I propose a practical method for grinding face gears on a general cylindrical internal gear grinding machine. My motivation is straightforward: face gears offer compact transmission layouts, convenient mounting and adjustment, and large transmission ratios, yet their precision manufacturing still depends heavily on specialized face gear grinding machines. That dependence raises tooling cost, reduces process flexibility, and slows the broader adoption of face gears in aviation, vehicle, and industrial drive systems. In my method, a disk-shaped grinding wheel is used as the generating tool, and the wheel carriage performs coordinated interpolation motions so that the disk wheel reproduces the motion of a generating gear. I establish a mathematical model of the grinding process, derive the face gear tooth surface equation including machine parameter errors, analyze how installation and geometric errors affect tooth surface topology, and verify the method through grinding experiments and gear measurement. The experimental results show that the maximum tooth surface deviation of the ground face gears is within a small tolerance band, and the correction of machine parameters effectively reduces the deviation. In my view, this method is important because it allows face gears to be ground without a dedicated face gear machine, which improves the推广 and cost profile of face gears.
Why face gears need a flexible grinding route. Face gears are used in applications where the axis intersection is orthogonal or near-orthogonal and where the pinion drives a face gear from the side. The geometry of face gears is different from that of conventional cylindrical gears and bevel gears, because the tooth surface is generated by a generating gear that meshes with the face gear under a specified shaft angle. Precision grinding of face gears is therefore a generating process in which the grinding wheel must represent one tooth of a generating gear. In many existing processes, the grinding wheel is dressed to an involute profile and then moved relative to the face gear according to the generating relationship. Worm grinding wheels can achieve high efficiency, but their dressing is difficult and time-consuming. Single-tooth generating grinding with a disk wheel is more flexible and has a simpler dressing principle, but its productivity is lower and is generally suitable for small-batch or medium-batch production. My method builds on the disk-wheel generating principle and adapts it to a general cylindrical internal gear grinding machine, so that face gears can be produced on a machine that is already widely available for internal gear grinding.
Core idea. I use an involute disk-shaped grinding wheel whose axial profile corresponds to the profile of one tooth of the generating gear. The grinding process simulates the meshing between the generating gear and the face gear. The disk wheel is mounted on a wheel arm, and the wheel arm is carried by a wheel carriage. The wheel carriage can rotate about an A axis and can interpolate in the Y and Z directions. The workpiece is mounted on a rotary table and rotates about a B axis, while the table can move along the X direction to cover the face width. By combining these motions, the disk wheel generates the face gear tooth surface. The generating ratio between the generating gear and the face gear is
$$i_{cg} = \frac{\omega_g}{\omega_c} = \frac{\phi_g}{\phi_c} = \frac{Z_c}{Z_g}$$
where $$\omega_g$$ and $$\omega_c$$ are the angular velocities of the face gear and the generating gear, $$\phi_g$$ and $$\phi_c$$ are their rotation angles, and $$Z_g$$ and $$Z_c$$ are their tooth numbers. This ratio is the foundation of the generating motion for face gears. In my model, the disk wheel does not simply rotate at a fixed position; instead, the wheel carriage center follows a circular path so that the involute profile of the disk wheel swings about the generating gear center. This is the key kinematic condition that makes a general internal gear grinding machine capable of generating face gears.
Notation used throughout my model. The main symbols are summarized in Table 1. I keep the notation consistent so that the kinematic equations, the tooth surface equations, and the error equations can be connected without ambiguity.
