Innovative 3D Printing of Mechanoluminescent Composites: Digital Light Processing Meets Machine Learning

Abstract

Mechanoluminescent (ML) materials, which refers to a class of material that emits light when subjected to external mechanical stimuli, have been drawing attention due to its unique multifunctionality, and potential applicability as a next-generation structural health monitoring technique. Nevertheless, the applicability of ML materials in real-world scenarios have been significantly confined due to its limitations, such as there is no universally accepted rules or framework for producing ML composites with high intensity, and the difficulty in producing the material in complex 3D shapes. As a breakthrough, here we present a novel approach where SrAl₂O₄:Eu²⁺, Dy³⁺ particle-based mechanoluminescent composite is produced via digital light processing (DLP)-based 3D printing, whose process parameters are optimized through a machine learning-based optimization algorithm. In this study, we adopt multi-objective Bayesian optimization (MBO) to optimize the three salient process parameters of DLP-based additive manufacturing; ML particle content, layer thickness, and cure ratio, to achieve both strong ML properties and short printing time. Gaussian process regression is used for the modeling of complex input-output relationship, and the training data is collected by performing actual experiments. As a result, the pareto-optimal process parameter solutions determined by MBO not only allowed us to produce the high-performance ML specimens in short time, but they also allowed us to empirically understand how the process parameters affect the end product’s ML property and the overall printing time. Furthermore, we validated the real-life applicability of our framework by applying the optimized DLP-based 3D printing framework to produce and test the ML-based stress sensors and ML-based mechanical components.