Low-Cost Laser-Acoustic PVC Identification System Based on a Simple Neural Network

Abstract

Desktop laser cutters are an affordable and flexible rapid-prototyping tool, but some materials cannot be safely processed. Among them is polyvinyl chloride (PVC), which users usually cannot distinguish from other, unproblematic plastics. Therefore, an identification system for PVC applicable in a low-cost laser cutter has been developed. For the first time, this approach makes use of the laser-ablative sound generated by a low-power laser diode. Using a capacitor microphone, a preprocessing algorithm and a very simple neural network, black PVC could be detected with absolute reliability under ideal conditions. With ambient noise, the accuracy dropped to 80%. A different color of the material did not influence the accuracy to detect PVC, but a susceptibility of the method against a color change was found for other materials. The ablation characteristics for different materials were recorded using a fast-framing camera to get a better insight into the mechanisms behind the investigated process. Although there is still potential for improvements, the presented method was found to be promising to enhance the safety of future desktop laser cutters.