Data-Driven printability mapping of laser powder bed fusion of Inconel 718 using AIMPOD

Printability mapping in laser powder bed fusion (LPBF) of Inconel 718 remains challenging because relative density is governed by coupled effects of laser power, scan speed, hatch spacing, layer thickness, beam size, powder characteristics, atmosphere, and printed geometry. Conventional trial-and-error process development and single-factor energy-density rules are often insufficient to identify stable processing windows across heterogeneous LPBF datasets. This study develops a data-driven printability framework focused exclusively on relative density using the Artificial Intelligence-based Manufacturing Process Optimization and Design (AIMPOD) workflow. A dataset of 241 LPBF conditions was analyzed, with relative density used as the response variable. The investigated window covered laser power from 90 to 400 W, scan speed from 100 to 2000 mm/s, hatch spacing from 0.05 to 0.25 mm, and relative density from 81.44% to 99.94%. Among the evaluated machine learning models, Gradient Boosting gave the best individual test performance, with R² ≈ 0.69 , RMSE ≈ 0.84 percentage points, and MAE ≈ 0.61 percentage points. The optimized ensemble improved the test performance to R²=0.72 and RMSE $$=0.79$$ percentage points. Explainability analysis identified scan speed, laser power, and hatch spacing as the dominant predictors. The developed AIMPOD framework provides an interpretable route for constructing LPBF printability maps of Inconel 718 and scales to data-driven manufacturing applications.