Construction of a Predictive Model for Assessing the Porosity of TiNi Materials Based on Sintering Parameters of Samples

Additive manufacturing of TiNi alloys has developed significantly in recent years. Such growth is due to the unique properties of the resulting designs − superelasticity, bending strength − stretching, corrosion, chemical and thermodynamic stability. In addition, three-dimensional (3D) printing technology significantly expanded the scope of use of this material in various applied areas from biomedicine (for printing implants and bone tissue engineering), to, for example, the automotive and aerospace industries. Despite the fact that 3D printing technology already has commercial solutions that allows producing complex-shaped structures, this area is in the stage of development. Of a particular interest is the optimization of printing process and the development of a quality control system (achieving the required parameters of porosity, roughness, minimizing defects such as cracks, contamination, delamination, surface and subsurface heterogeneities, surface quality, shrinkage, etc.). In this paper, an original non-destructive method for assessing the sintering temperature as an indirect measure of porosity of a material using optical coherence tomography (OCT) and machine learning methods is proposed. Random Forest Regression approach was used to predict depth maps and spatial statistics from the sintering temperature. The Random Forest model based on the first- and the second-order texture metrics of OCT images provided R ² of about 0.95 for sintering temperature prediction.