| Symbol | Meaning |
|---|---|
| Z_g | Number of teeth of the face gear |
| Z_c | Number of teeth of the generating gear |
| Z_1 | Number of teeth of the mating pinion |
| m_n | Normal module |
| alpha_k | Pressure angle |
| gamma | Shaft angle between pinion and face gear |
| R_1 | Inner radius of the face gear |
| R_2 | Outer radius of the face gear |
| r_b | Base circle radius of the generating gear |
| r_ps | Tip circle radius of the generating gear |
| E | Distance between the disk wheel center and the generating gear center |
| L | Length of the wheel arm |
| a | Sum of E and L |
| phi_c | Swing angle of the generating gear, equal to the wheel carriage swing angle |
| phi_g | Rotation angle of the face gear |
| theta_c | Involute angle parameter |
| theta_0 | Angle from the tool slot symmetry line to the involute starting point |
| u_c | Axial parameter of a point on the wheel surface |
| Delta y | Installation error of the wheel arm |
| Delta s | Installation error of the disk wheel |
| Delta l | Length error of the wheel arm |
| r_s | Position vector on the disk wheel surface |
| r_g | Position vector on the face gear tooth surface |
| e | Tooth surface deviation |
Generating principle with a disk wheel. The involute disk wheel is a tool whose cross-section is a single tooth profile of the generating gear. When the wheel swings about the generating gear axis and the face gear rotates according to the generating ratio, the envelope of the wheel positions forms the face gear tooth surface. I describe the generating relationship as follows. The wheel center is offset from the generating gear center by the distance E. The wheel arm length is L. The wheel carriage center moves along a circular arc, and the wheel itself is oriented so that its involute profile remains tangent to the generating motion. The swing angle of the generating gear is $$\phi_c$$, and the corresponding face gear rotation is $$\phi_g$$. The generating ratio is therefore fixed by the tooth numbers, as given in the first equation.
Machine structure and motion axes. The machine I use has five motion axes, namely X, Y, Z, A, and B, plus a spindle C for the disk wheel. The wheel carriage can move along Z and Y. The wheel arm is mounted on the wheel carriage and rotates about the A axis. The disk wheel is mounted on the wheel arm and rotates about the spindle C axis for cutting. The rotary table moves along the X axis. The workpiece is mounted on the B axis and rotates with the table. Table 2 lists the functions of these axes in my grinding method.
| Axis | Type | Function in the face gear grinding process |
|---|---|---|
| X | Linear | Moves the rotary table along the face width so that the full tooth width can be ground |
| Y | Linear | Interpolates the wheel carriage center during the generating swing |
| Z | Linear | Interpolates the wheel carriage center during the generating swing |
| A | Rotary | Swings the wheel arm and the disk wheel about the generating gear center |
| B | Rotary | Rotates the face gear workpiece according to the generating ratio |
| C | Spindle | Rotates the disk wheel for the cutting motion |
Generating motion equations. To make the disk wheel simulate the rotation of the generating gear, the wheel carriage rotates about the A axis while the wheel carriage center is interpolated in the Y and Z directions. The interpolation displacements are
$$D_Z = -(a – a \cos\phi_c) = -a(1-\cos\phi_c)$$
$$D_Y = a \sin\phi_c$$
where
$$a = E + L$$
The face gear rotation is synchronized with the wheel carriage swing by
$$\phi_g = i_{cg}\phi_c = \frac{Z_c}{Z_g}\phi_c$$
At the same time, the rotary table moves along the X direction to cover the face width. The X-axis stroke must be larger than the face gear tooth width. By combining the X, Y, and Z translations with the A and B rotations, the machine produces the generating motion required for grinding face gears. I emphasize that this generating motion does not require a special face gear machine tool; it uses the existing interpolation capability of a cylindrical internal gear grinding machine.
Mathematical model of the disk wheel surface. The axial profile of the disk wheel is made consistent with the generating gear profile. Therefore, I express the disk wheel tooth surface as an involute surface. The position vector of a point on the wheel surface is
$$r_s(\theta_c,u_c) = \begin{bmatrix}
-r_b[\cos(\theta_0+\theta_c)+\theta_c\sin(\theta_0+\theta_c)] – E \\
\pm r_b[\sin(\theta_0+\theta_c)-\theta_c\cos(\theta_0+\theta_c)] \\
u_c \\
1
\end{bmatrix}$$
In this expression, $$r_b$$ is the base circle radius of the generating gear, $$\theta_0$$ is the angle from the tool slot symmetry line to the involute starting point, $$\theta_c$$ is the involute angle parameter, and $$u_c$$ is the axial parameter of the wheel surface point. The double sign corresponds to the two sides of the wheel tooth slot. This surface is the geometrical foundation of my grinding model for face gears.
