Digital Light Processing (DLP) is a photopolymerization-based 3D printing technique known for its rapid fabrication speed, high resolution, and compatibility with a wide range of photocurable materials. By precisely adjusting resin compositions and mixing ratios, DLP enables tuneable mechanical properties, supporting applications from biocompatible hydrogels in medicine to ionic and hyperelastic elastomers for soft robotics.
Gray-scale Digital Light Processing (g-DLP), an advanced variant of DLP developed around 2016 from microelectromechanical systems and tissue engineering research, introduces pixel-level control over material properties. In g-DLP, variations in light intensity modulate the degree of monomer conversion and, consequently, the local crosslinking density. This allows continuous mechanical gradients to be printed directly from a single resin vat. The result is a cost-effective and versatile approach for fabricating structures with programmable mechanical behaviour, improved dimensional accuracy, and enhanced toughness – all within a single printing process.
However, g-DLP faces critical constraints. Photocurable resins offer limited property tunability, and structural optimization for complex geometries remains underexplored. Commercial resins typically impose a trade-off between viscoelastic damping and elastic modulus.
Polyurethane acrylate (PUA) resins (common DLP materials) contain dynamic covalent bonds that dissipate energy without excessive chain elongation, maintaining the low viscosity required for printing, yet their elastic moduli typically range from only a few MPa to a few hundred MPa, insufficient for mechanically demanding applications. Meanwhile, g-DLP requires spatially controlled elastic moduli through adjusted crosslinking to achieve robust designs.
Determining optimal placement and extent of these property gradients, along with corresponding grayscale values, necessitates advanced structural optimization – an ideal application for machine learning.
Meeting these challenges is Prof. Miso Kim, an alumnus of Massachusetts Institute of Technology, USA, and her team at Korea Advanced Institute of Science and Technology, Republic of Korea. They are using DLP printing technology to create mechanical metamaterials, producing ceramic composites for flexible sensing arrays and developing highly dense and precise ferroelectric ceramic structures, and they currently focus on g-DLP.
Prof. Kim’s team have developed a two-pronged solution. First, they created a new polyurethane acrylate resin system that dramatically expands the stiffness range — from 8.3 MPa all the way to 1.2 GPa — while keeping excellent damping properties. They achieved this by designing two building blocks: a soft segment with disulfide bonds and a hard segment based on hydroxyethyl acrylate. By mixing these in different ratios, they produced composites spanning a wide stiffness range while maintaining the low viscosity needed for DLP printing.
Second, they constructed a machine learning-driven multi-objective Bayesian optimization framework to generate gradient structures and corresponding grayscale masks for g-DLP printing. The optimization targets stress concentration reduction and effective stiffness enhancement. The adaptive framework employs a two-phase iterative approach: (i) weighted sum strategy for improved design generation, and (ii) Pareto front refinement to maximize the hypervolume. The iteratively generated solutions were then evaluated using Finite Element simulations to support both the optimization process and the failure behaviour evaluation of the printed structures.
To demonstrate real-world potential, the team applied their approach to artificial cartilage subjected to repeated compression and automotive bumpers tested under impact. Both applications showed significant mechanical improvements, validating the framework’s versatility.
Future directions include exploring functional resins for g-DLP beyond PUA systems, and optimizing gradient structures for time-dependent loading conditions to enhance adaptive responses under dynamic mechanical environments. Expanding the material alternatives while refining optimization algorithms could further broaden industrial applicability across sectors.
The integration of composite chemistry with artificial intelligence-driven structural optimization represents a significant advancement in additive manufacturing. This synergistic approach, combining molecular design, photopolymerization control, and computational optimization, provides a blueprint for next-generation 3D-printed materials with application-specific mechanical performance.
Reference: J. Nam, B. Chen, M. Kim, Machine Learning-Driven Grayscale Digital Light Processing for Mechanically Robust 3D-Printed Gradient Materials Advanced Materials (2025), DOI: 10.1002/adma.202504075
Featured Image Credit: Macey11 via Pixabay
