57 Prevention of the Major Industrial Accidents: About the Interest of Using Diagrams Influence in Risk Prevention Process (PSAM-0332)

The industry has, for long, implemented major risk prevention programs based mostly on technical improvements of the selected system. However the major accident investigations such as Challenger, Columbia, Paddington,… have underlined the human and organizational side of these major accidents.

Today, focusing only on the technical explanation of the accidents to take preventive measures would rather limit the effects of theses ones on the system; that is because the technical aspects are an important but only obvious part of the major industrial accidents.

Indeed, the industrial system can not be understood and separated from the stakeholders that make this system work, from the statutory and socio-economic environment where it evolves. This observation is also valid for the industrial accidents having an important impact or not and being rare or recurrent.

In this perspective, this paper will tackle at first the problematic of considering the complexity and the dynamic of change of the industrial system while choosing both technical and organizational measures to prevent major industrial accidents.

In the second part, we will discuss about the relevance of using both analytical and probabilistic risk assessment approaches to prevent major industrial accidents. This will be highlighted in the new regulatory context of industrial and natural risk prevention process in France after the promulgation of the july, 30^th 2003 law.

The third part is dedicated to these two concerns were studied considering the interest of using the “Bayesian Belief Network” in the process of industrial risk analysis and management. Indeed, the BBN offer the possibility of diagnosis through:

• exploitation of a graphic support to represent the influence relations which exists between the elements of a system;

• the possibility of seeing the influence, in partial information, of several states of a system.

However, this methodological approach comprises strong limits. This research project aims at highlighting these restrictions.

This paper report the first conclusions of our three years research program aiming at providing a framework to improve the actual practices in risk analysis and risk management.