Treffer: An adaptive online learning scheme for anomaly detection in IIoT data streams under varying operating conditions.

High-dimensional data streams from the industrial Internet of Things (IIoT) provide promising opportunities for data-driven anomaly detection in production processes. However, conventional methods exhibit critical limitations in interpretability and effectiveness within modern flexible manufacturing, as they primarily focus on correlation-based offline analysis and lack adaptability to varying operating conditions. Particularly, the evolving nature of data distributions and inter-variable relationships within IIoT data streams poses the challenge of <italic>concept drift</italic>, potentially causing detection performance degradation. To address these, an online adaptive IIoT data stream analysis (OAIDA) scheme is proposed. First, a novel key process features (KPFs) selection approach via Bayesian network (BN) learning is developed for interpretable dimensionality reduction, which identifies features with causal relationships to product quality. It adopts a computationally tractable local causal inference strategy, improving the efficiency and industrial applicability of BN. Then, a drift-adaptive ensemble online learning approach is developed, which detects the production state by averaging the base learners’ predictions with performance-based weighted probabilities. Additionally, a proactive drift-detecting mechanism is integrated to capture causality changes, supporting KPFs updates within sliding windows. Experimental results validate the superiority of our scheme, and a case study of an automotive welding process illustrates its practicality. [ABSTRACT FROM AUTHOR]

Copyright of International Journal of Production Research is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)