Artificial Neural Network-Based Optimization of CO2 Laser Cutting Parameters for Beech Plywood and HDF: A Kerf Geometry Perspective

This study presents the results of Artificial Neural Networks (ANN) predictions with the aim of optimizing the process of beech plywood and HDF laser cutting. A survey is given of the results of predictions of cutting kerf parameters made by Artificial Neural Networks to cover a wide spread of CO2 laser parameters, as well as the results of experimental cutting with maximum laser power (P) equal to 135 W and maximum feed rate (v) equal to 20 mm/s. Validity of the best neural network was checked versus overfitting of the best neural networks, confirmed according to r value of the model (minimum 0.971), MAPE (%) (maximum 6.21 %) and compared with the results of other authors. The article also presents the effect of energy density values E on values of cutting kerf parameters and their variance. The results show that the optimal value of laser power (P) and feed rate (v) for beech plywood are (200-300 W; 10-15 mm/s), while for more dense and more homogenous high-density fibreboard (HDF) they are (300-500 W; 5-10 mm/s). Optimal energy densities (E) are then 133 MJ/m2 for beech plywood and 433 MJ/m2 for HDF. Similar as for other wooden materials, it follows that more dense species of wood should be cut with higher values of energy densities. The results can be applied to reduce the material and energy demands by optimizing the quality of cut with minimum symmetrical kerf widths.