6.7 Tank structure
So, there appeared to be no serious problems with the material, processing or fabrication, but could the design of the tank be queried? The structure is shown in section in Figure 4, Paper 4, comprised essentially of 12 mm single thickness panels, which have been buttressed by three extra hoops of material at the base, centre and the top of the structure. But is this the best way of resisting hydrostatic load?
Equation (9) shows a simple linear relation between hydrostatic pressure and height for a given fluid (Figure 77). If that is the case, should not the thickness increase gradually with distance from the top? A dam increases in thickness from top to base to resist the water pressure, and so the same principle should apply to any fluid reservoir.
By adding hoops, the top hoop could be redundant, and the lower hoops might not be sufficient to resist the greater pressures towards the base of the structure. Such an hypothesis would explain why the failure occurred in a lower, unreinforced panel. It would not explain why only two welds failed from pinholes, but the calculation above shows the single thickness panel is having to resist a large hoop stress. Doubling the wall thickness would halve the applied stress, while having three panels here would give a stress of only about 1.15 MNm−2. Even if pinholes occurred in the weld, failure would be much less likely with such substantial lowering of the hoop stress.
So, the general conclusion of the stress analysis was that the design itself was faulty. To resist hydrostatic pressure, the tank should have been designed like a dam, rather than like a barrel (Figure 10, Paper 4).
Why should just two of the four welds have shown cracks? The answer came when the welding stage was inspected directly. The hoop of panels for such tanks is made sequentially by hot fusion welding, that is, by melting the surfaces of two panels and pushing them together. This is fine for three of the welds where the flat panels are joined, but difficult for the final joint when the ends have to be brought together by bending the sheet into a cylinder (Figure 9, Paper 4). It would certainly explain the lower quality of one weld, and the poor quality of another weld was probably caused by similar problems in bringing two large flat sheets together. The quality of such welds is tested for through-the-thickness holes using a spark tester, a method that will not detect partial pinholes. One rather disturbing aspect of the process is that the hoop so formed is under a bending stress, so the outer surface of the final wall will be in tension. This will, of course, make failure more likely, and is akin to the frozen-in strain problem of the radiator reservoir described previously. In this case, there is a frozen-in stress rather than strain.
It was possible to calculate the effect from simple bending theory, as described in Section 5, Paper 5 (page 222). The effect probably added about 1.5 MN m−2 to the hoop stress, so that the above calculation under-estimated the stress. This changes the effect of the stress concentration, reducing it to about 3, enhancing the effect of less serious pinholes.