Spiral bevel gears are used where smooth, high-load transmission is required between intersecting shafts, including aerospace, automotive, and energy equipment. Their curved teeth engage gradually and distribute load over several contact points, but the final grinding operation strongly influences surface integrity. Residual stress is one of the most important outcomes of that operation: compressive stress can retard fatigue-crack initiation, while tensile stress can promote crack opening, reduce load capacity, and shorten service life.
The local action of an abrasive wheel is difficult to observe directly. A wheel contains many randomly shaped and randomly spaced grains, each acting as a microscopic cutting tool. A single-grain model reduces this complexity to a controlled local contact problem, making it possible to isolate the effects of grinding speed, depth of cut, and grain rake angle on the stress field.
Grinding mechanics and the residual-stress profile
Material removal by an abrasive grain proceeds through three overlapping stages. In the rubbing stage, the grain contacts the tooth surface elastically and slides over it; friction generates heat but little material removal. In the ploughing stage, the grain penetrates and pushes material sideways, producing plastic flow, a grooved surface, and small ridges along the groove. In the chip-formation stage, the deformed material shears and flows ahead of the grain, and a chip is removed.
These stages create a coupled mechanical and thermal field. The grain exerts a local compressive load, while friction and plastic work raise the temperature. After the grain passes and the workpiece cools, the final residual stress depends on the competition between plastic compression, thermal expansion, and elastic recovery.
Under the reported grinding conditions, the residual stress in the grinding direction follows a characteristic profile: the outermost layer is in compression, the maximum compressive stress lies below the surface, the stress then becomes tensile with increasing depth, and finally decays toward zero. The subsurface peak is not accidental. The top layer is directly loaded and partly constrained by the underlying material, while the subsurface region experiences a combination of mechanical compression and thermal loading. Deeper into the workpiece, the mechanical influence weakens and elastic recovery dominates, leaving a tensile tail.
Finite-element representation
The single-grain model represents the abrasive grain as a rigid cone with a height of 0.07 mm, an included cone angle of 90°, and an edge radius of 0.01 mm. The workpiece is a local tooth-surface segment measuring 0.4 mm × 0.3 mm × 0.15 mm. This reduction is justified by the scale difference between a grain and a spiral bevel gear: over the contact length of one grain, the curved tooth surface can be approximated as a local plane.
The workpiece material is AISI 4340 steel, treated as an elastoplastic body. Its flow behavior is described by the Johnson-Cook constitutive model, and chip formation is represented with the Johnson-Cook failure criterion. The grain-workpiece interface uses a penalty friction model with a friction coefficient of 0.3. Grinding is solved with Abaqus/Explicit; after the tool is removed, cooling is solved with Abaqus/Standard. Residual stress is extracted as the normal stress in the grinding direction, S11, after cooling. For each layer, 170 equidistant nodes are sampled and averaged to reduce local mesh-induced scatter.
| Property | Value |
|---|---|
| Density | 7,830 kg/m³ |
| Elastic modulus | 2 × 105 MPa |
| Yield strength | 792 MPa |
| Tensile strength | 980 MPa |
| Poisson’s ratio | 0.29 |
| Thermal conductivity | 38 W/(m·°C) |
| Johnson-Cook A | 792 MPa |
| Johnson-Cook B | 510 MPa |
| Johnson-Cook n | 0.26 |
| Johnson-Cook C | 0.014 |
| Johnson-Cook m | 1.03 |
| Failure parameters D1–D5 | 0.05, 3.44, −2.12, 0.002, 0.61 |
A three-grain model with randomly distributed grains was also constructed using the same material properties, analysis steps, mesh size, and contact constraints. This provides a bridge between the isolated single-grain case and the collective action of a real grinding wheel.
Parameter sensitivity
Single-factor simulations were used to separate the effects of grinding speed, grinding depth, and grain rake angle. The tested ranges and the observed ranking are summarized in Table 2.
| Factor | Tested range | Observed effect on residual stress | Influence rank |
|---|---|---|---|
| Grinding speed, vs | 10–25 m/s | Both compressive and tensile peaks rise slightly; the effect is the weakest of the three factors | 3 |
| Grinding depth, ap | 0.01–0.04 mm | Stress increases with depth as contact area, force, and plastic deformation grow | 2 |
| Grain rake angle, γ | −20° to −50° | Stress increases as the rake angle becomes more negative and ploughing dominates | 1 |
The rake angle has the strongest influence. At a grinding speed of 25 m/s and a depth of 0.01 mm, changing the rake angle from −20° to −50° produces a clear rise in residual stress. A more negative rake angle shifts the interaction from clean cutting toward ploughing. The grain pushes material ahead and to the sides instead of removing it efficiently, so the plastic zone expands and the internal stress accumulation becomes larger. This is a micro-mechanical explanation for the dominant role of grain shape and orientation.
