Optimization of print fidelity, efficiency and texture of 3D printed gluten-free snacks using artificial neural network-genetic algorithm modeling

Abstract

3D food printing (3DFP) is emerging as a transformative technology for producing customized, nutritionally tailored foods, with growing potential for commercial applications in the health-oriented and gluten-free markets. In this study, gluten-free 3D-printed snacks were developed from proso millet flour, almond protein, and yeast protein, and the influence of infill percentage (I), extrusion multiplier (EM), printing speed (PS), and layer height (LH) on dimensional fidelity, texture, and process efficiency was systematically evaluated. EM was the dominant driver of dimensional accuracy (r ≈ 0.80 for both length and width), reflecting its critical role in controlling material flow and deposition precision. Post-processing preserved shape accuracy, although limited height variation occurred due to vapor-induced puffing. Denser structures exhibited lower baking loss (31.17–43.77 %), confirming density as the key determinant of moisture release. Hardness (3.38–38.62 N) was predicted with high accuracy using an artificial neural network–genetic algorithm (ANN–GA) model (R² = 0.968–0.981, RMSE = 1.271–1.903). A precise interplay between EM, I, PS, and LH was identified as essential for simultaneously optimizing dimensional stability, textural quality, and process efficiency. Sensitivity analysis identified EM (60.89 %) and I (27.18 %) as key positive contributors to textural quality, whereas LH (–8.86 %) and PS (–3.08 %) had negative impacts. ANN–GA optimization reduced printing time by 33 % (to 3 min) while maintaining product quality (CQ = 4.67), demonstrating an effective tool for balancing speed, fidelity, and structural stability.