Towards microstructure control in forging and rolling: combining AI with process models for closed-loop property control

被引:0
作者
Reinisch, Niklas [1 ]
Idzik, Christian [1 ]
Bailly, David [1 ]
机构
[1] Rhein Westfal TH Aachen, Inst Met Forming, Intzestr 10, D-52072 Aachen, Germany
关键词
metal forming; property control; process control; reinforcement learning; fast process models; microstructure; Umformtechnik; Eigenschaftsregelung; Prozessregelung; Reinforcement Learning; Schnelle Prozessmodelle; Mikrostruktur; MECHANICAL-PROPERTIES; PRODUCT PROPERTIES; OPTIMIZATION; PARAMETERS;
D O I
10.1515/auto-2023-0232
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Metal forming processes like open-die forging or hot rolling are well-established for the production of key components in various industries. Nevertheless, the control of the final microstructure and hence mechanical properties is not yet common. To achieve this, the authors propose and discuss a control concept based on reinforcement learning, fast process models (FPM) and an "operator in the loop" approach. The concept is explained and tested using deviating initial ingot temperatures as idealized process disruptions. RL algorithms are trained for both processes and transferred into the controllers that are connected to a simulative environment based on FPM. Within this framework, the online adaption is possible in similar to 2 s in rolling and 4-6 s in forging. This highlights the concepts suitability to be used for property control in hot metal forming. Umformprozesse wie das Freiformschmieden oder das Walzen sind etablierte Verfahren zur Herstellung von Komponenten mit exzellenten Eigenschaften f & uuml;r verschiedenste Industriezweige. Trotzdem ist die Regelung finaler Werkstoffeigenschaften wie der Mikrostruktur noch nicht & uuml;blich. Um diesem Ziel n & auml;her zu kommen, stellen die Autoren ein Regelungskonzept vor, das auf Reinforcement Learning, schnellen Prozessmodellen (FPM) und einem "operator in the loop"-Ansatz basiert. Das Konzept wird erl & auml;utert und durch die Annahme der idealisierten Prozessst & ouml;rung ,,abweichende Ausgangstemperatur im Werkst & uuml;ck" getestet. RL-Algorithmen werden f & uuml;r beide Prozesse trainiert und in die Regler & uuml;bertragen, die mit einer auf den FPM basierenden Simulationsumgebung verbunden sind. In diesem Rahmen ist die online Prozessadaption beim Walzen in similar to 2 Sekunden und beim Schmieden in 4 bis 6 Sekunden m & ouml;glich. Dies unterstreicht die Eignung des Konzepts f & uuml;r die Eigenschaftsregelung in der Warmumformung.
引用
收藏
页数:11
相关论文
共 28 条
[1]   Closed-loop control of product properties in metal forming [J].
Allwood, J. M. ;
Duncan, S. R. ;
Cao, J. ;
Groche, P. ;
Hirt, G. ;
Kinsey, B. ;
Kuboki, T. ;
Liewald, M. ;
Sterzing, A. ;
Tekkaya, A. E. .
CIRP ANNALS-MANUFACTURING TECHNOLOGY, 2016, 65 (02) :573-596
[2]   A soft sensor for property control in multi-stage hot forming based on a level set formulation of grain size evolution and machine learning [J].
Bambach, Markus ;
Imran, Muhammad ;
Sizova, Irina ;
Buhl, Johannes ;
Gerster, Stephan ;
Herty, Michael .
ADVANCES IN INDUSTRIAL AND MANUFACTURING ENGINEERING, 2021, 2
[3]  
Bertsekas D.P., 2019, Reinforcement learning and optimal control
[4]   A new multi-objective genetic algorithm applied to hot-rolling process [J].
Chakraborti, N. ;
Kumar, B. Siva ;
Babu, V. Satish ;
Moitra, S. ;
Mukhopadhyay, A. .
APPLIED MATHEMATICAL MODELLING, 2008, 32 (09) :1781-1789
[5]   Mapping the Global Flow of Steel: From Steelmaking to End-Use Goods [J].
Cullen, Jonathan M. ;
Allwood, Julian M. ;
Bambach, Margarita D. .
ENVIRONMENTAL SCIENCE & TECHNOLOGY, 2012, 46 (24) :13048-13055
[6]   Reinforcement learning for industrial process control: A case study in flatness control in steel industry [J].
Deng, Jifei ;
Sierla, Seppo ;
Sun, Jie ;
Vyatkin, Valeriy .
COMPUTERS IN INDUSTRY, 2022, 143
[7]   Reinforcement Learning Versus Model Predictive Control: A Comparison on a Power System Problem [J].
Ernst, Damien ;
Glavic, Mevludin ;
Capitanescu, Florin ;
Wehenkel, Louis .
IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS PART B-CYBERNETICS, 2009, 39 (02) :517-529
[8]  
Gamal O., 2021, INT J MECH ENG ROBOT, V10, P349, DOI [10.18178/ijmerr.10.7.349-356, DOI 10.18178/IJMERR.10.7.349-356]
[9]   Relations between Model Predictive Control and Reinforcement Learning [J].
Goerges, Daniel .
IFAC PAPERSONLINE, 2017, 50 (01) :4920-4928
[10]   Coupling of an analytical rolling model and reinforcement learning to design pass schedules: towards properties controlled hot rolling [J].
Idzik, C. ;
Kraemer, A. ;
Hirt, G. ;
Lohmar, J. .
JOURNAL OF INTELLIGENT MANUFACTURING, 2024, 35 (04) :1469-1490