Spiral bevel gears, including the hypoid variant used when the pinion axis is offset, transmit torque between intersecting shafts through curved teeth with a localized contact pattern. Their flanks are conjugate only in a small neighborhood, so the contact ellipse, load sharing, and transmission error change with torque and mounting position. Loaded tooth contact analysis (LTCA) is used to predict these changes before prototypes are cut. In that chain, the finite element (FE) model is often the weakest link: a small loss of flank accuracy or a distorted mesh can shift the predicted contact pattern as much as a change in cutter settings.
The Cost of the Conventional CAD-to-FEA Chain
A conventional LTCA workflow proceeds in four stages: tooth-surface coordinates are calculated in mathematical software, a solid model is built and assembled in CAD, the assembly is imported into an FE package, and the contact problem is prepared and solved. Each transfer introduces its own approximation. CAD surface reconstruction, format translation, and import into the solver can lose geometric precision or deform the tooth flank. For a component whose contact pattern is controlled at the micrometre scale, that is not a minor bookkeeping issue.
The workflow is also inefficient for parameter studies. Spiral bevel gears are already series products, and their design is increasingly handled through parameterized models, standardized assumptions, and automated analysis templates. If a change in cutter radius, tilt angle, modification depth, material data, or boundary condition requires the CAD assembly and FE preprocessing to be rebuilt by hand, sensitivity analysis and optimization become impractical. Parametric CAD in UG or SOLIDWORKS solves part of the problem by generating the solid, but the FE model still has to be prepared in another environment, and the data exchange remains.
ABAQUS offers a different route. Its scripting interface exposes CAE operations to Python and extends the language with more than 500 commands for geometry creation, assembly, meshing, interaction definition, load application, job submission, and postprocessing. A gear model can therefore be generated directly in the solver environment from the same tooth-surface equations used for the cutting simulation. The model becomes a scripted object: change a parameter, rerun the script, and rebuild the mesh and boundary conditions without repeating the CAD chain.
Cutting Kinematics Define the Flank
Spiral bevel gears are generated on a hypoid milling machine in which the cutter head and cradle reproduce a virtual generating gear, often called the crown gear. The cutter head represents one tooth of that virtual gear. As the cutter and workpiece rotate at a defined ratio, the cutter envelope cuts the flank in the same way that a mating gear would sweep through the blank. The wheel in the demonstration pair is cut by the formate method, without a generating roll; the pinion is cut by the duplex helical method, which adds a controlled helical motion during generation.
For the pinion, the cutter rotates through an angle \(\theta\), and the workpiece rotates through
$$R_b=\frac{N_t}{N_w},\qquad \psi_1=R_b\theta,\qquad R_a=\frac{N_p}{N_w},\qquad \psi=R_a\varphi+\psi_1.$$
Here \(N_t\) is the number of cutter-head starts or blade pairs, \(N_w\) is the workpiece tooth count, \(N_p\) is the generating-gear tooth count, \(\varphi\) is the cradle angle, \(R_a\) is the roll ratio, and \(\psi\) is the workpiece rotation. In duplex helical cutting, the bed position is not fixed; it changes with the generating motion according to
$$X_b=X_{b0}-H R_a \psi,$$
where \(X_b\) is the instantaneous bed position, \(X_{b0}\) is the bed position without helical motion, and \(H\) is the helical motion coefficient. The cutter profile includes inner and outer blades and a tip arc. The active flank is the envelope of the cutter family under the generating motion; the fillet is generated by the tip arc and merged with the active flank to close the tooth slot.
The same formulation applies to the wheel with the formate settings. The wheel has no cradle-generated motion in the usual sense, so its flank follows directly from the cutter profile and machine settings: radial cutter position, cutter radius, blade angles, machine installation angle, horizontal and vertical wheel positions, and bed position. These are not free stylistic choices. They determine the spiral angle, pressure angle, flank curvature, and ease-off that produce the loaded contact pattern.
