Page 197 - Failure Analysis Case Studies II

182

The third factor, the frozen-in strain, produced a visible widening of the original crack of ca 0.5
mm over a diameter of approximately 50 mm. This is equivalent to a strain of approximately 1%,
or approximately 0.5 kN load, equivalent to a stress of approximately 20 MN m-2 by interpolation
on the load-extension curve. So the total resultant (pessimistic) stress is given very approximately
by the equation

Totalstressm(25 x 1.55)+20 = 59MNmP2 (4)
This value may be compared with the best experimental estimate of about 80 MN mP2, and with
a lower bound of about 55 MN m-2 for the inherent strength of the composite nylon material.
The argument may thus be summarised. The combined effects of a geometric stress concentrator
at the corner of the adjacent buttress and cracks (either present as a void or at the surface of a
cold slug or weld line just below the corner), effectively magnified the real stress experienced by
the material by some 25 times. The material of the tank was also in a state of strain produced by
cold moulding, so that an extra component of about 20 MN mP2 must be added to the magnified
stress, giving a total stress of about 59 MN m-’. This value is comparable with the mean strength
measured for the material, and exceeds the lowest value actually obtained. It thus becomes possible
to see why cracks were initiated near the buttress corner and grew intermittently with each
successive pressurisation of the tank. Crack growth would, of course, have accelerated with each
successive exposure, since the length of the crack and hence the magnification at the tip would
have grown in step. The last event would probably have been the worst, and the event which
directly caused catastrophic leakage of cooling fluid, before the crack re-stabilised, owing to
relaxation of the frozen-in strain (Figs 3 and 6).
It is finally important to point out that no allowance has been made in this calculation for the
extra stresses imposed on the tank through the buttresses and lugs by fitment of fans, motors,
cowls etc. Similar considerations apply to the inlet and outlet pipes, especially as they will be
stressed by fitment of connecting tubes. All such add-ons will of course exacerbate the situation.
The possibility that the rogue moulding was produced during the warm-up period of the injection
moulding machine, remains the most likely cause of the failure.

6. Conclusions

1. A failed radiator tank has been examined in detail for the origin and causes of its rapid
catastrophic fracture on a new car. The crack was brittle in nature and had started at or near a
corner buttress. It propagated in several steps, probably corresponding to intermittent use of
the car, and exposure of the radiator to a normal, expected hydrostatic internal pressure of 25
psi. This pressure is equivalent to a sidewall stress of only about 1.55 MN mP2, well within the
ca 80 MN md2 strength of the material.
2. This benign stress was magnified, however, by a combination of three factors, two of which are
related to the moulding conditions under which the product was made, and the third is related
to the design geometry of the tank. The first factor was the presence of cold slugs of unmelted
or partly melted material in the outer sidewall, probably caused by moulding into a cold tool
or using too cool a melt. A larger, similar slug was discovered near the sprue, but had not led
to failure since it was not near to the second factor involved, a geometric stress concentrator.