motor drives 8-12 high-speed rolling mill stands. Only by adjusting the deformation parameters can the change of rolling piece speed be formed, so as to adjust the tension between stands, otherwise the rolling mill will vibrate!

Design and calculation of three groups of more than 100 sets of models was initially done with the data from German metal forming research institute, and than used data in USA Morgan. The accurate model calculations in the pass design were verified and improved based on projects of South Korea POSCO and China TISCO, etc. The metal forming research institute where the team leader studied for Ph.D. established the four stands of continuous rolling mill which can reach the rolling speed of 70 m/s in the 1980s. After decades of research and development, there are more than 40 doctoral theses; Morgan (now merged with Siemens as Primetal) has carried out laboratory rolling for five years, and based on a large amount of data collected by the company from customers all over the world for more than 100 years. Based on this, three groups of more than 100 sets of models were developed, including material deformation, force energy, microstructure and mechanical properties.

3.  Level 2 and New-Generation Level 2 Model

The team improved the Level 2 by integrating the metallographic model into the Level 2. Firstly, the materials were optimized and classified according to chemical composition rather than purpose, and the problem of continuously increasing new variety management was solved. In particular, combined with the principle of metallography, separate models are designed for each variety (including each steel grade, each size specification, each production mode and each blank size), and a set of models are designed for each variety in high, medium and low temperature areas for the high accuracy of the model. Thus, thousands to tens of thousands of models can be designed for each plant. At present, these data models for various varieties have been successfully applied to Evraz Oregon Steel of the United States and NISCO of China. For example, the products of high-strength steel 3500mm wide and 5mm thick of Evraz Oregon Steel of the United States have been improved from almost daily defective products to no similar defective products for half a year (customer evaluation during half a year's return visit), and various high-strength and difficult rolling varieties of NISCO have been rolled smoothly.

More about this model system can be seen in the in-depth discussion.

4.  Finite Difference Method (FDM) Model Based On Temperature Field Calculation

The finite difference method calculation of a complete set of temperature simulation from rolling to water cooling in cooling. Based on the finite difference method, the calculation model of temperature, rolling and water cooling temperature used in Morgan for many years is further optimized, which can accurately calculate the temperature field of rolled piece at any time in nearly 20 roughing and intermediate rolling passes, air cooling between passes, subsequent water cooling, 8-12 pass high-speed rolling and Stelmor process. The model is established by using various temperature measurement data collected by Morgan for more than 100 years, so as to ensure the accuracy of the calculation process. Morgan collected a total of more than ten books and thousands of pages in the field.

Morgan (now merged with Siemens) has carried out laboratory rolling for five years, and based on a large amount of data collected by the company from customers all over the world for more than 100 years. Based on this, I developed three groups of more than 100 sets of models, including material deformation, force energy, microstructure and mechanical properties.

The calculation of Morgan's Stelmo production line (appeared in University Textbooks) fully takes into account the temperature distribution in the steel coil and the difference between the core and surface temperature of the steel wire rod.

The simulation is carried out according to the production process data recorded on site. The simulation results are basically consistent with the measured results of production records, and there is a diagram of comparative data. In the figure, the calculated and measured values of surface temperature during more than ten passes of hot rolling and subsequent water cooling are compared. In production, only the surface temperature can be measured.

5.  More Than 100 Models Of Emerging Manufacturing Industry

More than 100 models have been established based on the data of Germany and the United States during the Morgan period; Besides, after entering the emerging industry, it has also developed more than 100 sets of models of emerging industries, such as slurry preparation, coating, baking, rolling, slitting, winding and other models related to lithium battery manufacturing, as well as models of various sections of PI film production. All projects of the team start with model development. Through machine learning based on online data, the model is extremely accurate on the corresponding production line, and then develop intelligent systems according to various needs.

6.  Material Deformation Model For Die Design

Taiyuan Iron and steel company once asked me to run the online software. The on-site personnel compared the data obtained from the operation with the on-site measured data. The consistency found surprised the personnel of TISCO! There are also two sets of simulation software for wire and bar design simulation. The first set allows non designers to design field data, and only the terms of field operation appear on the interface; The second set takes into account all possible influencing factors. Both sets are based on a large number of model operations, hiding behind the complex model prediction. There are also a large number of force and energy parameter calculation models, and there are dozens of such software on Metal Pass Internet alone.

The verified models work excellently for Die design, which employs the large number of established process models.

List of manufacturing model development and consulting projects