Coordinate transformations. I define four coordinate systems. The wheel coordinate system S_s is fixed to the disk wheel. The wheel carriage moving coordinate system S_p is fixed to the wheel carriage and rotates about the generating gear center. The machine fixed coordinate system S_n is fixed to the machine. The face gear coordinate system S_g is fixed to the rotary table and rotates with the face gear. The transformation from the wheel system to the wheel carriage system is
$$M_{p,s} = \begin{bmatrix}
0 & 0 & 1 & 0 \\
0 & 1 & 0 & 0 \\
-1 & 0 & 0 & -L \\
0 & 0 & 0 & 1
\end{bmatrix}$$
The transformation from the wheel carriage system to the machine fixed system is
$$M_{n,p}(\phi_c) = \begin{bmatrix}
1 & 0 & 0 & 0 \\
0 & \cos\phi_c & \sin\phi_c & a\sin\phi_c \\
0 & -\sin\phi_c & \cos\phi_c & -a(1-\cos\phi_c) \\
0 & 0 & 0 & 1
\end{bmatrix}$$
The transformation from the machine fixed system to the face gear initial system is
$$M_{m,n} = \begin{bmatrix}
1 & 0 & 0 & 0 \\
0 & 1 & 0 & 0 \\
0 & 0 & 1 & a \\
0 & 0 & 0 & 1
\end{bmatrix}$$
The transformation from the face gear initial system to the face gear moving system is
$$M_{g,m} = \begin{bmatrix}
\cos\phi_g & \sin\phi_g & 0 & 0 \\
-\sin\phi_g & \cos\phi_g & 0 & 0 \\
0 & 0 & 1 & 0 \\
0 & 0 & 0 & 1
\end{bmatrix}$$
Using these transformations, the disk wheel surface is represented in the face gear coordinate system as
$$r_g(\phi_c,u_c,\theta_c) = M_{g,m} M_{m,n} M_{n,p} M_{p,s} r_s(\theta_c,u_c)$$
This equation gives the family of wheel surfaces as seen from the face gear. The envelope of this family is the generated face gear tooth surface. In my derivation, the mathematical form of the generated surface is consistent with the tooth surface produced by shaping, which means that the disk-wheel grinding method has no theoretical generating error when the machine is correctly set.
Meshing equation and tooth surface solution. For the generating process, the wheel surface and the face gear tooth surface must satisfy the meshing equation. I write the meshing equation as
$$f_1(\theta_c,u_c,\phi_c) = \left( \frac{\partial r_g}{\partial \theta_c} \times \frac{\partial r_g}{\partial u_c} \right) \cdot \frac{\partial r_g}{\partial \phi_c} = 0$$
This equation states that the normal to the wheel surface is perpendicular to the relative velocity at the contact point. For a point on the face gear tooth surface, I also use the projection constraints. If the point has a rotational projection coordinate pair (R, Z), then
$$f_2(\theta_c,u_c,\phi_c) = r_{gx}^2 + r_{gy}^2 – R^2 = 0$$
$$f_3(\theta_c,u_c,\phi_c) = r_{gz} – Z = 0$$
For a specified point on the face gear tooth surface, I solve the nonlinear system formed by the meshing equation and the two projection equations. The unknowns are $$\theta_c$$, $$u_c$$, and $$\phi_c$$. Once these parameters are obtained, the tooth surface coordinate is determined. This procedure allows me to compute the theoretical face gear tooth surface for any grid point on the working region.
Machine parameter errors. In real grinding, the machine settings are not identical to the theoretical settings. I consider three important parameter errors: the installation error of the wheel arm, denoted $$\Delta y$$; the installation error of the disk wheel, denoted $$\Delta s$$; and the length error of the wheel arm, denoted $$\Delta l$$. These errors change the relative position and orientation between the disk wheel and the generating gear center, and therefore they modify the generated face gear tooth surface. To include these errors, I replace the theoretical transformation matrices with error-contaminated matrices. The wheel-to-carriage transformation becomes
$$M’_{p,s} = \begin{bmatrix}
0 & 0 & 1 & 0 \\
0 & 1 & 0 & \Delta s \\
-1 & 0 & 0 & -L + \Delta l \\
0 & 0 & 0 & 1
\end{bmatrix}$$
The carriage-to-machine transformation becomes
$$M’_{n,p}(\phi_c) = \begin{bmatrix}
1 & 0 & 0 & 0 \\
0 & \cos\phi_c & \sin\phi_c & (a+\Delta l)\sin\phi_c + \Delta y \\
0 & -\sin\phi_c & \cos\phi_c & -(a+\Delta l)(1-\cos\phi_c) \\
0 & 0 & 0 & 1
\end{bmatrix}$$
With these matrices, the face gear tooth surface containing machine parameter errors is
$$r_{g2}(\phi_c,u_c,\theta_c) = M_{g,m} M_{m,n} M’_{n,p} M’_{p,s} r_s(\theta_c,u_c)$$
This error model is the basis of my sensitivity analysis and machine parameter correction. I use it to predict how each error changes the face gear tooth surface topology before I adjust the actual machine.
