Evaluation of direct ink write processing parameter window via machine learning

Abstract

Direct ink writing 3D printing offers significant advantages for fabricating soft materials. It is a process of forming solid parts from extruded line structures, so the uniformity and stability of the line is critical for successful printing. This study mainly focuses on the process parameter window for achieving uniform and stable printed lines. Four machine learning models (support vector machine, backpropagation neural network, decision tree, and K-nearest neighbour) are built to predict the success of printing. These models exhibited robust predictive capabilities, particularly the backpropagation neural network and support vector machine models, both achieving an accuracy of 0.9091. Utilizing these models, the reasonable processing parameter window and the relationships among printing parameters were analysed. The results indicate that high printing speeds require large extrusion pressure to achieve a high success rate. An increase in nondimensional nozzle height enhances the possibility of successful printing with small pressure and low speed. The predicted processing parameters derived from the machine learning model were applied to print representative structures, showing good prediction performance.