Upon arrival at the erection site, an AISI type 316L stainless steel tank intended for storage of fast breeder test reactor coolant (liquid sodium) exhibited cracks on its shell at two of four shell/nozzle fillet-welded joint regions. The tank had been transported from the manufacturer to the erection site by road, a distance of about 800 km (500 mi). During transport, the nozzles were kept at an angle of 45 deg to the vertical because of low clearance heights in road tunnels. The two damaged joints were unsupported at their ends inside the vessel, unlike the two uncracked nozzles. Surface examination showed ratchet marks at the edges of the fracture surface, indicating that loading was of the rotating bending type. Electron fractography using the two-stage replica method revealed striation marks characteristic of fatigue fracture. The striations indicated that the cracks had advanced on many “mini-fronts,” also indicative of nonuniform loading such as rotating bending. It was recommended that a support be added at the inside end of the nozzles to rigidly connect with the shell. In addition to avoiding transport problems, this design modification would reduce fatigue loading that occurs in service due to vibration of the nozzles during filling and draining of the tank.

Four truck cross members intended for use in heavy-duty transport trucks were investigated. Two of the members had cracked on a prototype vehicle and two had been fatigue tested in the laboratory. The cross members were fabricated from SAE 950X plate and consisted of a formed channel section and an internal fillet-welded diaphragm. Sections from each of the cross members were subjected to a complete analysis, including chemical analysis, magnetic particle testing, mechanical testing, scanning electron microscope/fractography, and metallography. The primary mode of failure was found to be fatigue cracking that initiated at the toes of the fillet welds. Secondary fatigue cracking occurred at the torque rod mounting holes. Failure was attributed to cyclic stresses at the weld toes that exceeded the lowered fatigue strength at this location. A design change that eliminated the fillet welds alleviated the problem.

A system of carbon steel headers, handling superheated water of 188 deg C (370 deg F) at 2 MPa (300 psi) for automobile-tire curing presses, developed a number of leaks within about four months after two to three years of leak-free service. All the leaks were in shielded metal arc butt welds joining 200 mm (8 in.) diam 90 deg elbows and pipe to 200 mm (8 in.) diam welding-neck flanges. A flange-elbow-flange assembly and a flange-pipe assembly that had leaked were removed for examination. Investigation (visual inspection, hardness testing, chemical analysis, magnetic-particle testing, radiographic inspection, and 2% nital etched 1.7x views) showed varying IDs on the assemblies and supported the conclusions that the failures of the butt welds were the result of fatigue cracks caused by cyclic thermal stresses that initiated at stress-concentrating notches at the toes of the interior fillet welds on the surfaces of the flanges. Recommendations included using ultrasonic testing to identify the appropriate joints and then replacing them. Special attention to accuracy of fit-up in the replacement joints was also recommended to achieve smooth, notch-free contours on the interior surfaces.

An amusement ride failed when a component in the ride parted, permitting it to fly apart. The ride consisted of a central shaft supporting a spider of three arms, each of which was equipped with an AISI 1040 steel secondary shaft about which a circular platform rotated. The main shaft rotated at about 12 rpm and the platforms at a speed of 20 rpm. The accident occurred when one of the secondary shafts on the amusement ride broke. The point of fracture was adjacent to a weld that attached the shaft to a 16 mm thick plate, which in turn bore the platform support arms. Investigation (visual inspection, 0.4x magnification, and stress analysis) supported the conclusion that a likely cause for the fatigue failure was the combination of residual stresses generated in welding and centrifugal service stresses from operation that were accentuated by areas of stress concentration at the undercut locations. Without the excessive residual stress, the shaft dimensions appeared ample for the service load. Recommendations included applying the fillet weld with more care to avoid undercutting. The residual stresses could be minimized by pre-weld and post-weld heat application.

