| Abstract Scope |
Surface signatures are found on objects fabricated by various materials processing methods. This is especially apparent in three-dimensional (3D) printing, where an object is built upon stacking of materials, resulting in the surface finish that often deemed as unappealing. However, useful information can be extracted from these build surfaces. Here, we propose to utilize the surface signature in 3D printing to inversely predict the paramount processing-microstructure-property (PMP) relationships with machine learning. By creating a novel framework of tetrahedral relationships: surface-processing-microstructure-property (S-PMP), we can optimise the printing process parameters much faster than the classical statistic method. We also demonstrate a powerful toolbox that rapidly retrieves PMP data with high fidelity by surface image inputs. These findings show the capability of machine learning in expanding the materials library, reducing waste, and revolutionising materials studies in 3D printing. The developed S-PMP framework can be implemented across and beyond various 3D printing methods. |