Computer vision approach for robotic polishing application using artificial neural networks

Polishing is a highly skilled manufacturing process with a lot of constraint and interaction with environment. Unfortunately, manual polishing process takes time consuming of the total manufacturing time and takes a significant cost. On top of that, undesired working condition exists due to dust and noise and next it is quite difficult to find skilled technicians. Therefore, it is necessary to develop an automation system on the polishing process. One of the automation systems that are often developed by the industry is the use of robots. The goal is to handle the repetitive work that humans are not able to do so. In general, the purpose of polishing is to get the uniform surface roughness distributed evenly throughout part´s surface. This research combines computer vision and robotics systems to overcome the problems arising in the polishing process. Vision system is used to measure surface defect that divide into several categories. The surface data is learned using artificial neural networks then give the decisions to move the actuator on robot. Parameters that developed in this system are force and time polishing which has a significant effect on the polishing process. Therefore, this system studies the characteristics of surface defects before the given action with different value of force and polishing time, and then compared with surface defects after given the action. Results obtained show that it is possible to obtain surface parameters using vision-based methods with a certain limit of accuracy. However, there are some advantages using this system, including faster polishing time, simpler quality inspection, and more evenly surface roughness result compared with manual polishing. The overall results of this research would encourage further developments to achieve robust computer vision technique for robotic polishing application.