Mohanraj Murugesan, Saurabh Gairola, R. Jayaganthan, Palaniappan Ramu
Metal Additive Manufacturing (AM) offers significant opportunities for fabricating lightweight components with unique properties, but challenges remain in predicting the mechanical performance of alloys under elevated temperature and varying strain-rate conditions. The hot tensile behavior of additively manufactured AlSi9Cu3 alloy was systematically investigated using a Zwick-Roell Z100 universal testing machine, addressing a gap in the literature where no prior ML-based or experimental studies exist for this alloy at elevated temperature. Tests were performed from room temperature to 270 °C under 0.001 to 0.1 s −1 strain rates, providing a comprehensive dataset for constitutive and Machine Learning (ML) modeling. Constitutive models, including the Johnson–Cook, modified Johnson–Cook, and modified Zerilli–Armstrong, were developed to describe the material’s thermomechanical response. While these models captured certain trends at specific conditions, their predictive capability was limited under strong thermal softening. To address these limitations, ML approaches were employed, with multi-layer perceptron and support vector regression models trained on the experimental data. These ML frameworks provided higher accuracy and generalization across both interpolation and extrapolation conditions, demonstrating their capability to overcome the limitations of constitutive formulations. Fractography revealed a transition from mixed-mode fracture with retained melt pool features at lower temperatures to ductile, recrystallization-assisted failure at elevated conditions. Overall, the integration of experimental testing, constitutive modeling, and ML-based prediction provides new insights into the deformation behavior of AM AlSi9Cu3 under hot tensile loading. The findings highlight the limitations of constitutive models and underscore the potential of ML for reliable, data-driven prediction of mechanical performance in metal AM.