In my research, I focus on a line-contact spiral bevel gear pair and on the numerical control (NC) gear machining method required to realize its theoretical tooth surfaces. The conventional design and correction framework for spiral bevel gears is largely based on local conjugate contact, which was developed under the motion constraints of mechanical gear cutting machines. When modern five-axis NC machine tools and general machining centers are used, those constraints no longer dominate the manufacturing process. Therefore, I re-examine the problem from the perspective of line-contact conjugate meshing and develop a systematic gear machining method that includes tooth surface modeling, tooth surface discretization, geometric topology deviation evaluation, machine setting compensation, finite element contact verification, and kinematic transformation to a five-axis CNC machining center.
The main idea of my work is that the gear tooth surface is treated as the preferentially determined surface. I first establish the generating motion model of the gear and derive its machined tooth surface. Then I derive the pinion theoretical tooth surface that is line-contact conjugate with the gear machined surface. Because this theoretical pinion surface cannot be produced directly by a conventional mechanical cradle-type gear cutting machine, I introduce an additional cutting motion through NC gear machining. In particular, I modify the roll ratio of the traditional tilt cutter method by using a second-order and third-order modification polynomial. This creates a new pinion gear machining method. I then compare the pinion machined surface with the pinion theoretical surface, build a geometric topology deviation model, and solve a correction model between tooth surface deviations and machine tool adjustment parameters. Finally, I convert the corrected pinion cutting motion from a virtual cradle-type machine to a dual-rotary-table five-axis machining center and verify the whole process by NC simulation.
2. Mathematical Model of the Line-Contact Conjugate Tooth Surfaces
I begin with the gear, because the gear tooth surface is set as the known or preferentially machined surface. The gear is generated by a face-milling cutter head on a traditional cradle-type gear cutting machine. The cutter head rotates about its own axis and forms a conical cutting surface. At the same time, the cradle rotates and generates the gear tooth surface by enveloping. The relative motion between the cutter and the gear blank determines the final gear tooth surface. This is the foundation of the subsequent gear machining calculation.
2.1 Cutter Cutting Surface
In the cutter coordinate system, I express the cutter cutting surface position vector and unit normal vector. Let \(u\) and \(\theta\) be the cutter surface parameters, \(r_0\) the nominal cutter radius, \(W\) the cutter point width, and \(\alpha\) the cutter profile angle. The position vector of the cutter surface is
$$
\mathbf{r}_t(u,\theta)=
\begin{bmatrix}
\left(r_0 \pm \left(\frac{W}{2}-u\sin\alpha\right)\right)\cos\theta\\
\left(r_0 \pm \left(\frac{W}{2}-u\sin\alpha\right)\right)\sin\theta\\
-u\cos\alpha\\
1
\end{bmatrix}.
$$
The corresponding unit normal vector is
$$
\mathbf{n}_t=
\begin{bmatrix}
\cos\alpha\cos\theta\\
-\cos\alpha\sin\theta\\
\pm\sin\alpha
\end{bmatrix}.
$$
In the above expressions, the upper sign is used for the outer cutting edge, which machines the concave side of the gear, and the lower sign is used for the inner cutting edge, which machines the convex side. This sign convention is used consistently in my gear machining model.
2.2 Gear Machined Tooth Surface
By transforming the cutter surface from the cutter coordinate system to the gear coordinate system through the machine tool coordinate systems, I obtain the gear machined tooth surface and its unit normal vector:
$$
\mathbf{r}_2=\mathbf{M}_{2g}\mathbf{M}_{ga}\mathbf{M}_{ab}\mathbf{M}_{b0}\mathbf{M}_{0t}\mathbf{r}_t,
$$
$$
\mathbf{n}_2=\mathbf{L}_{2g}\mathbf{L}_{ga}\mathbf{L}_{ab}\mathbf{L}_{b0}\mathbf{L}_{0t}\mathbf{n}_t.
$$
The meshing equation between the generating gear and the gear blank is
$$
f_{2c}=\mathbf{n}_0\cdot\mathbf{v}_{2c}=0,
$$
where \(\mathbf{n}_0\) is the unit normal vector in the fixed machine coordinate system, and \(\mathbf{v}_{2c}\) is the relative velocity between the generating gear and the gear blank. The relative velocity components in the fixed machine coordinate system are written as
$$
\begin{aligned}
v_{2c,x}&=\omega_c y-i_{2c}\omega_c(y-E_2)\sin\delta_{M2},\\
v_{2c,y}&=\omega_c x+i_{2c}\omega_c\left[(x-X_2)\sin\delta_{M2}-z\cos\delta_{M2}\right],\\
v_{2c,z}&=-i_{2c}\omega_c(y-E_2)\cos\delta_{M2}.
