Prediction of Surface Roughness and Hardness on Multi 3D Printers with Machine Learning

Additive Manufacturing (AM), known as 3D Printing, is the “process of joining materials to make objects from 3D model data, usually layer upon layer.” Quality control in 3D printing is challenging, particularly in controlling surface roughness and hardness values of printed objects. This research aims to create predictive modeling using machine learning algorithms to predict surface roughness and hardness values. Data obtained from experimental testing of various printing parameters will be used as a dataset to train and test the predictive model. The results of this study show that there are differences in surface roughness and hardness values from variations of 3D printers. For the Ender 3 V3 KE 3D Printer, the average surface roughness and hardness values are $9.83 \mu \mathrm{~m}$ and 72.41 HD, respectively. Meanwhile, for the Flashforge Creator Pro KE 3D Printer, the average surface roughness and hardness values are $7.02 \mu \mathrm{~m}$ and 73.2 HD, respectively. To predict surface roughness with the highest accuracy, the XGBoost method has the lowest percentage error based on MAPE at 15.16%, while for hardness values, the Random Forest method has the lowest percentage error based on MAPE at 3.34%. The implementation of the developed predictive model is applied in a UI system to optimize printing parameters to achieve the desired quality in terms of surface roughness and hardness.