Spiral bevel gears in the Gleason system transmit power between intersecting shafts with high contact ratio, low noise, smooth motion and strong load capacity, which is why they are standard in automotive, aerospace and marine drivetrains. Their flanks are cut, heat treated and then ground. Grinding is the last operation, so the geometry, residual stress state and surface integrity it leaves behind are the ones that enter service. The process is also difficult to observe: local temperatures, forces and grit wear inside the contact zone are not accessible in production, and the tooth surface is curved and continuously generating, which complicates direct measurement.
Where grinding sits in the manufacturing chain
The usual route for a spiral bevel gear is tooth milling, heat treatment and grinding. Heat treatment distorts the tooth form, and grinding removes that distortion while bringing the flank to its final accuracy. The kinematics resemble the meshing of the workpiece with an imaginary generating crown gear: the grinding wheel and the workpiece move in a coordinated generating relationship, and the abrasive grains on the wheel act as a dense set of cutting edges. Each grain removes a chip only a few micrometres thick, but the accumulated action of many grains produces the finished surface.
That apparently simple description hides a coupled problem. Plastic deformation and friction generate heat within a very small volume; the heat softens the material, which changes the flow stress and therefore the force; the force changes the contact pressure and the wear of the abrasive. A single-grain model can capture the basic chip-formation mechanism, but a real wheel cuts with many grains whose stress and thermal fields overlap. The work summarized here addresses that gap with a paired-grit finite-element model applied to a local tooth-flank coupon of the gear material.
Grit-scale removal: rubbing, ploughing and cutting
Material removal by an abrasive grain passes through three stages. In the rubbing stage the grain touches the flank at a shallow depth and produces mainly elastic deformation. As the depth increases, the normal and tangential forces grow and the material yields; ridges form on both sides of the track and a groove is ploughed into the surface. Friction and plastic work raise the temperature sharply at this stage. With further depth, the temperature reaches the point where material shears along a plane and flows ahead of and beside the grain as a chip. The transition between these stages depends on depth of cut, wheel speed, grit geometry and the thermophysical properties of the workpiece.
The model represents a grain as a cone with a height of 70 µm and an apex angle of 108°, a geometry that reproduces the strongly negative rake angle of an abrasive edge. Two identical grains are bonded together at their upper halves and move side by side, so that the second grain cuts material that the first has already deformed and heated. A 60# grit size is used, with grain spacing set at 150 µm and 220 µm to represent different degrees of interference.
Finite-element representation
The workpiece is a 2 mm × 1 mm × 1 mm coupon cut from the tooth flank region, assigned the properties of 25CrMo4 gear steel: elastic modulus 202 GPa, shear modulus 78 GPa, Poisson’s ratio 0.25 and density 7.8 g/cm³. The abrasive is an SG alumina wheel, modelled as a rigid body. The coupon is fixed at its base, and the grains move tangentially across the top surface. A shear friction law with a coefficient of 0.6 is applied at the contact, and the top surface and the grits are treated as heat-exchange surfaces with a convection coefficient of 0.02 N/(s·mm·°C). The initial temperature of both bodies is 20 °C.
Because the flank material undergoes large strain, high strain rate and heating during cutting, the workpiece is described by the Johnson–Cook constitutive relation:
$$\sigma = \left( A + B \varepsilon_p^{\,n} \right) \left( 1 + C \ln \frac{\dot{\varepsilon}}{\dot{\varepsilon}_0} \right) \left[ 1 – \left( T^* \right)^m \right]$$
$$T^* = \frac{T – T_r}{T_m – T_r}$$
Here \(\sigma\) is the flow stress, \(\varepsilon_p\) the equivalent plastic strain, \(\dot{\varepsilon}\) the plastic strain rate and \(\dot{\varepsilon}_0\) the reference strain rate; \(T\), \(T_r\) and \(T_m\) are the local, room and melting temperatures. The parameters used for 25CrMo4 are \(A = 120\) MPa, \(B = 891\) MPa, \(C = 0.02\), \(n = 0.2\) and \(m = 0.64\). Chip separation is handled by the Johnson–Cook shear failure criterion, which is appropriate for this class of large-deformation simulation.
The mesh is graded from a global element size between 0.05 mm and 0.003 mm to a refined zone on the ground surface with elements of 0.02 mm and 0.001 mm, giving 162,339 elements and 36,389 nodes. The analysis uses a Lagrangian incremental formulation with adaptive remeshing, which avoids the severe element distortion that otherwise stops convergence when the grain presses into the surface. Direct iteration is used for the solution, and the step size is set to one tenth of the smallest element, 0.001 mm per step, over 2,000 steps with output every fifth step.
