STACKING FAULT ENERGY PREDICTION USING THERMODYNAMIC CALCULATIONS AND MACHINE LEARNING MODEL FOR COCRNIAL MEDIUM ENTROPY ALLOY

• Main Author: Bagastantya Widiastika, Nyoman
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/67906
• Institution: Institut Teknologi Bandung

After the discovery of high-entropy alloys in 2004, many researches on these alloys began. Many industries have benefited from the discovery of this alloy, such as the transportation, energy, and health industries. In Indonesia, industrial development and research of medical devices is one of the national priority programs to overcome dependence on imported medical devices. The superior mechanical and physical properties of high and medium entropy alloys, especially CoCrNiAl, have the potential to be developed as biomaterials for medical devices. Setting the SFE value of high and medium entropy alloys can affect the mechanical properties of the alloy through its deformation mechanism. To determine the value of the alloy SFE on certain parameters, simulations with thermodynamic calculations and machine learning were carried out to increase the effectiveness of the research. This study aims to determine the effect of composition and temperature of the alloy, determine the best machine learning model, and provide suggestions and guidelines for the design of CoCrNiAl medium entropy alloys. The simulation begins by determining the parameters used for thermodynamic calculations taken from several literatures. After that, the parameters and thermodynamic calculations are inputted into the MATLAB program for calculations. The results of these calculations are then used as a database that will be supplied to two machine learning models, namely decision trees and random forests, to be studied. After studying the data provided, the model is then asked to predict the SFE value both to test data from the database and independent test data from the literature. The performance and accuracy of the model are then measured using the calculation of the mean average error (MAE) and also the confusion matrix diagram to determine the best model. Based on the simulations that has been done, as the temperature increases, the SFE value will also increase in each simulated composition. Based on thermodynamic calculations, at a temperature of 10 K: the addition of Co atoms can reduce the SFE value from 77.45 mJ/m2 to 35.5 mJ/m2, the addition of Cr atoms can reduce the SFE value from 109.87 mJ/m2 to 0.36 mJ /m2, the addition of Ni atoms can increase the SFE value from 30.65 mJ/m2 to 60.23 mJ/m2, and the addition of Al atoms can increase the SFE value from -41.33 mJ/m2 to 85.07 mJ/m2. The best model obtained from the two models used is the random forest model with an MAE value of 0.29612 and an accuracy of 99.73%. To design a CoCrNiAl alloy that has a medium value of SFE, it can follow the range of at-Al 14% – 31.67%, at-Ni 5% – 28.67%, at-Cr 26.67% – 34%, at-Co 22% – 31.67% at a temperature of 300 K.