A 208 cm (82 in.) ID steel aqueduct (ASTM A572, grade 42, type 2 steel) fractured circumferentially at two points 152 m (500 ft) apart in a section above ground. A year later, another fracture occurred in a buried section 6.4 km (4 mi) away. Both pipes fractured during Jan at similar temperatures and pressures. The pipe had a 24 mm wall thickness, and the hydrostatic head was 331 m (1085 ft). The air temperature was approximately -13 deg C (9 deg F), the water temperature approximately 0.6 deg C (33 deg F), and the steel temperature approximately -4 deg C (25 deg F). The pipe had been shop-fabricated in 12 m (40 ft) lengths, then shop welded into 24 m (80 ft) lengths. Field assembly was with bell-and-spigot joints. Investigation (visual inspection and Charpy V-notch testing) supported the conclusion that brittle fracture of the aqueduct pipe was attributed to a combination of stress concentrations at the toes of the fillet welds due to poor welding technique, including shop welds made without preheat, and a brittle condition of the steel at winter temperatures. Recommendations included revised welding techniques, installation of expansion joints, and the use of steel plate rolled from fully killed ingots.

The impeller of a 4 ft. diam extraction fan driven by a 120 hp motor at 1,480 rpm. disrupted suddenly. The majority of the vanes had become detached where they were welded to the plates. At other locations, separation of the vanes was accompanied by tearing of the adjacent plate, failure being initiated at the weld fillets of the inner end of the vanes. An unusual feature was that the blades disclosed regions having a pronounced striated and stepped appearance. The etched microstructure was typical of a low carbon rolled plate having the usual banded appearance. A cross section through the fillet welds and zone showed lamellar tearing, which confirmed that failure had occurred in weld metal adjacent to the fusion face of the fillet to the vane. Results of the investigation indicated that the primary cause of failure of the impeller was the development of fatigue cracks from the unwelded roots of the fillet welds, by which the vanes were attached to the supporting plates. The impeller would have shown increased resistance to fatigue crack initiation if the T joint between the vanes and plates had been of the full penetration type.

Weld-decay and stress-corrosion cracking developed in several similar all-welded vessels fabricated from austenitic stainless steel. During a periodic examination cracks were revealed at the external surface of one of the vessels. External patch welds had been applied at these and several other corresponding locations. Cracks visible on the external surface developed from the inside in a region close to the toe of the internal fillet weld to the deflector plate, and another deep crack associated with a weld cavity is visible slightly to the right of the main fissure. Microscopic examination revealed that precipitation of carbides at the grain boundaries had taken place in the vicinity of the cracks, but that the paths of the cracks were not wholly intergranular. Conditions present in the vicinity of the internal fillet weld must have been such as to favor both inter- and transgranular cracking. It is probable that the heating associated with the repair welds made from time to time also contributed to the trouble. The transgranular cracks, however, were indicative of stress-corrosion cracking.

On 23 Dec 1997, a portion of the main ore conveyor at a large mine collapsed onto a highway and shut down mine operations. The conveyor structure that collapsed was supported by a steel truss spanning 185 ft. Truss failure occurred just as the conveyor transport rate was increased to 8,260 tph. Under this total loading, which was only slightly above the regular operating condition, a poorly designed and fabricated transition joint in the west lower chord failed, thereby overloading other key structural members and causing the entire truss to collapse. Another contributing cause of the collapse was the transition joint welds, where the fracture originated. They were made with undersized fillet welds, 20% smaller than specified on the original fabrication drawing. Because of the poorly designed joint detail and the deficient welds, both of which concentrated stress and strain in the low ductility direction of the transition joint plate, lamellar tearing of plate material occurred at the boxed I-beam fillet weld attachment. Brittle fracture of this joint precipitated global collapse of the truss structure.

After ten years of satisfactory operation, economizer-tube failures occurred in a large black liquor recovery boiler for a paper mill. The economizer contained 1320 finned tubes. Two fins ran longitudinally for most of the tube length and were attached by fillet welding on one side. The economizer tube leaks occurred at the end of the fin near the bottom of the economizer. A sample from a tube that had not failed showed heavy pitting attack on the inside of the tube, probably due to excess oxygen in the feedwater. Penetrant testing revealed numerous longitudinal cracks on the inside in the area of the fin tip. Cracking at the end of the fin-to-tube fillet weld was noted. The results indicate the failures were due to corrosion fatigue whose stresses were primarily thermally induced. A temporary solution included inspecting all tubes with shear-wave ultrasonics. Tubes with the most severe cracking were ground and repair welded. The square corners of the fins were trimmed back with a gradual taper so that expansion strains would be more gradually transferred to the tube surface. Water chemistry was closely evaluated and monitored, especially with regard to oxygen content.

