Predicting the extrudability of complex food materials during 3D printing based on image analysis and gray-box data-driven modelling

Material extrudability is a prerequisite for printing 3D structures with extrusion-based food printing. Food materials with different shear-thinning behaviors were printed to develop predictive models for extrudability. A dataset of 131 unique combinations of materials and printing parameters was collected. Image analysis was employed to rapidly quantify extrudability as determined by mode width, line height and width consistency of line filaments. The relation between the printing pressure and the volumetric flow rate followed a power-law relation, which characterizes the extent of shear-thinning of the food materials. Both regression and classification models were trained and tested using the random forest algorithm. The model performance indicated that extrudability can be predicted with moderate to high accuracy by using rheological measurements and printing parameters as inputs. The predictive workflow developed in this study provides a framework to quantitatively assess and predict extrudability for 3D printing of complex food materials. Industrial relevance: Extrusion-based 3D printing has been applied to customize food designs and can potentially enable personalized nutrition. To achieve this, we need to be able to effectively print complex food materials with high accuracy. Variations in composition and rheological properties of complex food materials make achieving proper extrudability not trivial at this moment, and this limits their applicability to extrusion-based food printing. Here, we developed an image analysis tool to measure line filament extrusion of complex food materials that vary in shear-thinning behaviors. We then built predictive models to estimate material extrudability based on material's shear-thinning properties and printing parameters. Data-driven prediction of material's extrudability can help avoid excessive trial-and error experiments in future