Prediction of tensile performance for 3D printed photopolymer gyroid lattices using structural porosity, base material properties, and machine learning

• Published in: Materials & design Vol. 232; p. 112126
• Main Authors: Peloquin, Jacob, Kirillova, Alina, Rudin, Cynthia, Brinson, L.C., Gall, Ken
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.08.2023; Elsevier
• Subjects: Additive manufacturing; Diverse materials; Lattice structures; Machine learning; Mechanical performance; Diverse materials; Lattice structures; Additive manufacturing; Mechanical performance; Machine learning
• ISSN: 0264-1275; 1873-4197
• DOI: 10.1016/j.matdes.2023.112126

[Display omitted] •The ability of a simple machine learning model to predict mechanical properties of 3D printed lattices is demonstrated.•The effect of printing defects on photopolymer lattice sample mechanical and geometric properties is highlighted.•A framework is presented for including both structure and material info into property predictions of 3D printed lattices.•Kernel ridge regression is used to make accurate mechanical property predictions using a small amount of training data. Advancements in additive manufacturing (AM) technology and three-dimensional (3D) modeling software have enabled the fabrication of parts with combinations of properties that were impossible to achieve with traditional manufacturing techniques. Porous designs such as truss-based and sheet-based lattices have gained much attention in recent years due to their versatility. The multitude of lattice design possibilities, coupled with a growing list of available 3D printing materials, has provided a vast range of 3D printable structures that can be used to achieve desired performance. However, the process of computationally or experimentally evaluating many combinations of base material and lattice design for a given application is impractical. This research proposes a framework for quickly predicting key mechanical properties of 3D printed gyroid lattices using information about the base material and porosity of the structure. Experimental data was gathered to train a simple, interpretable, and accurate kernel ridge regression machine learning model. The performance of the model was then compared to numerical simulation data and demonstrated similar accuracy at a fraction of the computation time. Ultimately, the model development serves as an advancement in ML-driven mechanical property prediction that can be used to guide extension of current and future models.