The shaft was 45 mm diameter, 40CrMo, hardened to 35 HRC, carrying a chain sprocket on a conveyor drive. The motor was 11 kW, the reducer output was 850 Nm. The shaft had a standard parallel key, 14×9 mm, 80 mm long, DIN 6885. It broke after fourteen months. The fracture surface was clean and flat, with beach marks that started at the keyway corner. A classic fatigue failure, and the starter crack was exactly where these always start.
The customer’s first reaction was to blame the material. “The steel was bad.” The steel wasn’t bad. The keyway was doing what keyways do.
What a keyway does to a shaft
A keyway cuts a rectangular notch into the shaft. At the corner of that notch, the stress from torque and bending concentrates. The theoretical stress concentration factor Kt at a keyway corner is between 2 and 3 for a shaft in torsion, depending on the radius at the corner. A sharp corner, one cut by a standard broach with no radius, sits at the high end.
The fatigue stress concentration factor Kf is lower than Kt because steel has some notch sensitivity, but for a hardened shaft the difference shrinks. At 35 HRC, q is around 0.8, so Kf lands near 0.8 x (Kt – 1) + 1. With Kt of 2.5, Kf is about 2.2. Meaning the alternating stress the shaft sees is more than twice what the plain section would see. The keyway is effectively asking the shaft to carry double the load, locally.
The numbers on the failed shaft
For a 45 mm shaft under 850 Nm torque, the nominal shear stress is about 48 MPa:
τ = 16T / (π d³) = 16 x 850,000 Nmm / (π x 45³) = 47.6 MPa
With Kf = 2.2, the local stress at the keyway corner is around 105 MPa shear. If the drive also puts a bending moment on the shaft (sprocket overhung load does), add that stress on top. The combined equivalent stress at the corner was well above the fatigue limit of the material at this hardness.
The fatigue limit for 40CrMo at 35 HRC in torsion is roughly 220-260 MPa. The nominal stress was 48 MPa — safe, with margin. The local stress after concentration, plus the overhung bending from the sprocket, put the corner right at or above the endurance limit. Fourteen months of 24/7 running, and the crack started.
The fixes, in the order we recommended them
First, radius the keyway. A broached keyway comes out sharp-cornered. Specifying a 0.5 mm radius at the keyway corners (cut with a ball-end mill after broaching, or better, specify a keyway with radius per DIN 6885/1 for highly loaded shafts) drops Kt from 2.5 to about 1.8. That alone buys back a third of the lost fatigue strength.
Second, move the keyway out of the high-bending zone. The sprocket was mounted at the very end of the shaft, so the keyway sat right at the point of maximum bending stress. Moving the sprocket 20 mm inboard, or keying a longer hub and letting the keyway start past the bending peak, changes the picture a lot.
Third, consider what the key is actually for. In a conveyor drive that reverses direction under load, the key carries the full torque alternation. A parallel key is fine for steady loads. For reversing drives, a spline or a serration distributes the load over the whole circumference. The stress concentration at a spline root is around 1.5, not 2.5. But splines cost more to cut, so the conversation usually ends with “use a bigger shaft instead.”
Fourth, and often the cheapest: go up one shaft diameter. A 50 mm shaft at the same torque has nominal shear of about 35 MPa instead of 48. Combined with the same keyway, the local stress drops below the endurance limit. Sometimes the answer to a keyway fatigue problem is to stop fighting the keyway and make the shaft bigger.
What actually fixed this one
The customer wasn’t going to change the reducer or the sprocket. So the shaft was remade with a 1 mm radius at the keyway ends, the sprocket hub was extended so the key bore further into the hub (the keyed length went from 80 to 100 mm, cutting the surface pressure on the key in half), and the shaft was re-specified at 50 mm diameter with the same keyway. The drive has been running three years without a crack. The cost was a remachined shaft and a new sprocket bore.
One thing we did not do was add a second key 180 degrees apart. Two keys don’t share the load evenly — manufacturing tolerance means one key takes most of the torque. Two keys are a way to double the stress concentration for the illusion of extra capacity. Don’t.
Keyways are the natural birthplace of shaft fatigue cracks. The stress concentration at the corner is real, it’s around 2x, and no amount of good steel fixes it. Radius the corners, keep the keyway out of the bending zone, or go up a shaft size. The broken shaft was doing exactly what the numbers said it would. The steel was fine. The design wasn’t.
