Failure Analysis Of Machine Shafts

As the industrial arena grows more sophisticated, it seems as though operations are confronting fewer and fewer broken machine shafts. When shafts DO break, however, there are almost always as many theories regarding the suspected culprits as there are people involved.

Fig. 1 The appearance of an overload failure depends on whether the shaft material is brittle or ductile. Whether related to motors, pumps or any other types of industrial machinery, shaft failure analysis is frequently misunderstood, often being perceived as difficult and expensive. For most machine shafts, however, analysis should be relatively straightforward. That’s because the failure typically provides strong clues to the type and magnitude of forces on the shaft and the direction they acted in: The failed parts will tell exactly what happened.

There are only four basic failure mechanisms: corrosion, wear, overload and fatigue. The first two—corrosion and wear—almost never cause machine-shaft failures and, on the rare occasions they do, leave clear evidence. Of the other two mechanisms, fatigue is more common than overload failure. (NOTE: Keep in mind that many times corrosion will act in conjunction with fatigue loading to cause a shaft failure.) This article will focus on failures resulting from overload and fatigue factors.

Overload failures
Overload failures are caused by forces that exceed the yield strength or the tensile strength of a material. As depicted in Fig. 1, the appearance of an overload failure depends on whether the shaft material is brittle or ductile. 

No shaft materials are absolutely brittle or absolutely ductile. The shafts used on almost all motors, reducers and fans are low- or medium-carbon steels and relatively ductile. As a result, when an extreme overload is placed on these materials, they twist and distort. The bent shaft shown in Photo 1 has been grossly overloaded by a torsional stress.

Important Note: When Was the Failure Force Applied? 

In diagnosing which mechanism caused the failure, a critical point to remember is that overload failures are generally caused by a single load application, while fatigue failures are always the result of a load applied repeatedly over many cycles. This means if the shaft failed as a result of an overload, the force that caused the failure was applied the instant before the shaft broke. Conversely, if fatigue was the culprit, the initial force may have been applied millions of cycles before the final failure occurred.

There are occasional cases when a ductile shaft will fail in a somewhat brittle manner. Photo 2 shows an example of this situation—i.e., what happened when a 200 hp, 3600 RPM motor suddenly stopped running. The result was a huge torsional stress and a cracked shaft. But because the material is ductile, the angle of the crack it is not at the 45° position shown in Fig. 1, and there is obvious distortion of the keyway. When ductile materials are grossly overloaded very rapidly, they tend to act in a brittle manner.

Fortunately, brittle fractures of machine shafts are extremely rare. Like all brittle fractures, they are characterized by a relatively uniform surface roughness—the crack travels at a constant rate, and surface features called “chevron marks” are evident. Photo 3 shows the brittle fracture of the input shaft of a large reducer that was dropped. The “chevron marks” are the fine ripples on the surface that all point just to the left of the keyway. 

Occasionally, a portion of a machine shaft will be case-hardened to reduce the wear rate. (NOTE: Case-hardening is usually done solely for wear-resistance purposes.) Photo 4 shows the case-hardened splined section of a hydraulic pump shaft, including its hardened case, the ring around the circumference with a very different texture than the majority of the shaft and “chevron marks” that point to the origin of the damage. Based on how this fracture grew straight across the shaft, the cause could have been related to either bending or tension. Its relatively uniform surface, though, would indicate that this fracture is of a brittle nature—which also means it was caused by a single force application. Furthermore, since it’s impossible to put significant tension on a spline, the analyst could safely say that a single bending force caused the failure. 

Fatigue failures
Fatigue is caused by cyclical stresses, and the forces that cause fatigue failures are substantially less than those that would cause plastic deformation. Confusing the situation even further is the fact that corrosion will reduce the fatigue strength of a material. The amount of reduction is dependent on both the severity of the corrosion and the number of stress cycles. 

Once they are visible to the naked eye, cracks always grow perpendicular to the plane of maximum stress. Figure 2 shows the fracture planes caused by four common fatigue forces. Because the section properties will change as the crack grows, it’s crucial for the analyst to look carefully at the point where the failure starts to determine the direction of the forces. For example, while it is common for torsional fatigue forces to initiate a failure, the majority of the crack propagation could be in tension. That’s because the shaft has been weakened and the torsional resonant frequency has changed.

The condition or roughness of the fracture surface is one of the most important points to look at in analyzing a failure because of the difference between overload failures and fatigue failures. With overload failures—because the crack travels at a constant rate—the surface is uniformly rough. Fatigue-induced cracks, however, travel across the fracture face at ever-increasing speeds. As a result, the typical fatigue fracture face is relatively smooth near the origin(s) and ends in a comparatively rough final fracture.

