Anomaly Detection in a Production Line: Statistical Learning Approach and Industrial Application
Abstract
This paper explores industrial engineering, particularly focusing on discrete processes and emphasizing real-time control of Production Lines within these processes. A critical component of this control involves the incorporation of a dashboard system, essential for providing workshop managers with valuable insights into the estimated time required for each Production Order (PO) to progress through the remaining stations in the production line.
The key contribution of this work is the conception and development of a novel mathematical model, applied to real-world industrial data, capable of detecting anomalies within the production line. These anomalies are defined as deviations from expected timeframes. Constructed using Statistical Learning techniques and Information Theory, the model can be integrated within the dashboard framework, offering prompt identification of anomalies and ensuring optimal performance and efficiency of the production process.