\end{aligned}
$$
Here, \(i_{2c}\) is the gear machining roll ratio, \(\delta_{M2}\) is the gear blank mounting angle, \(E_2\) is the vertical offset, \(X_2\) is the axial setting, and \(B_2\) is the床位 setting. The gear tooth surface model is therefore expressed as
$$
\Sigma_2:
\begin{cases}
\mathbf{r}_2=\mathbf{r}_2(u_2,\theta_2,\phi_2,\lambda_2),\\
f_{2c}(u_2,\theta_2,\phi_2,\lambda_2)=0.
\end{cases}
$$
The gear machine setting vector is
$$
\lambda_2=[S_2,q_2,\delta_{M2},X_2,E_2,B_2,i_{2c}]^T.
$$
This vector contains the radial cutter position, angular cutter position, blank mounting angle, axial setting, vertical offset,床位, and roll ratio. I compute these values by following the conventional generating method and the standard cutter selection procedure. The gear machining process is thus fully determined by \(\lambda_2\) and the cutter parameters.
2.3 Pinion Theoretical Tooth Surface in Line Contact
Once the gear machined tooth surface is known, I establish the meshing coordinate system of the gear pair. The gear rotates about its axis, and the pinion rotates about its axis. For the spiral bevel gear pair studied here, the shaft angle is \(90^\circ\), and the offset is zero. The gear tooth surface is transformed into the pinion coordinate system by
$$
\mathbf{r}_1’=\mathbf{M}_{10}\mathbf{M}_{0g}\mathbf{M}_{g2}\mathbf{r}_2,
$$
$$
\mathbf{n}_1’=\mathbf{L}_{10}\mathbf{L}_{0g}\mathbf{L}_{g2}\mathbf{n}_2.
$$
The meshing equation for line-contact conjugate motion is
$$
f_{12}=\mathbf{n}_0\cdot\mathbf{v}_{12}=0.
$$
The relative velocity components between the pinion and the gear are
$$
\begin{aligned}
v_{12,x}&=\omega_2 i_{12} y-\omega_2 y,\\
v_{12,y}&=\omega_2 x-\omega_2 i_{12} x+\omega_2 z,\\
v_{12,z}&=-\omega_2 y.
\end{aligned}
$$
After simplification with the gear angular velocity set to unity, I obtain
$$
\begin{aligned}
v_{12,x}&=i_{12}y-y,\\
v_{12,y}&=x-i_{12}x+z,\\
v_{12,z}&=-y.
\end{aligned}
$$
Therefore, the pinion theoretical tooth surface is
$$
\Sigma_1′:
\begin{cases}
\mathbf{r}_1’=\mathbf{r}_1′(u_2,\theta_2,\phi_2),\\
f_{12}(u_2,\theta_2,\phi_2)=0.
\end{cases}
$$
This surface is line-contact conjugate with the gear machined surface. It is the target surface that the pinion gear machining process must reproduce as closely as possible.
2.4 Boundary Lines of the Conjugate Tooth Surfaces
To determine the effective contact region, I calculate two types of boundary lines on the conjugate tooth surfaces: the undercut boundary line and the meshing boundary line. The undercut boundary line is also called the first-type boundary line. It is the envelope of the family of instantaneous contact lines on the pinion theoretical surface. Outside this line, the pinion surface may be undercut. The conditions for the undercut boundary line are
$$
\begin{cases}
f=0,\\
\Psi=0,
\end{cases}
$$
where
$$
\Psi=
\begin{vmatrix}
\mathbf{r}_u & \mathbf{r}_\theta & \mathbf{v}\\
f_u & f_\theta & f_t\\
D & \cdots & \cdots
\end{vmatrix}.
$$
The meshing boundary line, or second-type boundary line, separates the meshing region from the non-meshing region. Its conditions are
$$
\begin{cases}
f=0,\\
f_t=0.
\end{cases}
$$
By solving these equations, I obtain a series of boundary points on the gear tooth surface and then map them onto the pinion theoretical surface. These boundary lines are used later to discard invalid discrete points from the pinion theoretical surface and to check whether the corrected pinion machined surface has a safe undercut boundary.
3. New Gear Machining Method for the Pinion
The pinion theoretical tooth surface is a line-contact conjugate surface. In general, a traditional cradle-type gear cutting machine with a tilt cutter mechanism cannot produce this surface exactly, because the mechanical machine has only a limited set of adjustable motions. I therefore develop a new gear machining method that uses the flexibility of NC motion. The method is based on the tilt cutter method, but I modify the machining roll ratio during the generating motion. In other words, the relationship between the cradle rotation and the pinion rotation is no longer a constant ratio; it becomes a variable ratio described by a second-order and third-order polynomial.
