Self-Learning Monitoring and Control of Manufacturing Processes Based on Rule Induction and Event Processing

Manufacturing enterprises are trying to cope with turbulent market situations by enhancing their existing monitoring and control of manufacturing processes. Enterprise integration within and across the enterprise can assist to realize the aforementioned goal. Further, event processing (EP) techniques can be employed to monitor and control manufacturing processes in real-time. Rules derived from stored process data using the knowledge discovery in databases process can be managed in an EP engine as event patterns. Nonetheless, rule identification is usually an offline activity whereas the control of manufacturing processes is a real-time activity. Consequently, the rule identification process should be transformed from an offline activity to an online or (near) realtime activity. In the contribution, a methodology is presented to overcome the previously mentioned drawback. Machine learning (i.e., rule induction) methods are used to automatically adapt the existing set of event patterns. The implementation of the presented methodology has been started in a casting enterprise.