Advanded Methods for Machine and Process Monitoring

Process monitoring systems are being successfully utilized to detect process errors in cutting operations such as e.g. tool breakage or raw parts deviating from their nominal geometry. Conventional techniques follow a teach-in approach making use of previously recorded data. In single item production, teach-in based methods fail due to nonexistent reference data. Alternatively, a simulation- based monitoring method is proposed in this work. It allows for single item monitoring of turning, milling and drilling processes. A unique online cutting simulation system enables process parallel derivation of confidence limits without the need for a CAM system. Thus, the parameterization effort is limited to a minimum. To that, the monitoring reference for error- sensitive process signals is estimated based on the process simulation using a regression model. A self- learning friction- and acceleration compensation system has been implemented in order to extract the process specific components of measured spindle currents, which are sensitive to process errors. This allows signal monitoring of accelerating machine axes, for the first time. Additionally, self-optimizing multi sensor strategies are studied with regard to the industrial application in series monitoring minimizing manual parameterization effort.