Three examples of corrosion-fatigue cracking from the toes of substantial fillet welds applied to seal-leaking riveted seams in steam accumulators are described. In the first case, this practice resulted in a disastrous explosion; in the second, which involved two identical vessels at the same location, cracking in course of development was discovered during internal inspection. Microscope examination of several specimens cut to intersect a crack showed it to be typical of corrosion-fatigue; it was in the form of a broad fissure, contained oxide deposits, and the termination was blunt-ended. The two cases not only serve to illustrate the danger of applying fillet welds to seal the lap edges of riveted seams, but point to the inadvisability of employing riveted construction for vessels intended for service under conditions involving frequent pressure and thermal fluctuations, as it is extremely difficult to maintain the tightness of riveted seams under these conditions. Such vessels are now almost exclusively of all-welded construction

Handles welded to the top cover plate of a chemical-plant downcomer broke at the welds when the handles were used to lift the cover. The handles were fabricated of low-carbon steel rod; the cover was of type 502 stainless steel plate. The attachment welds were made with type 347 stainless steel filler metal to form a fillet between the handle and the cover. The structure was found to contain a zone of brittle martensite in the portion of the weld adjacent to the low-carbon steel handle; fracture had occurred in this zone. The brittle martensite layer in the weld was the result of using too large a welding rod and too much heat input, melting of the low-carbon steel handle, which diluted the austenitic stainless steel filler metal and formed martensitic steel in the weld zone. Because it was impractical to preheat and postheat the type 502 stainless steel cover plate, the low-carbon steel handle was welded to low-carbon steel plate, using low-carbon steel electrodes. This plate was then welded to the type 502 stainless steel plate with type 310 stainless steel electrodes. This design produced a large weld section over which the load was distributed.

Radial cracking occurred adjacent to 11 vanes in a 19-vane impeller operating in a chemical plant environment. The impeller vanes were fillet welded to both the disk and the cover Cracks were next to the fillet welds and near the cover outer diameter They generally did not extend to the outer diameter. The entire impeller surface was tested by the dry magnetic particle method. Visual and microstructural examinations revealed intergranular cracking. Energy-dispersive spectroscopy of corrosion products contained in the cracks disclosed the presence of chlorine and sulfur The failure was attributed to stress-corrosion cracking caused by a corrosive atmosphere.

An anchorage plate which fractured was one of a pair used as intermediate members through which the boom suspension ropes were attached to the jury-mast of an excavator. Failure of the plate released the ropes on one side of the boom, resulting in extensive damage to the latter and also bending of the other anchorage plate. The anchorage plates were 23 x 9 in. and had been flame-cut from mild steel plate. Collars were fillet-welded on each side at both ends to provide extra bearing area for the pins. Holes had then been flame-cut slightly under size and bored to final dimensions. The plates were given a slight set after flame-cutting to provide a more direct line of pull for the ropes. The fracture surface was bounded by narrow lips, indicative of shear failure. Failure of the anchorage plate was attributed to cracks present at the junctions of the fillet welds, and deficient notch-ductility of the material from which the plates were made.

A suspension bushing separated from the disk on the rear wheel suspension of a racing vehicle while under operation on an express highway, causing the wheel to detach from the car. Visual inspection showed fresh turning grooves at four built-up fillet welds on the torus of the outside of the disk. The welding heat and rapid cooling caused the base material of the disk – already with a martensitic structure – to harden throughout the torus, both of which caused cracking in the inner and outer fillets at the transition from bushing to disk. Visual inspection of the other rear wheel showed similar stress cracks in the hardened base material of the transition region as well as the same four welds. Rough finishing and the sharp-edged formation of the cross section transition may have also contributed to the failure. The results of the investigation suggested that machine shops neither execute nor permit repair-welds on highly stressed machine parts and especially vehicle components.

A welded ferritic stainless steel heat exchanger cracked prior to service. The welding filler metal was identified as an austenitic stainless steel and the joining method as gas tungsten arc welding. Investigation (visual inspection, SEM images, 5.9x images, and 8.9x/119x images etched with Vilella's reagent followed by electrolytic etching in 10% oxalic acid) supported the conclusion that the heat exchanger cracked due to weld cold cracking or postwelding brittle overload that occurred via flexure during fabrication. The brittle nature of the weld was likely due to a combination of high residual stresses, a mixed microstructure, inclusions, and gross grain coarsening. These synergistic factors resulted from extreme heat input during fillet welding. Recommendations included altering the welding variables such as current, voltage, and travel speed to substantially reduce the heat input.

