History

A PVDF lined GRP pressure vessel containing 250 litres of concentrated sulphuric acid failed catastrophically after 5 years in service. The damage created by high velocity vessel fragments and by the release of acid was extensive. The integrity of similar tanks at the site could no longer be relied upon and the manufacturing operation was suspended pending an investigation.

The operating conditions were as follows:

• Normal working pressure = 0.2 MPa,
• Normal acid temperature = 50 °C,
• Acid concentration ranged from 50 – 80%.

The vessel construction was as follows:

• Horizontal cylinder with domed ends,
• PVDF liner of 6 mm thickness,
• Helically wound inner structural laminate of E glass fibre and epoxy resin,
• Additional hoop reinforcement in the barrel region using hoop windings of E glass, and epoxy.,

The weight of the vessel was carried by two rubber lined saddle supports.

Inspection and analysis

The fracture surfaces of the epoxy laminate were generally rough due to extensive fibre pull-out. Two evidently contiguous fragments contained planar fracture surfaces which were essentially smooth, these being about 20 cm in length. The PVDF liner was generally distorted and torn but exhibited no evidence of brittle fracture. Fragments were reassembled. The fracture initiation site (A) as shown in Figure 5.13 was traced to that part of the cylindrical section of the vessel adjacent to one of the supports, this being the location of the planar smooth laminate fracture.

A detailed inspection of the planar fracture surfaces revealed that they were smooth to about 90% of the laminate thickness. This would suggest that the fracture initiated at the outside surface and developed parallel to the hoop windings.

Detailed inspection of other similar vessels revealed multiple circumferential microcracking in the external surface adjacent to each saddle support.

Figure 5.13 Location of fracture initiation on GRP pressure vessel

Failure diagnosis

The circumferential planar fracture surfaces within the hoop windings could be the conventional response to overstressing in the axial direction. The laminate is simply splitting along its weakest plane. However the almost complete local absence of fibre pull in the helically wound part of the laminate could not be explained by overstressing alone. It would suggest that the crack has penetrated the laminate to about 90% of its thickness by slow growth at modest stresses and has then suffered fast fracture. Slow crack growth in GRP is most often linked with acid stress corrosion of the glass fibres. It is mainly for this reason that glass reinforced laminates in contact with acids are protected by a resin rich barrier layer or (as in this case) a solid acid resistant thermoplastic liner.

Chemical analysis of the planar fracture surfaces revealed the presence of calcium sulphate. This is known to be the most abundant reaction product resulting from sulphuric acid attack on E glass fibres. To check whether acid contact following rupture could explain the presence of the sulphate the analysis was repeated on fast fracture surfaces. No calcium sulphate was detected.

The unusual inference in this case is that the corrosion appears to have developed as a result of contact with acid on the external surface of the vessel. The probable cause was eventually traced to a defective flange coupling. Maintenance records revealed that this had been replaced 6 months before the failure because of leakage. The location of the flange could have soaked the external surface of the vessel in the region where cracks first developed.

Therefore the failure sequence was concluded to be as follows:

1. Resin microcracking on the external hoop wound surface due to excessive stress alone. The saddle width was insufficient to carry the weight of vessel and its contents without locally exceeding the resin or resin glass bond strength. Microcracking observed in other similar vessels suggests that this was a basic design error.
2. The microcracks were in contact with sulphuric acid due to leakage. This would accelerate resin cleavage crack growth in the hoop layers.
3. Eventually the acid would have access to the stressed helix wound fibres. The fibres would suffer from stress corrosion cracking at modest stresses with local stresses increasing as the crack develops.
4. When the crack length exceeds a critical length, fast fracture occurs due to fibre pullout and stress alone.
5. The PVDF liner tears.

Lessons and consequences

1. The weakest cracking plane in a laminate is parallel to the fibre direction. Although in this case the pressure stresses have been efficiently accommodated, the local bending stresses due to support and constraint have been given inadequate attention. This is a common oversight that explains why the vast majority of laminated structures that fail do so at joints, changes in section and other local geometrical features where bending stresses maximise.
2. Once cracked, a stressed laminate is susceptible to stress corrosion cracking if the crack is in contact with an acid.
3. In most industrial environments involving acid transport and containment the risk of external acid contact should not be ignored. The painting of external surfaces with flexibilised epoxy is a cost effective safety measure. In addition, the use of an acid resistant grade of reinforcement is recommended.
4. The remaining vessels were deemed to be safe provided the surface microcracks were removed and the abraded surface recoated with resin, also with the proviso that the saddle widths were doubled.