Machine learning-aided lattice optimization for ultra-lightweight 3D-printed aligners
Poom Narongdej, Bradley Yuhasz, Tianmin Kong, Diego Rojas, Qidi Peng, Ehsan Barjasteh
Abstract
The widely used vat photopolymerization, such as stereolithography (SLA) and digital light processing (DLP), is a type of 3D printing to create solid orthodontic models for thermoforming clear aligners that generates a significant amount of plastic waste, posing a considerable environmental challenge. This study addresses this issue by investigating the use of additively manufactured photopolymer lattice structures for orthodontic thermoforming models, aiming to optimize physical strength while minimizing material usage and waste. To accelerate the design and optimization process, a novel AI-based model, specifically a model-free voting ensemble, was developed to predict the most significant physical- and mechanical-property parameter of various lattice structures. Eleven lattice geometries were evaluated experimentally for mechanical and physical properties, including weight, modulus of elasticity, ultimate tensile and compressive strengths, strains, and specific energy absorption (SEA). The AI model’s predictions were validated with experimental results, demonstrating high accuracy. Results indicate that the Octet, Iso-Truss, and Weaire-Phelan (WP) structures are top performers, with WP offering the highest compressive SEA, though it was excluded due to its weight. Octet and Iso-Truss emerged as ideal candidates, demonstrating substantial weight savings of up to 59.86 % and resistance to high pressures with minimal deformation, positioning them as efficient alternatives to traditional solid models. The factor importance analysis from the AI model highlighted the modulus as the critical parameter, with the experimental results further validating the analysis. The integration of lattice structures and this predictive AI model offers a promising, resource-efficient approach to manufacturing thermoforming models for dental aligners, reducing material usage, waste, weight, processing time, and production costs.