Prediction of nanoindentation creep behavior of tungsten-containing high entropy alloys using artificial neural network trained with Levenberg-Marquardt algorithm

This paper describes the synthesis of tungsten-containing high-entropy alloys (HEAs). The synthesis method involves a powder metallurgy process, and spark plasma sintering (SPS) is used to compact the powder. XRD and SEM analyses of synthesized HEAs showed that the main phases formed were body -centered and face-centered cubic phases, and a sigma phase was also observed after sintering at 900 celcius. Furthermore, nanoindentation showed that the introduction of tungsten in HEA resulted in high hardness and elastic modulus, which ranged from 8.31 to 13.57 GPa and 197.21-209.43 GPa respectively. The in-dentation creep behavior was ascertained at room temperature. The HEAs exhibited a significant bench-mark after the addition of a specific amount of W for further investigation because of their lower creep rate. Experimental creep displacement data were used for modeling by artificial neural networks (ANNs) in which the training has been performed by the Levenberg-Marquardt algorithm. The experimental creep displacement data and the ANN model predictions have an excellent agreement. The ANN model is reliable and can accurately forecast the room temperature creep behavior of HEAs. HEAs are promising candidates for use in elevated and wear-resistance applications, owing to their unique combination of high hardness and high creep resistance. (c) 2023 Elsevier B.V. All rights reserved.

Automated summary

This paper describes the synthesis of tungsten-containing high-entropy alloys (HEAs) using a powder metallurgy process and spark plasma sintering (SPS) to compact the powder. XRD and SEM analyses showed the main phases formed were body-centered and face-centered cubic phases, with a sigma phase observed after sintering at 900 celcius. The introduction of tungsten in HEAs resulted in high hardness and elastic modulus, and the experimental creep displacement data were accurately predicted by an artificial neural network (ANN) model.