Page 101 - Failure Analysis Case Studies II

86

the first penetration of the crack to the interior of the tank, or some time afterwards, when the
crack was examined by firemen, inspectors and workmen. The in situ crack could be opened by
hand pressure against one of the adjacent panels, and it is likely that inspections simply extended
the size of the brittle crack. It is important to note that this hypothesis would explain why the
initial size of the jet was only “. . .8 or 9 inches.. .” wide. This visual estimate corresponds not
unreasonably with the actual 134 mm (5.25 in) length of the central white zone. Alternatively,
other crack arrest lines in the proximity of the white zone give measurements of 7.25 in (1 84 mm)
and 8.75 in (222 mm), which may indicate that the critical crack had grown somewhat after the
first penetration of the tank. The crack would have then been wetted by the caustic soda, and
further crack growth enhanced through the lowering of surface energy.

4. Re-examination of the failed tank

Since the mechanism of failure of the panel had now been identified as brittle fracture through
slow crack growth from pre-existing flaws (pits) within the weld zone, it was now thought essential
to determine why only one of the four vertical welds had failed in this way. The problem could be
tackled in two ways. Firstly, was it true that the other panels were completely unaffected by the
high loading stresses? Examination on site had been necessarily made under poor lighting
conditions, and some of the welds were physically difficult to access. The analysis of fracture clearly
indicated that catastrophic failure must have been preceded by the slow growth of hairline cracks
from defects on the outer surface of the weld, so that close examination of the other welds should
reveal similar, but sub-critical features either if they had been welded under the same conditions,
or if there was a more general problem with the basic design of this tank. It was therefore decided
to re-visit the site before the tank was destroyed totally, and inspect the other welds directly.
Secondly, the behaviour of good versus poor weld material had been roughly tested by comparing
the behaviour of sections cut from the two large panels. Both specimens fractured in a brittle
manner along the weld, the fracture surface from the poor weld showing a distinctly greater
number of internal defects than the good weld material. So it was natural to place such a result on
a more systematic basis by cutting tensile specimens for rigorous analysis. Further, tests of each
weld and surrounding panel would be required to ascertain whether or not there were any material
differences between ’good’ and ‘poor’ welds which could explain the differences between them.
The prime objective of a second visit to the site was to inspect the final two vertical welds which
had not been examined earlier, for hairline cracks. The first weld examined was that one lying
partly under the HazChem warning notice (Fig. 2), and it showed an irregular bead smoothing
mark, but otherwise no hairline cracks were detected. Fine grade talcum powder was smeared over
the weld to highlight any cracks. The second weld proved more difficult to access, since it lay on
the side of the tank nearest the outside wall of the factory, and was therefore in almost complete
darkness. The bund wall was also very close to it, so making access extremely difficult. There
appeared to be numerous linear features along the weld zone, so the whole weld was taken away
for closer examination.

4.1. Confirmatory microscopy
Inspection of the two welds extracted with a small part of the adjacent panels confirmed the
results obtained in the factory. The whole weld sections showed numerous hairline cracks in the