Capturing Connectivity and Causality in Complex Industrial Processes (SpringerBriefs in Applied Sciences and Technology) - Softcover

Über die Autorin bzw. den Autor

The authors jointly have extensive research experience in modeling, control, and monitoring of complex industrial processes. In particular, they have worked on industrial projects in oil and petrochemical sectors to address safety, alarm, and fault diagnosis issues from operating plants. Moreover, they have conducted research in the related areas on capturing connectivity and causality using process data and various forms of process knowledge; their research results have been published in international journals, benefiting the automation community. Realizing the importance of capturing connectivity and causality in real-world problems, and summarizing their knowledge and understanding on various approaches currently available, the authors have made a great effort in presenting this brief as an introduction, a survey, and also a tutorial on this seasoned topic.

Von der hinteren Coverseite

This brief reviews concepts of inter-relationship in modern industrial processes, biological and social systems. Specifically ideas of connectivity and causality within and between elements of a complex system are treated; these ideas are of great importance in analysing and influencing mechanisms, structural properties and their dynamic behaviour, especially for fault diagnosis and hazard analysis. Fault detection and isolation for industrial processes being concerned with root causes and fault propagation, the brief shows that, process connectivity and causality information can be captured in two ways:

· from process knowledge: structural modeling based on first-principles structural models can be merged with adjacency/reachability matrices or topology models obtained from process flow-sheets described in standard formats; and

· from process data: cross-correlation analysis, Granger causality and its extensions, frequency domain methods, information-theoretical methods, and Bayesian networks can be used to identify pair-wise relationships and network topology.

These methods rely on the notion of information fusion whereby process operating data is combined with qualitative process knowledge, to give a holistic picture of the system.