Integrated in-line transmission NIR monitoring and PLSR-based composition prediction for starch-based 3D food printing.

Three-dimensional (3D) food printing enables personalized and sustainable fabrication, but extrusion stability and shape fidelity remain sensitive to formulation and processing conditions, while quality is still commonly evaluated off-line. In this study, in-line near-infrared (NIR) spectroscopy was integrated into extrusion through a custom transmission-mode nozzle for real-time monitoring during deposition. Separate concentration series of corn starch, κ-carrageenan, and D-sorbitol were evaluated to determine their individual effects on printability and structural quality. Partial least squares regression (PLSR) enabled accurate in-line prediction of all three components. Among the tested concentrations, 20.0 wt% corn starch showed the clearest preservation of the intended internal architecture in the representative X-ray CT cross-section, whereas CT-visible irregularities, including unintended voids within nominally deposited regions, were more evident overall above 22.0 wt%. A κ-carrageenan concentration of 0.5 wt% best preserved the designed internal openings, whereas concentrations above 0.8 wt% showed a greater tendency toward coalescence between adjacent deposited strands. D-sorbitol improved moisture retention, but concentrations over 4.0 wt% caused visible lateral inclination of the printed towers, indicating reduced post-deposition self-support. The in-line predictions tracked intentionally introduced concentration disturbances and supported measurement-guided manual intervention during printing, demonstrating the feasibility of real-time compositional monitoring and providing a basis for future automated feedback control.