Face gears occupy a distinctive position among geared transmissions because they combine a compact envelope, a low sensitivity to axial mounting displacement, and a comparatively light structural configuration. When power must be transmitted between intersecting axes, a face gear pair offers an advantage that a bevel gear pair cannot match, since the cylindrical pinion that meshes with the face gear carries no axial force. Despite these merits, the complex topography of the face gear tooth surface makes its manufacturing equipment and process chain cumbersome, and this difficulty has kept face gears from being adopted as widely as their kinematic advantages would suggest. In the work presented here, I address the precision control and the tooth surface formation mechanism of face gears produced by hobbing, and I develop a systematic modeling and analysis framework that links the kinematics of a six-axis CNC machine tool to the geometry of the generated surface.
The mathematical description of face gears was first placed on a firm theoretical footing through the classical theory of gearing, and a large body of subsequent research has extended the forming theory and the contact behavior of face gears toward many different configurations. Eccentric face gear pairs have been modeled to reveal their transmission principle, modified face gear surfaces produced by grinding have been analyzed for their contact performance, and bidirectional modification strategies have been established for very large face gear blanks, together with spline surfaces fitted from measured tooth coordinates. Computer-aided design approaches have also been introduced to raise the accuracy and the efficiency with which the face gear tooth surface can be solved. Early investigations concentrated primarily on form-generating methods, which established the mathematical model and the forming theory of the face gear. As numerical control technology matured, the generating method gradually became a principal direction for face gear manufacturing, because generating motion reproduces the spatial rolling of the gear and therefore produces conjugate flanks with a theoretical accuracy higher than that of purely form-generating processes.
Hobbing realizes the tooth surface by means of a conjugate rolling motion between the cutting tool and the blank. It offers high forming efficiency, good machining flexibility, and the possibility of achieving high accuracy even on hardened tooth surfaces. Coordinate systems for non-orthogonal face gear hobbing have been constructed to analyze the spatial motion relationship between the tool and the workpiece and the associated tooth surface formation mechanism, an improved rolling method based on a disk-shaped grinding wheel has been proposed to achieve high-precision modification of the face gear flank, and both milling and shaving approaches have been applied to face gear manufacturing. Taken together, however, most studies of face gear pairs focus on the generation of the special flank geometry, and research devoted to the modification of the face gear tooth surface remains comparatively scarce. To improve the machining accuracy and the error control of face gears, I propose an improved modeling and analysis method for face gear hobbing.
1. Six-Axis CNC Hobbing Model
In my formulation, the tooth surface of a face gear is obtained from the machining motion itself. The machining of a face gear comprises the rotation of the hob, the rotation of the blank, and the feed motion of the hob along the radial direction of the gear. In the kinematic scheme I adopt, the linear axes X, Y and Z represent the axial, the tangential, and the radial feed directions respectively, while the rotary axes A, B and C represent the tool spindle, the face gear spindle, and the workpiece fixture rotation axis. The hob and the face gear blank are mounted on the A axis and the B axis respectively. During machining, the C axis is first adjusted to a prescribed position, the hob is then moved to the correct location through the coordinated motion of the Y and Z axes, and finally the hob is driven along the X axis to perform the feed motion.
Two principal motions therefore coexist during machining, namely the feed motion and the rotational motion of the hob, and together they constitute a two-parameter enveloping system. When the hob rotates counterclockwise about the axis $Z_{so}$ through an angle $\phi_s$, the face gear blank rotates counterclockwise about the axis $Z_2$ through an angle $\phi_2$. Guided by this relationship, I construct the following coordinate frames: the frame $S_s$–$O_sX_sY_sZ_s$ rigidly connected to the hob center, the auxiliary frames $S_{sr}$–$O_{sr}X_{sr}Y_{sr}Z_{sr}$, $S_{so}$–$O_{so}X_{so}Y_{so}Z_{so}$, $S_{21}$–$O_{21}X_{21}Y_{21}Z_{21}$ and $S_{20}$–$O_{20}X_{20}Y_{20}Z_{20}$, and the frame $S_2$–$O_2X_2Y_2Z_2$ rigidly connected to the face gear.
