1  DESIGN OF CoCrFeMnNi HIGH-ENTROPY ALLOY WITH STACKING-FAULT ENERGY MODELLING USING THERMODYNAMIC CALCULATION AND MACHINE LEARNING FINAL PROJECT This Manuscript is Submitted as a Requirement for Bachelor Degree in Metallurgical Engineering, Faculty of Mining and Petroleum Engineering, Bandung Institute of Technology Arranged by: DWINA MAHIRA NAHDI 12520058 UNDERGRADUATE PROGRAM IN M ETALLURGICAL ENGINEERING FACULTY OF MINING AND PETROLEUM ENGINEERING BANDUNG INSTITUTE OF TECHNOLOGY 2024 ii  VALIDATION SHEET DESIGN OF CoCrFeMnNi HIGH-ENTROPY ALLOY WITH STACKING-FAULT ENERGY MODELLING USING THERMODYNAMIC CALCULATION AND MACHINE LEARNING FINAL PROJECT DWINA MAHIRA NAHDI 12520058 Bandung, August 2024 Has been approved for Undergraduate Metallurgical Engineering ITB By: Tria Laksana Achmad, S.T., M.T., Ph.D. Supervisor iv  DESAIN PADUAN ENTROPI T INGGI CoCrFeMnNi DENGAN PERMODELAN STACKING FAULT ENERGY MENGGUNAKAN PERHITUNGAN TERMODINAMIKA DAN MACHINE LEARNING ABSTRAK Paduan telah berevolusi secara progresif. Paduan entropi tinggi,yang terdiri dari setidaknya lima elemen dengan masing-masing berkisar antara 5—35 at.% dan ∆S mix ≥ 1,5R, menawarkan sifat mekanik yang menjanjikan. CoCrFeMnNi, paduan entropi tinggi yang pertama kali ditemukan, dikenal karena sifat mekaniknya yang sangat baik. Paduan ini cocok untuk aplikasi suhu tinggi di aerospace dan nuklir, elektronik, kesehatan, dan coating pada suhu kamar, serta tangki penyimpanan LNG pada suhu rendah. Kekerasan merupakan sifat penting dari paduan yang nilainya berkaitan dengan stacking-fault energy (SFE). Optimalisasi komposisi paduan berdasarkan SFE di berbagai suhu memerlukan eksperimen yang masif sehingga diperlukan pendekatan yang lebih efisien. Studi ini bertujuan untuk memberikan panduan desain paduan CoCrFeMnNi berdasarkan hasil analisis SFE melalui perhitungan termodinamika dan pendekatan machine learning yang sebelumnya hanya dilakukan pada komposisi dan temperatur terbatas. Pada studi ini, SFE dari paduan CoCrFeMnNi dihitung secara termodinamika menggunakan persamaan Olsen-Cohen dalam MATLAB pada rentang komposisi 5—40 at.% dan suhu dari 1 hingga 1200 K. Hasil yang diperoleh dianalisis untuk mempelajari pengaruh komposisi elemen, suhu, dan kontribusi magnetik terhadap SFE dan digunakan sebagai input untuk model machine learning—random forest (RF), support vector machine (SVM), dan neural network (ANN)—yang dilatih di Google Colaboratory. Root mean square error (RMSE) dan accuracy digunakan untuk menentukan model terbaik, sedangkan features importance dan spearman correlation digunakan untuk mengidentifikasi pengaruh dan korelasi suhu serta komposisi terhadap SFE. Hasil analisis menghasilkan rekomendasi desain praktikal untuk paduan CoCrFeMnNi. RF, SVM, and ANN dapat memprediksi nilai SFE dengan baik. ANN merupakan model terbaik dengan nilai RMSE terendah sebesar 1,02 dan akurasi tertinggi sebesar 98,8%. SFE menunjukkan korelasi positif dengan peningkatan suhu dan kontribusi magnetik sangat tinggi pada suhu rendah tetapi menurun signifikan seiring dengan peningkatan suhu. Penambahan unsur Fe dan Ni meningkatkan nilai SFE, sedangkan penambahan unsur Co, Cr, dan Mn menurunkan nilai SFE. Dengan demikian, komposisi yang direkomendasikan pada suhu kamar (300 K) untuk mencapai deformasi twinning terdiri dari 5—40 at.% Co, 5—37 at.% Cr, 13—40 at.% Fe, 5—36 at.% Mn, dan 5—40 at.