Abstract

This ENBIS invited session addresses recent developments in statistical and data-driven methods for monitoring industrial and applied processes under modern data conditions. The increasing availability of high-frequency sensor data, continuous data streams, and heterogeneous information sources has substantially changed how processes are observed, monitored, and controlled across industrial, healthcare, environmental, and service systems. At the same time, these developments introduce significant challenges for traditional Statistical Process Monitoring (SPM), particularly in the presence of high dimensionality, complex temporal dependence, non-linear dynamics, and evolving operating regimes.
The session brings together complementary methodological perspectives that reflect current directions in process monitoring, with a focus on approaches that integrate statistical modelling with machine learning, Bayesian inference, and adaptive learning strategies. A central theme is the design of monitoring methods that operate sequentially and in real time, supporting timely detection of changes, anomalies, and emerging patterns in streaming data. Particular emphasis is placed on methods that enhance diagnostic capability and decision support in settings where labelled data are limited, costly to obtain, or only partially available.
Several contributions consider the integration of Artificial Intelligence, Machine Learning, and Advanced Data Analytics within classical monitoring frameworks, illustrating how data-driven techniques can complement control-charting procedures and extend their applicability to high-dimensional and dynamic processes. Other contributions focus on adaptive learning and online classification strategies aimed at the efficient allocation of limited labelling resources while maintaining effective monitoring performance. Bayesian online approaches are also discussed as principled tools for detecting structural changes and regime shifts in complex time series, particularly in sensor-driven applications.
Across the session, emphasis is placed on methods motivated by real-world monitoring problems and suitable for deployment in operational environments. Application areas include industrial manufacturing processes, sensor-based monitoring systems, and other data-intensive settings where robustness, scalability, and interpretability are essential. By linking methodological advances with practical monitoring challenges, the session aligns closely with ENBIS’s mission to promote statistically sound and practically relevant solutions for business and industrial applications.