Artificial neural network helping in designing elastoplastic properties of alloys
A paper saw light in Metals.
Amorphous alloys are widely used in machine manufacturing and other industires, says research lead, Professor Anatolii Mokshin, "They are used in seismic sensors, manometer membranes, speedometers, accelerometers, and torque sensors in automaking, clock springs, weights, tire cords, cores of high-frequency transformers, etc. Amorphous metal alloys are of interest because of their durability, such as failure resistance, elasticity, and hardness. That's why there is always a need for modeling instruments to find alloys with desired durability properties."
A neural network has been presented by the team to analyse empirical data on properties of chemical elements. Among such data are atomic mass, charge, and electronegativity.
"As part of suggested methodology, an alloy is 'assembled' from various elements with their ratios determined. The neural network then analyses information on chemical and physical properties of the elements comprising such alloys, and determines mechanical properties of the alloy. All that only takes a couple of minutes," adds Associate Professor Bulat Galimzyanov.
According to the scientist, as a result of such studies it was found that the strength of amorphous alloys based on Cr, Fe, Co, Ni, Nb, Mo and W with additives of semimetals ( Be, B, Al, Sn, etc.) and non-metals (Si and P, etc.), as well as lantanoids (La, Gd, etc.) is higher than alloys based on other elements.
Engineer Maria Doronina explains, "Data on over 50 thousand amorphous alloys was used to teach the neural network. The network doesn't require information about the coordinates and speeds of atoms, which frees us from a very labor-intensive process of stucture analysis. The mean relative error is about 12 percent, which is a good result."
More information:
Neural Network as a Tool for Design of Amorphous Metal Alloys with Desired Elastoplastic Properties
www.mdpi.com/2075-4701/13/4/812
Provided by Kazan Federal University
