PSI - Issue 13

Hans-Jakob Schindler / Procedia Structural Integrity 13 (2018) 398–403 Author name / Structural Integrity Procedia 00 (2018) 000 – 000

401

4

3.2. Lower shelf On the lower shelf, the material exhibits no plasticity prior to cleavage initiation, so K Ic -testing according to ASTM E 399 is adequate. Unfortunately, for some (non-understandable) reasons, there is a note in E399 saying that this standard should not be applied to ferritic steel in the DBT-range or below. Instead, the user is directed to E1921 or E1820. The latter includes evaluation of J Ic for brittle behavior, which can be transferred to K JIc . Thus, formally, there is no way anymore to get a standard K Ic for structural steels, which is strange. Actually, there is no good physical reason for not to use E399 for structural steels on the lower shelf. Testing based on E399 is relatively simple, so there is no strong need for an alternative or for estimates. Correlation with CV-data are suggested in the literature for this range, too. However, these empirical correlation formulas are of limited accuracy, since for KV-values on the lower shelf, there are several influencing factors and sources of errors in KV. Therefore, direct evaluation of K Ic by E399 is preferable. 3.3. Ductile-to-brittle transition Ductile-to-brittle transition (DBT) is the most important toughness range for structural steels. In this range, FT in terms of K Jc rises by about one order of magnitude within a temperature range of about 100°C, so it is strongly temperature-dependent. Furthermore, it exhibits inherently a large scatter and a high sensitivity to influencing factors such as crack-tip constraints, loading rate and specimen size. According to ASTM E1921, K Jc (T) in the DBT-range is characterized by just one parameter, the so-called reference temperature T 0 . The latter is determined by a statistical procedure based on Weibull-statistics. At least six “valid” tests with standard CT or SENB specimens are required for a standard T 0 . In practice, about 8 -10 specimens should be available to get a valid T 0 . Despite of the enormous testing effort to get a valid T 0 , it is affected by a large uncertainty. ASTM E1921 contains several disclaimers concerning the precision of T 0 , by mentioning biases and differences in T 0 depending on specimen shape, specimen size and loading rate. For pre-cracked CV-specimens (PCCV), a mean difference of 10°C (compared with CT-specimens) is mentioned in E1921 as a corresponding bias. However, there are publications (e.g. Nanstadt et al. (2009)) showing that the difference can be up to 45°C. Furthermore, ASTM E1921 mentions a possible influence of specimen size, without providing numbers. Recently, the author witnessed an unpublished large test series on a RPV-steel, where T 0 from PCCV-specimens was found to be 55°C lower than the one from 1T-CT-specimens and the difference between 25 mm and 10 mm thick CT-specimens was as much as 45°C. Furthermore, as shown by Schindler and Kalkhof (2013), the loading rate can influence T 0 significantly, depending on the test demperature. Thus, the possible accumulation of these effects in some cases can lead to an uncertainty of T 0 in the order of 50°C, which is unacceptable for most practical purposes. Considering this uncertainty on one hand and the huge testing effort associated with standard determination of T 0 on the other, it is usually preferable to estimate T 0 from the empirical relation with CV impact data such as T 0 = T 41J –  T 0/41J with  T 0/41J = 24°C (3) T 41J denotes the temperature where the CV-impact energy reaches the level of 41 J. Eq. (3) has proven to be quite accurate and reliable (Sokolov and Nanstadt (1999)), although, for theoretical reasons, the temperature shift  T 0/41J is expected to depend somewhat on the yield stress. The uncertainty of T 41J is much smaller than the one of T 0 , if the latter is determined by relatively small specimens such as pre-cracked CV-specimens. Ironically, eq. (3), which is provided in ASTM E 1921 in order to estimate a suitable test temperature for T 0 determination, often is more accurate than the T 0 subsequently obtained by the standard procedure. Considering, on one hand, the relatively minor or even absent benefit in terms of accuracy of standard FT testing, compared with correlations from CV-testing, and the much higher testing effort on the other, performing standard K Ic or J Ic -tests is often not worthwhile. Estimates from CV-data are usually sufficient for the practical needs. However, there are situations where improved FT-data are required or would be beneficial. After all, the classical CV-test was introduced more than 100 years ago just intuitively as a simple test to characterize toughness qualitatively. Interpreted as a FT test, it does not fulfill a single validity criterion of standard FT-testing, by far. Still, 4. Suggestions for more suitable fracture toughness testing standards