Page 283 - Handbook of Materials Failure Analysis

1 Introduction 279

investigate these effects on the response of two different types of fracture mechanics
specimens in the upper-shelf as well as the DBTT regime. The following issues and
their solution in the context of use of nonlocal damage models have been discussed.
The dependence of fracture resistance behavior and scatter of fracture toughness in
the DBTT regime on the following parameters has been studied, that is,
- Crack depth (shallow crack vs. deep crack);
- Specimen geometry (compact tension “CT” vs. single-edged-notched-bend
“SEB”);
- Specimen size (1T, 2T, and 4T CT specimens);
- Specimen thickness (keeping all other dimension unchanged)
- Boundary conditions (symmetric models vs. full model)
For simulation of probability of fracture in the DBTT regime, the combined nonlocal
Rousselier’s damage model [17–19] and Beremin’s cleavage fracture model [20–22]
was used. It was shown that local damage models are not able to predict several of
these effects including the important aspect of the use of symmetric boundary
conditions in FE analysis.
Almost all the specimens (especially, the fracture mechanics specimens) are sym-
metric to the expected crack plane in the mode-I loading case. In FE calculations, this
symmetry is usually exploited and hence, only the half of the specimen is included in
the geometrical model. In local damage-mechanics calculations, the damage tends to
localize at the integration points. As a consequence, the predicted crack moves
through the interior of the element (and not along the element borders as expected).
This is in contradiction to experimental observation where the crack propagation
path is on the element borders (and not through the element) in a symmetrically
loaded specimen, unless there are heterogeneities in material property and micro-
defect distribution in certain planes or directions. It was noted that there is a major
difference between the results of local and nonlocal models when the symmetric
boundary conditions are used.
In the DBTT regime, fracture toughness tests were conducted according to
ASTM E-1921 standard [23] and the variation of reference temperature T 0 with
respect to specimen size and geometry was studied. It was observed that the nonlocal
damage models are able to satisfactorily predict the fracture resistance behavior as
observed in the experiments and the effects of specimen size, geometry, crack-depth,
loading and boundary conditions on the J-R curve in the upper-shelf and probability
of fracture in the DBTT regime were also accurately predicted. It was also observed
that the reference temperature T 0 does not have a fixed value as used in the master-
curve approach [6]. The variation of T 0 for various types and sizes of specimens
were predicted by the FE simulation and these results were compared with those
of experiment. The advantage of nonlocal models is that the predicted results are
not dependent upon the mesh size near the crack-tip. It is also able to accurately
predict the effect of specimen geometry, size, loading and boundary conditions on
the fracture resistance behavior.
Dissimilar metal welds impose a challenge to the engineers concerned with the
structural integrity assessment of these joints. This is because of the highly