Grinding depth ranks second. At a speed of 25 m/s and a rake angle of −20°, increasing the depth from 0.01 mm to 0.04 mm increases the contact area between the grain and the tooth surface. The grinding force rises, the plastic deformation zone extends deeper, and the volume of material removed per unit time increases. The result is a stronger compressive peak and a stronger tensile tail.
Grinding speed has the smallest effect. At a depth of 0.04 mm and a rake angle of −20°, raising the speed from 10 m/s to 25 m/s increases residual stress, but the change is modest. Two mechanisms compete. A higher speed brings more grains into contact per unit time, which tends to increase the thermal load. At the same time, the undeformed chip thickness per grain becomes smaller, which tends to reduce the force on each grain. Under the tested conditions, these effects largely offset each other, leaving speed as a secondary control variable.
Single-grain and multi-grain behavior
The three-grain simulation preserves the same qualitative stress profile as the single-grain case: compression near the surface, a subsurface compressive peak, a tensile transition, and decay toward zero. However, the maximum compressive stress is lower in the multi-grain case. Adjacent grains do not act independently. A grain that produces compression in one region can be surrounded by other grains that produce tensile fields, and the superposition partially cancels the compressive peak.
This result is important for interpreting single-grain models. The single-grain approach is effective for ranking parameters and revealing trends, but it overestimates the compressive peak relative to the three-grain case. Absolute stress values should therefore be treated as trend-level predictions unless they are calibrated against multi-grain simulations or physical measurements.
Response-surface model and process optimum
Fifteen simulation runs were arranged over the parameter ranges and fitted with quadratic response-surface models. One model represents the peak compressive stress magnitude, S1, and the other represents the peak tensile stress, S2. Analysis of variance shows that both models are significant at P < 0.001, and the F-value ranking confirms the single-factor result: rake angle > grinding depth > grinding speed. Normal residual plots place the data close to the fitted line, supporting the adequacy of the quadratic approximation within the tested window.
For gear performance, the optimization objective is not simply to maximize compression. It is to maximize the beneficial compressive stress while minimizing the detrimental tensile stress. Let S1 be the peak compressive magnitude and S2 the peak tensile stress. The multi-objective problem is
$$
\max S_1,\qquad \min S_2
$$
subject to
$$
10 \le v_s \le 25\ \text{m/s},\quad 0.01 \le a_p \le 0.04\ \text{mm},\quad -50^\circ \le \gamma \le -20^\circ
$$
The response-surface solution is a grinding speed of 25 m/s, a grinding depth of 0.01 mm, and a grain rake angle of −48°. At this point, the predicted peak compressive stress is 638.6 MPa and the predicted peak tensile stress is 24.9 MPa.
| Case | vs (m/s) | ap (mm) | γ (°) | S1 (MPa) | S2 (MPa) |
|---|---|---|---|---|---|
| Single-factor peak compression | 25 | 0.025 | −50 | 706.2 | 34.2 |
| Response-surface optimum | 25 | 0.01 | −48 | 638.6 | 24.9 |
The optimum is a compromise rather than a monotonic maximum. Compared with the single-factor peak-compression case, it reduces the compressive magnitude by about 68 MPa but lowers the tensile stress by about 9 MPa. This trade-off matters because tensile residual stress is the more damaging component for fatigue. The selected point also lies near the −50° boundary of the rake-angle range, which suggests that a wider range could shift the optimum and that the boundary should be examined in后续 work.
Engineering implications and limitations
For spiral bevel gears, the practical process window suggested by the model is a shallow grinding depth, a relatively high grinding speed within the thermal limits of the material, and a wheel topography that produces an effective negative rake angle. Shallow depth limits the force and the depth of the plastic zone. Higher speed can improve productivity, but it also raises heat generation, so burn control and cooling must be considered separately. The rake angle is not usually set as a machine parameter; it is controlled by grain shape, wheel specification, dressing, and wear. This makes wheel selection and dressing strategy the practical levers for the rake-angle effect.
The model has clear boundaries. It uses a local planar workpiece segment rather than the full curved tooth surface, a conical grain rather than a real irregular abrasive particle, and AISI 4340 steel rather than every gear steel in service. The stress output is the S11 component in the grinding direction, not the complete residual-stress tensor. The three-grain model is still a strong simplification of a real wheel, which contains many grains, pores, bond material, and wear flats. The comparison against additional parameter sets is a numerical check; it does not replace physical grinding trials with residual-stress measurement.
Within those limits, the analysis provides a coherent mechanism-based picture. Grinding-induced residual stress in spiral bevel gears is controlled primarily by grain geometry and engagement, secondarily by depth of cut, and only weakly by speed over the tested range. The subsurface compressive peak is beneficial, but the tensile tail must be kept low. A response-surface optimization around this trade-off identifies a useful process point, while the gap between single-grain and multi-grain predictions points to the next step: validation on actual spiral bevel gear surfaces using diffraction or hole-drilling measurements, together with multi-grain models that include realistic wheel topography.