Good Practices for a Better Risk Estimation

Risk studies for external safety comprise technological scenarios of loss of containement, where a risk analyst strives to model them to obtain a risk value for plant hazards. Most of the time, plant complexity under study is low, with a sparse population in the vicinity, and yielding a low risk for that enterprise. On the other hand, there are studies on complex industrial sites with high density of equipments, bearing huge inventories and surrounded by high populated communities. These cases require a more specialized approach to make an enterprise viable. This paper presents a discution on variables not usually explored, and which if properly treated, may provide gains in assessing the safety level of an industrial plant, not overly conservative, facing a governamental risk criteria for external safety.

In an usual QRA, a very simple weather representation is considered, being selected up to two weather scenarios: one for day time and another for night time, resulting in savings of CPU-time, however magnifying the calculated risk. For the intrincate cases just mentioned above, a more refined meteorological representation is developed through sensitivity studies, demonstrating effectiveness, and weather time-serie data are treated on the way to promote a better meteorogogical data representation. The weather at a given location depends on many factors: topography of landscape, atmosphere behavior, and season of the year. A finer representation of the frequencies of windspeeds and atmospheric stability classes (from the raw time-series) can better approximate reality. The point is: how fine a representation needs to be, in order to payoff a reduction in risk, with realizable computing time.

Another drive along the same lines, advocates to explore pipeline failure rates in databases, to account for significant advances in pipeline technology in the last years, as well as the actual mechanisms which a given pipeline is subjected to, such as corrosion, third-party interference, material & construction and natural causes. Data from CONCAWE and EGIG are considered on this assessment for a better pipeline failure representation and conclusions are drawn. However a more detailed approach is possible to be implemented even not threated on this paper, what should be by a structural reliability analysis.

As conclusion, as deep as risk analysis goes better resolution is gathering for industrial site as well from the variables affecting the risk.