In fracture toughness test standard BS EN ISO 15653 for weld and HAZ specimens, there is limited guidance about how the weld width, and the weld metal to parent yield strength mismatch ratio can affect the fracture toughness of materials, nor how to interpret the results should these parameters fall outside the permitted limits given in the Standard. This research was carried out to evaluate how the values of J determined from fracture toughness test results are affected by the weld width and weld strength mismatch, by using FEA models of single edge notched bend (SENB) specimens.
Fracture toughness results from specimens notched into the weld centreline, of different mismatch ratios and weld widths, are compared with homogeneous materials specimens. The results show that for any mismatch ratio, fracture toughness for welds wider than 20mm are similar to homogeneous material. However, the fracture toughness decreases as the weld become narrower than 20mm. When this variation is taken as a percentage between a homogeneous material and a welded specimen, J varies for each weld width following the same trend for a specific mismatch ratio, independently of the material strength. To enable prediction of the fracture toughness behaviour of welded specimens, equations have been developed for comparison to homogeneous weld metal or homogeneous parent metal, as a function of the weld mismatch ratio and the weld width.
The real question is whether the Standard methods can nonetheless determine weld metal fracture toughness accurately in overmatched welds of different widths. From comparisons of J values extracted directly from the contour integral, and J values calculated using the standard equation (but based on load-CMOD data), it was found that standard methods vary by less than 5% in the majority of the cases. The exception was for the combination of mismatch ratio above 1.46 and welds narrower than 20mm. Therefore, it could be said that the Standard BS EN ISO 15653 is conservative with the overestimation percentage of 10%, since all model cases within the Standard limits of mismatch of 1.5, the largest over-estimation was only 6%, and typically less than 4%.