Return to Failure Analysis Case Histories

Fatigue Cracking of Steam Line Expansion Bellow

ENVIRONMENT:

EQUIPMENT:

MATERIAL:

FAILURE MODE:

Summary

Failure of the bellows occurred first as a fatigue crack at the weld joining the lower end of the bellows to the pipe.  Failure then propagated as ductile tears perpendicular to the cracked weld.  Once the bellows opened up enough to relieve the internal pressure, cracking continued to propagate as chloride stress corrosion cracks until the bellows was removed from service. 

Background 

The subject bellows was part of an expansion joint assembly in a superheated steam line operating at 750°F.  This assembly was oriented in a vertical position, and the failed bellows was the upper of two bellows in the overall assembly.  There was an internal flow guide on the inside of this bellows in the form of a portion of straight stainless steel tube welded to the pipe at the bottom edge of the bellows and open at its top.  The purpose of this flow guide was to minimize steam turbulence inside the bellows. 

This tubular internal guide had one or more drain holes drilled near its lower end to prevent steam condensate from accumulating between the guide and the bellows.  The drain holes were a fraction of an inch above the joining circumferential weld, which meant that a small amount of condensate could still accumulate there.  Plant personnel were concerned that such an accumulation might have somehow led to the failure of the bellows. 

There was also an external protective shield, in the form of a piece of steel pipe welded to an external ring just above the bellows and open on its lower end.  This shield was larger than the bellows and stood out from it on all sides so as not to contact the bellows directly.  The question was also raised as to whether this shield might have contributed to the failure. 

The bellows was made of Type 304 stainless steel and was in service for three years. 

Findings

Visual examination revealed that the bellows had cracked along the weld that had joined the bottom end of the bellows to its mating pipe.  This crack extended approximately half way around the circumference of the bellows.  The remainder of the circumference had been cut with an abrasive disk to remove the bellows.  It was not clear whether the crack originally extended only about half way around the circumference, or whether it had been longer but part of it was obliterated by the cutting operation. 

When viewed under a low-power microscope, the circumferential crack had a “woody” fracture surface typical of fatigue failures of stainless steel welds.  This is illustrated in Figure 1.  No clear initiation point for this crack was found, and no mechanical damage was seen that might have caused high local stresses to initiate such a crack.

Initiating from the circumferential crack were several other cracks, mostly smaller in magnitude than the circumferential crack and starting perpendicular to it.  These include: 

Crack Observation #1

A long crack ran roughly perpendicular to the circumferential crack for a distance of about six inches, and then itself became circumferential in nature for a distance of roughly three inches.  At this point, it turned and ran roughly along the axis of the pipeline. 

On close examination it was seen that this second crack, where it took off from the circumferential crack, was on a plane approximately 45° relative to the surfaces of the bellows.  Unlike the woody appearance of the first crack, this one had a smoother surface typical of simple tensile failure.  After traveling perpendicular to the circumferential crack for about an inch, this second crack split into two cracks in a “Y” pattern.  One leg of this “Y” continued on while the other leg propagated only a short distance., Figure 2. 

The last inch or so of the crack went off at an angle.  This last portion of the crack showed a distinct branching pattern.  Microscopic examination of this crack in a plane cut through the bellows wall showed that the crack propagation in this case was from the outside in. 

A third crack, also starting perpendicular to the major circumferential crack, was observed.  This crack initially occurred on a 45° plane as did the second crack, but, after traveling roughly two inches, made an abrupt 90° turn in a circumferential direction.  Shortly thereafter it split into two cracks.  One continued in the circumferential direction and the other propagated in a more axial direction.

Three additional short cracks were observed.  Two of these three cracks, like the last inch of the second crack described in “1” above showed a more wandering nature with much branching.   

Visual examination of the various crack faces showed that the major circumferential crack was discolored (oxidized), whereas the other cracks were less discolored and, in some cases, had a bright metal surface.  The oxidation of the primary crack was evident in Figures 1. 

The last inch of the second crack (item 1 above) that included the wandering, branched characteristics, was cut out, mounted, and examined at higher magnification.  The branching nature was apparent in Figure 53  In the higher magnification view in Figure 4 it was seen that the crack was transgranular.  This branching, transgranular pattern is typical of chloride stress corrosion cracking of stainless steels.

The mounted specimen was placed in the SEM and an EDS semi-quantitative microprobe analysis done at several locations in and away from the crack.  A summary of the analyses was compiled in Table 1. 

Table 1.

EDS Analysis (WT%)

In crack

Near crack

Iron

55

68

Chromium

20

20

Nickel

2

9.2

Silicon

7

1.0

Chloride

14

--

A sample of the metal adjacent to the primary circumferential crack was cut and polished and microhardness measurements were made in the base metal, the weld metal, and at the weld/base metal interface.  In all cases the readings fell within the range of 92-98 Rockwell B. 

Another sample was cut from the bellows, mounted, and electrolytically etched with oxalic acid to look for carbide precipitation along grain boundaries.  The grain structure was normal for this alloy and no grain boundary carbide precipitation was seen. 

Discussion

The “woody” appearance of the circumferential crack is typical of intergranular fatigue of wrought stainless steel or interdendritic fatigue of stainless weld metal.  When such alloys are exposed to elevated temperatures above about 1000°F, a second phase called “sigma” phase slowly forms along grain boundaries.  This metallic phase is extremely brittle and has been blamed for similar failures of stainless steel components in the past.  Sigma phase is also quite hard, however, and its presence can usually be detected by making microhardness measurements of the alloy.  In this case there was no such apparent increase in hardness, as evidenced by the very normal measurements obtained in and adjacent to the fracture zone.  

One other known cause for fatigue being intergranular is precipitation of a brittle carbide phase at the grain boundaries resulting from what is called “sensitization” of the alloy due to extended exposure to elevated temperatures.  The 750°F operating temperature of this bellows is somewhat low for sensitization of Type 304 stainless steel.  Indeed, no such grain boundary precipitation was found.  This particular failure, therefore, was simply a result of high cyclic stresses along the weld caused by vibration in the line.  The edge of the circumferential fillet weld in this case is an area of high localized stresses, and is the area most likely to develop such fatigue cracks given cyclic loading of the part. 

Corrosion fatigue is perhaps a more accurate description of the failure.  Corrosion fatigue simply refers to mechanical fatigue that occurs in a medium somewhat more corrosive than dry air.  Steam certainly qualifies as such a medium, especially since chloride was discovered in the crack..  To have corrosion fatigue does not mean that there is visible corrosion of the metal.  Rather, it simply means that the fatigue crack develops more rapidly than it would have if the part were only exposed to dry air.  Corrosion fatigue usually cannot be determined after the fact, as the crack morphology appears identical to other mechanical fatigue.  The only way to show that corrosion fatigue is a factor would be to run controlled laboratory fatigue tests of the part in question in both dry air and in the actual medium of exposure – steam in this case.  The extent to which failure occurs more rapidly in the non-air medium determines how susceptible that part is to corrosion fatigue in the given environment.  In this case, the presence of steam is a given, and the role that it may have played in accelerating this particular failure is academic. 

web@corrosionlab.com