6.4 Subcritical cracks
So the picture that emerged from the fracture surface was of a weld failure. But did this mean the weld itself was faulty, was the material sub-standard, or had the material been attacked by the caustic soda? The design of the tank might also be at fault.
Further inspection of the tank was needed to check the other welds. Those welds at the same level would have been subject to the same hydrostatic pressure from the caustic soda.
The reason for examining those welds is that the pressure in the tank at the time of failure is given by the equation
where p is the hydrostatic pressure of the contents, ρ the density of the fluid, h the height from the top fluid surface and g the acceleration due to gravity. Pressure therefore varies with height, as shown in Figure 77, for fluids of varying density. In this case, as the density is fixed, and g is a constant, the only factor controlling pressure is the vertical distance from the surface of the liquid to the weld height. If a vertical weld had failed at a particular height, there could be evidence of other cracks at nearby welds at the same height.
The hunch proved correct, but only in one of the four possible welds (W4 as shown in Figure 1, Paper 4). The weld was obscured by its position, making inspection difficult (looking for a dark crack on black material in the dark); although when fine talc was used to dust the weld, it revealed several subcritical cracks (Figure 7, Paper 4).
But what did it indicate? It might show the welding was faulty, but did not explain the uncracked welds (W2 and W3). Were there any other clues to the cause? Another factor could be overloading: was there any distortion visible in the tank? After all, the distortion visible in the radiator reservoir described in the previous case study was a significant clue to that problem.
If Figure 3, Paper 4 is re-examined, such distortion does indeed exist. It is best seen by tipping the page and viewing the edges of the far side and the crack at an angle. Both features can be seen to be bowed outwards, the cracked weld showing the greater effect. The effect was confirmed by direct measurement of the circumference at several heights: a difference of 20 mm in a total of 8.55 m being measured for the centre of the cracked weld compared with a higher panel. Although the effect was small, it indicated possible overloading, producing creep of the panel material.