Towards computer vision feedback for enhanced CNC machining

A vision-based sensing utility is proposed and developed to enhance the performance of CNC milling machines. Two cameras are attached to the machine spindle to provide visual information on work-piece surface quality while machining. This information is processed and processing outcome is used to improve the surface roughness of the machined part by adjusting the employed machining parameters. To facilitate on-line monitoring and control, computer programs are developed to manipulate the source G-code file in order to enable real time integration of the gained feedback information. The manipulated G-code program is then executed by the machine controller to allow adaptive control of the work-piece machining. The developed system automatically determines which camera is required to provide scene data for each step of tool path movement. Relative camera is then activated just as the machine completes the cutting of each tool path segment. Acquired vision data is processed to evaluate resulting surface roughness and accordingly a decision is made to either maintain the implied machining parameters or adjust their values to beneficially suit the machining task and result in better surface quality. The developed approach is tested and verified experimentally. Promising results are acquired that encourage further development towards full integration of CNC-vision systems.