Novel Analytical and Machine Learning Framework Predicting LPBF Heat Treatment Effects on Inconel ™ 718 Hardness

In Laser Powder Bed Fusion (LPBF), precise control over the melting of metal powders is essential for producing high-performance components from materials like Inconel ™ 718. This study aims to develop a robust analytical and machine learning framework for predicting the hardness of LPBF-fabricated Inconel ™ 718 after undergoing heat treatment and surface polishing. Utilizing particle swarm optimization and genetic algorithms, we established the relationships between various input parameters and the resulting hardness, measured by the Vickers hardness test (HV20). We implemented five distinct regression models, including support vector machine, Gaussian process regression (GPR), single-layer and deep-layer artificial neural networks (ANNs), and random tree. The innovative rollback technique was applied to enhance the precision of the output data. Among the models assessed, GPR and ANNs yielded the highest performance, with R², NSE, and KGE values between 0.97 and 0.99, alongside reduced MAPE, MAE, and RMSE on test data, confirming their efficacy as predictive tools. Furthermore, the Kruskal-Wallis test and Taylor diagram were employed to systematically evaluate model performance. This model not only streamlines the prediction process but also propels advancements in additive manufacturing utilizing LPBF technology.