The stimulus-driven theory of probabilistic dynamics as a framework for probabilistic safety assessment J.M. IZQUIERDO Consejo de Seguridad Nuclear Justo Dorado 11, E - 28040 Madrid jmir@csn.es Fax: +34-913460496 P.E. LABEAU ∗ Universit´ e Libre de Bruxelles (CP165/84) Av. F.D. Roosevelt 50, B - 1050 Brussels pelabeau@ulb.ac.be Fax: +32-2-6504534 Abstract (ID: 0414) The current version of the Theory of Probabilistic Dynamics (TPD) falls short in mod- eling essential aspects of PSAs, because automatic or manual actions that induce safety oriented events curbing undesired time evolutions, are always ”stimulated” by particular features of the situation. Typical stimulus examples are initiation safeguards setpoints or control room alarms, but many others, as the influence of organizations in the operator interventions are also to be accounted for. To incorporate the concept of stimulus covering all those circumstances into the TPD was the purpose of two recent papers that describe the resultant Stimulus-Driven Theory of Probabilistic Dynamics (SDTPD). However, the theory is still general and does not introduce the familiar concept of sequences. This sum- mary paper shows how sequences may also be incorporated into SDTPD in order to better understand PSA. A set of equations derived from this new extension is able to provide exceedance frequency and sequence probability under the same assumptions as those of current PSA methods. SDTPD is also presented in a more refined version with a symmet- ric treatment of stimulus activation and disactivation events, and allows the probabilities of the next event to depend on the state of all stimulus. 1 Introduction The PSA methodology is widely used in the nuclear industry as a systematic process to in- tegrate partial elements into a coherent risk analysis. One of its challenges is the consistent treatment of the dynamics of the evolution of accidents and transients (accident analysis), and its interface with systems features that are critical to the outcome of accidents (system reliability). A rigorous theory supporting this consistency is still in need, as pointed out in many places (Bley et al., 1992, Siu, 1994). Promising candidates include the Theory of Prob- abilistic Dynamics(TPD) (Devooght and Smidts, 1992, Smidts and Devooght, 1992, Devooght and Smidts, 1994). However, its current version falls short in modeling essential aspects of PSAs, because in reality automatic or manual actions that induce safety oriented events curb- ing undesired time evolutions , are always ”stimulated” by particular features of the situation. ∗ Research Associate, National Fund for Scientific Research (Belgium), corresponding author 1