A typical plain bending fatigue failure is depicted in Fig. 3. The crack started at the origin and slowly grew across the Fatigue Zone (FZ). When it reached the boundary of the Instantaneous Zone (IZ) the crack growth rate increased tremendously and the crack traveled across the IZ at approximately 8000 ft/sec. During the period of growth across the FZ, there may be changes in the loading on the shaft, which result in changes in the surface that appear as progression marks.

Rotational loads or plane bending… 
For a fatigue failure to occur, the forces must have been applied many times. There are low-cycle failures but most industrial fatigue failures we’ve seen involve more than 1,000,000 load cycles. A valuable feature of fatigue-failure interpretation is that the crack growth, i.e., the surface appearance, tells how the load was applied. If the crack grows straight across the shaft (as shown in Fig. 3), the force that caused the failure must have been a bending load operating in a single plane. 

Figures 4 and 5, however, show examples of rotating bending. The difference between these two failures is that the shaft in Fig. 4 has a single origin, while the fracture in Fig. 5 has multiple origins. Looking at the two sketches, we see the IZ of Fig. 4 is the larger of the two—which indicates that the load on the shaft when it failed was greater than that on Fig. 5. The analysis also shows that, even though Fig. 5 was less heavily loaded, it had many more fracture origins, an indication of a high stress concentration, such as a shaft step with a very small radius. The ratchet marks are the planes between adjacent crack origins and grow perpendicular to the crack propagation.

Some examples of plane and rotating bending fatigue diagnosis are shown in Photos 5 and 6.

Photo 5 shows a 200 hp, 1180 RPM motor shaft that failed in less than a day. No progression marks means the fatigue load was constant. The instantaneous zone is relatively large, indicating the shaft was heavily loaded. Cracking started at numerous locations around the shaft, pointing to rotating bending as the cause. So many ratchet marks concentrated on the top and bottom of the photo make us suspect the shaft may not have been straight. Inspection, though, would indicate the root cause was associated with the belt drive.
In fact, the sheaves were worn so badly that the belts were riding in the bottom of the grooves. This situation approximately doubled the shaft bending stress.

The drive shaft in Photo 6 was on a steel-mill elevator. The surface is smoothest near the root of the keyway and became progressively rougher as the crack grew across the shaft. Numerous progression marks surrounding the tiny IZ and the change in surface condition about 40% of the way across the shaft from the IZ suggest something changed during the crack growth or that the elevator was not used for an extended period. These features are indicative of a slow-growing failure—and the fact that fretting corrosion may have substantially reduced the fatigue strength. 

Torsional fatigue failures… 
Until the advent of variable speed drives (VSDs), torsional fatigue failures were rare: Equipment designers could anticipate operating speeds and excitation frequencies and engi-neer around them. The purpose of a VSD is to allow operation at a wide range of speeds. That, unfortunately, has led to many motor and driven-shaft failures due to torsional-fatigue factors. While the most common torsional fatigue cracks start at the sharp corner (stress concentration) at the bottom of the keyway when couplings are poorly fitted, another common appearance is the diagonal shaft crack (like that shown in Fig. 2).

Photo 7 reflects the battered end of a motor shaft with a terrible (loose) coupling fit that let the hub repeatedly drive the key against the side of the keyway until a fatigue crack developed. (It’s not uncommon to see cases where the crack has propagated entirely around the shaft, leaving only a stub on the shaft.) 

Photo 8 shows both halves of the torsional-fatigue failure of a fan shaft in a plant that had recently changed to a VSD. The 45° angle to the central axis is a sure sign of torsional stresses, and the change in surface roughness across the shaft indicates the cause was fatigue forces.

Torsional fatigue stresses frequently go unnoticed (until too late) because personnel don’t understand what they’re looking at. For example, both of the pump shafts shown in Photo 9
failed due to torsional fatigue aggravated by a reduction in strength caused by corrosion. Some might look at the fracture face of the shaft on the right and think it was caused by rotating bending. Closer examination of the many ratchet marks shows they are at a 45° angle to the centerline of the shaft—a positive indication of torsional fatigue stresses with numerous origins. (Note that the ratchet marks seen in Photo 5 have straight sides, an indication that they were caused by bending forces.) 

Words of caution on interpreting the clues
While the oldest part of a fatigue failure typically has the smoothest surface—at least 98% of the time—it’s still crucial to look carefully at the failed part in the area of the origin: The shaft surface will describe the force. 

One of the greatest takeaways from this article is that a crack always grows perpendicular to the plane of maximum stress. Many times, we’ve seen shafts where the originating force was torsion with a short angular crack, but the majority of crack propagation was in bending—fooling inspectors into thinking that bending was the primary force. Don’t let yourself be taken in this way. MT

Neville Sachs is a Senior Consulting Engineer with the Sachs Salvaterra & Associates division of Applied Technical Services, Inc., a firm specializing in nondestructive testing and technical support services for improved plant and equipment reliability. Email: sachscracks@att.net.