Abrasive wear is computed with the Usui model, which relates wear to contact pressure, sliding velocity and interface temperature:
$$W = \int a \, p \, V \, e^{-b/T} \, dt$$
where \(W\) is the wear volume, \(p\) the interface pressure, \(V\) the sliding velocity, \(T\) the interface temperature, and \(a\) and \(b\) are calibration coefficients.
Simulation matrix
Four paired-grit simulations were run with the control-variable method, using the first as the reference case. One parameter was changed at a time so that the effects of speed, depth and spacing could be separated.
| Run | Wheel speed (m/s) | Depth of cut (µm) | Grain spacing (µm) |
|---|---|---|---|
| 1 | 30 | 20 | 150 |
| 2 | 50 | 20 | 150 |
| 3 | 30 | 50 | 150 |
| 4 | 30 | 20 | 220 |
Temperature response of the flank
The highest flank temperature in the reference case, at 30 m/s and 20 µm depth with 150 µm spacing, was 819 °C. Raising the wheel speed to 50 m/s at the same depth increased it to 950 °C. Raising the depth to 50 µm at 30 m/s increased it to 1,050 °C. Changing the grain spacing from 150 µm to 220 µm produced no significant change in the temperature field.
These trends follow from the balance between heat generation and heat removal. A higher wheel speed removes more material per unit time and does more frictional work per unit time, so the heat input rate rises even though each grain spends less time in the cut. A larger depth of cut increases the contact area and the normal pressure, which raises the friction force and the plastic work; it also places the heat source deeper and lengthens the conduction path, so the heat is carried away more slowly. The result is a much stronger temperature sensitivity to depth than to speed. At 50 µm the computed peak is already in the range associated with grinding burn, and the slower cooling makes thermal damage more likely. Adequate coolant delivery and controlled depth are therefore essential, not optional, process conditions.
The paired-grit model adds an effect that a single-grain model cannot show. The two grits deform and heat overlapping volumes of material, and the thermal fields interact. In the simulation, regions of the flank that were not in direct contact with either grain still exceeded 500 °C, and the cooling rate after the pass depended on the local distance from both grit tracks. This is consistent with the non-uniform heat flow observed on ground gear flanks. It matters because tempering and residual-stress gradients can extend beyond the visible contact marks and alter the subsurface state of the tooth.
Force history and the effect of grain interference
The force signal has a characteristic shape. As the grains enter the material, both normal and tangential components rise steeply to a peak. The normal force is consistently higher than the tangential force, by a factor of roughly 1.8, and the two follow nearly the same history. During steady cutting the force fluctuates around its peak within a band, and as the grains leave the surface it falls to zero. The dominance of the normal component reflects the large negative rake of the conical grit: material is pushed sideways and ahead of the edge, which reduces the resistance in the cutting direction. Some individual steps show outlier fluctuations, most likely from distorted elements during remeshing; these affect the trend negligibly.
At the same depth and spacing, increasing the speed from 30 m/s to 50 m/s lowered the mean forces slightly. The normal force fell from 17.4040 N to 16.8783 N and the tangential force from 9.8559 N to 9.3735 N. The mechanism is thermal softening: the higher interface temperature reduces the flow stress of the workpiece, so less force is needed to produce the same deformation. The change is modest, less than 5%, which indicates that the thermal softening effect partly offsets the higher strain rate at the faster speed.
| Wheel speed (m/s) | Mean normal force (N) | Mean tangential force (N) |
|---|---|---|
| 30 | 17.4040 | 9.8559 |
| 50 | 16.8783 | 9.3735 |
Depth of cut acts much more strongly. Increasing the depth from 20 µm to 50 µm raises both force components markedly, because the uncut chip thickness and the contact area both grow, and the material removed per unit time increases. The force history also becomes more irregular, which is consistent with the deeper groove and the larger volume of ploughed material.
