Switching supervisory control in fusion laser cutting with vision-based process state detection

Future unmanned laser cutting machines will require the capability to automatically detect and adapt to process states avoiding critical defects. In the fusion laser cutting of metals with low thickness (1-3 mm) the process can rapidly evolve from successful separation to plasma-dominated or loss of cut due to uncontrolled factors. Thus industrial systems set the cut velocity conservatively to ensure separation while avoiding low quality or incomplete cuts. This work proposes a novel switching supervisory control architecture to increase process productivity by regulating the velocity in real-time accordingly to the detected condition. A coaxial camera-based vision system was employed to develop a classification framework capable of a low latency process state identification during the cutting of 2 mm thick mild steel. The control logic allowed high quality cutting of 2 mm thick mild steel with a productivity increase of 10-33% hence validating the architecture for autonomous machines.