| Symbol | Physical meaning | Role in the model |
|---|---|---|
| $S_s$ | Frame rigidly attached to the hob | Definition of the generating worm surface |
| $S_{sr},\,S_{so}$ | Auxiliary frames of the hob spindle | Transformation of the hob rotation angle $\phi_s$ |
| $S_{21},\,S_{20}$ | Auxiliary frames of the blank | Transformation of the blank rotation angle $\phi_2$ |
| $S_2$ | Frame rigidly attached to the face gear | Final expression of the generated flank |
| $X,\,Y,\,Z$ | Linear feed axes | Axial, tangential and radial feed |
| $A,\,B,\,C$ | Rotary axes | Tool, face gear and fixture rotation |
| $\phi_s$ | Hob rotation angle | First generating parameter |
| $\phi_2$ | Blank rotation angle | Second generating parameter |
| $l_s$ | Feed motion parameter | Independent kinematic variable |
2. Hob Tooth Surface Equations
According to the face gear machining principle, the generating gear is an involute cylindrical gear. The tooth profile of an involute hob in its axial section coincides with an involute curve, and for this reason I select an involute hob as the cutting tool. The left flank of the basic worm of the hob and its normal vector are expressed in the frame attached to the hob. Let $u_s$ be the tooth direction parameter, $\theta_s$ the involute parameter, $\theta_{0s}$ the reference angle of the base circle involute, $p$ the helix parameter, $\alpha$ the pressure angle and $r_{bs}$ the base circle radius of the worm. The left flank then reads
$$ \mathbf{r}_s(u_s,\theta_s)=\begin{bmatrix} r_{bs}\left[\sin(\theta_s+\theta_{0s}+pu_s)-\theta_s\cos(\theta_s+\theta_{0s}+pu_s)\right] \\[4pt] -r_{bs}\left[\cos(\theta_s+\theta_{0s}+pu_s)+\theta_s\sin(\theta_s+\theta_{0s}+pu_s)\right] \\[4pt] u_s \end{bmatrix} $$
and the corresponding normal vector is obtained from the cross product of the two first partial derivatives,
$$ \mathbf{n}_s(u_s,\theta_s)=\frac{\partial \mathbf{r}_s}{\partial u_s}\times\frac{\partial \mathbf{r}_s}{\partial \theta_s}=\begin{bmatrix} p\sin\theta_s-u_s\sin\alpha\cos\theta_s \\[4pt] p\cos\theta_s+u_s\sin\alpha\sin\theta_s \\[4pt] u_s\cos\alpha \end{bmatrix}. $$
Here the sign convention of the helix parameter is important, because $p>0$ corresponds to a right-handed worm whereas $p<0$ corresponds to a left-handed worm. The handedness directly influences the direction in which the contact lines sweep across the face gear flank, and therefore it also influences where singular points are likely to appear. Building on the hobbing motion described above, the face gear tooth surface $\mathbf{r}_2$ can be written as
$$ \mathbf{r}_2(u_s,\theta_s,\phi_s,\phi_2,l_s)=\mathbf{M}_{s2}\left(\phi_s,\phi_2,l_s\right)\cdot\mathbf{r}_s(u_s,\theta_s)\cdot $$
In this expression, $\mathbf{M}_{s2}$ denotes the coordinate transformation matrix that maps a point expressed in the frame $S_s$ attached to the hob into the frame $S_{20}$ attached to the face gear. Because the transformation depends simultaneously on $\phi_s$, $\phi_2$ and $l_s$, the resulting surface is a two-parameter envelope embedded in a three-parameter family.
