Potential role of machine learning techniques for modeling the hardness of OPH steels

0560556 - ÚFM 2023 RIV NL eng J - Journal Article
Khalaj, O. - Ghobadi, M. - Zarezadeh, A. - Saebnoori, E. - Jirková, H. - Chocholaty, O. - Svoboda, Jiří
Potential role of machine learning techniques for modeling the hardness of OPH steels.
Materials Today Communications. Roč. 26, MAR (2021), č. článku 101806. ISSN 2352-4928. E-ISSN 2352-4928
R&D Projects: GA ČR(CZ) GA17-01641S
Institutional support: RVO:68081723
Keywords : Oxide Precipitation Hardened (OPH) steels * Hardness * Heat treatment * Artificial neural network (ANN) * anfis * Fe-Al-O
OECD category: Materials engineering
Impact factor: 3.662, year: 2021
Method of publishing: Limited access
https://www.sciencedirect.com/science/article/pii/S2352492820328178?via%3Dihub

Oxide Precipitation Hardened (OPH) alloys are a new generation of Oxide Dispersion Strengthened (ODS) alloys which have outstanding mechanical properties based on using appropriate heat treatment (HT). The production consists of mechanical alloying, which leads to a ductile matrix and hard oxide dispersion, however, the initial state shows a fine grain structure and basic mechanical properties. The composition, production process parameters, and HT affect the hardness of the OPH. In order to obtain a better understanding of the hardness of OPH alloys, three machine learning techniques were developed using ANN, ANFIS and SVMR to simulate the hardness. Moreover, the importance and intensity of the impact of each parameter on the hardness of OPH alloys were discussed. Based on the experimental results achieved by the authors, the composition of OPH alloys (Al, Mo, Fe, Cr, Ta, Y, and O), HT conditions, and Mechanical Alloying (MA) was used to train the models as inputs. The validity of the models was measured by different statistical criteria such as R, k, k', m, n, and Rm. The mean absolute error for prediction of the hardness values at the test set was about 32 HV (ANFIS model), 37 HV (ANN model), and 44 HV (SVMR model). The results demonstrated that the ANFIS model predicts more accurately than the other methods. The sensitivity analysis and the influence of valid parameters were studied for the ANFIS model. It was revealed that HT temperature has a great effect on the hardness of the OPH alloys.
Permanent Link: https://hdl.handle.net/11104/0333426