3.1 Modified Roll Motion
I express the modified roll motion as
$$
\phi_1=i_{1c}(\phi_c-\phi_{c0})+C_0(\phi_c-\phi_{c0})^2+D_0(\phi_c-\phi_{c0})^3,
$$
and the corresponding angular velocity relation as
$$
\omega_1=i_{1c}\omega_c\left[1+2C_0(\phi_c-\phi_{c0})+3D_0(\phi_c-\phi_{c0})^2\right].
$$
Here, \(i_{1c}\) is the initial pinion machining roll ratio, \(C_0\) is the second-order modification coefficient, and \(D_0\) is the third-order modification coefficient. The initial values of \(C_0\) and \(D_0\) are set to zero, and they are later determined by the tooth surface correction calculation. This modified roll motion is the key additional cutting motion in my gear machining method.
3.2 Pinion Machined Tooth Surface
Using the coordinate systems of the traditional cradle-type machine with a tilt cutter mechanism, I transform the cutter surface to the pinion coordinate system. The pinion machined tooth surface is
$$
\mathbf{r}_1=\mathbf{M}_{1g}\mathbf{M}_{ge}\mathbf{M}_{ed}\mathbf{M}_{d0}\mathbf{M}_{0Q}\mathbf{M}_{Qc}\mathbf{M}_{c\beta}\mathbf{M}_{\beta b}\mathbf{M}_{bJ}\mathbf{M}_{Ja}\mathbf{M}_{at}\mathbf{r}_t,
$$
and its unit normal vector is
$$
\mathbf{n}_1=\mathbf{L}_{1g}\mathbf{L}_{ge}\mathbf{L}_{ed}\mathbf{L}_{d0}\mathbf{L}_{0Q}\mathbf{L}_{Qc}\mathbf{L}_{c\beta}\mathbf{L}_{\beta b}\mathbf{L}_{bJ}\mathbf{L}_{Ja}\mathbf{L}_{at}\mathbf{n}_t.
$$
The meshing equation for the pinion machining process is
$$
f_{1c}=\mathbf{n}_0\cdot\mathbf{v}_{1c}=0,
$$
where the relative velocity components are
$$
\begin{aligned}
v_{1c,x}&=\omega_c y-i_{1c}\omega_c(y-E_1)\sin\delta_{M1},\\
v_{1c,y}&=\omega_c x+i_{1c}\omega_c\left[(x-X_1)\sin\delta_{M1}-z\cos\delta_{M1}\right],\\
v_{1c,z}&=-i_{1c}\omega_c(y-E_1)\cos\delta_{M1}.
\end{aligned}
$$
The pinion machined tooth surface is therefore
$$
\Sigma_1:
\begin{cases}
\mathbf{r}_1=\mathbf{r}_1(u_1,\theta_1,\phi_c,\lambda_1),\\
f_{1c}(u_1,\theta_1,\phi_c,\lambda_1)=0.
\end{cases}
$$
The pinion machine setting vector is
$$
\lambda_1=[Q,\beta,J,I,\delta_{M1},X_1,E_1,B_1,i_{1c},C_0,D_0]^T.
$$
This vector includes the cradle angle, eccentric angle, cutter rotation angle, cutter tilt angle, blank mounting angle, axial setting, vertical offset,床位, roll ratio, and the two modification coefficients. By adjusting these parameters, I can make the pinion machined surface approach the pinion theoretical surface. This is the core of the proposed gear machining strategy.
4. Tooth Surface Discretization and Geometric Topology Deviation
To compare the pinion machined surface with the pinion theoretical surface, I convert both continuous tooth surfaces into discrete point sets. I divide the tooth surface into a grid along the tooth height direction and the tooth length direction. I use an axial-section projection method to constrain the grid points. For each grid point, I solve a nonlinear system consisting of the surface equations and the meshing equation. This gives the coordinates of the discrete tooth surface points.
4.1 Grid Generation
Let the tooth surface be divided into \(m\) rows in the tooth height direction and \(n\) columns in the tooth length direction. The boundary points of the projection are determined from the outer cone distance, face width, addendum, dedendum, pitch cone angle, and root cone angle. The coordinates of the boundary points are
$$
\begin{aligned}
P_1:&\quad
\begin{cases}
x_{P1}=(R_e-b)\cos\delta_1-(h_{ae}-b\tan\theta_a)\sin\delta_1,\\
y_{P1}=(R_e-b)\sin\delta_1+(h_{ae}-b\tan\theta_a)\cos\delta_1,
\end{cases}\\
P_2:&\quad
\begin{cases}
x_{P2}=R_e\cos\delta_1-h_{ae}\sin\delta_1,\\
y_{P2}=R_e\sin\delta_1+h_{ae}\cos\delta_1,
\end{cases}\\
P_3:&\quad
\begin{cases}
x_{P3}=R_e\cos\delta_1+h_{fe}\sin\delta_1,\\
y_{P3}=R_e\sin\delta_1-h_{fe}\cos\delta_1,
\end{cases}\\
P_4:&\quad
\begin{cases}
x_{P4}=(R_e-b)\cos\delta_1+(h_{fe}-b\tan\theta_f)\sin\delta_1,\\
y_{P4}=(R_e-b)\sin\delta_1-(h_{fe}-b\tan\theta_f)\cos\delta_1.
