Resumo

ROQUE-VILLALONGA, Gabriel  e  CAMARAZA-MEDINA, Yanán. Empirical modeling of thermal conductivity for a group of steels. Dyna rev.fac.nac.minas [online]. 2022, vol.89, n.224, pp.156-164.  Epub 30-Mar-2023. ISSN 0012-7353.  https://doi.org/10.15446/dyna.v89n224.103879.

The relationship between chemical composition and working temperature of the steel are not linear with the thermal conductivity, so empirical models are proposed for its prediction. Measurements were made on 32 rolled and annealed AISI steel markings. The K-Nearest Neighbor machine learning algorithm was used; in addition, a neural network was trained using the RStudio software, specifically the caret library, to obtain an empirical model that allowed predicting, with an adequate level of uncertainty, the thermal conductivity in the temperature range from 0-800℃. The model was tested with a group of values reserved for this purpose, obtaining low levels of uncertainty. The best results are obtained by training a neural network with 25 neurons in the hidden layer and a regularization value of 0,001, obtaining an error of 5,4% and an RMSE of 0,0228.

Palavras-chave : steel; empirical modelling; thermal conductivity; machine learning.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )