STACKING FAULT ENERGY PREDICTION AND CLASSIFICATION USING THERMODYNAMIC CALCULATIONS AND MACHINE LEARNING FOR FECRNIAL MEDIUM ENTROPY ALLOY DESIGN

• Main Author: Farhan Syahputra H, Muhammad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/69089
• Institution: Institut Teknologi Bandung

Metals and alloys have evolved from simple to complex composition to improving functions and performance. After the discovery of high-entropy alloys (HEA) in 2004, many studies and researches on these alloys. The rapid progress of our social economy brings forward new demands and challenges to the engineering industry, such as transportation and energy industries. It becomes critical to develop of new material with exceptional mechanical properties such as ductility and strength. Stacking Fault Energy (SFE) value effects the mechanical properties of the alloy through its deformation mechanism. SFE value of alloy on certain parameter can be determine by simulations with thermodynamic simulations and machine learning. This study aims to determine the effect of composition and temperature of the alloy, determine the best machine learning model, and provide guidelines for the design of FeCrNiAl medium entropy alloy. The simulation begins by determining parameters and collecting thermodynamic data taken from several literatures. Then, parameters and thermodynamic calculations are inputted into MATHLAB program for calculation. The result is SFE value, then it comes as database for machine learning models, namely logistic regression model, random forest model and support vector machine model. The data from database was divided into training data and test data with ratio 80:20. Training data are trained for machine learning models. The test data and independent data predict the classification of SFE value. The performance of machine learning model is measured using precision score, recall score, accuracy score and with ROC-AUC curve. Based on simulation results, it is known that changes in composition and temperature of the alloy will affect the SFE value of medium entropy alloy. As the temperature increase, the SFE value will also increased in each simulated composition. Based on thermodynamic calculation, at temperature of 270 K: the addition of Al can increase SFE value from 27.92 to 120.4 mJ/m2, the addition of Cr can reduce SFE value from 146.31 to 52.23 mJ/m2, the addition of Ni can increase SFE value from 70.91 to 104.65 mJ/m2, the addition of Fe can reduce SFE value from 103.10 to 79.49 mJ/m2. The best model obtained from three machine learning model is support vector machine model, with accuracy 0.999, macro precision score 0.9914, macro recall score 0.9995 and AUC score 1. To design FeCrNiAl medium entropy alloy at temperature 300 K, the range of each element: at-Al 5-25%, at-Cr 22-35%, at-Ni 5-35%, at-Fe 15-35%. The example of composition FeCrNiAl medium entropy alloy is Fe25, Cr 35, Ni25, Al 15 with SFE value is 38,71 mJ/m2.