A detailed fracture mechanics evaluation is the most accurate and reliable prediction of process equipment susceptibility to brittle fracture. This article provides an overview and discussion on brittle fracture. The discussion covers the reasons to evaluate brittle fracture, provides a brief summary of historical failures that were found to be a result of brittle fracture, and describes key components that drive susceptibility to a brittle fracture failure, namely stress, material toughness, and cracklike defect. It also presents industry codes and standards that assess susceptibility to brittle fracture. Additionally, a series of case study examples are presented that demonstrate assessment procedures used to mitigate the risk of brittle fracture in process equipment.

This article focuses on characterizing the fracture-surface appearance at the microscale and contains some discussion on both crack nucleation and propagation mechanisms that cause the fracture appearance. It begins with a discussion on microscale models and mechanisms for deformation and fracture. Next, the mechanisms of void nucleation and void coalescence are briefly described. Macroscale and microscale appearances of ductile and brittle fracture are then discussed for various specimen geometries (smooth cylindrical and prismatic) and loading conditions (e.g., tension compression, bending, torsion). Finally, the factors influencing the appearance of a fracture surface and various imperfections or stress raisers are described, followed by a root-cause failure analysis case history to illustrate some of these fractography concepts.

Two examples involved brittle fracture promoted by small fatigue cracks owing to welding deficiencies. Other parts involving inadequate welding were a ski-wheel axle flange, ski fitting (brackets), and undercarriage shock strut stub assembly. In an attach fitting for an engine mount, weld cracks (severe stress concentrations) formed during repair welding. Cracks were severely oxidized. The main cause was incorrect repair and inadequate inspection of the fitting. In a cast CrNi alloy ski wheel axle, brittle fatigue failure emanated from welding cracks (notches). These welding cracks formed during the fabrication of the axle mounting plate. So-called all-purpose electrodes were used. Thus, the main cause for failure was a manufacturing deficiency-fatigue failure developed because of improper welding during fabrication of the axle. The proper electrode should have been used.

After only a short time in service, oil-fired orchard heaters made of galvanized low-carbon steel pipe, 0.5 mm (0.020 in.) in thickness, became sensitive to impact, particularly during handling and storage. Most failures occurred in an area of the heater shell that normally reached the highest temperature in service. A 400x etched micrograph showed a brittle and somewhat porous metallic layer about 0.025 mm (0.001 in.) thick on both surfaces of the sheet. Next to this was an apparently single-phase region nearly 0.05 mm (0.002 in.) in thickness. The examination supported the conclusion that prolonged heating of the galvanized steel heater shells caused the zinc-rich surface to become alloyed with iron and reduce the number of layers. Also, heating caused zinc to diffuse along grain boundaries toward the center of the sheet. Zinc in the grain boundaries reacted with iron to form the brittle intergranular phase, resulting in failure by brittle fracture at low impact loads during handling and storage. Recommendation included manufacture of the pipe with aluminized instead of galvanized steel sheet for the combustion chamber.

A railway tank car developed a fracture in the region of the sill and shell attachment during operation at -34 deg C (-30 deg F). On either side of the sill-support member, cracking initiated at the weld between a 6.4 mm thick frontal cover plate and a 1.6 mm thick side support plate. The crack then propagated in a brittle manner upward through the side plate, through the welds attaching the side plate to a 25 mm (1 in.) thick shell plate (ASTM A212, grade B steel), and continued for several millimeters in the shell plate before terminating. Other plates involved were not positively identified but were generally classified as semi-killed carbon steels. Investigation (visual inspection, hardness testing, chemical analysis, Charpy V-notch testing, and drop-weight testing) supported the conclusions that the fracture was initiated by weld imperfections and propagated in a brittle manner as a result of service stresses acting on the plate having low toughness at the low service temperatures encountered. Recommendations included that the specifications for the steel plates be modified to include a toughness requirement and that improved welding and inspection practices be performed to reduce the incidence of weld imperfections.

