Article: Laser cutting of wood: experimental study for machine learning Journal: International Journal of Mechatronics and Manufacturing Systems (IJMMS) 2024 Vol.17 No.3 pp.319 - 347 Abstract: The use of CO2 lasers for cutting wood patterns has become a common practice. The aim of this work is the quantitative correlation of the laser cut surface characteristics such as roughness, flatness and burning, with the process parameters. For this purpose, cuts were made using a CO2 laser, with a maximum power of 3 kW. The way in which the type of wood, power of the laser, feed, focusing distance of the laser beam and pressure of the auxiliary gas affect cutting quality was studied quantitatively for 4 types of wood, namely oak, Swedish pine, particle board and fiberboard (MDF). Cutting quality was quantified through kerf geometry, i.e., initial, neck, maximum and mean width, variation of kerf width through its length, as well as ripples along the thickness. Based on the experimental results, a set of feedforward type artificial neural networks (ANNs) were trained as predictors of part quality. The models are sufficiently accurate with generalisation error from 5% to 15% despite the restricted training dataset. Inderscience Publishers - linking academia, business and industry through research

Title: Laser cutting of wood: experimental study for machine learning

Authors: George-Christopher Vosniakos; Ioanna Kyriopoulou; Panorios Benardos; Dimitrios Pantelis

Addresses: School of Mechanical Engineering, National Technical University of Athens, Athens, GR-15773, Greece ' School of Naval Architecture and Marine Engineering, National Technical University of Athens, Athens, GR-15773, Greece ' School of Mechanical Engineering, National Technical University of Athens, Athens, GR-15773, Greece ' School of Naval Architecture and Marine Engineering, National Technical University of Athens, Athens, GR-15773, Greece

Abstract: The use of CO₂ lasers for cutting wood patterns has become a common practice. The aim of this work is the quantitative correlation of the laser cut surface characteristics such as roughness, flatness and burning, with the process parameters. For this purpose, cuts were made using a CO₂ laser, with a maximum power of 3 kW. The way in which the type of wood, power of the laser, feed, focusing distance of the laser beam and pressure of the auxiliary gas affect cutting quality was studied quantitatively for 4 types of wood, namely oak, Swedish pine, particle board and fiberboard (MDF). Cutting quality was quantified through kerf geometry, i.e., initial, neck, maximum and mean width, variation of kerf width through its length, as well as ripples along the thickness. Based on the experimental results, a set of feedforward type artificial neural networks (ANNs) were trained as predictors of part quality. The models are sufficiently accurate with generalisation error from 5% to 15% despite the restricted training dataset.

Keywords: laser cutting; wood; process planning; kerf; waviness; roughness; artificial neural networks.

DOI: 10.1504/IJMMS.2024.143337

International Journal of Mechatronics and Manufacturing Systems, 2024 Vol.17 No.3, pp.319 - 347

Received: 02 Jun 2024
Accepted: 30 Aug 2024
Published online: 13 Dec 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Laser cutting of wood: experimental study for machine learning

Keep up-to-date