We present first results of our research on closed loop decision support of an industrial laser welding process fusing on inline process monitoring data to improve quality-relevant properties in the production line. By using position dependent synchronization of the recorded signals we are able to precisely correlate the online measured properties of the welding process in time and space. A data-driven self-learning system (SLS) evaluates the temporal evolution of the correlation, between the observed process data and failure events. Over time, the system’s knowledge evolves up to a point when it becomes sufficiently accurate to forecast process behaviour automatically. In our setup, we record the intensity of the spectrally resolved light emitted from the process zone. We also record an image of the process zone which was taken coaxially to the laser beam. After pre-processing the image a threshold method is used to determine the statistics of the weld seam width within a predefined region of interest. Both measurements are triggered by a position dependent signal sent out from the rotary axis the assembled elements are mounted on. The extracted data as well as process parameters are stored in a database for a deferred evaluation by the SLS. Results show a good agreement between the SLS-forecasted process behaviour and the measured quality of the welds. The developed system can be used in parallel to the real process, to automatically predict weld quality as a form of decision support.