IJPEM-GT
AI-Based Mechanical Properties Prediction of Sintering-Aging Combined Effect in Metal-Material Extrusion
양대종, 문승기/Nanyang Technological University
- Keywords : Metal-Material Extrusion, 17-4PH stainless steel, Sintering and aging, Process parameter optimization, Multi-layer perceptron
- Metal-Material Extrusion (M-MEX) offers geometric flexibility and cost-effectiveness for 17-4PH stainless steel components, but complex processing interactions hinder mechanical property prediction and industrial adoption. This study introduces a comprehensive framework to predict mechanical properties of 17-4PH stainless steel components fabricated via M-MEX. The proposed framework comprises three phases. First, systematic physical characterization quantifies the effects of sintering and aging on density, grain size, and tensile performance. Second, polynomial regression-based data augmentation expands the experimental dataset and captured non-monotonic strength trends below the threshold sintering temperature at 1330 °C and monotonic decreases above the same threshold temperature. Third, a two-stage AI architecture predicts intermediate properties (density and grain size) and final outcomes (yield strength and ultimate tensile strength) using a group-specific Multi-Layer Perceptron. Model performance metrics include coefficient of determination (R2) of 0.987 and root mean square error (RMSE) below 15 MPa for both strength measures. Three-dimensional response surfaces reproduce the observed trade-offs among densification, -ferrite formation, grain-boundary strengthening, and over-aging. The proposed framework highlights narrow processing windows necessary for peak mechanical performance. The proposed framework reduces reliance on costly experiments and supports multi-objective parameter selection for optimized component design. This approach demonstrates that combining data augmentation with neural network modeling can efficiently tailor M-MEX processing parameters for high-performance 17-4PH stainless steel components.