A neural network model for high entropy alloy design

Abstract

A neural network model is developed to search vast compositional space of high entropy alloys (HEAs). The model predicts the mechanical properties of HEAs better than several other models. It’s because the special structure of the model helps the model understand the characteristics of constituent elements of HEAs. In addition, thermodynamics descriptors were utilized as input to the model so that the model predicts better by understanding the thermodynamic properties of HEAs. A conditional random search, which is good at finding local optimal values, was selected as the inverse predictor and designed two HEAs using the model. We experimentally verified that the HEAs have the best combination of strength and ductility and this proves the validity of the model and alloy design method. The strengthening mechanism of the designed HEAs is further discussed based on microstructure and lattice distortion effect. The present alloy design approach, specialized in finding multiple local optima, could help researchers design an infinite number of new alloys with interesting properties.

Machine Learning Design for High-Entropy Alloys: Models and Algorithms

“…J. Wang et al [73] developed ensemble NN models and algorithms to test input data in order to design HEAs with a higher yield strength (YS) and ultimate tensile strength (UTS) (see Figure 1). They collected 501 data points from previous studies to be used for NN model training and validation.…”

Lattice Distortion Promotes Incipient Plasticity in Multiprincipal Element Alloys

“…The second-nearestneighbor modified embedded-atom method (2NN MEAM) interatomic potentials were employed to simulate indentations on the M-1−M-10 systems. 30,31 To assess any possible errors associated with parametrization, additional nanoindentation simulations on the E-1, E-2, E-1-AA, and E-2-AA systems were performed by EAM potentials 32,33 and average-atom (AA) potentials. 34 The effective Young's modulus, E s , for each system was obtained by fitting the F−h curves based on the classical linear elastic Hertzian contact mechanics relation 26…”

“…No amorphous phases were found in these systems (Figure S3). The second-nearest-neighbor modified embedded-atom method (2NN MEAM) interatomic potentials were employed to simulate indentations on the M-1–M-10 systems. , To assess any possible errors associated with parametrization, additional nanoindentation simulations on the E-1, E-2, E-1-AA, and E-2-AA systems were performed by EAM potentials , and average-atom (AA) potentials …”