When the meshing equations for the hobbing process are formulated, the motions of the A axis and the X axis of the numerical control machine tool must be treated as independent variables. Consequently, two meshing equations must be satisfied at the same time in addition to the tooth surface equation. These two meshing equations are
$$ f^{(1)}=\left(\frac{\partial \mathbf{r}_2}{\partial \theta_s}\times\frac{\partial \mathbf{r}_2}{\partial u_s}\right)\cdot\frac{\partial \mathbf{r}_2}{\partial \phi_2}=0, $$
$$ f^{(2)}=\left(\frac{\partial \mathbf{r}_2}{\partial \theta_s}\times\frac{\partial \mathbf{r}_2}{\partial u_s}\right)\cdot\frac{\partial \mathbf{r}_2}{\partial l_s}=0. $$
The simultaneous solution of the surface equation together with these two meshing equations determines the family of contact points that generates the working flank of the face gear. The two meshing equations are what distinguish my treatment from a single-parameter generating model, and they are also what make the analysis of interference and of singular points meaningful for the six-axis configuration.
3. Modification Design of the Hob
When a face gear is machined on a six-axis CNC machine tool, the hob behaves as a helical gear with a large helix angle and a small number of teeth. The hob contacts the blank at its end face during face gear machining, and this contact configuration differs fundamentally from the contact configuration encountered when a cylindrical gear is machined. As a consequence, when the basic worm of the hob performs a spatial conjugate motion with the face gear, pronounced interference arises at several locations near the inner radius of the face gear, and the interference causes excessive undercutting of the tooth flank. Further machining verification has shown that when a standard involute hob is used on a six-axis CNC machine tool, the flank becomes severely distorted by interference, both the addendum and the dedendum regions suffer shape warping, and the resulting geometry deviates considerably from the theoretical tooth profile.
To characterize this phenomenon quantitatively, I evaluate the velocity of a contact point on the face gear blank. Whether a singular point exists on the flank, and therefore whether undercutting will occur, can be judged from the velocity of the contact point. For an arbitrary contact point $m$, the velocity in the frame $S_2$ and the velocity in the frame $S_s$ satisfy
$$ \mathbf{v}_m^{(2)}=\mathbf{v}_m^{(s)}+\mathbf{v}_s^{(\phi_s)}\frac{d\phi_s}{dt}+\mathbf{v}_s^{(l_s)}\frac{dl_s}{dt}, $$
where the motion of the point relative to the hob frame itself decomposes into the two generating parameters according to
$$ \mathbf{v}_m^{(s)}=\frac{\partial \mathbf{r}_s}{\partial u_s}\frac{du_s}{dt}+\frac{\partial \mathbf{r}_s}{\partial \theta_s}\frac{d\theta_s}{dt}\cdot $$
In these relations, $\mathbf{v}_s^{(\phi_s)}$ denotes the angular velocity component of the velocity of point $m$ and $\mathbf{v}_s^{(l_s)}$ denotes the feed velocity component of that velocity, while $t$ stands for the machining time. A singular point appears on the tooth surface when the velocity of the contact point vanishes in the frame $S_2$, that is, when
$$ \mathbf{v}_m^{(2)}=\frac{\partial \mathbf{r}_s}{\partial u_s}\frac{du_s}{dt}+\frac{\partial \mathbf{r}_s}{\partial \theta_s}\frac{d\theta_s}{dt}+\mathbf{v}_s^{(\phi_s)}\frac{d\phi_s}{dt}+\mathbf{v}_s^{(l_s)}\frac{dl_s}{dt}=\mathbf{0}\cdot $$
