Anomaly detection and productivity analysis for cyber-physical systems in manufacturing
Miguel Saez, Francisco Maturana, Kira L. Barton, Dawn M. Tilbury · 2017
Cyber-physical Systems (CPS) which can be defined by the integration of a physical process with network communication and computing is a key concept for smart manufacturing. Analysis of complex machines as CPS can support anomaly detection and diagnosis by providing the required data to model different operational conditions. In this work, we develop a hybrid model of manufacturing machines to estimate operational state based on machine functionality, dynamics, and interactions. We used the identification of a Global Operational State (GOS) to help partition, provide context to signals, and monitor the time of sub-tasks in a manufacturing process. Partitioning a signal based on GOS can improve anomaly detection and diagnosis. Monitoring duration of sub-tasks in the process can provide a detailed insight of cycle time and improve productivity analysis. The proposed approach was implemented using data from an automotive assembly plant to detect backlash and parts slipping in a conveyor system by monitoring energy signature and state variables.