Prediksi Sifat Mekanik Aluminium Berdasarkan Unsur Kimia Paduan Menggunakan Model Machine Learning

Abstract

This study designs and compares optimal machine learning models for predicting the mechanical properties of alloyed aluminum with code 1050 (A91050) based on the percentage composition of chemical elements using the software RapidMiner with decision tree and Random Forest algorithms. The aim is to develop a data-driven predictive model with high accuracy to minimize the need for physical testing on aluminum with various compositional variations. The machine learning modeling in this study involves nine input variables, comprising chemical elements such as Al, Mg, Zn, Ti, Cu, Mn, Cr, Fe, Si, and two output or target variables, YS and TS (yield strength and tensile strength). Additionally, a Heatmap correlation is employed to observe the correlations between the chemical elements and the mechanical properties of the alloyed aluminum. The comparison of these algorithms reveals that Random Forest (RF) outperforms other algorithms in predicting YS with a Mean Absolute Error (MAE) of 7.157, Root Mean Square Error (RMSE) of 11.248, and a coefficient of determination (SC) of 0.977. On the other hand, Random Forest (RF) also exhibits better performance in predicting TS with an MAE of 29.296, RMSE of 42.382, and SC value of 0.443.