\end{cases}
\end{aligned}
$$
After the boundary points are obtained, I interpolate the points on the four boundary curves and then calculate the interior grid points by intersecting the row and column lines. The projection-grid relation is
$$
\begin{cases}
z_p=u_c,\\
x_p^2+y_p^2=x_c^2+y_c^2.
\end{cases}
$$
Combining this relation with the pinion meshing equation gives a nonlinear system in the unknowns \(u\), \(\theta\), and \(\phi_c\). I solve this system by a numerical solver and obtain all grid point coordinates.
4.2 Geometric Topology Deviation Model
Let \(\Sigma_1’\) be the pinion theoretical tooth surface and \(\Sigma_1\) the pinion machined tooth surface. Because the two surfaces may have a spatial offset, I first rotate the theoretical surface about the pinion axis so that its midpoint coincides with the midpoint of the machined surface. The rotation transformation is
$$
\begin{bmatrix}
x_0\\y_0\\z_0\\1
\end{bmatrix}
=
\begin{bmatrix}
\cos\Delta\theta & -\sin\Delta\theta & 0 & 0\\
\sin\Delta\theta & \cos\Delta\theta & 0 & 0\\
0 & 0 & 1 & 0\\
0 & 0 & 0 & 1
\end{bmatrix}
\begin{bmatrix}
x_1’\\y_1’\\z_1’\\1
\end{bmatrix}.
$$
For any point \(M_1\) on the pinion machined surface, the deviation between the two surfaces is defined along the normal direction. The vector form is
$$
\Delta\delta(\xi_1,\lambda_1)=(\mathbf{r}_1-\mathbf{r}_0)\cdot\mathbf{n}_1(\xi_1,\lambda_1).
$$
The numerical form is
$$
\Delta=[x_0-x_1,\ y_0-y_1,\ z_0-z_1]
\begin{bmatrix}
n_x\\n_y\\n_z
\end{bmatrix}.
$$
This deviation model is the basis for the parameter compensation calculation. The smaller the deviation, the closer the pinion machined surface is to the theoretical line-contact surface.
5. Machine Setting Compensation and Correction Algorithm
I keep the cutter parameters fixed and correct only the machine tool adjustment parameters. The tooth surface deviation is therefore treated as a function of the pinion machine setting vector \(\lambda_1\). For \(k\) discrete points, the deviation vector is
$$
\mathbf{d}(\lambda_1)=
\begin{bmatrix}
d_1(\lambda_1), & d_2(\lambda_1), & \cdots, & d_k(\lambda_1)
\end{bmatrix}^T.
$$
I formulate the correction problem as a least-squares optimization problem:
$$
\min_{\lambda_1} f(\lambda_1)=\frac{1}{2}\mathbf{d}(\lambda_1)^T\mathbf{d}(\lambda_1).
$$
To solve this problem, I use a Levenberg-Marquardt algorithm with a trust-region strategy. The iteration formula is
$$
\lambda_{k+1}=\lambda_k+\Delta\lambda_k,
$$
$$
(G_k+\mu_k E)\Delta\lambda_k=-g_k,
$$
where
$$
g_k=J_k^T d_k,
$$
$$
G_k=J_k^T J_k.
$$
Here, \(J_k\) is the Jacobian matrix of the deviation vector with respect to the machine setting parameters, \(E\) is the identity matrix, and \(\mu_k\) is the damping coefficient. The trust-region ratio is
$$
\eta_k=\frac{\Delta f}{\Delta q},
$$
and the damping coefficient is updated according to the value of \(\eta_k\).
In my improved correction strategy, I first correct a subset of parameters,
$$
\rho_1=[Q,\beta,E_1,i_{1c},C_0,D_0]^T,
$$
while keeping the cutter point surface and the blank position in the axial direction within a controlled range. I solve the reduced least-squares problem
$$
\min_{\rho_1} f(\rho_1)=\frac{1}{2}\mathbf{d}(\rho_1)^T\mathbf{d}(\rho_1)
$$
until the maximum deviation is reduced to approximately \(100\,\mu m\). Then I use the updated deviation as the initial condition and correct the full parameter vector \(\lambda_1\). This two-stage correction improves numerical stability and avoids unreasonable machine setting values.
