5.1 Fracture toughness is expressed in terms of an elastic-plastic stress intensity factor, K_Jc, that is derived from the J-integral calculated at fracture.

5.2 Ferritic steels are microscopically inhomogeneous with respect to the orientation of individual grains. Also, grain boundaries have properties distinct from those of the grains. Both contain carbides or nonmetallic inclusions that can act as nucleation sites for cleavage microcracks. The random location of such nucleation sites with respect to the position of the crack front manifests itself as variability of the associated fracture toughness (16). This results in a distribution of fracture toughness values that is amenable to characterization using the statistical methods in this test method.

5.3 The statistical methods in this test method presume that the test materials are macroscopically homogeneous such that both the tensile and toughness properties are uniform. The fracture toughness evaluation of nonuniform materials is not amenable to the statistical analysis methods employed in the main body of this test method. For example, multipass weldments can create heat-affected and brittle zones with localized properties that are quite different from either the bulk material or weld. Thick section steel also often exhibits some variation in properties near the surfaces. An appendix to analyze the cleavage toughness properties of nonuniform or inhomogeneous materials is currently being prepared. In the interim, users are referred to (6-8) for procedures to analyze inhomogeneous materials. Metallographic analysis can be used to identify possible nonuniform regions in a material. These regions can then be evaluated through mechanical testing such as hardness, microhardness, and tensile testing to compare with the bulk material. It is also advisable to measure the toughness properties of these nonuniform regions distinctly from the bulk material.

5.4 Distributions of K_Jc data from replicate tests can be used to predict distributions of K_Jc for different specimen sizes. Theoretical reasoning (12), confirmed by experimental data, suggests that a fixed Weibull slope of 4 applies to all data distributions and, as a consequence, standard deviation on data scatter can be calculated. Data distribution and specimen size effects are characterized using a Weibull function that is coupled with weakest-link statistics (17). An upper limit on constraint loss and a lower limit on test temperature are defined between which weakest-link statistics can be used.

5.5 The experimental results can be used to define a master curve that describes the shape and location of median K_Jc transition temperature fracture toughness for 1T specimens (18). The curve is positioned on the abscissa (temperature coordinate) by an experimentally determined reference temperature, T_o. Shifts in reference temperature are a measure of transition temperature change caused, for example, by metallurgical damage mechanisms.

5.6 Tolerance bounds on K_Jc can be calculated based on theory and generic data. For added conservatism, an offset can be added to tolerance bounds to cover the uncertainty associated with estimating the reference temperature, T_o, from a relatively small data set. From this it is possible to apply a margin adjustment to T_o in the form of a reference temperature shift.

5.7 For some materials, particularly those with low strain hardening, the value of T_o may be influenced by specimen size due to a partial loss of crack-tip constraint (5). When this occurs, the value of T_o may be lower than the value that would be obtained from a data set of K_Jc values derived using larger specimens.

5.8 As discussed in 1.3, there is an expected bias among T_o values as a function of the standard specimen type. The magnitude of the bias may increase inversely to the strain hardening ability of the test material at a given yield strength, as the average crack-tip constraint of the data set decreases (19). On average, T_o values obtained from C(T) specimens are higher than T_o values obtained from SE(B) specimens. Best estimate comparison indicates that the average difference between C(T) and SE(B)-derived T_o values is approximately 10 °C (2). However, individual C(T) and SE(B) datasets may show much larger T_o differences (3, 20, 21), or the SE(B) T_o values may be higher than the C(T) values (2) . On the other hand, comparisons of individual, small datasets may not necessarily reveal this average trend. Datasets which contain both C(T) and SE(B) specimens may generate T_o results which fall between the T_o values calculated using solely C(T) or SE(B) specimens.

