Issue |
Metall. Res. Technol.
Volume 116, Number 2, 2019
|
|
---|---|---|
Article Number | 201 | |
Number of page(s) | 12 | |
DOI | https://doi.org/10.1051/metal/2018053 | |
Published online | 14 January 2019 |
Regular Article
Modelling of viscosity of fluorine-free mold fluxes using neural network
1
School of Materials Science and Engineering, UNSW Australia,
2052
Sydney, Australia
2
School of Metallurgy and Material Engineering, Chongqing University of Science and Technology,
401331
Chongqing, PR China
3
College of Mechanics and Materials, Hohai University,
211100
Nanjing, PR China
4
Steelmaking Research Department, Baosteel Group Corporation Research Institute,
201900
Shanghai, PR China
* e-mail: j.q.zhang@unsw.edu.au
Received:
3
April
2018
Accepted:
31
May
2018
Viscosity is an important property of mold fluxes for steel continuous casting. However, direct measurement of viscosity of multi-component systems in a broad range of temperatures and compositions is an onerous work and has some limitations. This paper developed a model using the back propagation (BP) neural network to describe the viscosity of fluorine-free mold fluxes. The BP neural network model was developed and validated using 70 experimental values of viscosity of fluorine-free mold fluxes CaO-SiO2-Al2O3-B2O3-Na2O-TiO2-MgO-Li2O-MnO-ZrO2; 51 of them were used for developing the neural network model and the rest 19 viscosity data for the model validation. Calculated viscosities were in a good agreement with the experimental data. Based on the developed model, the effects of temperature and composition on the viscosity of fluorine-free fluxes were predicted and discussed.
Key words: mold flux / viscosity / neural network modelling
© EDP Sciences, 2019
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.