6. Numerical Example of the Line-Contact Gear Pair
I apply the proposed method to a spiral bevel gear pair with a tooth number ratio of \(11/32\). The gear pair is designed with standard taper. The main geometric parameters are listed in Table 1. The gear generating parameters are listed in Table 2. These values are used to compute the gear machined surface and the pinion theoretical surface.
| Parameter | Pinion | Gear |
|---|---|---|
| Number of teeth \(z\) | 11 | 32 |
| Hand of spiral | Left | Right |
| Module \(m\) / mm | 6 | 6 |
| Normal pressure angle \(\alpha\) / deg | 20 | 20 |
| Mean spiral angle \(\beta\) / deg | 35 | 35 |
| Shaft angle \(\eta\) / deg | 90 | 90 |
| Face width \(b\) / mm | 32 | 32 |
| Outer cone distance \(R_e\) / mm | 101.5135 | 101.5135 |
| Working tooth height \(h_k\) / mm | 10.17 | 10.17 |
| Addendum \(h_a\) / mm | 7.23 | 4.062 |
| Dedendum \(h_e\) / mm | 2.94 | 8.352 |
| Pitch cone angle \(\delta\) / deg | 18.9704 | 71.0296 |
| Face cone angle \(\delta_a\) / deg | 23.6738 | 66.3262 |
| Root cone angle \(\delta_f\) / deg | 16.679 | 73.321 |
| Parameter | Value |
|---|---|
| Nominal cutter radius \(r_{02}\) / mm | 95.25 |
| Cutter point width \(W_2\) / mm | 3 |
| Inner cutter profile angle \(\alpha_{2i}\) / deg | 22 |
| Outer cutter profile angle \(\alpha_{2o}\) / deg | -18 |
| Blank mounting angle \(\delta_{M2}\) / deg | 66.3262 |
| Radial cutter position \(S_2\) / mm | 83.9129 |
| Angular cutter position \(q_2\) / deg | 68.4074 |
| 床位 \(B_2\) / mm | 0 |
| Axial setting \(X_2\) / mm | 0 |
| Vertical offset \(E_2\) / mm | 0 |
| Roll ratio \(i_{2c}\) | 1.0539 |
Using these parameters, I discretize the gear machined surface and the pinion theoretical surface into a \(9\times 5\) grid. Some representative coordinates are given in Table 3 and Table 4. I then generate the discrete point models and the tooth profile curves. The pinion theoretical surface contains an undercut region near the root, so I discard the points outside the undercut boundary before the deviation analysis.
| Index | Gear concave \(x\) | Gear concave \(y\) | Gear concave \(z\) | Gear convex \(x\) | Gear convex \(y\) | Gear convex \(z\) |
|---|---|---|---|---|---|---|
| 1 | 37.8697 | -54.3929 | -21.0280 | 29.3581 | -59.4206 | -21.0280 |
| 2 | 36.8410 | -54.3720 | -22.7725 | 29.7935 | -58.5314 | -22.7725 |
| 3 | 35.8641 | -54.3041 | -24.5170 | 30.1685 | -57.6630 | -24.5170 |
| 4 | 34.9499 | -54.1846 | -26.2615 | 30.4865 | -56.8159 | -26.2615 |
| … | … | … | … | … | … | … |
| 45 | 25.4892 | -89.7350 | -40.8983 | 21.1967 | -90.8448 | -40.8983 |
| Index | Pinion concave \(x\) | Pinion concave \(y\) | Pinion concave \(z\) | Pinion convex \(x\) | Pinion convex \(y\) | Pinion convex \(z\) |
|---|---|---|---|---|---|---|
| 1 | 25.4222 | -8.9298 | -64.2435 | 26.7814 | 2.9643 | -64.2435 |
| 2 | 24.1719 | -7.1260 | -64.8432 | 25.1617 | 1.3965 | -64.8432 |
| 3 | 22.7714 | -5.6252 | -65.4429 | 23.4551 | 0.2010 | -65.4429 |
| 4 | 21.2392 | -4.5038 | -66.0425 | 21.7035 | -0.5865 | -66.0425 |
| … | … | … | … | … | … | … |
| 45 | 24.0840 | 16.4377 | -96.9080 | 21.0528 | 20.1743 | -97.3204 |
For the pinion gear machining calculation, the initial machine settings of the pinion concave surface are listed in Table 5. The pinion concave surface is discretized into a \(9\times 7\) grid. After solving the correction model, the partially corrected parameters are listed in Table 6, and the fully corrected parameters are listed in Table 7. The initial tooth surface deviations and the deviations after correction are given in Table 8 and Table 9. The maximum deviation after correction is generally less than \(2\,\mu m\), which indicates that the proposed gear machining method and compensation algorithm are effective.
