A stainless steel expansion bellows was utilised in an airconditioning system at the “Tygervalley Mall” shopping centre. The bellows was used to compensate for linear movement of the pipes associated with inevitable thermal expansion. Failure of the bellows occurred which allowed massive leakage from the system.  The failed bellows is shown in figure 1 where the burst part can be seen on the right hand side.

Investigation

Visual inspection of the bellows was undertaken followed by microstructural analysis of the strip making up the convolution at the failure site.  The bellows is classed as an unrestrained type of bellows.  It consists of a 145 mm length of convoluted 0.5 mm 304 stainless steel welded to a carbon steel flange assembly. The flange is 10 mm thick. An inner pipe is welded to the flange on the one side and extends 140 mm longitudinally into the bellows as shown in the schematic in Figure 2.  The pipe within the bellows is a design feature meant to prevent turbulent effects in the system as a result of fluid flow in contact with the convolutions.  The fracture was found to occur circumferentially on the crown of the second convolution from the flange assembly.

Figure 2: Schematic of bellows assembly

A site inspection was undertaken to ascertain any additional information with regard to the arrangement and mechanical loading of the bellows. Figure 3 is a schematic of the piping system layout relevant to the bellows and Table 1 lists the fluid flow parameters in the system.  The bellows is situated upstream of an overhead 900 elbow which extends to a fixed point 30 m upstream. The bellows inline pipe is unrestrained at both ends and thus any loading as a result of linear thermal expansion is considered to extend the bellows in a lateral fashion. The failure occurred on the elbow side of the bellows facing in the direction of the perpendicular pipe.  Figure 4 shows a photograph taken of the bellows/pipe installation and it can be seen that the bellows and pipe is covered with lagging to insulate the system.  The flanges of the bellows was reported to us to be on the right hand side of the pipe.

Figure 3: Schematic of piping arrangement

Figure 4: Photograph of bellows installation (restricted section)

Table 1: Fluid flow parameters in the pipeline

Results

A microstructural investigation was completed on a section through the bellows and the flange at the weld. No evidence of a heat affected zone could be found in the region of the bellows failure. Further, no evidence of inclusions was found in the material. The microstructure is characteristic of a cold worked austenitic stainless steel. Spectrographic analysis was conducted on the bellows material which is shown in table 2 together with the AISI 304 specification for this steel. Table 2 shows the results of the SA compared with the AISI standard, which indicates that the material is within the specification for an AISI 304 stainless steel.

Table 2: Compositional analysis of bellows material

Scanning electron microscopy (SEM) was conducted on the fracture surface.  Figure 5 shows the fracture surface and indicates that the crack probably propagated slowly and the surface was progressively corroded as the crack propagated.  This indicates that a crack propagated slowly over time, probably by corrosion assisted fatigue. 

Figure 4 - Scanning electron microscope image of the fracture surface

Discussion

The results of this investigation indicate that failure was not a consequence of fatigue or stress corrosion cracking. If this were the case additional cracks would be observed in other areas on the bellows. Further there was no evidence of corrosion product within the bellows. Some pitting corrosion could be observed but was mild and superficial. Consequently failure by corrosion was ruled out. The corrosion that could be seen on the fracture surface is not an indication that corrosion as such caused the failure. The corrosion observed was a result of the progression of a crack by another means. The fact that failure was localised to one convolution of the bellows together with the morphology of the fracture surface was an indication of some form of point load at this surface. The role of the hanger was considered but then ruled out since the hanger is positioned away from the flange rather than on the bellows side of the flange.

In terms of the extraneous loading on the bellows, several possibilities were considered including failure as a result of fatigue via vibration, fatigue via linear thermal expansion or via a water hammer effect. Water hammer was excluded as a source of failure since pump shut off is ramped down preventing the formation of a pressure pulse sufficient to be a cause of failure. Vibration was also excluded since the fluid flow speed was within the design parameters and insufficient to set up any vibrational mode on the bellows material. The pipe installation was assembled at ambient conditions meaning that maximum linear thermal expansion over the temperature range is 7 - 8 mm in either direction. With regard to the geometry, an 8 mm lateral movement of the bellows will cause an impingement of the inner collar on the convolution, between which there is a clearance of 3 – 5 mm. In an unrestrained situation the bellows is designed for up to 35 mm of lateral movement. The installation of the inner pipe causes the bellows to become stiff, the same stiffness of the pipe, but since it is not welded to the flange nearest the t-piece, the bellows bends about the free end of the pipe. This causes all the load associated with the normal expansion and contraction of the system to be concentrated on one convolution of the bellows. A similar installation in way of repairs will probably result in a similar failure.