Title: A deep learning model for the accurate prediction of the microstructure performance of hot rolled steel
Authors: Bin-bin Wang; Yong Song; Jing Wang
Addresses: University of Science and Technology Beijing, Engineering Technology Research Institute, Beijing 100083, China ' University of Science and Technology Beijing, Engineering Technology Research Institute, Beijing 100083, China ' University of Science and Technology Beijing, Engineering Technology Research Institute, Beijing 100083, China
Abstract: The prediction of microstructure performance can guide the adjustment of parameters during hot rolling. Scholars from all over the world has developed physical metallurgical models of rolling process based on the physical and thermodynamic characteristics of strip steel, but the prediction accuracy of the model is greatly affected by the complex production environment. In recent years, neural network method is used to build the prediction model of organisational performance. However, the prediction accuracy and robustness of the single hidden layer neural network model are poor. Deep learning method is introduced in this paper to establish the prediction model of hot rolling microstructure performance in this paper. The application results show that compared with the traditional model, the prediction accuracy of the hot rolled steels yield strength, tensile strength and elongation increased by 3.46%, 2.35%, and 5.11%, respectively. [Submitted 13 March 2019; Accepted 27 October 2019]
Keywords: auto encoder; deep learning; hot rolled steel; microstructure prediction; steel properties.
International Journal of Manufacturing Research, 2021 Vol.16 No.3, pp.262 - 279
Received: 13 Mar 2019
Accepted: 27 Oct 2019
Published online: 04 Oct 2021 *