Prediction of Poly-methyl-methacrylate Laser Milling Process Characteristics Based on Neural Networks and Fuzzy Data

Laser milling is a recent technology adopted in rapid prototyping to produce tool, mould and polymer-based microfluidic devices. In this process, a laser beam is used to machine a solid bulk, filling the area to be machined with a number of closely spaced parallel lines. Compared to traditional machining, this method has some advantages, such as: greater flexibility of use, no mechanical contact with the surface, a reduction in industrial effluents, a fine accuracy of machining, even with complex forms, and the possibility to work different kind of materials. While it is relatively easy to predict the depth of the area worked, the surface roughness is more difficult to predict due to the materials behaviors at microscopic level. This is truer when polymer processing is considered due to the local thermal effects. The paper addresses the application of an artificial neural network computing technique to predict the depth and the surface roughness in laser milling tests of poly-methyl-methacrylate. The tests were carried out adopting a CO2 laser working in continuous and pulsed wave mode. The obtained results showed a good agreement between the model and the experimental data. As a matter of fact, despite the thermal degradation that occurs on the PMMA surface, neural network processing offers an effective method for the prevision of roughness parameters as a function of the adopted process parameters.