A large crack developed at a girder-truss joint area of the Fremont bridge in Portland, OR, on 28 Oct 1971. It occurred during a positioning procedure involving a junction piece welded to a girder, starting as a brittle fracture and terminating in plastic hinges in the girder web welds. The arch rib top plate, as it met the main girder, formed a composite beam of A588/A36 composition. Investigation showed the original design of the failed component called for an angle of high geometric stress concentration (90 deg with no radius) in a region of substantial transverse weld joints. While the material met chemical and mechanical property requirements, tests showed it had low fracture toughness and critical-sized flaws oriented normal to the principal stress in the failed junction piece. Fabrication procedures resulted in high residual stresses and a metallurgical notch at the radius in the junction piece. Stresses induced during jacking (the procedure used to raise bridge components into position) applied the stresses in the critical radius that triggered the cracking.

The specially designed sand-cast low-alloy steel jaws that were implemented to stretch the wire used in prestressed concrete beams fractured. The fractures were found to be macroscale brittle and exhibited very little evidence of deformation. The surface of the jaws was disclosed by metallographic examination to be case carburized. The case was found to be martensite with small spheroidal carbides while the core consisted of martensite plus some ferrite. The fracture was revealed to be related to shrinkage porosity. Tempering was revealed to be probably limited to about 150 deg C by the hardness values (close to the maximum hardness values attainable) for the core. It was interpreted that the low tempering temperature used may have contributed to the brittleness. The procedures used for casting the jaws were recommended to be revised to eliminate the internal shrinkage porosity. Tempering at a slightly higher temperature to reduce surface and core hardness was recommended.

A roll assembly consisting of a forged AISI type 440A stainless steel sleeve shrink fitted over a 4340 steel shaft and further secured with tapered keys on opposite ends was crated and shipped by air. Upon arrival, the sleeve was found to have cracked longitudinally between the keyways. A roll manufacturer had successfully used the above procedure for many years to make them. Analysis (visual inspection; 150x micrograph of sections etched with a mixture of 2 parts HNO3, 2 parts acetic acid, and 3 parts HCI; electron microscopy; and stress testing) supported the conclusion that superficial working of the metal, probably insufficient hot working, produced a microstructure in which the carbide particles were not broken up and evenly distributed. As a result, the grains were totally surrounded with brittle carbide particles. This facilitated the formation of a crack at a fillet in the keyway. Crack growth was rapid once the crack had initiated, causing brittle fracture to occur.

A section from a stop-block guide fell to the floor on a crane runway after it failed. A brittle crystalline-type break was disclosed by examination of the fracture surface. The point of initiation was in a hardened heat-affected layer that had developed during flame cutting and welding. The metal was identified to be 1020 steel. It was indicated by the coarse as-rolled structure (grain size of ASTM 00 to 4) of the base metal that the weldment (stop block and guide) had not been normalized. The brittle failure was evaluated to have been initiated at a metallurgical and mechanical notch produced by flame cutting and welding. As corrective measures, fully silicon-killed 1020 steel with a maximum grain size of ASTM 5 were used to make new stop-block weldments. The weldments were normalized at 900 deg C after flame cutting and welding to improve microstructure and impact strength. All flame-cut surfaces were ground to remove notches.

Chain link, a part of a mechanism for transferring hot or cold steel blooms into and out of a reheating furnace, broke after approximately four months of service. The link was cast from 2% Cr austenitic manganese steel and was subjected to repeated heating to temperatures of 455 to 595 deg C (850 to 1100 deg F). Examination included visual inspection, macrograph of a nital-etched specimen from an as-received chain link 1.85x, micrographs of a nital-etched specimen from an as-received chain link 100x/600x, normal microstructure of as-cast standard austenitic manganese steel 100x, micrograph of a nital-etched specimen that had been austenitized 20 min at 1095 deg C (2000 deg F) and air cooled 315x, and micrograph of the same specimen after annealing 68 h at 480 deg C (900 deg F) 1000x). Investigation supported the conclusions that the chain link failed in a brittle manner, because the austenitic manganese steel from which it was cast became embrittled after being reheated in the temperature range of 455 to 595 deg C (850 to 1100 deg F) for prolonged periods of time. The alloy was not suitable for this application, because of its metallurgical instability under service conditions.