Differentiating the two meshing equations with respect to each of the four variables $u_s$, $\theta_s$, $\phi_s$ and $l_s$ yields a homogeneous linear system in which four unknowns are connected by five equations,
$$ f^{(1)}_{u_s}\frac{du_s}{dt}+f^{(1)}_{\theta_s}\frac{d\theta_s}{dt}+f^{(1)}_{\phi_s}\frac{d\phi_s}{dt}+f^{(1)}_{l_s}\frac{dl_s}{dt}=0, $$
$$ f^{(2)}_{u_s}\frac{du_s}{dt}+f^{(2)}_{\theta_s}\frac{d\theta_s}{dt}+f^{(2)}_{\phi_s}\frac{d\phi_s}{dt}+f^{(2)}_{l_s}\frac{dl_s}{dt}=0. $$
Combining the singularity condition with these two differentiated meshing equations produces a system of five homogeneous relations among four unknowns. The coefficient matrix $\mathbf{M}_c$ of that system assembles the partial derivatives of the hob surface coordinates together with the velocity components and the meshing derivatives,
$$ \mathbf{M}_c=\begin{bmatrix} \dfrac{\partial x_s}{\partial u_s} & \dfrac{\partial x_s}{\partial \theta_s} & v_{xs}^{(\phi_s)} & v_{xs}^{(l_s)} \\[8pt] \dfrac{\partial y_s}{\partial u_s} & \dfrac{\partial y_s}{\partial \theta_s} & v_{ys}^{(\phi_s)} & v_{ys}^{(l_s)} \\[8pt] \dfrac{\partial z_s}{\partial u_s} & \dfrac{\partial z_s}{\partial \theta_s} & v_{zs}^{(\phi_s)} & v_{zs}^{(l_s)} \\[8pt] f^{(1)}_{u_s} & f^{(1)}_{\theta_s} & f^{(1)}_{\phi_s} & f^{(1)}_{l_s} \\[8pt] f^{(2)}_{u_s} & f^{(2)}_{\theta_s} & f^{(2)}_{\phi_s} & f^{(2)}_{l_s} \end{bmatrix}=\begin{bmatrix} \mathbf{x}_1 \\ \mathbf{x}_2 \\ \mathbf{x}_3 \\ \mathbf{x}_4 \\ \mathbf{x}_5 \end{bmatrix}, $$
where $v_{xs}^{(\phi_s)}$, $v_{ys}^{(\phi_s)}$, $v_{zs}^{(\phi_s)}$ and $v_{xs}^{(l_s)}$, $v_{ys}^{(l_s)}$, $v_{zs}^{(l_s)}$ are the components of the velocities $\mathbf{v}_s^{(\phi_s)}$ and $\mathbf{v}_s^{(l_s)}$ along the $x$, $y$ and $z$ directions respectively.
A non-trivial solution of the system exists when all five fourth-order determinants of the coefficient matrix vanish. Denoting the relevant determinants by $\Delta_1$, $\Delta_2$ and $\Delta_3$, the condition takes the compact form
$$ \Delta_1^2+\Delta_2^2+\Delta_3^2=0, $$
with the individual determinants built from the row vectors as $\Delta_1=\left[\mathbf{x}_1\;\mathbf{x}_2\;\mathbf{x}_3\;\mathbf{x}_4\right]^{T}$, $\Delta_2=\left[\mathbf{x}_2\;\mathbf{x}_3\;\mathbf{x}_4\;\mathbf{x}_5\right]^{T}$ and $\Delta_3=\left[\mathbf{x}_1\;\mathbf{x}_3\;\mathbf{x}_4\;\mathbf{x}_5\right]^{T}$. This condition can be regarded as a functional relation
$$ f(u_s,\theta_s,\phi_s,\phi_2,l_s)=0. $$
Solving this relation together with the tooth surface equation determines the position of the limit line $L$ on the hobbed flank. The limit line therefore reads
$$ \begin{cases} \mathbf{r}_s=\mathbf{r}_s(u_s,\theta_s), \\[4pt] f(u_s,\theta_s,\phi_s,\phi_2,l_s)=0. \end{cases} $$
Modifying the hob along the limit line $L$ provides an effective means of eliminating the undercutting that a standard hob would produce. The modification design procedure I follow consists of several stages. First, I derive the hob tooth surface equation and the limit line equation. Second, I compute the point cloud of the involute basic worm and, in a similar manner, the point cloud of the limit line $L$. Third, I import the basic worm point cloud into three-dimensional modeling software to build the hob model and to obtain the unmodified involute hob. Fourth, I constrain the hob geometry using the point cloud of the limit line and carry out the modification design, thereby producing the improved hob.
