Year: 2026
Type: article
Abstract: Manufacturing systems increasingly generate heterogeneous, high-dimensional data, yet many existing artificial intelligence (AI) frameworks remain problem-specific, weakly interpretable, and insufficiently adapted to temporal dependence or material-specific variability. This limits the reliable deployment of AI for process optimization, quality prediction, and future autonomous decision-making in digital manufacturing. To address this gap, this study develops AIMPOD (Artificial Intelligence for Manufacturing Process Optimization and Design), an explainable AI framework that unifies data conditioning, intelligent preprocessing, multi-model learning, ensemble optimization, and interpretability within a single manufacturing-aware workflow.The framework is demonstrated on two distinct case studies: time-aware energy prediction in steel manufacturing and material-aware relative density prediction in laser-beam powder bed fusion of metals (PBF-LB/M). For temporally ordered industrial data, AIMPOD applies leakage-aware chronological validation, while for heterogeneous additive manufacturing data, it adopts material-split modelling to preserve alloy-specific process–structure relationships. Across the steel case, nonlinear and ensemble models achieve test \(R^2\) values between approximately 0.98 and 0.99, while the optimized ensemble reduces RMSE to about 0.63 kWh, improving on the best individual model by roughly 20–25%. In the PBF-LB/M case, pooled materials’ dataset modelling reaches about \(R^2_{\text {test}} \approx 0.76\) , whereas material-specific learning improves performance to about 0.87 for 18Ni300 and 0.78 for CuCrZr. The results demonstrate that AIMPOD enhances predictive robustness while preserving interpretability and providing an adaptive pathway toward explainable and agent-ready manufacturing intelligence.
Source: Scientific Reports
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: gold
Estimated APC paid: $2,690
Funder: King Saud University
Award: G602733861
OpenAlex ID: W7220952436