Finite Element Analysis of the Deformation of Crushing Teeth of Cycloid Gears

• 1
  Why the crushing teeth of cycloid gears bend before they break
• 2
  The three force components that drive deformation
• 3
  How grinding speed, depth and feed change the result
• 4
  When FEA is worth it and when it is not
• 5
  Turning the model into machining decisions
• 6
  Cutting parameters and their effect on tooth deformation

A cycloid tooth is not a rigid body. The profile follows an epicycloid or hypocycloid curve, and the wall behind the working flank is thin compared with a spur gear of the same module. When the grinding wheel or shaving cutter pushes into that flank, the load has nowhere to go except through a short, tapered section of steel.

That geometry is why the crushing teeth of cycloid gears deform at loads that would barely register on a helical gear. The bending stiffness drops as the contact point moves toward the tooth tip. The tip is where the section modulus is lowest and where the radius of curvature is smallest.

Finite element analysis turns this into numbers. You mesh the tooth, apply the measured or calculated grinding force, and read the displacement field. The output is not a single value. It is a map: which flank moves, by how much, and in which direction.

The practical question is always the same. Does that displacement push the finished profile outside the tolerance band? If the answer is yes, the process needs changing before the part is cut, not after.

Grinding force is usually split into three components. The normal force FN acts perpendicular to the radial direction, pressing the wheel or cutter into the flank. The tangential force FT acts along the cutting direction and drives material removal. The axial force FA acts along the gear axis.

FN does most of the damage. It is the component that bends the tooth and compresses the subsurface layer. FT matters for heat and for wheel wear, but its contribution to elastic deflection is smaller because it acts closer to the tooth root, where the section is thicker.

FA becomes important when the wheel is fed along the face width. A large axial component pushes the tooth sideways and can twist thin-rim cycloid discs. On a disc with a rim under about 5 mm, that twist is often the largest single displacement in the model.

Model all three components if you are chasing micron-level accuracy. For a first-pass check, FN alone gets you within roughly 10 to 15 percent of the full result on a standard cycloid profile.

Once the force is known, the model becomes a study in cutting parameters. Grinding speed has a modest effect on elastic deformation. At 50 m/s the chip is thinner than at 25 m/s, so the specific cutting energy rises, but the total normal force usually changes by less than a few percent on a dressed wheel.

Grinding depth is the parameter that moves deflection most. Going from a 10 μm to a 30 μm depth of cut roughly triples the removed volume per pass, and the normal force tracks that increase. Deformation scales with it. If a finish pass is creeping past tolerance, halving the depth is the fastest fix.

Axial feed speed behaves in a way that surprises people. As the feed rate rises, the uncut chip thickens, the grinding force climbs, and the elastic deformation of the tooth grows with it. The deformation does not fall away. It follows the force.

The interaction matters more than any single number. A fast axial feed with a shallow depth can produce less deflection than a slow feed with a deep cut, because the normal force stays lower. Model the combination, not the parameters one at a time.

FEA earns its keep on thin-rim cycloid discs, on profiles with a small radius of curvature at the tip, and on any part where the tolerance band is under ±0.01 mm. In those cases, a two-hour model can save a scrapped batch.

It is less useful on thick, rigid discs with generous tolerances. If the predicted deflection is a tenth of the tolerance band, the model is confirming what a shop already knows. Spend the time on fixturing instead.

The main limitation is input quality. Grinding force depends on wheel grade, dressing condition, coolant, and material hardness. A model fed with a guessed force is a guess with a mesh on top.

Measure the force on a test piece when you can. A dynamometer reading on the actual wheel and material is worth more than a fine mesh with the wrong load.

A displacement map is only useful if it changes a setup sheet. The usual conclusions are straightforward. If the tip deflects outward, the finish pass needs a smaller depth and a spring pass. If the root deflects, the fixture is not supporting the disc well enough.

Thermal deformation is often confused with elastic deformation. The two look similar in a dial indicator reading, but they need different fixes. Elastic deflection disappears when the load is removed. Heat-driven growth does not, until the part cools.

For a cycloid disc ground to Ra 0.8–1.6 μm and ±0.005 mm, we run the finish pass in two stages and let the part stabilize before final inspection. That is a process control decision, not a modeling one, but the model is what tells you the split is needed.

If the geometry is complex enough that grinding alone cannot hold the profile, we machine the disc on a simultaneous 5-axis center and grind only the working flank. That keeps the cutting force off the thin sections for most of the cycle.