% Ni. Komposisi paduan juga dapat merujuk pada rumus CrFeNi (100-x-y)CoxMny, FeMnNi(100-x-y)CoxCry, dan CoCrFe(100- x-y) NixMny. Co16Cr36Fe16Mn16Ni16 dan Co19Cr24Fe19Mn19Ni19 (tinggi Cr), serta Co 19Cr19Fe19Mn24Ni19 dan Co18Cr18Fe18Mn28Ni18 (tinggi Mn), adalah contoh komposisi menarik dengan SFE lebih rendah daripada komposisi equiatomik dan deformasi twinning pada suhu kamar. Kata kunci: paduan entropi tinggi, CoCrFeMnNi, stacking-fault energy, perhitungan termodinamika, machine learning v  DESIGN OF CoCrFeMnNi HIGH ENTROPY ALLOY WITH STACKING FAULT ENERGY MODELLING USING THERMODYNAMIC CALCULATION AND MACHINE LEARNING ABSTRACT Alloys have evolved to support civilization with improved properties and performance. High-entropy alloys (HEAs), containing at least five elements each ranging from 5 to 35 at.% and a ∆S mix ≥ 1.5R, show promising mechanical properties. CoCrFeMnNi HEA, the first ever found HEA, is known for its excellent mechanical properties, making it suitable for high-temperature aerospace and nuclear applications, room-temperature electronics, biomedic, and coating, and low-temperature LNG storage. Hardness, an essential alloy property, is related to stacking-fault energy (SFE). Optimizing alloy composition based on SFE across wide temperatures demands extensive experimentation, necessitating a more efficient approach. This study aims to provide CoCrFeMnNi HEA design guidelines by analyzing SFE for wide compositions and temperature ranges through thermodynamic calculation and machine learning approaches, which have not been previously conducted. In this study, SFE of CoCrFeMnNi HEAs were thermodynamically calculated using the Olsen-Cohen equation in MATLAB across a compositional range of 5—40 at.% and temperatures from 1 to 1200 K. The results were analyzed to investigate the effect of element composition, temperature, and magnetic contributions on SFE and used as inputs for machine learning models—random forest (RF), support vector machine (SVM), and neural network (ANN)— trained in Google Colaboratory. Root mean square error (RMSE) and accuracy were used to determine the best model, while feature importance and Spearman correlation identified the influence and correlation of temperature and composition on SFE, leading to practical design recommendations for CoCrFeMnNi HEAs. RF, SVM, and ANN were applicable for predicting SFE. Among the models investigated, ANN emerged as the best model, with the lowest RMSE value of 1.02 and the highest accuracy of 98.8%. Overall, SFE increased with increasing temperature. Magnetic contributions are extremely high at low temperatures but decrease significantly as temperature increases. The addition of Fe and Ni elements increases SFE values, while the addition of Co, Cr, and Mn elements decreases SFE values. Therefore, the recommended compositions of CoCrFeMnNi HEA at room temperature (300 K) to achieve twinning deformation consists of 5—40 at.% Co, 5—37 at.% Cr, 13—40 at.% Fe, 5—36 at.% Mn, and 5—40 at.% Ni. The composition of the HEAs can also refer to CrFeNi (100-x-y)CoxMny, FeMnNi(100-x-y)CoxCry, and CoCrFe (100-x-y)NixMny formulas. Co 16Cr36Fe16Mn16Ni16 and Co19Cr24Fe19Mn19Ni19 (high Cr), as well as Co 19Cr19Fe19Mn24Ni19 and Co18Cr18Fe18Mn28Ni18 (high Mn), are examples of attractive compositions with lower SFEs than the equiatomic composition and deformation by twinning at room temperature. Keywords: high-entropy alloy, CoCrFeMnNi, stacking-fault energy, thermodynamic calculation, machine learning.