From Surface Points to a Solid Model
The tooth-surface equations are continuous, but the FE model needs a bounded surface. The active flank is therefore trimmed and divided into an \(m\times n\) grid along the face-width and tooth-height directions. Each grid point is located by intersecting the tooth surface with a line through the pitch cone and transforming the result into the pinion coordinate system. In a local pitch-cone frame \((X,Y)\), the transformation to the pinion-axis frame \((X_1,Y_1)\) can be written as
$$X_1=X\cos\delta_m-Y\sin\delta_m+A_m\cos\delta_m-Z_p,\qquad Y_1=X\sin\delta_m+Y\cos\delta_m+A_m\sin\delta_m,$$
where \(\delta_m\) is the pitch cone angle, \(A_m\) is the cone distance, and \(Z_p\) is the distance from the cone apex to the crossing point. The corresponding point on the tooth surface is then obtained from the position vector and the pinion axis direction. Because the surface equations are nonlinear in the machine settings, a rotational projection scheme supplies the initial guess and a Newton iteration solves for the cradle angle and cutter phase at each grid point. The loop continues until the full \(m\times n\) point set is obtained.
The point cloud is then fitted into a family of curves, the curves are fitted into a surface, and the tooth slot is closed into a solid. Boolean subtraction from the gear blank produces one tooth space; circular arraying produces the gear. The important point is that the solid is created inside ABAQUS from the same discrete points that define the cutting geometry. There is no intermediate CAD format and no separate surface reconstruction step between the tooth-surface equations and the FE mesh.
Demonstration Pair and Finite-Element Setup
The demonstration pair has 7 pinion teeth and 43 wheel teeth, a 90° shaft angle, and a 25.4 mm pinion offset. The offset places the pair in the hypoid branch of the spiral bevel family, but the modeling procedure is the same. The wheel module is 6.860 mm. The principal geometry is listed below; the pinion offset and shaft angle are single-value properties of the pair.
| Parameter | Pinion | Wheel |
|---|---|---|
| Number of teeth | 7 | 43 |
| Module (mm) | — | 6.860 |
| Face width (mm) | 43.512 | 40.000 |
| Pinion offset (mm) | 25.400 | — |
| Shaft angle (°) | 90.000 | — |
| Mid-spiral angle (°) | 49.000 | 38.770 |
| Hand | Left | Right |
| Outer cone distance (mm) | 151.744 | 150.465 |
| Outer pitch diameter (mm) | 58.945 | 294.98 |
| Outer addendum (mm) | 8.347 | 1.205 |
| Outer dedendum (mm) | 2.524 | 9.666 |
| Pitch cone angle (°) | 11.200 | 78.588 |
| Face cone angle (°) | 14.299 | 78.932 |
| Root cone angle (°) | 10.862 | 75.433 |
| Parameter | Pinion | Wheel |
|---|---|---|
| Radial cutter position (mm) | 102.805 | 117.492 |
| Cutter tilt (°) | 3.582 | — |
| Cutter swivel (°) | 196.300 | — |
| Vertical wheel position (mm) | 23.114 | — |
| Machine installation angle (°) | 11.427 | 75.433 |
| Horizontal wheel position (mm) | −1.988 | 9.652 |
| Bed position (mm) | 4.813 | — |
| Roll ratio | 6.070 | — |
| Angular cutter position (°) | 65.581 | 70.977 |
| Helical motion coefficient (mm·rad−1) | −1.475 | — |
| Cutter radius (mm) | 111.882 | 114.300 |
| Inner blade profile angle (°) | 25.000 | 25.000 |
| Outer blade profile angle (°) | 20.000 | 20.000 |
| Cutter point width (mm) | 2.536 | 3.300 |
For the contact analysis, the wheel is represented by an eight-tooth sector to reduce computational cost. The pinion and wheel are assembled at their standard mounting position, meshed with C3D8R brick elements, and defined as a surface-to-surface contact pair, with the pinion flank as the master surface. Coupling constraints connect the gear bodies to reference points on their axes. Torque is applied to the wheel, and the analysis extracts contact pressure, contact area, and rotation history. The model is solved as a quasi-static problem so that the load-dependent elastic contact can be isolated from inertial effects.
Geometric Fidelity Compared with CAD Import
The scripted model was compared with a model built from the same discrete points in SOLIDWORKS and then imported into ABAQUS. The comparison was made at three levels: mesh quality, single-flank geometry, and single-tooth solid geometry. The imported model produced irregular elements and local mesh defects; merging the circular array at the same tolerance was also less clean than in the scripted model. The scripted model meshed regularly because the surface topology and the array operation were controlled by the same script.