| Stage | Operation | Output |
|---|---|---|
| 1 | Derive the hob flank equation and the normal vector | Analytical surface description of the basic worm |
| 2 | Derive the limit line relation from the singularity condition | Functional relation $f(u_s,\theta_s,\phi_s,\phi_2,l_s)=0$ |
| 3 | Discretize the basic worm surface and the limit line | Two sets of point cloud data |
| 4 | Reconstruct the geometry in three-dimensional modeling software | Unmodified involute hob model |
| 5 | Apply the limit line point cloud as a constraint | Modified hob model |
4. Tooth Surface Comparison Model
Because the geometry of a face gear is complex, an accurate discrete modeling method is indispensable. In my implementation, the procedure proceeds as follows. I first input the basic parameters, namely the module $M$, the number of teeth of the generating gear $N_s$, the number of teeth of the face gear $N_2$, the pressure angle $\alpha$ and the shaft angle $\gamma$, and I solve for the associated intermediate parameters. I then obtain the tooth direction coordinate $y_i$ by discretization and determine the relation between the parameter $u_s$ in the profile equation and the tool rotation angle $\theta_s$. Next, using the addendum parameter $u_{\max}$ and the coordinate $y_2$, I solve the tooth surface equation for the transition curve coordinates. Finally, I substitute the pair $(y_2,z_2)$ into the tooth surface equation and solve for $x_2$. The resulting point cloud reveals both the working flank and the transition flank, and it makes the two characteristic defects of an unmodified design immediately visible, namely the sharpening of the tooth tip and the undercutting of the root region.
To verify the rationality of a flank machined with a modified hob, I evaluate its deviation from the ideal flank, which allows the consistency of the machined geometry with the theoretical tooth profile to be assessed and the feasibility of the machining method to be judged. In the actual machining of a face gear, machining errors cannot be avoided, and it is therefore necessary to analyze the influence of tool installation errors on the tooth surface. Since the involute hob simulates the conjugate generating motion between a virtual production gear and the face gear, the study of hob machining errors can be replaced by the study of the machining errors of the virtual production gear. When the virtual gear performs the conjugate generating motion with the face gear blank, the machining errors are mainly manifested as an angular error $\Delta\alpha$, an offset error $\Delta E$, a radial error $\Delta r$ and a displacement error $\Delta\sigma$. The tooth surface with the machining errors introduced is then written as
$$ \mathbf{r}_2′(\Delta E,\Delta r,\Delta\alpha,\Delta\sigma)=\mathbf{M}_{s2}\left(\Delta E,\Delta r,\Delta\alpha,\Delta\sigma\right)\mathbf{r}_s, $$
where the error-augmented coordinate transformation matrix takes the form
$$ \mathbf{M}_{s2}(\Delta E,\Delta r,\Delta\alpha,\Delta\sigma)=\begin{bmatrix} \cos\Delta\alpha\cos\Delta\sigma & \cos\Delta\alpha\sin\Delta\sigma & \sin\Delta\alpha & 0 \\[4pt] -\sin\Delta\sigma & \cos\Delta\sigma & 0 & \Delta r \\[4pt] -\sin\Delta\alpha\cos\Delta\sigma & -\sin\Delta\alpha\sin\Delta\sigma & \cos\Delta\alpha & \Delta E \\[4pt] 0 & 0 & 0 & 1 \end{bmatrix}. $$
This matrix is convenient because each error term appears in a separate position, so the influence of the individual errors can be isolated in the numerical experiments.
