A cognitive approach to monitor and control focal shifts in laser beam welding applications

Even though widely used, laser beam welding faces a strong demand for improved monitoring and control capabilities. In this work the occurring focal shift which leads to faulty welding results will be explained. An approach is taken to monitor this effect caused by the heading of the optics. By means of a camera sharing the same optical components as the laser beam, the deviation of the optimal focal point is obtained through application of sharpness operators. Moreover, a certain measure of light intensity allows one to determine whether the focal plane is in front or behind the optimal focal point. By inclusion of further process parameters and an estimation of the process quality given by a cognitive monitoring system, a large manifold is built up. By application of linear and non-linear dimensionality reduction algorithms, the manifold is reduced to significant features fed to a fuzzy kNN-classifier. Its outcome suggests the required change in z-direction of the optical system with a behavior similar to the human eyes. The overall system is taught by an expert during its initialization and later saves production down times and faulty welding procedures.