Reactor Cooling Water Expansion Joint Bellows: The Role of the Seam Weld in Fatigue Crack Development


 The secondary cooling water system pressure boundary of Savannah River Site reactors includes expansion joints utilizing a thin-wall bellows. While successfully used for over thirty years, an occasional replacement has been required because of the development of small, circumferential fatigue cracks in a bellows convolute. One such crack was recently shown to have initiated from a weld heat-affected zone liquation microcrack. The crack, initially open to the outer surface of the rolled and seam welded cylindrical bellows section, was closed when cold forming of the convolutes placed the outer surface in residual compression. However, the bellows was placed in tension when installed, and the tensile stresses reopened the microcrack. This five to eight grain diameter microcrack was extended by ductile fatigue processes. Initial extension was by relatively rapid propagation through the large-grained weld metal, followed by slower extension through the fine-grained base metal. A significant through-wall crack was not developed until the crack extended into the base metal on both sides of the weld. Leakage of cooling water was subsequently detected and the bellows removed and a replacement installed.

a thin-wall bellows.
While successfully used for over thirty years, an occasional replacement has been required because of the development of small, circumferential fatigue cracks in a bellows convolute.
One such crack was recently shown to have initiated from a weld heat-a2"fected zone liquation microcrack. The crack, initially open to the outer surface of the rolled and seam welded cylindrical bellows section, was closed when cold torming of the convolutes placed the outer surface in residual compression.
However, the bellows was placed in tension when installed, and the tensile stresses reopened the microcrack. This five to eight grain diameter microcrack was extended by ductile fatigue processes. Initial extension was by relatively ra:pid propagation through the large-grained weld metal. followed by slower extension through the fine-grained base metal.
A significant through-wall crack was not developed until the crack extended into the base metal on both sides of the weld.
Leakage of cooling water was subsequently detected and the bellows removed and a replacement installed.
1Materials Technology Section, Savannah River Technology Center Westinghouse Savannah River Company, Aiken, SC

BACKGROUND
The Savannah River Site (SRS) has, five nuclear material production reactors (C, K, L, P and R) which were designed and built in the early 1950 In 1958, the joints were modified to reduce the tendency for fatigue and sinde that time, approximately seven leaks have been detected in expansion joints in process water systems.
In each case, the leaking expansion joint was replaced before reactor operations were resumed. Nuclear materials production at Savannah River was temporarily halted in 1989 so that safety upgrades could be made to the reactors systems.
The restart of the production reactors included hydraulic testing of the primary and secondary cooling systems of the K-reactor. During this testing, a crack developed in a convolute of an expansion joint in the secondary cooling water system. This crack provided a path for river water to leak from the bellows and necessitated both the replacement of, and a failure analysis for, the expansion joint. This paper presents the metallurgical evaluation portion of that failure analysis.

