Cyber-physical systems, such as connected machine tools in a factory shop floor, provide a vast amount of data regarding the machine itself, the tool or the process. Together with the continuously increasing computing power, recent advances in the field of machine learning enable powerful mechanisms to handle information retrieval from those large and complex data sets during runtime of the machine tool or during the process. Those algorithms can significantly contribute to monitor the machine’s dynamics and predict future states and events. If this prediction is fed back to the system model itself, the closed loop forms a self-identifying Digital Twin of the machine and provides powerful methods to monitor machine and tool conditions, predict appropriate maintenance intervals or provide cloud-based process analysis.