Three bearing bosses from the cover of scrap shears were sent in for examination. They had torn off the base plate to which they had been welded by fillet welds all around. Two of these were examined. They showed entirely the same symptoms. The bosses had broken away on three sides along the welds. The cleaved fractures in the burned notches propagated partially above and partially below several incipient cracks which may have been fatigue fractures. Metallographic sections showed that the fractures had occurred either at the burned notches near the transition from the weld to the sheet, or else they ran in the sheet material next to the weld. The quality of the welds could not be judged because the opposite fracture pieces to which they adhered had not been sent in. It was concluded that the breakaway of these bosses was at least favored by overheating and hardening.

Leakage was identified around a coupling welded into a stainless steel holding tank that stored condensate water with low impurity content. The tank and fitting were manufactured from type 304 stainless steel. The coupling joint consisted of an internal groove weld and an external fillet weld. Cracking was found to be apparent on the tank surface, adjacent to the coupling weld. Chlorine, carbon, and oxygen in addition to the base metal elements were revealed by energy-dispersive x-ray spectrometric analysis. A great number of secondary, branching cracks were evident in the weld, heat-affected zone, and base metal. The branching and transgranular cracking was found to emanate primarily from the exterior of the tank. It was concluded that the tank failed as a result of stress-corrosion cracking that initiated at the exterior surface as aqueous chlorides, especially within an acidic environment, have been shown to cause SCC in austenitic stainless steels under tensile stress.

A spindle made of hardenable 13% chromium steel X40 Cr13 (Material No. 1.4034) that was fastened to a superheated steam push rod made of high temperature structural steel 13Cr-Mo44 (Material No. 1.7335) by means of a convex fillet weld, fractured at the first operation of the rod directly next to the weld bead. Investigation showed that the fracture of the superheated steam push rod spindle was caused by hardening and hardening crack formation in the weld seams and adjoining areas. It would have been preferable to avoid welding near the cross sectional transitions altogether in consideration of the crack sensitivity of high hardenability steels. If for some reason this was not possible, then all precautions should have been taken that are applicable to the particular steel, such as preheating, slow cooling and stress relief tempering after welding. The selection of an austenitic additive material should have been considered because it could have equalized stresses due to its high elongation. Most probably, however, a material of lower hardenability should have been selected for the spindle if high operating properties were of paramount importance.

The Pandia digester is a long cylindrical vessel which uses alkaline sulfite liquor to cook sawdust for pulping. The inlet cone was fabricated from AISI 304L stainless steel with E308 welds. Typical liquor concentration was approximately 80% NaOH, 20% Na2SO3 with chloride concentrations at 2 grams per liter. The operating pressures in the inlet cone were up to 1.2 MN/sq m (170 psig). The inlet cone had developed leaks within a year of service. Liquid penetrant inspection showed significant through-wall cracking near the fillet welds joining the bottom flange and side wall and the butt welds. Metallographic specimens were prepared from the welds to examine the microstructure and nature of the cracks. The cooking liquor at the inlet cone contained over ppm chlorides and was aggressive to 304 stainless steel. The cracking was identified as chloride-induced SCC. The inlet cone was replaced with an Inconel clad carbon steel inlet cone to combat the SCC.

When a large LPG low-carbon steel storage tank was put into service for the first time and filled beyond the proof testing level, a brittle fracture crack initiated at a fillet weld between a stiffener ring and the wall. The crack propagated to a length of 5.5 m and arrested. Analysis showed that the plates satisfied the criteria of BS 4741. It was concluded that the cause of crack initiation was the lack of a mouse hole at the junction between the stiffening ring and the wall of the tank. The tank was repaired and put back in service. When it was filled beyond the proof test level, again a brittle crack was initiated at a horizontal weld defect and propagated vertically, destroying the tank and the liquefaction plant. The initiation site was a thumbnail elliptical crack in a horizontal weld, having a depth of 1.5 mm, and a length of 4.5 mm. This showed that as late the mid-1970s, misunderstanding of brittle fracture led to the wrong design and construction of an LPG storage tank. The best design specification is to use a correlation between LAST, the Lowest Anticipated Service Temperature, and the DBTT measured by either Charpy tests or DTT.