FAILURE ANAYSIS TECHNIQUE
A crack in the third of sixteen convolutes of an expansion joint in the secondary cooling water system was discovered during restart testing of K-reactor. The circumferential crack, initially detected because of water accumulation in a sump, was approximately thirteen centimeters (five inches) long and was centered on an autogenous gas tungsten arc (GTAW) seam weld in the bellows, Figure 2. Testing was discontinued as soon as the leak was detected and the leaking bellows was located by a walkdown el" the piping system.
The installed length of the expansion joint and tie rods were measured before the bellows and the associated portions of the expansion joint, Figure  3, were removed from the cooling water system.
The crack, even though it was approximately thirteen centimeters (five inches) long, was not opened significantly and closed tightly when the bellows was removed from the expansion joint assembly.    in the HAZ of the autogenous GTAW seam weld, and are defined as HAZ liquation cracks [1]. The grain morphology in the weld fusion zone was typical of a procedure utilizing a "high" welding speed and producing a teardrop-shaped weld pool [2], Figure 6. Large grains grew epitaxiatly from the base material to the weld centerline at a constant angle with respect to the weld axis. This microstructural evidence indicates that the GTAW process was utilized effectively and the overall heat input was lower than for a slow welding speed (elliptical weld pool shape), which helped to minimize the size of the liquation zones, Figure 7. The liquation cracks initiated on the outer surface of the cylinder and penetrated approximately five to eight grains into the weld HAZ.
Cold forming of the convolutes closed the microcrack responsible for the failure by placing the outer surface of the bellows in residual compression at the convolute intrados. A metallographic cross section of a potential microcrack is shown in Figure 8.
Such microcracks were only observed near the outer surface of the bellows, which was the top surface of the weld.
This may be attributed to a steeper temperature gradient on the back side of the weld from a chill block, fixture or purge gas.
A steep temperature gradient would reduce the width of the partially melted region (liquid at grain boundaries) in the HAZ and the likelihood of cracks occurring [31.  The fracture surface topography was typical of fatigue, Figure 9.
Striations were apparent throughout the surface and, based on striation spacing, the fatigue fracture process was very rapid across the large grained microstructure typical of the weld fusion zone. Once the crack bridged the weld fusion zone, it grew circumferentially along the intrados of the convolute (outer surface), at approximately equal rates on either side of the weld.
Based on the shape of the partially through-wall thumb nail crack, the predicted shape of the through-wall crack indicated it was approximately one to three centimeters (0.5 to 1.25 inches) long when it penetrated the wall of the bellows.
At this point, the crack opening was not significant and little or no water leaked from the secondary cooling water system. Subsequent crack growth increased the crack opening and when the crack length reached approximately thirteen centimeters (five inches), the leak rate through the bellows was sufficient for detection through accumulation in a sump which was instrumented for such detection. Stains on the fracture surface of the thirteen centimeter (five inch) crack were also consistent with a lack of significant leakage from the initial through-wall crack portion. The surface of the initial through-wall crack portion was covered with a light redish-brown film, indicating increased age relative to other portions of the crack.
Additionally, on either side of the weld, stains were evident beyond the initial through-wall crack region at a depth about haffway through the wall, Figure 10.
A in the weld HAZ, suggest that the crack growth rate was slow enough to assure that a full circumferential fatigue crack would not develop before leaking was detected. Three types of flat, reduced section test samples were prepared: base metal samples, samples which had the seam welds parallel to the specimen axis, and samples which had the seam welds perpendicular to the specimen axis.
Fabrication processes caused the strength of the bellows to be anisotropic.
The ultimate strength of the steel in the direction parallel to the seam weld was 621 MPa (90,000 psi) while the ultimate strength perpendicular to the seam weld was approximately 586 MPa (85,000 psi). The seam weld was stronger than the base material and when tests were conducted on samples which had the seam weld perpendicular to the specimen axis, most of the deformation was remote from the weld and fracture took place in the base metal.
The ultimate strength of these samples was slightly decreased by tl,e welding process because of the constraint '_ imposed on the base metal on either side of the weld. This base metal.. ......... ., provided the weak link in the deformation and fracture processes and the ultimate strength of these samples was about 614 MPa (89,000 psi). " When the seam weld was parallel to the specimen axis, the measured strength was increased because the specimen configuration forced deformation and fracture to occur in the weld metal.
The ultimate strength of these samples was 752 MPa (109,000 psi).
Optical and scanning electron microscopy of the fractured test samples failed to reveal any evidence of fracture initiation in the liquated regions of the weld heat-affected zones. Ali samples failed by microvoid coalescence processes and even the samples which had the seam weld parallel to the specimen axis displayed no evidence of failure related to intergranular microcracks, Figure  These tensile data, coupled with metallographic evidence that liqua_ion induced microcra_.ks are common to seam welds, demonstrate that the microcracks should not be a major problem if the bellows are installed properly.
These data are also supported by the successful, long term operation of numerous expansion joint bellows of similar design also known to contain microcracks in weld heat-affected zones.

CONCLUSIONS
The metallurgical evaluationsof the failedexpansion joint bellows demonstrated that failure was by fatigue crack growth from intergranularmicrocracks which resultedfrom liquationcracking in the weld HAZ.
However, the GTAW procedurewas not at faultbecause the microcracksdid not degrade the strengthof the bellows. The root cause of the premature failure of the bellows was improper installation which placed the bellows convolute in tension and significantly reduced the fatigue life of the system.
Even under these conditions, the fatigue crack developed slowly and was detected by cooling water leakage, thus preventing the development of conditions favoring the onset of rapid fracture in the bellows convolute.
Microstructural evaluation revealed that the GTAW parameters used would minimize the heat input and size of the liquation cracks for the process.
In addition to proper installation and minimizing the number of welds, the fatigue life of these expansion joints could be improved by using a high energy density welding process such as laser beam welding to further reduce heat input .
These steps would reduce the number and size of potential microcracks to act as fatigue crack initiation sites.