Autor: |
Koleva-Petrova, M., Kulina, H., Dobrev, G., Gocheva-Ilieva, S. |
Předmět: |
|
Zdroj: |
AIP Conference Proceedings; 2023, Vol. 2953 Issue 1, p1-12, 12p |
Abstrakt: |
The study of steels is essential for material science and a wide range of engineering applications. We applied two powerful ensemble machine learning methods-Random Forests and CART-Ensembles and Bagging to predict the hardness of alloy steels. The objective was to determine the dependence of hardness on the chemical composition, the product's diameter, the temperatures and type of treatment. Highly efficient models with these four groups of predictors have been constructed, cross-validated and evaluated. Additionally, the initial dataset was randomly divided into two 50:50 subsamples. One subsample was used for training and the other as a test. It has been shown that the sample size does not affect significantly the quality of models. The constructed models predict the hardness data up to 96-98%, and for the holdout test sample-up to 86%. This approach demonstrates great opportunities for preliminary assessment of hardness with application in practice. [ABSTRACT FROM AUTHOR] |
Databáze: |
Complementary Index |
Externí odkaz: |
|