| Parameter | Initial value |
|---|---|
| Cutter tip radius \(r_{01}\) / mm | 89.058 |
| Cutter profile angle \(\alpha_1\) / deg | -14 |
| Blank mounting angle \(\delta_{M1}\) / deg | 18.0067 |
| Cradle angle \(Q_0\) / deg | 139.4683 |
| Eccentric angle \(\beta\) / deg | 44.0167 |
| Cutter rotation angle \(J\) / deg | 35.1618 |
| Cutter tilt angle \(I\) / deg | 21.0119 |
| 床位 \(B_1\) / mm | 6.0354 |
| Axial setting \(X_1\) / mm | -2.343 |
| Vertical offset \(E_1\) / mm | 3.8146 |
| Roll ratio \(i_{1c}\) | 3.0996 |
| Second-order coefficient \(C_0\) | 0 |
| Third-order coefficient \(D_0\) | 0 |
| Parameter | Value after partial correction |
|---|---|
| Cradle angle \(Q_0\) / deg | 144.5055 |
| Eccentric angle \(\beta\) / deg | 44.4760 |
| Vertical offset \(E_1\) / mm | -4.7196 |
| Roll ratio \(i_{1c}\) | 3.1055 |
| Second-order coefficient \(C_0\) | -0.0106 |
| Third-order coefficient \(D_0\) | -0.0062 |
| Parameter | Value after full correction |
|---|---|
| Blank mounting angle \(\delta_{M1}\) / deg | 18.0488 |
| Cradle angle \(Q_0\) / deg | 144.4485 |
| Eccentric angle \(\beta\) / deg | 44.5051 |
| Cutter rotation angle \(J\) / deg | 35.0784 |
| Cutter tilt angle \(I\) / deg | 21.0714 |
| 床位 \(B_1\) / mm | 6.0355 |
| Axial setting \(X_1\) / mm | -2.3421 |
| Vertical offset \(E_1\) / mm | -4.7192 |
| Roll ratio \(i_{1c}\) | 3.1090 |
| Second-order coefficient \(C_0\) | -0.0138 |
| Third-order coefficient \(D_0\) | -0.0137 |
| Point | Initial deviation / mm | Point | Initial deviation / mm | Point | Initial deviation / mm |
|---|---|---|---|---|---|
| 1 | 0.8207 | 13 | 1.2246 | 26 | 0.4398 |
| 2 | 0.9342 | 15 | 0.1851 | 27 | 0.6623 |
| 3 | 1.0532 | 16 | 0.3060 | 29 | -0.4083 |
| 4 | 1.1820 | 17 | 0.4327 | 30 | -0.2797 |
| 5 | 1.3311 | 18 | 0.5694 | 31 | -0.1450 |
| 6 | 1.5168 | 19 | 0.7260 | 32 | 0 |
| 8 | 0.4979 | 20 | 0.9419 | 33 | 0.1651 |
| 9 | 0.6151 | 22 | -0.1172 | 34 | 0.3880 |
| 10 | 0.7379 | 23 | 0.0075 | 36 | -0.6873 |
| 11 | 0.8705 | 24 | 0.1382 | 37 | -0.5549 |
| 12 | 1.0232 | 25 | 0.2790 | 38 | -0.4163 |
| Point | Deviation after correction / \(\mu m\) | Point | Deviation after correction / \(\mu m\) | Point | Deviation after correction / \(\mu m\) |
|---|---|---|---|---|---|
| 1 | 0.6 | 13 | 3.5 | 26 | -1.5 |
| 2 | -0.5 | 15 | 2.0 | 27 | 0.5 |
| 3 | -1.5 | 16 | 1.3 | 29 | 1.5 |
| 4 | -2.1 | 17 | 0.2 | 30 | 1.8 |
| 5 | -1.3 | 18 | -1.0 | 31 | 1.1 |
| 6 | 5.0 | 19 | -1.6 | 32 | 0 |
| 8 | 1.6 | 20 | 1.7 | 33 | -1.2 |
| 9 | 0.6 | 22 | 2.0 | 34 | -0.3 |
| 10 | -0.5 | 23 | 1.8 | 36 | 0.7 |
| 11 | -1.5 | 24 | 0.8 | 37 | 1.5 |
| 12 | -1.5 | 25 | -0.5 | 38 | 1.4 |
I also check the undercut boundary of the corrected pinion machined surface. The undercut curve of the machined surface lies outside the theoretical undercut curve, which means that the corrected gear machining parameters do not introduce undercut into the effective contact region. This confirms that the compensation strategy is geometrically safe.
7. Three-Dimensional Modeling and Finite Element Contact Analysis
After obtaining the gear machined surface and the corrected pinion machined surface, I generate their discrete point sets and import them into a three-dimensional modeling environment. I connect the points along the tooth length and tooth height directions to form tooth profile curves. Then I fit the curves into complete tooth surfaces, add the root cone surfaces, and establish the tooth groove models. By subtracting the tooth grooves from the gear blanks and arraying them around the axes, I obtain the complete three-dimensional models of the pinion and the gear. The assembled gear pair is checked for global interference, and no interference is found. This validates the tooth surface equations and the modeling procedure.
