About this Abstract |
Meeting |
MS&T24: Materials Science & Technology
|
Symposium
|
Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process
|
Presentation Title |
Study on Thermal Cracks in Steel Slab Using Neural Networks Model to Predict Impact Absorption Energy |
Author(s) |
Kyung-Chul Cho, Gi-beom Kim, Sang-Hum Kwon, Chang-Hee Yim, Dae-Geun Hong |
On-Site Speaker (Planned) |
Kyung-Chul Cho |
Abstract Scope |
In this study, a machine-learning model was developed to predict impact absorption energy (EIA) of steels by considering data about the temperature and alloying elements, then a criterion to predict the occurrence of thermal cracks in slab produced in an operating continuous-casting plant was established by analyzing several cases of crack occurrence. The optimized Deep Neural Network algorithm was the most accurate, with root mean square error of 10.82 J. At TS = 250 °C, cracking did not occur if the steel had predicted EIA > 175 J. Thermal cracking of the slab was caused by high ductile-to-brittle transition temperature and low EIA due to the effect of precipitation of Nb and Ti, and to matrix structure caused by high C content of steel components. |