Tooth surface deviation definition. To evaluate the ground face gear, I compare the theoretical tooth surface with the measured tooth surface. I define a grid on the working tooth surface. The grid has five rows and nine columns. The tooth tip and tooth root are shrunk by five percent of the tooth height, and the large end and small end are shrunk by ten percent of the tooth width. The grid points are therefore located inside the working region and avoid the transition surface, which does not participate in meshing. Table 3 shows the grid arrangement conceptually.
| Grid feature | Value in my evaluation |
|---|---|
| Number of rows | 5 |
| Number of columns | 9 |
| Tip shrinkage | 5 percent of tooth height |
| Root shrinkage | 5 percent of tooth height |
| Large-end shrinkage | 10 percent of tooth width |
| Small-end shrinkage | 10 percent of tooth width |
For each grid point, I compute the theoretical position vector $$r_g(i,j)$$ and the theoretical normal vector $$n_g(i,j)$$. The measured position vector on the actual tooth surface is $$R_v(i,j)$$. I choose the center point of the grid as the reference point, so that the measured point and the theoretical point coincide at that location and the deviation there is zero. The deviation at any grid point is the distance by which the actual tooth surface departs from the theoretical surface along the normal direction:
$$e(i,j) = [r_g(i,j) – R_v(i,j)] \cdot n_g(i,j)$$
where the normal vector is obtained from the cross product of the two partial derivatives of the theoretical surface:
$$n_g(i,j) = \frac{\partial r_g}{\partial \theta_c} \times \frac{\partial r_g}{\partial u_c}$$
This definition gives a signed deviation. Positive and negative values indicate opposite sides of the theoretical surface. I use this deviation field to quantify the effect of machine parameter errors on face gears.
Influence of wheel arm installation error. The wheel arm installation error $$\Delta y$$ mainly changes the pressure angle of the face gear tooth surface. When $$\Delta y$$ increases, the pressure angle on the left tooth side decreases, while the pressure angle on the right tooth side increases. When $$\Delta y$$ decreases, the opposite occurs: the left tooth side pressure angle increases, and the right tooth side pressure angle decreases. This behavior is important because it means that a single installation error can create an asymmetric deviation pattern between the two sides of the same face gear tooth. In my sensitivity analysis, I therefore treat $$\Delta y$$ as the primary control parameter for pressure angle correction.
Influence of disk wheel installation error. The disk wheel installation error $$\Delta s$$ mainly changes the helix angle or tooth trace direction of the face gear. When $$\Delta s$$ increases, the face gear tooth trace tends to take a right-hand orientation. When $$\Delta s$$ decreases, the tooth trace tends to take a left-hand orientation. This error does not primarily produce a symmetric pressure angle change; instead, it twists the contact pattern along the tooth width. For face gears, the tooth trace direction is closely related to the contact path and the load distribution, so controlling $$\Delta s$$ is essential for achieving a centered contact pattern.
Influence of wheel arm length error. The wheel arm length error $$\Delta l$$ mainly produces a twist deviation and also changes the pressure angle. The effect is strongest near the small end and the tooth root. When $$\Delta l$$ increases, the overall pressure angle of both left and right tooth sides increases. When $$\Delta l$$ decreases, the overall pressure angle of both sides decreases. Because $$\Delta l$$ affects both sides in the same direction for pressure angle while also introducing twist, it is an important parameter for correcting combined pressure angle and twist errors. Table 4 summarizes the qualitative effects of the three machine parameter errors on face gears.