For the finite element contact analysis, I simplify the assembly by keeping only part of the gear teeth and the full pinion. I assign a commonly used carburizing steel, 20CrMnTi, to both gears. The material properties are listed in Table 10. I create a surface-to-surface contact pair with the pinion surface as the master surface and the gear surface as the slave surface. The tangential behavior is frictionless, and the normal behavior is hard contact. The analysis steps are given in Table 11. I apply a resisting torque of \(10\,N\cdot m\) and an angular velocity of \(0.5\,rad/s\) to the gear pair.
| Property | Value |
|---|---|
| Elastic modulus / MPa | \(2.06\times 10^5\) |
| Poisson’s ratio | 0.3 |
| Density / \(t/mm^3\) | \(7.85\times 10^{-9}\) |
| Step | Time length / s | Increment size / s | Maximum number of steps |
|---|---|---|---|
| Step-1 | 0.01 | 0.01 | 10000 |
| Step-2 | 1.2 | 0.04 | 1000 |
The finite element results show that at every instantaneous meshing state, the contact region is a line rather than a point. Multiple teeth participate in meshing at the same time, which indicates a favorable contact ratio. This result verifies that the proposed line-contact gear machining method and the tooth surface correction calculation are correct. I also observe stress concentration near the tooth tips, which suggests that further tooth surface modification may be needed in future work.
8. NC Machining Transformation to a Five-Axis Machining Center
The corrected pinion cutting motion is more complex than the motion of a conventional cradle-type machine. Therefore, I implement it on a dual-rotary-table vertical five-axis machining center. This is a typical and flexible platform for advanced gear machining. The machine has three linear axes \(X\), \(Y\), and \(Z\) and two rotary axes \(A\) and \(C\). I establish the kinematic model of the machine and derive the transformation between the cutter coordinate system and the blank coordinate system.
The machine coordinate system \(\mathbf{S}_0\) is fixed to the machine bed. The cutter coordinate system \(\mathbf{S}_t\) is attached to the cutter head center but does not rotate with the cutter. The blank coordinate system \(\mathbf{S}_g\) is attached to the pinion blank. The transformation matrix from the cutter system to the blank system is
$$
\mathbf{H}_{gt}=
\begin{bmatrix}
\cos c & -\sin c\cos a & \sin c\sin a & U\\
\sin c & \cos c\cos a & -\cos c\sin a & V\\
0 & \sin a & \cos a & W\\
0 & 0 & 0 & 1
\end{bmatrix},
$$
where
$$
\begin{aligned}
U&=-z\sin a\sin c-y\sin a\cos c+x\cos c,\\
V&=z\sin a\cos c-y\sin a\sin c-x\sin c,\\
W&=z\cos a+y\sin a+h-H.
\end{aligned}
$$
Here, \(a\) is the rotation angle of the \(A\)-axis, \(c\) is the rotation angle of the \(C\)-axis, \(x\), \(y\), and \(z\) are the linear axis positions, \(h\) is the distance from the \(A\)-axis to the \(C\)-axis table plane, and \(H\) is the distance from the pinion cone apex to the \(C\)-axis table plane.
The relative pose between the cutter and the blank must be the same in the virtual cradle-type machine and in the five-axis machining center. Therefore,
$$
\mathbf{H}_{gt}=\mathbf{M}_{1t}.
$$
The matrix \(\mathbf{M}_{1t}\) is determined by the corrected pinion machine settings and the cradle rotation angle. By expanding both sides of this equation, I obtain the motion expressions of all five NC axes as functions of the cradle angle \(\phi_c\):
$$
\begin{aligned}
a&=f_a(\phi_c)=\cos^{-1}\left(f_{33}(\phi_c)\right),\\
c&=f_c(\phi_c)=\tan^{-1}\left(\frac{f_{13}(\phi_c)}{f_{23}(\phi_c)}\right)+\text{periodic correction},\\
x&=f_x(\phi_c),\\
y&=f_y(\phi_c),\\
z&=f_z(\phi_c).
\end{aligned}
$$
The corresponding NC commands are
$$
\begin{aligned}
A&=-f_a(\phi_c),\\
C&=-f_c(\phi_c),\\
X&=-f_x(\phi_c),\\
Y&=-f_y(\phi_c),\\
Z&=f_z(\phi_c).
\end{aligned}
$$
Because the \(A\) and \(C\) solutions come from inverse trigonometric functions, multiple periodic solutions exist. I select the solution that keeps the \(C\)-axis table facing the operator and that makes the \(C\)-axis motion vary monotonically around its initial value. This ensures safe and visible gear machining.
9. NC Program Generation and Simulation
To realize the continuous cutting motion on the five-axis machine, I discretize the cradle rotation into a finite number of instantaneous states. I choose an increment of \(0.125^\circ\) for the cradle angle. For each state, I calculate the corresponding \(X\), \(Y\), \(Z\), \(A\), and \(C\) values using the derived expressions. The NC program then drives the machine through all states in sequence. The \(X\), \(Y\), \(Z\), and \(A\) motions simulate the rotating cradle, while the \(C\) motion simulates the variable-speed pinion rotation determined by the modified roll polynomial.
