STATISTICAL ANALYSES FOR PROBABILISTIC ASSESSMENTS OF THE REACTOR PRESSURE VESSEL STRUCTURAL INTEGRITY: BUILDING A MASTER CURVE ON AN EXTRACT OF THE 'EURO' FRACTURE TOUGHNESS DATASET, CONTROLLING STATISTICAL UNCERTAINTY FOR BOTH MONO-TEMPERATURE AND MULT

摘要

Assessing the structural integrity of a nuclear Reactor Pressure Vessel (RPV) subjected to pressurized-thermal-shock (PTS) transients is extremely important to safety. In addition to conventional deterministic calculations to confirm RPV integrity, Electricite de France (EDF) carries out probabilistic analyses. Probabilistic analyses are interesting because some key variables, albeit conventionally taken at conservative values, can be modeled more accurately through statistical variability. One variable which significantly affects RPV structural integrity assessment is cleavage fracture initiation toughness. The reference fracture toughness method currently in use at EDF is the RCCM and ASME Code lower-bound K_(IC) based on the indexing parameter RT_(NDT). However, in order to quantify the toughness scatter for probabilistic analyses, the master curve method is being analyzed at present. Furthermore, the master curve method is a direct means of evaluating fracture toughness based on K_(JC) data. In the framework of the master curve investigation undertaken by EDF, this article deals with the following two statistical items: building a master curve from an extract of a fracture toughness dataset (from the European project "Unified Reference Fracture Toughness Design curves for RPV Steels") and controlling statistical uncertainty for both mono-temperature and multi-temperature tests. Concerning the first point, master curve temperature dependence is empirical in nature. To determine the "original" master curve, Wallin postulated that a unified description of fracture toughness temperature dependence for ferritic steels is possible, and used a large number of data corresponding to nuclear-grade pressure vessel steels and welds. Our working hypothesis is that some ferritic steels may behave in slightly different ways. Therefore we focused exclusively on the basic french reactor vessel metal of types A508 Class 3 and A 533 grade B Class 1, taking the sampling level and direction into account as well as the test specimen type. As for the second point, the emphasis is placed on the uncertainties in applying the master curve approach.