| Observation | CAD import | Direct ABAQUS script |
|---|---|---|
| Mesh quality | Irregular elements and local defects; poorer merge after circular array | Regular mesh; clean merge at the same tolerance |
| Single-flank comparison | Vertices coincided; mid-edge deviations from a few nanometres to about 90 nm | Reference surface generated directly from the point set |
| Single-tooth solid comparison | Key-point deviations of about 0.07–1.93 μm | Reference solid generated directly from the surface |
The deviations are small in absolute terms, but LTCA contact patterns are sensitive to micrometre and sub-micrometre flank shape. The comparison is a model-to-model check, not an absolute metrology result: it quantifies the error introduced by the CAD exchange, not the error of the cutting model itself. The direct model is not exact either. Its accuracy still depends on the density of the discrete point set and on ABAQUS surface fitting. The advantage is that the error path is shorter and under script control. The discrepancy also grew with geometric complexity, from a single flank to a solid tooth, which is consistent with the accumulation of surface-fitting and translation errors in the CAD route.
Loaded Contact Pattern
The contact analysis was run at 10, 300, 600, and 900 N·m on both the drive and coast flanks. At 10 N·m, the contact patch is narrow and elongated. The flanks are locally conjugate and deform little under light load, so the instantaneous contact ellipse has a short major axis. As torque increases, the patch spreads outward and the major axis grows. At the highest load, edge contact becomes visible. This is the expected consequence of elastic deformation in a localized-contact gear pair: the teeth bend and the contact area migrates toward the edges of the theoretical bearing pattern.
The high-pressure zone remains near the middle of the flank, and no severe abnormal contact was observed in the loaded zone. The practical concern is the edge contact at high torque, which raises sensitivity to mounting error, misalignment, and stress concentration. A contact pattern that looks acceptable at low load can therefore become a durability issue at high load if the flank relief and mounting stiffness are not matched to the operating torque.
Transmission Error Under Load
Transmission error (TE) is the difference between the actual driven-gear rotation and the rotation predicted by the tooth ratio. For pinion 1 and wheel 2,
$$\varepsilon=(\varphi_2-\varphi_{2,0})-\frac{N_1}{N_2}(\varphi_1-\varphi_{1,0}).$$
Here \(\varphi_1\) and \(\varphi_2\) are the instantaneous rotations, the subscript 0 denotes the initial position, and \(N_1/N_2=7/43\) is the theoretical ratio. The angles are in radians; \(\varepsilon\) is converted to arcseconds for comparison. The numerical TCA solves the unloaded contact path from the same tooth-surface equations, while LTCA includes the elastic deformation of the teeth and the contact. The two curves have a phase offset because the initial contact point in the FE model is not exactly the design reference point. The TCA curve was shifted along the abscissa for comparison.
At 10 N·m, the LTCA curve is slightly below the TCA curve, with a similar opening size and period. That level of agreement supports the use of the scripted model for LTCA. As torque increases to 300, 600, and 900 N·m, the curve shifts farther downward and its peak-to-peak amplitude decreases. The downward shift is caused by elastic deformation: higher torque produces a larger mean lag. The amplitude decrease is caused by load sharing among multiple tooth pairs and by reduced impact at mesh entry and exit, which makes the motion smoother under heavy load.
The TE curve also shifts to the right with increasing load. In a locally conjugate pair, the contact ratio near the reference point is low, so the deformation at the reference point is relatively large. The maximum TE therefore tends to move toward the reference point as the load rises. At the same torque, the coast-side TE amplitude is larger than the drive-side amplitude. The drive side is consequently the smoother operating direction under load, while the coast side is more sensitive to impact and noise.
What the Scripted Workflow Changes
The main benefit is not a single gear model but a repeatable model generator. Once the tooth-surface solver and the FE template are scripted, a change in cutter radius, tilt angle, blade angle, modification depth, or load case becomes a parameter edit rather than a new CAD project. That matters for spiral bevel gears because their design is already series-based and their loaded behavior is controlled by a large set of coupled machine settings. Parameterized FE models allow sensitivity studies, uncertainty analysis, and optimization to be run on the same geometric basis.
The scope of the present analysis is a single 7×43 pair under quasi-static, isothermal conditions with a fixed mounting position. It isolates the geometric and elastic effects that dominate LTCA at moderate speed. It does not resolve high-speed dynamic effects, lubricant film formation, thermal distortion, or manufacturing scatter. Those require transient, lubricated, or coupled thermo-mechanical models. Within its scope, direct scripted construction removes a data-transfer step, keeps the FE mesh traceable to the cutting parameters, and makes loaded contact analysis practical enough for design iteration.