A Refreshing Take: Analysing Accident Scenarios through Causal Network Topology Metrics

Abstract

Accident causation investigation and even more hazard scenario identification are troubled by the complexity of interactions between three elements in a process facility: People, Plant and Procedures. Interactions are of various nature, such as physical change and information transfer, all influencing the process. To facilitate investigation the digraph network was applied as the most flexible visual aid to describe a causal structure. Such structure consists of nodes and edges representing an event or condition in the accident scenario and a causal link respectively. Attributing the nodes and edges to the type of interaction, numbers of the same type can be counted, and so two metrics are developed:  The P3 Interaction Contribution (PIC). This is the proportion of nodes and edges associated with an interaction between People, Plant and Procedures.  The Average Edge Weight. This relates to the proportion of events in the scenario that are associated with the logical AND gate conjunction from its causes (incident nodes), where the event requires more than one simultaneous cause. The technique was tried on four CSB accident descriptions. Interesting differences are seen. Also, in view of a paper accepted to be published in Safety Science the approach seems quite helpful in process hazard analysis.