Machine Learning-Based Prediction of Textural Properties and Nonlinear Regulatory Pattern Analysis of 3D-Printed...

The precision of 3D-printed food is dictated by the macroscopic textural stability of the dough system during extrusion. In this study, we investigated the nonlinear regulatory effects of Konjac Glucomannan (KGM) concentration and printing pressure on the textural properties of 3D-printed dough. Using a space-filling experimental design (n = 30), Support Vector Regression (SVR) and Gaussian Process Regression (GPR) models were developed to map the complex interactions between formulation and process variables. The results indicated that KGM concentration and printing pressure exhibit significant nonlinear coupling effects on hardness, cohesiveness, and chewiness. After 4-fold cross-validation and systematic hyperparameter optimization, the SVR model demonstrated satisfactory interpolative predictive performance within the investigated parameter space, achieving Rp2 values of 0.990 for gumminess and 0.987 for chewiness, while the GPR model effectively characterized the predictive uncertainty. Furthermore, the model predicted a favorable processing region (0.5-0.8% KGM and 4.0-4.6 bar) within the investigated design space. This research provides a quantitative, data-driven framework for the formulation pre-optimization of 3D-printed dough under specific experimental settings.