Before finishing the pinion, I also perform rough cutting for both the gear and the pinion. The gear rough cutting uses the forming method, and the pinion rough cutting uses the generating method. The rough cutting cutter point width is reduced to leave finishing allowance. For the gear, the rough cutter point width is \(W_2’=2.2\,mm\). For the pinion, the rough cutter point width is \(W_1’=1.4\,mm\), the inner profile angle is \(22^\circ\), the outer profile angle is \(18^\circ\), and the rough roll ratio is \(i_{1c}’=3.07371\). The cradle angle range for the gear finishing and pinion rough cutting is \(40^\circ\). For the pinion finishing, I determine the cradle angle range from the discrete tooth surface calculation, and I use a range of approximately \(-25^\circ\) to \(8^\circ\) for the concave side.
The NC program for the pinion concave surface is organized as follows. First, the cutting motion model is converted from the virtual cradle-type machine to the five-axis machining center. Second, the tool and the machine axes are moved to the initial safe position, and the program variables are initialized. Third, the tool path and workpiece path are generated for each instantaneous state. Fourth, the tool and blank move from the start position to the final position, and one tooth flank is completed. Fifth, the blank is indexed to the next tooth slot. Sixth, when the number of machined teeth equals the total number of teeth, the cutting process ends, and all axes return to the safe position. Seventh, the cutter is changed to machine the other flank, and the same procedure is repeated.
I use a FANUC 0M numerical control milling system as the target controller. The simulation is performed in a CNC simulation environment. The machine tool model, blank models, and cutter models are imported as STL files. The cutter library contains the rough and finishing cutter heads for both the gear and the pinion. The blank models are generated from the gear pair geometric parameters. During the simulation, no collision or interference occurs between the cutter, the machine components, and the blank. The simulated gear and pinion profiles are complete, and no obvious tooth surface defects are observed. This verifies the tooth surface modeling, the kinematic transformation, and the NC program generation for the proposed gear machining method.
10. Discussion and Conclusions
I have developed a complete line-contact spiral bevel gear machining methodology for NC manufacturing. The main contributions and conclusions of my work are summarized as follows.
First, I established a mathematical model of line-contact conjugate spiral bevel gear tooth surfaces. The gear machined surface is treated as the known surface, and the pinion theoretical surface is derived as the line-contact conjugate surface. I also derived the two types of boundary lines, namely the undercut boundary line and the meshing boundary line, to determine the effective contact region. This provides a rigorous geometric basis for the subsequent gear machining correction.
Second, I proposed a new pinion gear machining method. The method is based on the tilt cutter method, but the machining roll ratio is modified by second-order and third-order polynomials. This additional cutting motion gives the machine tool enough freedom to approximate the line-contact theoretical pinion surface. I built the cutting motion model on a traditional cradle-type machine, derived the pinion machined surface equation, and compared it with the theoretical surface. The comparison was formulated as a geometric topology deviation model.
Third, I developed a machine setting compensation strategy. The tooth surface deviations are treated as functions of the machine setting parameters. I used a two-stage Levenberg-Marquardt algorithm to minimize the deviation sum of squares. In the first stage, a subset of parameters is corrected to reduce the maximum deviation to about \(100\,\mu m\). In the second stage, the full parameter vector is corrected. For the numerical example, the final deviations are generally below \(2\,\mu m\), which demonstrates the effectiveness and accuracy of the compensation method.
Fourth, I built three-dimensional models of the pinion and gear from the discrete tooth surface points. The assembled gear pair passed the global interference check. I then performed finite element contact analysis. The results show that the gear pair maintains line contact at every instantaneous meshing state, and multiple teeth participate in meshing simultaneously. This confirms that the line-contact design and the corrected gear machining parameters are correct.
Fifth, I transformed the corrected pinion cutting motion from the traditional cradle-type machine to a dual-rotary-table five-axis machining center. By equating the relative pose of the cutter and the blank in the two machines, I derived the motion expressions for the \(X\), \(Y\), \(Z\), \(A\), and \(C\) axes as functions of the cradle angle. I discretized the continuous cutting process and generated NC programs. The simulation results show that the machine axes follow the programmed motion without interference, and the resulting tooth profiles are complete. This validates the entire NC gear machining workflow.
In future work, I plan to extend the method to hypoid gears and to spiral bevel gears with different shaft angles and offsets. I also intend to investigate other additional cutting motions, such as helical modification, and to combine them with multi-pass or segmented cutting strategies. In addition, tooth surface modification should be studied to reduce the stress concentration observed near the tooth tips and to improve load capacity and durability. The proposed line-contact gear machining framework can serve as a foundation for high-performance spiral bevel gear manufacturing on modern NC machine tools.