Intelligent modeling and prediction of CO 2 laser cutting performance in FFF-printed thermoplastics using machine learning algorithms

This paper investigates the laser cutting performance regarding CO 2 of four fused filament fabrication-printed thermoplastics, namely polylactic acid (PLA), carbon fiber reinforced PLA (PLA-CF), acrylonitrile styrene acrylate (ASA), and polyethylene terephthalate glycol (PETG). We investigate kerf open deviation, kerf angle, bottom heat-affected zone, and material removal rate. A total of 72 trial runs were conducted with various states of material type, plate thickness, laser power, and cutting speed. The resulting experimental data were then fed into several machine learning algorithms to assess and compare their predictive abilities: linear regression, decision tree, random forest, CatBoost, support vector regression, k-nearest neighbors, and multi-layer perceptron. Of all the machine learning algorithms, random forest and multi-layer perceptron performed best with high R 2 in conjunction with low error metrics for all response variables. The study also shows that PLA-CF achieves the highest material removal rate, while ASA and PETG yield high dimensional stability with minimum kerf and thermal distortion. Hence, the research study has made it clear that, with machine learning algorithms, the laser cutting performance can be modeled for additively manufactured polymers, and the processing parameters can be optimized for laser cutting; thus, this will enhance the smart manufacturing approaches for the processing of polymers.