The T-section cross member of the lifting sling failed in service while lifting a 966 kg (2130 lb) load. The L-section sling body and the cross member were made of aluminum alloy 5083 or 5086 and were joined by welding using aluminum alloy 4043 filler metal. The fracture was found by visual examination to have occurred at the weld joining the sling body and the cross member. Inadequate joint penetration and porosity was revealed by macrographic examination of the weld. Lower silicon content and a higher magnesium and manganese content than the normal for alloy 4043 filler metal were found during chemical analysis. It was revealed by examination of the ends of the failed cross member that a rotational force that had been applied on the cross member caused it to fracture near the sling body. It was concluded that brittle fracture at the weld was caused by overloading which was attributed to the misalignment of the sling during loading. Aluminum alloy 5183 or 5356 filler metal was recommended to be used to avoid brittle welds.

During the lifting of a piece of machinery by means of an overhead travelling crane the hook fractured suddenly. The load was attached to the hook by means of fiber rope slings and rupture occurred in a plane which appeared to coincide with the sling loop nearest to the back of the hook. The rated capacity of the crane was 15 tons. At the time of the mishap it was being used to lift one end of a hydraulic cylinder with a total weight of about 27 tons. Fracture was of the cleavage type throughout. There was no evidence of any prior deformation of the material in the vicinity, nor was there any indication of a pre-existing crack or major discontinuity at the point of origin. A sulfur print suggested the hook had been forged from a billet cogged down from an ingot of semi-killed steel. Failure of this hook was attributed to strain-age embrittlement of the material at the surface of the intrados.

Several thousands of new 16 mm diam alloy steel sling chains used for handling billets failed by chain-link fractures. No failures were found to have occurred before delivery of the new chains. It was observed that the links had broken at the weld. It was found that all failures had occurred in links having hardness values in the range of 375 to 444 HRB. It was revealed by the supplier that the previous hardness level of 302 to 375 HRB was increased to minimize wear which made the links were made notch sensitive and resulted in fractures that initiated at the butt-weld flash on the inside surfaces of the links. A further reduction in ductility was believed to have been caused by lower temperatures during winter months. Thus, the failure was concluded to have been caused in a brittle manner caused by the notch sensitivity of the high hardness material at lower temperatures. The chains were retempered to a hardness of 302 to 375 HRB as a corrective measure and subsequently ordered chains had this hardness as a requirement.

A hook, which was marked for a safe working load of 2 tons, failed while lifting a load of approximately 35 cwts. Fracture took place at the junction of the shank with the hook portion, at which no fillet radius existed. Except for an annular region round the periphery, which was of a smooth texture, the fracture was brightly crystalline indicative of a brittle failure. Microscopic examination showed the material was a low-carbon steel in the normalized condition; no abnormal features were observed. The basic cause of failure was the presence of a fatigue crack at the change of section where the shank joined the hook portion. To minimize the possibility of fatigue cracking, it was recommended that a generous radius be provided at the change of section.

A 58.4 cm (23 in.) diam heavy-duty brake drum component of a cable-wound winch broke into two pieces during a shutdown period. Average service life of these drums was two weeks; none had failed by wear. The drums were sand cast from ductile iron. During haul-out, the cable on the cable drum drove the brake drum, and resistance was provided by brake bands applied to the outside surface of the brake drum. Friction during heavy service was sufficient to heat the brake drum, clutch mount, and disk to a red color. Examination of the assembly indicated that the brake drum would cool faster than its mounts and would contract onto them. Brittle fracture of the brake drum occurred as a result of thermal contraction of the drum web against the clutch mount and the disk. The ID of the drum web was enlarged sufficiently to allow for clearance between the web and the clutch mount and disk at a temperature differential of up to 555 deg C (1000 deg F). With the adoption of this procedure, brake drums failed by wear only.

Splined rotor shafts (constructed from 1151 steel) used on small electric motors were found to miss one spline each from several shafts before the motors were put into service. Apparent peeling of splines on the induction-hardened end of each rotor shaft was revealed by visual and stereo-microscopic examination. One tooth on each shaft was found to be broken off. It was revealed by metallographic examination of an unetched section through the fractured tooth that the fracture surface was concave and had an appearance characteristic of a seam. Partial decarburization of the surface was revealed after etching with 1% nital. The presence of a crack, with typical oxides found in seams at its root, was disclosed by an unetched section through the shaft in an area unaffected by induction heating. The etched samples revealed similar decarburization as was noted on the fracture surface of the tooth. It was concluded that the seam had been present before the shaft was heat treated and these seams acted as stress raisers during induction hardening to cause the shaft failure. It was recommended that the specifications should specify that the shaft material should be free of seams and other surface imperfections.

