par Devooght, Jacques 
;Smidts, Carol-Sophie
Référence Nuclear science and engineering, 112, 2, page (101-113)
Publication Publié, 1992-10
;Smidts, Carol-Sophie Référence Nuclear science and engineering, 112, 2, page (101-113)
Publication Publié, 1992-10
Article révisé par les pairs
| Résumé : | During an accident, components fail or evolve within operating states because of operator actions. Physical variables such as pressure and temperature vary, and alarms appear and disappear. Operators diagnose the situation and effect countermeasures to recover the accidential sequence in due time. A mathematical modeling of the complex interaction process that takes place between the operating crew and the reactor during an accident is proposed. This modeling derives from a generalization of the theory of continuous event trees developed for hardware systems to a mixture of human and hardware systems. Such a generalization requires extension of the evolution equations built under the Markovian assumption to semi-Markovian processes because dead times as well as nonexponential distributions must be modeled. Operator and reactor states have transitions due to their own evolution (dQ00, dQRR) or to their mutual influence (dQ0R, dQR0). The correspondence between the estimates yielded by current human reliability models and the transition rates required as input data by the model is given. This model should be seen as a mold in which most existing human reliability models fit. |
