Publication Details

Technical Paper TP09

Time Dependency of Accidents and Response in Petrochemical Plants

Author: S Medonos, Petrellus Limited. Proceedings of the International Conference on Risk, Quality and Reliability (RQR2007) at the Technical University of Ostrava, September 20-21, 2007, Ostrava, Czech Republic.


Processes when an accident happens in a petrochemical plant are time-dependent. This can be illustrated by an accidental leak of flammable fluid which may be ignited when (in time) flammable vapours or cloud form, which results in explosion and / or fire, which may escalate to subsequent explosions and fires when (in time) new leaks form resulting from the effects of the initial explosion / fire.

As these processes are time-dependent, they must be reflected in the emergency response planning of the plant. This very often does not happen. As for example, escape and evacuation plot plans are prepared without knowledge of potential escalation of initial accidents. The information on the possible modes of escalation is very superficial despite the fact that it is available from observations, tests and analyses.

Emergency response plans should be based on the likelihood of how the events that are to be responded on develop. The mathematical description of the event development is normally in the form of Event Trees, but Event Trees are by definition steady state and cannot represent time-dependent events.

Attempts have been made in design risk assessments of process plants to make Event Trees at least quasi-static, i.e. Event Tree branches were made to reflect time intervals representing defined sub-events. This resulted in improved quality of information from Event Tree analyses, but the Event Trees were of several thousand branches, which made them unpractical to use. There is a call for development of technology in this aspect, which research organisations and academia can carry out. However, the results of this development should be aimed at practical applications in the petrochemical industry, where computational tools must be fast in application, in order to follow the rapid pace of the design process.

Return to the Publications page.