We demonstrate a neural network that predicts the physical properties of steels based on the composition and heat treatment. The neural network model was trained from a library of experimental data from 1000 alloys.
The first panel below contains the composition and heat treatment of a standard commercial steel. Click predict to get the neural network’s estimates for the yield stress, ultimate tensile strength, and elongation of this steel.
Now try adding a hardener, for example 4% silicon, to the composition, and re-predict the properties: you should see the ultimate tensile strength increase.
You can also click here to use this technology to optimise the yield stress, ultimate tensile strength, and elongation of the steel.
This same technology was used to understand nickel alloys where the composition covered 20 elements, 5 heat treatment parameters, and predicted 11 material properties. Click here to read more about this study.