An advantage of the Master Curve method is that it makes possible the use of Charpy-size and even smaller specimens for a valid determination of frac­ture toughness and T0. The number of tests always has to be determined so that the minimum required confidence level is achieved for the estimation.

ASTM E1921 describes a special weighting system to ensure a sufficient confidence level for the analysis. The final check can be made only after the value of T0 is rather well known, when the ade­quacy of tests can be determined from the condition У]ni ri > 1, where Гі is the number of valid data in the valid temperature range і and ni is the weight factor of this range. Normally, the required number can be tentatively determined only after some tests have been conducted, because the optimal test tempera­ture range is not known beforehand. With very small specimens, the final number of tests needed for a

.

valid estimate will likely be larger than the given minimum of six because of the smaller validity win­dow, which is reduced with decreasing specimen size and the material yield strength (see Figure 7). Exam­ples of possible small and miniature SE(B) specimen geometries for fracture toughness testing are com­pared in Figure 12.

As mentioned previously, a commonly used speci­men configuration for irradiated RPV steels is the full Charpy-size geometry with 10 x 10 mm cross­section. For most applications, this geometry provides a sufficiently large validity window (Figure 9), because as few as six specimens may be sufficient for a valid estimate. From the present experience for irradiated steels with different size specimens, Charpy square or half Charpy rectangular SE(B) or 0.4 T or 0.5 T C(T) geometries are generally optimum when the amount or form of the test material is limited. In some cases, even smaller test specimens may be required due to very small amounts of test material.

A comparison made between T0 and scatter estimates from test results measured with miniature specimens (i. e., smaller than the Charpy-size) shows that the definitions of scatter and the measuring capacity (specimen size) criterion (eqn [14]) apply even to miniature specimens that are of 3 x 4 mm and

3.3 x 3.3 mm cross-section SE(B).15 The results indi­cate no bias between the T0 estimates measured with the miniature specimens compared to the overall
mean values of T0, which is shown in Figure 13 for the Charpy-size specimen and three subsize SE(B) specimens. The comparison demonstrates, in all respects, applicability ofthe miniature size specimens for the fracture toughness estimation using the Mas­ter Curve approach. In some datasets (e. g., on grade 15 Kh 2 MFA and on the International Atomic Energy Agency (IAEA) reference material JRQ), the smallest specimens (3 x 4 and 5 x 5 mm) produced some low T0 values, which most likely were caused by macroscopic inhomogeneity encountered with such small specimen dimensions. An example of miniature specimen results for the A508 Cl. 3 steel FFA (a French steel grade) is presented in Figure 14, demonstrating nearly consistent fracture toughness data independent of the specimen size.

Another benefit of using the Master Curve approach is that the uncertainty associated with the T0 estimation can be determined and taken into account for assessing a conservative, realistic estimate for the lower bound fracture toughness. As discussed previously, ASTM E 1921 does not set limits for the specimen size or configuration; however, the mini­mum number of test results, dependent on test tem­perature relative to T0, is predefined to ensure an acceptable minimum confidence level for the esti­mate. If a larger uncertainty for the T0 estimation is accepted, the minimum number of specimens required can be reduced from that given in the stan­dard. Correspondingly, having less than the minimum

number of valid data in the test series does not necessarily invalidate the dataset, but it does result in a lower confidence level of the estimate. It is essential that the most realistic confidence level is estimated and taken into account in final integrity assessments.

Very small specimens (around 3 x 3 mm) tend to give slightly (1-3 °C) higher values of T0 compared to 10 x 10 mm specimens.15 This trend is likely due to the censoring procedure, which screens out pro­portionally more data from the upper tail of the dataset than from the lower tail. This screening affects both the scatter caused by possible material inhomogeneity and that from statistical outliers. The optimal test temperature range for miniature specimens has been proposed to be —50 °C < T — T0 <—20 °C.

Even though more specimens are needed when smaller specimens are tested, the consumption of test material becomes smaller, even if more than the minimum number of specimens were tested
for the T0 estimate. In this respect, the 5 x 5 mm specimen is the least material consuming SE(B) size (~12 specimens are needed for the standard estimate if sys = 500 MPa).15 Using the 5 x 5 or 3 x 4 mm specimen geometry, it is possible to prepare up to 8 or 12 subsize specimens from the halves of one tested 10 x 10 mm specimen (see Figure 15). When selecting specimen configuration, it should be noted that with deeply cracked specimens having the liga­ment size equal to or less than the thickness, the ligament is the primary dimension limiting the specimen-measuring capacity, not thickness. Also, with slim (reduced thickness) and very small speci­mens (like the 3 x 4 mm cross-section), it is recom­mended to side-groove the specimens to increase stress triaxiality near the surfaces.