Genetic algorithm optimization of magnetic properties of Fe-Co-Ni nanostructure alloys prepared by the mechanical alloying by using multi-objective artificial neural networks for the core of transformer. (September 2021)

• Record Type: Journal Article
• Authors: Zeraati, Malihe; Arshadizadeh, Razieh; Chauhan, Narendra Pal Singh; Sargazi, Ghasem
• Is Part Of: Materials today communications. Volume 28(2021)
• Publication Date: 2021-09
• Subjects: Genetic algorithm -- Artificial neural network -- Multi-objective -- Iron Cobalt Nickel -- Nanostructure alloys; Materials science -- Periodicals; 620.11
• DOI: 10.1016/j.mtcomm.2021.102653
• Languages: English
• ISSNs: 2352-4928

Abstract

Abstract: In this study, Fe-Co-Ni nanostructure ternary alloys were prepared by mechanical alloying and its magnetic and structural properties were also evaluated. Multi-objective artificial neural networks (ANN) and genetic algorithms (GA) have been used to optimize and improve the magnetic properties of products. The weight percentage of Fe, Co, and Ni as alloying element, milling times and speed annealing times and temperature as well as a ball to powder ratio (BPR) were selected as input parameters. Meanwhile, grain size, magnetization saturation (μs ) and coercivity (hc ) of Fe-Co-Ni nanostructure alloys were considered as output parameters. GA was introduced to the established models of multi-objective ANN. Proposed optimum condition as a candidate for transformer core is a combination of highest μs as (222.9) emu/g, lowest grain size as (9.6 nm) and hc as (5.9 Oe) with the root mean squared error (RMSE) lower than 0.9%. Furthermore, the sensitivity analysis results confirmed that the weight percentages of Ni, BPR, and the weight percentages of Ni and BPR are the most effective parameters on μs, hc and grain size respectively. Highlights: Fe-Co-Ni nanostructure alloys were prepared by mechanical alloying and properties are optimized by multi-objective artificial neural networks and genetic algorithm. Its magnetic properties are optimized by multi-objective artificial neural networks and genetic algorithm. Its optimized composition showed μs (222.9) emu/g, grain size (9.6 nm) and hc (5.9 Oe) with the root mean square error (MSER) lower than 0.9%.