Real-time process control by machine vision

Abstract

Quality assurance in laser processing is often provided by post process inspections with machine vision solutions. When process stability gain is needed, one-dimensional signals like diodes or pyrometers which measure the emitted radiation of the laser process may be applied to closed-loop-control of e.g. the laser power. Such signals mostly provide short latency and good signal to noise ratio, but represent only mean values of the whole processing area or parts of it.On the other Hand secondary process radiation can provide information about the process stability, but may not always show the cause for process imperfections. To overcome the lack of spatial resolution in closed loop control, this paper explains how real-time computer vision improves existing processes and may even enable new laser processes. Furthermore an example machine vision algorithm is presented which measures kinematic properties of the process zone in real time. Such signals provide reliable information of the process state by recovering the influences which cause the process to leave its stable state.The shown system can be adapted to most laser processes (e.g. welding, brazing, cutting, selective laser melting, etc.). For complexity reasons most examples in this paper will be based on remote laser welding only.Quality assurance in laser processing is often provided by post process inspections with machine vision solutions. When process stability gain is needed, one-dimensional signals like diodes or pyrometers which measure the emitted radiation of the laser process may be applied to closed-loop-control of e.g. the laser power. Such signals mostly provide short latency and good signal to noise ratio, but represent only mean values of the whole processing area or parts of it.On the other Hand secondary process radiation can provide information about the process stability, but may not always show the cause for process imperfections. To overcome the lack of spatial resolution in closed loop control, this paper explains how real-time computer vision improves existing processes and may even enable new laser processes. Furthermore an example machine vision algorithm is presented which measures kinematic properties of the process zone in real time. Such signals provide reliable information of the process state by reco...