No abstract

No abstract

No abstract

Being able to quickly detect anomalies and reason about their root causes in critical manufacturing systems can significantly reduce the analysis time to bring operations back online, thus reducing expensive unplanned downtime. Machine learning-based anomaly detection approaches often need significant amounts of labeled data for training and are challenging to scale for manufacturing deployments. A robust blended system dynamics and discrete event simulation physics-based modeling methodology is proposed for the task of automated anomaly detection. The blended model consists of discrete event simulation (DES) components for the discrete manufacturing process modeling, and system dynamics (SD) components for continuous variables. The methodology strikes a balance between the computational overhead for online monitoring and the level of details required to perform anomaly detection tasks. The implementation of models takes an object-oriented approach, allowing multiple components of a smart factory to be robustly described in a modular, extendable and reconfigurable manner. The proposed methodology is applied to and validated by data collected from a real commercial manufacturing plant. A production line is modeled with DES components and heat transfer is modeled with SD. The blended model is then utilized for anomaly detection. It is demonstrated that the model-based approach is effective not only for detecting but also explaining particular types of anomalies in a commercial discrete manufacturing system.

No abstract