There is strong interest from the nuclear industry to use the precracked Charpy single-edge notched bend, SE(B), specimen (PCVN) to enable determination of the reference temperature, T0, with reactor pressure vessel surveillance specimens. Unfortunately, for many different ferritic steels, tests with the PCVN specimen (10×10×55 mm) have resulted in T0 temperatures up to 25°C lower than T0 values obtained using data from compact, C(T), specimens. This difference in T0 reference temperature has often been designated a specimen bias effect, and the primary focus for explaining this effect is loss of constraint in the PCVN specimen. The International Atomic Energy Agency has developed a three-part coordinated research project (CRP) to evaluate various issues associated with the fracture toughness Master Curve for application to light-water reactor pressure vessels. One part of the CRP is focused on the issue of test specimen geometry effects, with emphasis on the PCVN bias. This topic area was organized in two parts, an experimental part and an analytical part with a view towards each part complementing the other. Within the analytical part, elastic plastic finite element methods are extensively used in order to access local stress and strain information that is the basic ingredient for most of the micromodels of cleavage fracture developed to date. In the framework of the international qualification and acceptance of such a tool for actual loss of constraint prediction, the validation of such tool is of prime importance. Therefore, a round robin exercise has been proposed and performed by ten laboratories from nine different countries. The round robin focuses on the modeling of realistic three-dimensional geometries containing shallow and deep crack. This round robin has been useful to qualify different finite element codes and to identify possible errors in the input file. The round robin demonstrates that errors in the input file can be easily introduced. Some remaining differences cannot be attributed to one particular finite element code or to actual errors. Those differences are attributed to the so called “user effect” which can only be reduced through in depth discussion and deep understanding of each finite element code. Independently of the used code and of relatively small user effect differences, it is found that shallow crack specimens are more sensitive to loss of constraint than deep crack specimens for a given specimen size. The difference in terms of reference temperature between the two geometries is evaluated to be about 40 °C. For a deep crack, loss of constraint is identified to appear at M values around 200. This value is larger than the one specified in current standard (M = 30). Increasing the M value to 200 will jeopardize the use of PCVN for the nuclear industry on the other hand bias introduced by M value of 30 is acceptable.
On the distribution of winners’ scores in a round-robin tournament
