Figure 4 shows another “tricky” part. We are looking at the fracture surface of a shaft loaded in torsion. We see the smear marks that we might think happened after the crack had completely fragmented the shaft, perhaps due to rubbing of the mating fracture face. There is no necking. The diameter is unchanged at the crack area. So, this is a brittle crack, right? Think again! This shaft was loaded in torsion, so the crack opening stresses, the normal stresses, act on helical planes that are at 45 degrees to the axis of the shaft. Brittle cracks are helical when they are created in torsion. Prove this to yourself by twisting a carrot or a piece of chalk. These are inherently brittle materials, and they will not usually break in any way but in a helical shape.
The planes where shear stresses can act in torsional loading are the disk-shaped transverse planes and the rectangular longitudinal (radial) planes. This steel was quite hard (HRC42), but it was highly stressed and twisted off in a ductile manner. In fact, if you look at the cylindrical surfaces, which may have longitudinal grinding marks for example, you can often see evidence of twisting, even though there is no obvious change in shape to the casual observer.
We conclude that when we see smear marks like this, (unless it was a fatigue crack caused by rotating bending, which has a totally different stress state and, in general, does not create such a flat fracture surface) they actually were created during the actual separation process, and this is a macro scale ductile part!
Understanding how a component fails is an important step in understanding why a component fails. In order to understand how a component cracks, it is important to understand what loading geometry or geometries could have been responsible for the fracture. It is equally important to understand how high the load was or how fast the component was loaded, and the basic loading geometries, including tension, compression, bending, torsion, contact stresses and direct shear. The failure analyst must strive to learn to “read” the fragment shapes to determine what loading geometry was actually present. This is a key to being able to properly determine whether the component was installed and used per the design intent.