| Machine parameter error | Dominant tooth surface effect | Directional behavior | Correction role |
|---|---|---|---|
| Delta y | Pressure angle deviation | Increase causes left pressure angle to decrease and right pressure angle to increase; decrease causes the opposite | Mainly used for asymmetric pressure angle correction |
| Delta s | Helix angle or tooth trace deviation | Increase produces a right-hand tooth trace; decrease produces a left-hand tooth trace | Mainly used for tooth trace and contact path correction |
| Delta l | Twist deviation and pressure angle deviation | Increase raises the overall pressure angle on both sides; decrease lowers it; strongest near the small end and root | Mainly used for twist and symmetric pressure angle correction |
Sensitivity matrix and correction calculation. The qualitative trends in Table 4 are useful, but for actual machine adjustment I need a quantitative correction. I therefore build a sensitivity matrix that relates the tooth surface deviation at each grid point to the three machine parameter errors. For each grid point (i,j), the sensitivity row is
$$S(i,j) = \begin{bmatrix}
\frac{\partial e(i,j)}{\partial \Delta y} & \frac{\partial e(i,j)}{\partial \Delta s} & \frac{\partial e(i,j)}{\partial \Delta l}
\end{bmatrix}$$
If the measured deviation vector is $$e_{measured}$$, then the correction vector for the three machine parameters is obtained from the least-squares relation
$$\Delta p = \begin{bmatrix}
\Delta y \\
\Delta s \\
\Delta l
\end{bmatrix} = -(S^T S)^{-1} S^T e_{measured}$$
This equation is the core of my machine parameter adjustment method. It uses the sensitivity of the face gear tooth surface to the three dominant errors and computes the correction amounts that best reduce the measured deviation. In practice, I may apply the correction iteratively because the relationship between the error parameters and the tooth surface deviation is not perfectly linear over large error ranges. However, for the small errors encountered after initial setup, the linear sensitivity model is sufficient and efficient.
Experimental setup for face gear grinding. I conducted grinding experiments on a cylindrical internal gear grinding machine. The machine configuration matches the kinematic model described above. The disk wheel was dressed to the involute profile of the generating gear. The face gear workpiece was mounted on the rotary table, and the generating motion was executed through the coordinated motions of the X, Y, Z, A, and B axes. After grinding, I measured the face gear on a Gleason gear inspection center. Table 5 lists the basic parameters of the face gear pair used in my experiment.
| Parameter | Value |
|---|---|
| Pinion tooth number Z_1 | 21 |
| Generating gear tooth number Z_c | 24 |
| Face gear tooth number Z_g | 120 |
| Shaft angle gamma | 90 degrees |
| Normal module m_n | 4.0 mm |
| Pressure angle alpha_k | 25 degrees |
| Face gear inner radius R_1 | 220 mm |
| Face gear outer radius R_2 | 270 mm |
The experimental procedure was as follows. First, I set the machine according to the theoretical generating motion and the nominal machine parameters. Second, I ground the face gear with the disk wheel. Third, I measured the tooth surface deviations on the Gleason inspection center. Fourth, I used the measured deviation field to calculate the corrections for $$\Delta y$$, $$\Delta s$$, and $$\Delta l$$. Fifth, I adjusted the machine parameters and ground the face gear again. Finally, I measured the corrected face gear and compared its deviation with the initial result. This procedure allows me to verify both the feasibility of grinding face gears on a general cylindrical internal gear grinding machine and the correctness of my machine parameter adjustment method.
Initial grinding results. The initial tooth surface deviation measurement showed that the grinding method could produce face gear teeth, but the deviation was relatively large. The left tooth surface had deviations from -87 micrometers to 176 micrometers. The right tooth surface had deviations from -98.6 micrometers to 166.8 micrometers. These deviations were mainly expressed as pressure angle and helix angle deviations. This pattern is consistent with my error analysis: the initial machine setup contained small installation errors and a wheel arm length error, and those errors produced pressure angle and tooth trace deviations across the face gear tooth surface. Table 6 summarizes the initial measurement.
| Tooth side | Minimum deviation (micrometers) | Maximum deviation (micrometers) | Dominant deviation type |
|---|---|---|---|
| Left | -87 | 176 | Pressure angle and helix angle |
| Right | -98.6 | 166.8 | Pressure angle and helix angle |
Correction amounts. Using the sensitivity matrix and the measured deviation field, I calculated the correction amounts for the three machine parameters. The corrections were $$\Delta y = 0.09$$ mm, $$\Delta s = 0.13$$ mm, and $$\Delta l = -0.32$$ mm. These values are small, which is typical for precision machine setup. The sign and magnitude of each correction correspond to the dominant deviation pattern. The correction of $$\Delta y$$ addresses the pressure angle asymmetry between the left and right tooth sides. The correction of $$\Delta s$$ addresses the tooth trace direction. The correction of $$\Delta l$$ addresses the twist and the overall pressure angle level. Table 7 lists the calculated corrections.
