Search for: All records

The development of new thermoplastic-based nanocomposites for, as well as using, 3D printing requires extensive experimental testing. One typically goes through many failed, or otherwise sub-optimal, iterations before finding acceptable solutions (e.g. compositions, 3D printing parameters). It is desirable to reduce the number of such iterations as well as exclude failed experiments that often require laborious disassembly and cleaning of the 3D printer. This issue could be addressed if we were able to understand, and ultimately predict ahead of experiments if a given material can be 3D printed successfully. Herein, we report on our investigations into forecasting the printing and resultant properties of polymer nanocomposites while encompassing both material properties and printing parameters, enabling the model to generalize across various thermoplastics and additives. To do so, nanocomposites of two different commercially available bio-based PLAs with varying concentrations of nanoclay (NC) and graphene nanoplatelets (GNP) were prepared using a twin-screw extruder. The thermal and rheological properties of the nanocomposites were analyzed. These materials were printed at varying temperature and flow using a pellet printer. The quality of the printing was evaluated by measuring weight fluctuation, internal diameter of cylindrical specimen, and surface uniformity. The interactions between material properties and printing parameters are complex but captured effectively by a machine learning model, specifically we demonstrate such a predictive model to forecast printability and, printing quality utilizing a Random Forest algorithm. Printability was predicted by developing a classification model with constraints based on the weight fluctuation (W ) of the printed sample w.r.t. the optimal print; defining ‘‘not printable’’ for −1.0 d W < −0.8 and ‘‘printable’’ for W e −0.8. The classification model for predicting printability, performed well with an accuracy of 92.8% and identified flow index and complex viscosity, contributing 52% to the model’s importance. Another model to predict W of the only on successful prints also showed strong performance, emphasizing the importance of viscoelastic properties, thermal stability, and printing temperature. For diameter change (Di), the Random Forest model identified flow consistency index, complex viscosity, and thermal stability as influential parameters, with crystallization enthalpy gaining increased importance, reflecting its role in crystallization and shrinkage. In contrast, the surface roughness average (RA) model had lower performance, yet revealed remarkable insights regarding the feature importance with crystallization enthalpy and complex viscosity being most significant.