An end-to-end AI framework for data-driven evaluation and optimisation of direct ink writing-based 3D printing

Abstract

3D food printing (3DFP) has emerged as a transformative technology in healthcare and food sectors, offering various benefits such as tailored nutrition and texture. The development of food inks with high printing quality is essential for Direct Ink Writing (DIW)-based 3DFP; however, existing methods largely rely on trial-and-error for ink optimisation and subjective visual inspection for printing quality assessment. In this study, we present an end-to-end artificial intelligence (AI) framework that integrates data-driven characterisation and machine learning (ML) to optimise food ink formulations for DIW-based 3DFP. A quantitative metric named total deviation for printing quality assessment is established based on 3D scanning and directionality analysis. A multimodal artificial neural network (ANN) model that processes both visual information and numeric data is trained to directly correlate compositional data with printing quality, which accurately predicts optimal formulations when combined with a genetic algorithm (GA) for search in high-dimensional compositional spaces. Experimental validations demonstrate good agreement between predicted and measured printing quality. Overall, this AI-enabled streamlined process enables data-driven optimisation of food inks with minimal experimental burden. The framework is applicable to other DIW-based 3D printing processes, offering a solution to accelerate formulation development.