Adaptive Direct Ink Writing via a Hybrid Physics–Machine Learning Framework with G-Code Optimizer

The system integrates machine learning algorithms with empirical printing data to dynamically adjust G-code commands, improving print success rates, dimensional accuracy, and material efficiency.

This research software addresses the significant challenge of parameter optimization in viscous material printing, where traditional trial-and-error approaches are time-consuming and material-intensive. The application implements four machine learning models (Neural Network, Random Forest, Gradient Boosting, Ensemble) trained on experimental printing data to predict optimal adjustments for six key parameters: Z-height, print speed, flow rate, layer height, quality prediction, and material saving.