From process data and alarm data, one can extract correlation between process variables, but these data are not enough for capturing causality information without the help of connectivity data because connectivity is a necessary condition and a basic source of causality. For a large-scale industrial process, connectivity-based process topology provides a qualitative overview and a schematic structure. This topology can be described mathematically and electronically by a signed directed graph (SDG) which is usually built based on process knowledge as described by piping and instrumentation diagrams (P&IDs). This graphical model has been widely used in root cause and hazard propagation analysis, in which both material flow and information flow are considered and expressed. Following is an example of a P&ID and the corresponding SDG of a real industrial process.

The conversion from P&IDs to SDGs can be potentially semi-automated based on the uniform format of P&IDs compatible with open standards such as XML (extensible markup language), RDF (resource description framework), or OWL (web ontology language). Therefore how to automatically capture the flow path information between process units from CAD or other tools is of great concern.

SDG modeling is a complex and experience-dependent task, and thus the resulting SDG should be validated by process data before being used for further analysis. Two validation methods have been taken into account. The first method is based on cross-correlation analysis of process data with assumed time delays. The resulting correlation coefficients can then be validated by examining the paths in SDGs of all the variable pairs and also comparing the signs with the directions of causal relations. The second method is based on transfer entropy, where the information transfer from one variable to another can be computed to validate the corresponding arcs in SDGs.

On the basis of the information fusion from process data and process connectivity, the validated SDGs can be used in fault detection and isolation, hazard assessment, and correlative alarm analysis by inferencing. One can search the upstream and downstream nodes based on graph traversal to find the possible reasons and consequences of the current or assumed symptom; consistent paths describe the fault propagation. If the model is described as an ontology using RDF and OWL, then query language such as SPARQL can be used for this inferencing and more interesting results can be obtained automatically by specific queries on the ontological model.