Page 97 - Failure Analysis Case Studies II

82

almost half the total capacity of some 30 t, and had only been delivered to the premises that day.
Pumping of the fluid from the tanker into the tank had been completed by about 09:15 h, and
nothing untoward had apparently been observed by operators working in the factory, either at
that time or later during the day before catastrophic fracture. It is worth observing that the corner
of the factory where the tank stood was poorly illuminated (Fig. 2). Sub-critical cracks would have
been difficult to spot, since they would simply have been fine hairline features against a black matte
background of the polypropylene material of the panel walls. There was no evidence of any leakage
of fluid before the accident near the critical crack-such leakage would have been relatively easy
to see, owing to the contrast between the white deposit and the black surface of the tank.

2.1. Site inspection

A visit to the site showed that failure had occurred by a single crack apparently running the
complete length of one of the four vertical welds directly visible on the exterior wall of the tank
(Fig. 3). The crack was brittle and it was whitish in appearance in one large region near the centre
of the panel. There was also extensive surface contamination from caustic soda crystals and sodium
carbonate. The results of measurement of the tank are shown in Figs 1 and 4. Its circumference
was also measured at two points. One was at a height about 1.5 m above the ground, approximately
level with the centre of the crack, and the other at a location well away from the crack, at the top
of the second buttress. The results were: circumference at crack level = 8.57 m; circumference at
top of second buttress = 8.55 m.
The difference of 2 cm thus represented creep of the wall, plus any crack opening that may have
occurred as a result of the fracture. Since the crack had only opened by about 1 mm, it was
concluded that creep of the material accounted for most of the increase in circumference, of about
0.23%. Some bulging of the failed panel was in fact detectable by eye, as shown in Fig. 3 (the effect
can be seen clearly by tilting the figure and looking along the edge of the tank). The tank had been
created at crack level by welding four 12 mm thick panels together, three being of roughly equal
contour length (2.43,2.48 and 2.45 m clockwise from the cracked panel, Fig. 4) the panel adjacent
to the crack having 1.21 m contour length. The visible tank structure at first buttress level (i.e.
below the crack) had been created in a similar way by welding three panels of about the same
contour length (2.47, 2.49 and 2.48 m clockwise, offset by 0.71 m from the cracked panel weld).
The final panel comprised a panel of contour length 1.31 m.
The surface of the tank was examined closely for any trace of surface damage, whether from
impact or abrasion, which might have accounted for the catastrophe. None of any kind was found,
the patterned exterior surface appearing pristine in condition, especially near the crack itself. If
impact had initiated failure, then it was here that might show surface damage (Fig. 3). Even small
impacts could damage the plastic, owing to its relative softness. It was decided to cut out several
areas of the tank with a circular saw for further analysis of the fracture surface.

3. Fractography
The fracture surface was revealed by estimating where the ends were situated, and then carefully
cutting through the double panels so as to leave a few mm of material holding the crack together