1.1 This test method covers the determination of a reference temperature, T_o, which characterizes the fracture toughness of ferritic steels that experience onset of cleavage cracking at elastic, or elastic-plastic K_Jc instabilities, or both. The specific types of ferritic steels (3.2.1) covered are those with yield strengths ranging from 275 to 825 MPa (40 to 120 ksi) and weld metals, after stress-relief annealing, that have 10 % or less strength mismatch relative to that of the base metal.

1.2 The specimens covered are fatigue precracked single-edge notched bend bars, SE(B), and standard or disk-shaped compact tension specimens, C(T) or DC(T). A range of specimen sizes with proportional dimensions is recommended. The dimension on which the proportionality is based is specimen thickness.

1.3 Median K_Jc values tend to vary with the specimen type at a given test temperature, presumably due to constraint differences among the allowable test specimens in 1.2. The degree of K_Jc variability among specimen types is analytically predicted to be a function of the material flow properties (1)2 and decreases with increasing strain hardening capacity for a given yield strength material. This K_Jc dependency ultimately leads to discrepancies in calculated T_o values as a function of specimen type for the same material. T_o values obtained from C(T) specimens are expected to be higher than T_o values obtained from SE(B) specimens. Best estimate comparisons of several materials indicate that the average difference between C(T) and SE(B)-derived T_o values is approximately 10°C (2). C(T) and SE(B) T_o differences up to 15°C have also been recorded (3). However, comparisons of individual, small datasets may not necessarily reveal this average trend. Datasets which contain both C(T) and SE(B) specimens may generate T_o results which fall between the T_o values calculated using solely C(T) or SE(B) specimens. It is therefore strongly recommended that the specimen type be reported along with the derived T_o value in all reporting, analysis, and discussion of results. This recommended reporting is in addition to the requirements in 11.1.1.

1.4 Requirements are set on specimen size and the number of replicate tests that are needed to establish acceptable characterization of K_Jc data populations.

1.5 T_o is dependent on loading rate. T_o is evaluated for a quasi-static loading rate range with 0.1< dK/dt < 2 MPa√m/s. Slowly loaded specimens (dK/dt < 0.1 MPa√m) can be analyzed if environmental effects are known to be negligible. Provision is also made for higher loading rates (dK/dt > 2 MPa√m/s).

1.6 The statistical effects of specimen size on K_Jc in the transition range are treated using weakest-link theory (4) applied to a three-parameter Weibull distribution of fracture toughness values. A limit on K_Jc values, relative to the specimen size, is specified to ensure high constraint conditions along the crack front at fracture. For some materials, particularly those with low strain hardening, this limit may not be sufficient to ensure that a single-parameter (K_Jc) adequately describes the crack-front deformation state (5).

1.7 Statistical methods are employed to predict the transition toughness curve and specified tolerance bounds for 1T specimens of the material tested. The standard deviation of the data distribution is a function of Weibull slope and median K_Jc. The procedure for applying this information to the establishment of transition temperature shift determinations and the establishment of tolerance limits is prescribed.

1.8 This test method assumes that the test material is macroscopically homogeneous such that the materials have uniform tensile and toughness properties. The fracture toughness evaluation of nonuniform materials is not amenable to the statistical analysis methods employed in the main body of this test method. Application of the analysis of this test method to an inhomogeneous material will result in an inaccurate estimate of the transition reference value T_o and non-conservative confidence bounds. For example, multipass weldments can create heat-affected and brittle zones with localized properties that are quite different from either the bulk material or weld. Thick section steel also often exhibits some variation in properties near the surfaces. Metallography and initial screening may be necessary to verify the applicability of these and similarly graded materials. An appendix to analyze the cleavage toughness properties of nonuniform or inhomogeneous materials is currently being prepared. In the interim, users are referred to (6-8) for procedures to analyze inhomogeneous materials.

1.9 This standard does not purport to address all of the safety concerns, if any, associated with its use. It is the responsibility of the user of this standard to establish appropriate safety and health practices and determine the applicability of regulatory limitations prior to use.