AI-driven modeling and prediction of mechanical properties of additively manufactured Al-6061/B4C composite

This study investigates the effects of friction stir processing on the mechanical and damping properties of Al-6061 aluminum alloy, reinforced with boron carbide (B₄C) nanoparticles. A CNC milling machine was used to conduct friction stir processing, varying key processing parameters such as rotational speed, feed rate, and the number of passes. The mechanical properties analyzed include ultimate tensile strength, yield strength, natural frequency, and damping ratio. An advanced machine learning approach was implemented using a long short-term memory model optimized with the sine cosine algorithm to predict the processed material’s attributes. The experimental findings demonstrate that friction stir processing significantly enhances damping characteristics due to grain refinement, with the highest damping efficiency observed at 1400 rpm. Higher rotational speeds resulted in a notable increase in yield strength, attributed to finer grain structures. The introduction of B₄C nanoparticles further improved damping properties. Additionally, the study found that an increased number of friction stir processing passes decreased shear modulus and natural frequency while increasing the loss factor and damping ratio. The developed machine learning model achieved high predictive accuracy, with R² values of 0.981 for the ultimate tensile strength, 0.991 for YS, 0.973 for natural frequency, and 0.995 for damping ratio. The special relativity search-optimized long short-term memory model outperformed other approaches, attaining R² values ranging from 0.961 to 0.998 during training and 0.919 to 0.992 during testing. These findings highlight the effectiveness of friction stir processing in enhancing material properties and the superior predictive capability of machine learning models in capturing the effects of processing parameters.