| Machine parameter | Calculated correction (mm) | Main purpose |
|---|---|---|
| Delta y | 0.09 | Reduce asymmetric pressure angle deviation |
| Delta s | 0.13 | Reduce tooth trace or helix angle deviation |
| Delta l | -0.32 | Reduce twist and overall pressure angle deviation |
Corrected grinding results. After applying the corrections, I ground and measured the face gear again. The corrected tooth surface deviations were significantly smaller. The left tooth surface deviation ranged from -24.6 micrometers to 23.8 micrometers. The right tooth surface deviation ranged from -21.4 micrometers to 17 micrometers. The maximum absolute deviation was 24.6 micrometers, and the maximum deviation was located away from the main contact region. This result confirms that the correction method is effective and that the proposed grinding method can produce face gears with acceptable precision on a general cylindrical internal gear grinding machine. Table 8 compares the initial and corrected deviations.
| Tooth side | Initial minimum (micrometers) | Initial maximum (micrometers) | Corrected minimum (micrometers) | Corrected maximum (micrometers) |
|---|---|---|---|---|
| Left | -87 | 176 | -24.6 | 23.8 |
| Right | -98.6 | 166.8 | -21.4 | 17 |
Interpretation of the experimental verification. The experimental results support three conclusions. First, the cylindrical internal gear grinding machine can execute the required generating motion for face gears when the disk wheel and the wheel carriage are properly controlled. Second, the mathematical model including machine parameter errors correctly predicts the dominant deviation patterns. Third, the sensitivity-based correction method can reduce the face gear tooth surface deviation to a small range. In my experiment, the maximum deviation after correction was 24.6 micrometers, which is a strong indication that the method is feasible for precision grinding of face gears. The fact that the maximum deviation moved away from the contact region is also favorable, because the contact performance of face gears is primarily determined by the central working region.
Comparison with existing face gear grinding routes. Existing precision manufacturing of face gears often relies on dedicated face gear grinding machines. Worm wheel grinding can provide high productivity, but the worm wheel is difficult to dress and the dressing process is inefficient. Single-tooth generating grinding with a disk wheel has a simpler dressing principle and is suitable for small-batch and medium-batch production, but it traditionally requires a machine that can realize the generating motion. My method makes use of a general cylindrical internal gear grinding machine. This machine already has the necessary translation and rotation axes, and with the correct interpolation model, it can reproduce the generating motion of face gears. Therefore, my method reduces the need for a dedicated face gear machine and lowers the manufacturing cost for face gears. Table 9 compares the main characteristics of my method with the general expectations of dedicated face gear grinding.
| Feature | Dedicated face gear grinding | My method on a cylindrical internal gear grinding machine |
|---|---|---|
| Machine requirement | Special face gear machine | General internal gear grinding machine |
| Tool type | Worm wheel or disk wheel | Disk-shaped involute wheel |
| Generating motion | Built into the dedicated machine | Realized by X, Y, Z, A, and B interpolation |
| Dressing principle | Can be complex for worm wheels | Relatively simple for a disk wheel |
| Production scale | Depends on the process | Suitable for small and medium batches |
| Cost implication | Higher due to special machine | Lower because general machine tools can be used |
Error analysis and machine setup strategy. My error analysis shows that the three machine parameter errors have distinct and partially separable effects on face gears. The wheel arm installation error $$\Delta y$$ is mainly an asymmetric pressure angle error. The wheel installation error $$\Delta s$$ is mainly a tooth trace error. The wheel arm length error $$\Delta l$$ is mainly a twist error combined with a symmetric pressure angle error. Because these effects are different, I can use a small number of measured grid points to estimate the corrections. In practice, I recommend the following setup strategy for face gears. First, set the machine according to the nominal generating parameters. Second, grind a test face gear. Third, measure the tooth surface deviation on a gear inspection center. Fourth, compute the sensitivity matrix for the measured grid. Fifth, solve for $$\Delta y$$, $$\Delta s$$, and $$\Delta l$$. Sixth, adjust the machine and grind the production face gear. This strategy is iterative if the initial deviations are large, but it converges quickly for small errors.
Mathematical consistency and theoretical error. An important point in my method is that the disk wheel is not an approximation of the generating gear; its axial profile is constructed as one tooth of the generating gear. The generating motion is also exact in the sense that the wheel carriage center follows the circular path required to swing the wheel about the generating gear center. Therefore, the proposed method has no theoretical generating error when the machine parameters are exact. The deviations observed in the experiment come from machine parameter errors, tool setting errors, thermal effects, wheel wear, and measurement uncertainty. This is why the error model and sensitivity correction are essential. The fact that the corrected maximum deviation is only 24.6 micrometers confirms that the theoretical model is consistent with the actual grinding process.