Grain spacing changes the stability of the force rather than its mean. With 150 µm spacing the variance of the normal force was 9.2639 N² and that of the tangential force 2.8441 N²; with 220 µm spacing the corresponding values fell to 7.7975 N² and 2.2797 N². Closer grains interfere more strongly: the stress field of the leading grain has not relaxed before the trailing grain arrives, so the force fluctuates over a wider range. That fluctuation is a source of grinding vibration and can affect chatter, wheel runout and the uniformity of the finished surface.
| Grain spacing (µm) | Variance of normal force (N²) | Variance of tangential force (N²) | Relative stability |
|---|---|---|---|
| 150 | 9.2639 | 2.8441 | Lower |
| 220 | 7.7975 | 2.2797 | Higher |
Wear of the abrasive
The wear pattern is consistent across all four cases. The main wear face lies on the side of the grit that faces the direction of relative motion, and because the grit is conical the wear scar spreads elliptically, growing faster as the pass continues. The characteristic wear depth was 1.72 µm in the reference case. At 50 m/s it rose to 2.14 µm, with the worn region occupying almost the same area. At 50 µm depth it rose to 2.5 µm and the wear zone extended outward. Changing the spacing to 220 µm had little effect on wear depth.
The ranking is therefore depth first, speed second and spacing third. This follows directly from the Usui formulation: pressure, sliding velocity and temperature all enter the wear integral, and depth raises pressure and temperature together, whereas speed raises velocity and temperature without increasing the contact pressure as strongly. Spacing changes the interaction between grains but not the mean load per grain, so its influence on wear is limited.
Checking the trends against ground teeth
The simulated trends were checked on ground flanks. Four groups of gears with comparable material, geometry and prior accuracy were ground under the conditions in Table 4, changing one variable at a time. In the fourth group, a fine-grained wheel replaced the coarse wheel; a finer grit corresponds to smaller spacing between active edges.
| Group | Wheel speed (m/s) | Depth of cut (µm) | Grit specification |
|---|---|---|---|
| 1 | 30 | 20 | Coarse |
| 2 | 50 | 20 | Coarse |
| 3 | 30 | 50 | Coarse |
| 4 | 30 | 20 | Fine |
The ground surfaces were examined with a ZEISS EVO scanning electron microscope. At the higher speed, clearly visible abrasive marks appeared on the flank. At the larger depth, those marks became deeper and localized depressed fracture zones developed. With the finer grit, the marks took on a folded, pleated appearance, which is the surface signature of stronger interference between closely spaced cutting edges. The direction of all three effects agrees with the simulation: speed increases the visibility of cutting marks, depth increases surface damage, and reduced spacing intensifies interference.
The comparison is qualitative rather than a direct measurement of temperature or force, but it supports the model’s ranking of the parameters and its prediction that interference is a real surface-forming mechanism rather than a numerical artefact.
What this means for production grinding
For spiral bevel gears, the practical message is that depth of cut is the dominant lever. It produces the largest rise in flank temperature, the largest increase in grinding force and the largest wear of the abrasive. A process window that keeps depth modest, distributes removal over more passes and delivers coolant effectively is therefore safer than one that relies on a single aggressive pass. The temperature penalty is not linear: a 2.5-fold increase in depth raised the peak temperature by about 28%, while a 67% increase in speed raised it by about 16%.
Speed presents a trade-off. Higher speed lowers the force per grit through thermal softening and can improve productivity, but it raises the interface temperature and accelerates abrasive wear. Where speed is increased, cooling and wheel-grade selection have to compensate. Grain spacing, by contrast, is a stability variable. Wider spacing reduces force variance and the associated vibration without strongly changing the mean thermal load or the wear depth. That makes the abrasive architecture — grain spacing, uniformity of distribution and grit size — a useful independent control for surface quality and dynamic stability, distinct from the thermal control exercised through depth and cooling.
Wear follows the same ranking as temperature. Because depth drives both, keeping depth under control also extends the interval between dressing operations, which in turn helps hold tooth accuracy and avoids the cost of frequent wheel conditioning.
Model scope and open questions
The model is deliberately local. It represents two idealized conical grits cutting a flat coupon, not a complete wheel engaging a curved tooth with generating motion. Bond material, grain randomness, elastic deflection of the wheel and machine, coolant flow and the full kinematic path are not included, and the friction coefficient is held constant. The Johnson–Cook and Usui parameters are material- and pair-specific, and the forces reported are for a grit pair rather than for an entire wheel, where many grains interact stochastically.
Within that scope the results are internally consistent and agree with the observed surface features. Extending the approach to a full generating simulation, with measured interface temperatures and forces for validation, would allow the local trends to be translated into shop-floor parameter windows for specific gear sizes and wheel specifications.