PREDICTING ALUMINA COMPOSITES’ MECHANICAL CHARACTERISTICS USING A MACHINE LEARNING APPROACH

Abstract

Obtaining the requisite properties in alloys is crucial problem in the production of aluminium components, requiring great deal of time and effort for investigation and experimentation. In this study, machine-learning technique utilizing Bayesian-fine tuned Adaptive Gated Recurrent Unit (B-AGRU) to forecast the mechanical characteristics of aluminium alloys is presented. Training and testing are conducted on dataset, which has undergone comprehensive preparation process that includes cleaning and Z-score normalization. Principal Component Analysis (PCA) is used for feature extraction to increase algorithmic efficiency. The GRU approach, which is implemented in Python, hardness and yield strength, leading in more accurate findings. When compared to standard methodologies, process saves significant time and energy, as evidenced by metrics such as RMSE-20%, MAE-10% and R-squared-97%. This study reveals B-AGRU-based machine learning as a feasible strategy for enhancing efficiency and sustainability in forecasting mechanical properties of aluminium alloys, paving the way for wider application in industrial sector.

Keywords

• aluminium alloys
• mechanical characteristics
• machine learning
• bayesian-fine tuned adaptive gated recurrent unit (b-agru)
• tensile strength
• hardness