Similar pits were also observed in the general area of fracture initiation, but not on the fracture plane (Fig. 4
). A metallographic section parallel to the length of the rotor blade spar was prepared through the pit shown in
and extending through the fracture surface (Fig. 5
). The appearance of the pit, especially in cross section, was not that of a corrosion pit. In the cross section, the material forming the pit contained voids and had a different appearance than the base alloy; this material appeared to have been melted and resolidified.
Fig. 4 Pit located about 0.25 mm (0.0l in.) from the fracture plane near the origin on the bottom surface of the main rotor blade.
Fig. 5 Longitudinal section through the pit shown in
and the fracture surface (horizontal at top), showing that the material forming the pit contained voids and had a microstructure different from the base material.
Two hypotheses were possible for the intense, local heating necessary to melt the material forming the pits without affecting the surrounding material—a lightning strike or electrical arcs. The latter was deemed possible: a search of the maintenance records had revealed that the fiberglass pocket at this location had recently been replaced. Replacement of a pocket requires curing of the epoxy bond with a contact electrical heater.
Several areas in the resolidified material in the cross section exhibited unusual microstructures. One of these areas was analyzed using an energy-dispersive X-ray (EDX) detector on a scanning electron microscope (SEM). The results of this analysis are shown in
; the major element in the spectrum was iron. The area analyzed is indicated by the arrow in
, which is a slightly higher-magnification view of the same cross section. An X-ray map was obtained of this same area to show the distribution of iron (Fig. 8
). This map indicated that all of the unusual-appearing microstructural areas in the resolidified material were areas of high iron concentration.
Fig. 6 EDX spectrum obtained from an anomalous-appearing area in the material forming the pit (arrow in
Fig. 7 Higher-magnification view of the cross section shown in
Fig. 5. Arrow indicates the location at which the spectrum shown in
Fig. 8 X-ray map showing the distribution of iron in the area shown in
Fig. 7. Note that the material forming the surface of the pit contained a considerable amount of iron, far in excess of that found in the base material.
The presence of these iron contamination concentrations was sufficient evidence to conclude that the source of the pits was an electrical arc from an iron object, not from lightning. Investigation of the heating elements used to cure the epoxy when replacing the pockets showed that their exterior was iron and that it was quite possible for the outer surface to become “live.” These pits acted as stress concentrations that were the origin sites for the fracture.
SEM examination of the fracture surface confirmed that the flat portion of the blade fracture was generated by fatigue crack growth. The load cycles in the main rotor blade of any helicopter can be generated from a variety of sources: the differences in relative air speed encountered during a revolution, the perturbations in air flow when the blade passes over the fuselage, the natural frequencies of the blade, and so forth. Because the main rotor blades are rotating wings giving lift to the aircraft, they are subjected to bending stresses during use, with the lower side normally in tension.
It was decided that the flight time between the loss of nitrogen pressure to the final separation of the blade could be estimated by counting the number of fatigue striations on the fracture surface between the point at which the fatigue crack penetrated the wall of the spar (allowing the nitrogen to escape) and the transition from fatigue crack growth to the final overload failure.
The striation counts were performed at two laboratories by examination of the fracture using both SEM and transmission electron microscope (TEM) replica techniques. The general appearances of the fatigue striations are shown in
. The results of the striation counts are given in
. Although a statistician might dwell on the variation in striation counts obtained, these results were considered to be in good agreement, especially considering that they were achieved by a sampling technique (rather than by actually counting every striation) on three different electron microscopes by three different investigators.
Fig. 9 SEM micrograph of the fracture surface of the crack at a location close to the point at which the crack penetrated the wall of the hollow extrusion, showing uniformly spaced fatigue striations. 3700×.
Fig. 10 SEM micrograph of the fracture surface of the crack at a location close to the point at which the transition from fatigue to overload occurred, showing nonuniform fatigue striation spacing (a largest striation followed by two or three small ones). 3700×.
Table 1 Striation counts and equivalent flight time
SEM Lab 1
SEM Lab 2
As shown in
, although the three different striation counts were in good agreement, an enormous disparity in the flight time resulted from differing theories regarding the number of load cycles experienced by the blade per revolution—about 1 h if there are four stress cycles per revolution and about 4 h if there is only one stress cycle per revolution. The results of strain gage tests of the blade in operation were obtained;
shows the results from the strain gage location that was near the fatigue crack area. These results indicated that there were four stress cycles per revolution of the blade; however, they varied greatly in magnitude. Some of these stress cycles were sufficiently small that they may not have caused fatigue crack growth, particularly in the initial stages of crack growth.
Fig. 11 Output of strain gages attached to an operating main rotor blade at the locations shown in
Fig. 2, indicating that in each revolution of the blade four tensile stress cycles of different magnitude were encountered.
To determine how many stress cycles per revolution would cause crack growth as a function of the crack length and to determine how long it would take for failure to occur, a computer simulation of the fatigue crack growth process was used. Fortunately, fatigue crack growth data were available (Ref 1
) for 6061-T651 alloy (Fig. 12
). The cyclic stress data and the handbook crack growth data were input into a computer program that was similar to the USAF “CRACKS” program, based on Foreman's equation (Ref 2
) for crack growth (below), using numerical integration and including the Willembourg/USAF crack retardation model.
Fig. 12 Cyclic crack growth rate as a function of the change in stress intensity for 6061-T651 aluminum alloy. Source:
Foreman's equation for crack growth is expressed as:
is the crack growth per cycle of stress, ΔK
is the change in stress intensity associated with that cycle of stress,
is the stress ratio,
is the plane-strain fracture toughness (28.6 MPa
, or 26 ksi
), and C and m are constants that can be determined from the
data for the material.
The results of the computer simulation showed that, given the properties of the 6061-T651 aluminum (Fig. 12
) and the stress cycles as shown in
, it would require slightly less than 1 h of flight operation for the fatigue crack to grow from the point at which it broke through the wall to the final overload failure. Additionally, the computer simulation showed that at short crack lengths only the highest stress cycle was active in causing crack growth, whereas as the crack lengthened, the other stress cycles started to play a role in crack growth. Ultimately, near the end of the fatigue crack growth, all four stress cycles were causing crack growth.
These computer results were confirmed by careful examination of the fatigue crack surface. Striation structure and spacings were found to vary as the crack grew. As shown in
, the striation spacing near the point of breakthrough (relatively short crack length) was very uniform, indicating that only one magnitude of stress was causing crack growth. As shown in
, the striation spacing near the transition to the final overload failure was very different; the striations were grouped—one large striation followed by two or three smaller ones, indicating that several stress cycles of different magnitude were now causing crack growth. These observations reflected the computer results with surprising accuracy.