After two weeks of operation, a poppet used in a check valve to control fluid flow and with a maximum operating pressure of 24 MPa (3.5 ksi) failed during operation. Specifications required that the part be made of 1213 or 1215 rephosphorized and resulfurized steel. The poppet was specified to be case hardened to 55 to 60 HRC, with a case depth of 0.6 to 0.9 mm (0.025 to 0.035 in.); the hardness of the mating valve seat was 40 HRC. Analysis showed that the fracture occurred through two 8 mm (0.313 in.) diam holes at the narrowest section of the poppet. The valve continued to operate after it broke, which resulted in extensive loss of metal between the holes. 80x micrograph and 4x macrograph of a 5% nital etched longitudinal section, and chemical analyses showed the poppet did fit 1213 or 1215 specs. However, hardness measurements showed surface hardness was excessive-61 to 65 HRC instead of the specified 55 to 60 HRC. Thus, the poppet failed by brittle fracture, and cracking occurred across nonmetallic inclusions. Recommendation was to redesign the valve with the poppet material changed to 4140 steel, hardened, and tempered to 50 to 55 HRC.

A sand-cast gray iron flanged nut was used to adjust the upper roll on a 3.05 m (10 ft) pyramid-type plate-bending machine. The flange broke away from the body of the nut during service. Analysis (visual inspection and 150x micrographs of sections etched with nital) supported the conclusions that brittle fracture of the flange from the body was the result of overload caused by misalignment between the flange and the roll holder. The microstructure contained graphite flakes of excessive size and inclusions in critical areas; however, these metallurgical imperfections did not appear to have had significant effects on the fracture. Recommendations included carefully and properly aligning the flange surface with the roll holder to achieve uniform distribution of the load. Also, a more ductile metal, such as steel or ductile iron, would be more suitable for this application and would require less exact alignment.

After less than 30 days in service, several cadmium-plated retaining rings, made of 4140 steel tubing and heat treated to 36 to 40 HRC, broke during operation that included holding components of a segmented fitting in place under a constant load. Photographic and 100x nital-etched micrographic examination showed a microstructure of tempered martensite with low inclusion content as well as a pit or burned spot on the outer area of the ring. The defect was approximately 0.18 mm (0.007 in.) deep and 0.5 mm (0.020 in.) in diam and had a hardness of 58 to 60 HRC. The base metal adjacent to the defect had a hardness of 36 to 40 HRC. Small cracks or fissures were also evident within the defect. Thus, the rings failed in brittle fracture as the result of an arc strike (or burn) on the surface of the ring. At the site of the arc strike, a small region of hard, brittle untempered martensite was formed as the result of an arc strike during the cadmium-plating operation. Fracture occurred readily when the ring was stressed. No recommendations were made.

During installation, a clamp-strap assembly, specified to be type 410 stainless steel-austenitized at 955 to 1010 deg C (1750 to 1850 deg F), oil quenched, and tempered at 565 deg C (1050 deg F) for 2 h to achieve a hardness of 30 to 35 HRC, and used for securing the caging mechanism on a star-tracking telescope, fractured transversely across two rivet holes closest to one edge of the pin retainer in a completely brittle manner. Comparison with a non-failed strap using microscopic examination, spectrographic analysis, and slow-bend tests showed that both fit the 410 stainless steel specs, but hardness and grain size were different. Reheat treatment of full-width specimens showed that coarse grain size (ASTM 2 to 3) was responsible for the brittle fracture, and excessively high temperature during austenitizing caused the large grain size in the failed strap. The fact that the hardness of the strap that failed was lower than the specified hardness of 30 to 35 HRC had no effect on the failure because that of the non-failed strap was even lower. Recommendation was that the strap should be heat treated as specified to maintain the required ductility and grain size.