Practical advantages for face gears. The practical advantages of my method are significant for the推广 of face gears. A general cylindrical internal gear grinding machine is more widely available than a dedicated face gear grinding machine. The disk wheel is easier to dress than a worm wheel. The generating motion is implemented by standard multi-axis interpolation, which many modern machine tools already support. The sensitivity-based correction method provides a systematic way to improve accuracy without repeated trial and error. For face gears used in aviation and vehicle transmissions, this combination of flexibility and accuracy is valuable. It allows manufacturers to produce face gears in small batches, prototype new face gear designs, and reduce the capital cost associated with dedicated face gear grinding equipment.
Limitations and sources of uncertainty. I also recognize the limitations of my current study. The error model focuses on three dominant machine parameter errors, but other errors can also affect face gears. These include spindle thermal drift, wheel wear, workpiece clamping deformation, table motion errors, and angular positioning errors of the rotary axes. The sensitivity matrix is based on a linearized relationship, so large corrections may require iteration. The experiment was performed on one machine and one face gear parameter set; additional tests with different modules, tooth numbers, and face widths would be useful. The measurement uncertainty of the gear inspection center also contributes to the observed deviation, especially when the deviation is on the order of tens of micrometers. Despite these limitations, the experimental result establishes the feasibility of the method and provides a foundation for further refinement.
Future improvements. In future work, I plan to extend the error model to include more machine errors and tool errors. I would like to incorporate thermal compensation and wheel wear compensation into the sensitivity model. I also plan to study the optimal placement of measurement grid points for face gears, so that a small number of measurements can identify the dominant error parameters. Another direction is to integrate the correction algorithm directly into the machine controller, so that the operator can measure a test face gear and obtain correction values automatically. I believe that these improvements will make the grinding of face gears on general cylindrical internal gear grinding machines even more practical and accurate.
Detailed equations for the implemented correction. For completeness, I summarize the numerical correction workflow in equation form. The measured deviation vector is assembled from the grid points as
$$e_{measured} = \begin{bmatrix}
e(1,1) \\
e(1,2) \\
\vdots \\
e(5,9)
\end{bmatrix}$$
The sensitivity matrix is assembled as
$$S = \begin{bmatrix}
S(1,1) \\
S(1,2) \\
\vdots \\
S(5,9)
\end{bmatrix}$$
Each row contains the partial derivatives with respect to the three machine parameters. The correction vector is then computed by
$$\Delta p = -(S^T S)^{-1} S^T e_{measured}$$
After the correction is applied, the residual deviation is
$$e_{residual} = e_{measured} + S \Delta p$$
In an ideal linear case, the residual deviation would be zero. In practice, the residual is small but not zero because of nonlinearity and measurement noise. I therefore check the residual after correction and, if necessary, perform another iteration. For the experimental face gears, one correction step was sufficient to reduce the maximum deviation to 24.6 micrometers.
Tooth surface topology and contact relevance. The tooth surface topology of face gears is directly related to their contact behavior. Pressure angle errors shift the contact pattern along the tooth profile direction. Helix angle errors shift the contact pattern along the tooth width. Twist errors change the orientation of the contact pattern and can cause edge contact if they are not controlled. In my error analysis, the three machine parameter errors produce these three types of topology changes. Therefore, by correcting $$\Delta y$$, $$\Delta s$$, and $$\Delta l$$, I can control the most important features of the face gear tooth surface. The experimental results show that the corrected face gear has a centered and acceptable topology, with the maximum deviation away from the contact region. This is an important outcome for face gears because their load capacity and vibration behavior depend strongly on contact pattern placement.
