PREDICTION OF STACKING FAULT ENERGY ON HIGH ENTROPY SUPERALLOY FENICRCOAL USING THERMODYNAMIC CALCULATION AND MACHINE LEARNING

• Main Author: Teja Sukma, Fauzi
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/74177
• Institution: Institut Teknologi Bandung

Most of the gas turbine components consists of coated nickel-based super alloys. The drawbacks are the relatively expensive manufacturing costs and price of Ni. Therefore, research is carried out to find new alloys as a substitute for super alloys. One of the alloys that has potential to match the performance of super alloys at a relatively lower price is the high entropy super alloy FeNiCrCoAl. The objective of this research is to study the effect of temperature and composition on the stacking fault energy of FeNiCrCoAl alloys with thermodynamic calculations and machine learning as suggestions for alloy design guidelines. The simulation is carried out by collecting the parameters used for thermodynamic calculations from the literature. These parameters are used to calculate the SFE value by utilizing the Matlab software. The results of these calculations become datasets for machine learning random forest models and decision trees. The dataset is divided into train data and test data with a ratio of 75:25. The model is trained using the train data while the test data is used to evaluate the performance of the trained model. Model performance is measured through several evaluation metrics such as root mean square (RMSE) and accuracy. Based on the simulations performed, an increase in temperature increases the value of SFE. At 300 K, addition of Ni atoms increased SFE from 89.34 mJ/m2 to 96.06 mJ/m2, addition of Al atoms increased SFE from 53.51 mJ/m2 to 131.41 mJ/m2, increased Cr decreased SFE from 107.89 mJ/m2 to 17.97 mJ/m2, increase in Fe increases SFE from 87.2 mJ/m2 to 98.3 mJ/m2 and increase in Co decreases SFE from 82.26 mJ/m2 to 63.18 mJ/ m2. Magnetic contribution to SFE occurs significantly at low temperatures. The best machine learning model is a random forest with an RMSE of 4.75 and an accuracy of 0.98. Design guide for HESA FeNiCrAlCo at 300 K, can follow the range of Ni 17-25 at%; Cr 24-35 at%; Al 5- 15 at%; Co 20-35 at%; Fe 15-25 at%.