Detecting Asset Cascading Failures Using Complex Network Analysis

Experienced process plant personnel observe that corrective maintenance work on one asset may often be followed by the same or connected assets within a short amount of time. This problem is referred to as cascading failure. Confirming if these events are chronic difficult given number and volume operation data. If can identified, preventative measures implemented prevent those cascades, eliminating unnecessary work. project uses complex network analysis identify where co-occurrence most frequent, in plant. Data drawn from over 50,000 orders for 5,655 pumps mining company five-year period. A produced connecting based frequency Beside advantages visualisation networks, method produces quantified measures, normalised degree, eigenvector centrality betweenness centrality, which used with significant impact other assets. Affected apparent communities network. identifies "super-spreaders": who experience lead pumps. The model tuned different time windows, example seven days. From insights, changes made operational, recording practices re-occurrence. Of particular note this data was occurrence self-loops certain prevalence hidden failures standby