Machine parameter sensitivity in table form. To make the sensitivity behavior easier to use, I can also express the expected deviation pattern in a qualitative matrix. Table 10 shows how each error affects the left and right sides of the face gear tooth at different regions. This table is not a replacement for numerical sensitivity, but it is useful for initial diagnosis of measured deviations.
| Error parameter | Left side pressure angle | Right side pressure angle | Tooth trace direction | Twist near small end or root |
|---|---|---|---|---|
| Delta y increases | Decreases | Increases | Weak effect | Weak effect |
| Delta y decreases | Increases | Decreases | Weak effect | Weak effect |
| Delta s increases | Weak effect | Weak effect | Right-hand tendency | Moderate effect |
| Delta s decreases | Weak effect | Weak effect | Left-hand tendency | Moderate effect |
| Delta l increases | Increases | Increases | Moderate effect | Increases |
| Delta l decreases | Decreases | Decreases | Moderate effect | Decreases |
Experimental validation of the mathematical model. The initial deviation measurement and the corrected deviation measurement together provide a validation of my mathematical model. The initial deviation was dominated by pressure angle and helix angle errors, which is exactly what the error model predicts for a combination of $$\Delta y$$, $$\Delta s$$, and $$\Delta l$$. The calculated corrections were 0.09 mm, 0.13 mm, and -0.32 mm. After applying these values, the left side deviation dropped from the range -87 to 176 micrometers to the range -24.6 to 23.8 micrometers. The right side deviation dropped from the range -98.6 to 166.8 micrometers to the range -21.4 to 17 micrometers. This improvement is substantial and confirms that the sensitivity-based correction works for face gears. The maximum absolute deviation of 24.6 micrometers is small enough for many practical face gear applications and demonstrates the feasibility of the proposed method.
Why the method is useful for face gear推广. The broader adoption of face gears has been limited by manufacturing cost and machine availability. My method directly addresses both issues. It does not require a dedicated face gear grinding machine. It uses a general cylindrical internal gear grinding machine, which is already used for internal gear production and is therefore more likely to be available in gear manufacturing shops. It uses a disk-shaped involute wheel, which is easier to dress than a worm wheel. It provides a mathematical model and a correction algorithm, so the process can be transferred to different face gear designs. These features make the method attractive for prototyping, small-batch production, and research on face gears. Over time, the method could also be extended to higher-volume production if the grinding cycle time and tool life are optimized.
Summary of the process chain. The complete process chain in my method is as follows. I define the face gear geometry and the generating gear. I calculate the generating ratio and the interpolated motions of the wheel carriage. I establish the disk wheel surface equation. I transform the wheel surface into the face gear coordinate system. I solve the meshing equation and projection equations to obtain the theoretical face gear tooth surface. I introduce the machine parameter errors into the transformation matrices. I compute the tooth surface deviation on a grid. I build the sensitivity matrix and solve for the correction values. I adjust the machine parameters. I grind the face gear with the disk wheel. I measure the face gear on a gear inspection center. I compare the measured deviation with the theoretical deviation. If the residual deviation is acceptable, the process is complete. If not, I repeat the correction step. This process chain is systematic and repeatable, which is important for precision manufacturing of face gears.
Conclusions drawn from my work. I have proposed and verified a method for grinding face gears on a general cylindrical internal gear grinding machine. The main conclusions are as follows. First, the kinematics of the cylindrical internal gear grinding machine can be organized to reproduce the generating motion of face gears. The wheel carriage swing and the Y and Z interpolation replace the dedicated generating motion of a special face gear machine. Second, the disk wheel surface equation and the coordinate transformation model provide a complete mathematical description of the generated face gear tooth surface. Third, the machine parameter error model, including the wheel arm installation error, the wheel installation error, and the wheel arm length error, explains the dominant tooth surface deviations. Fourth, the sensitivity matrix and least-squares correction method can calculate effective machine parameter corrections. Fifth, the experimental grinding and measurement results validate the method. The initial maximum deviation was reduced from more than 100 micrometers to 24.6 micrometers after correction. The corrected face gear has an acceptable tooth surface topology with the maximum deviation away from the contact region. Sixth, the method avoids the need for a dedicated face gear grinding machine, which is meaningful for the推广 and cost reduction of face gears.
Final perspective. In my view, the grinding of face gears on a general cylindrical internal gear grinding machine is a promising direction for precision face gear manufacturing. It combines the geometric accuracy of a generating process with the flexibility of a general multi-axis machine. The mathematical model and error correction method presented here can serve as a foundation for further research on face gear grinding, including thermal compensation, wheel wear compensation, in-situ measurement, and automated machine parameter identification. As face gears continue to find applications in aviation, vehicle, and industrial transmissions, methods that reduce manufacturing cost and increase machine availability will become increasingly important. My work shows that such methods are feasible and can produce face gears with small tooth surface deviations using a disk-shaped grinding wheel and coordinated machine interpolation.