To verify the rationality of the flank produced by the modified hob, I assess the difference between the machined surface and the ideal surface, which indicates how well the machining method reproduces the theoretical tooth geometry. As described in my comparison model, the tooth surface equation and the meshing equations are obtained both for the flank machined by the modified hob and for the hypothetical flank. Using the point cloud extraction method, I solve for the point cloud data of both the hobbed flank and the ideal flank. I then discretize the two surfaces uniformly along the $X_2$ direction and the $Z_2$ direction with an interval length $\Delta d$, which yields $i\times j$ corresponding point pairs, and I record their coordinates $\left[X(i,j),Y(i,j),Z(i,j)\right]$. Finally, I compute the difference $\sigma(i,j)$ between the corresponding points of the two surfaces. The deviation between the hobbed flank and the hypothetical flank is therefore determined by the relation
$$ \sigma(i,j)=\left|\left(\mathbf{r}_2-\mathbf{r}_2’\right)\cdot\mathbf{n}_2\right|, $$
in which $\mathbf{n}_2$ denotes the normal vector of the face gear tooth surface. The scalar projection of the difference vector onto the surface normal removes the tangential component of the point cloud mismatch arising from the discretization and retains only the physically meaningful normal deviation.
| Quantity | Definition | Purpose in the comparison |
|---|---|---|
| $\Delta d$ | Uniform discretization interval | Generates the corresponding point grid |
| $i,\,j$ | Grid indices along the two directions | Addresses each corresponding point pair |
| $X(i,j),\,Y(i,j),\,Z(i,j)$ | Recorded coordinates of the point pairs | Geometric input to the deviation computation |
| $\sigma(i,j)$ | Normal deviation between the two surfaces | Quantifies the machining consistency |
| $\mathbf{n}_2$ | Normal vector of the ideal flank | Projects the deviation onto the normal direction |
5. Computational Example and Parameters
To validate the model I have developed, I carried out a representative computation. The parameters of the face gear pair used in the example are collected in Table 4. The module is 3.175 mm, the number of teeth of the generating gear is 28, the number of teeth of the face gear is 160, the number of teeth of the cylindrical pinion is 26, the pressure angle is 27.5 degrees, and the shaft angle is 90 degrees. These values correspond to a realistic face gear pair in which the face gear is considerably larger than the mating pinion, a configuration typical of the applications for which face gears are attractive.
| Parameter | Symbol | Value |
|---|---|---|
| Module | $M$ | 3.175 mm |
| Number of teeth of the generating gear | $N_s$ | 28 |
| Number of teeth of the face gear | $N_2$ | 160 |
| Number of teeth of the cylindrical pinion | $N_1$ | 26 |
| Pressure angle | $\alpha$ | 27.5° |
| Shaft angle | $\gamma$ | 90° |
Based on the hob modification method described above, I designed and manufactured a hob intended for modification and then conducted hobbing experiments on a numerical control machine tool. The experimental procedure produced a clear improvement over the flank obtained before modification, and the surface quality was significantly enhanced. I then computed the difference between the machined flank and the theoretical flank. The results show that the deviation of the machined face gear flank increases in both its upper and its lower bounds as the pressure angle of the hob increases. For practical machining, therefore, a smaller pressure angle should be selected in order to improve the tooth surface accuracy.
| Pressure angle $\alpha$ | Observed tendency of the deviation band | Recommended use |
|---|---|---|
| 25.0° | Narrowest deviation band | Preferred for high-accuracy face gears |
| 27.5° | Intermediate deviation band | Balanced choice for general production |
| 30.0° | Widest deviation band | Less favorable for precision control |
6. Influence of Machining Errors on the Face Gear Flank
Following the error formulation established earlier, I selected ten uniformly distributed discrete points along the tooth width direction of the face gear and examined how the angular error $\Delta\alpha$ and the displacement error $\Delta\sigma$, varied from 0.02 rad to 0.10 rad, together with the offset error $\Delta E$ and the radial error $\Delta r$, varied from 0.02 mm to 0.10 mm, affect the difference between the machined flank and the ideal flank at those discrete points. This parametric study provides the quantitative basis on which machining tolerances can be assigned.
The results demonstrate that as the amplitude of each type of machining error increases, the flank deviation at the discrete points rises continuously, which indicates that machining errors cause an accumulation of the overall flank deviation. As the angular error $\Delta\alpha$ grows, the growth of the flank deviation gradually accelerates, and the outer radius position along the tooth width is more sensitive to the introduction of the angular error. For the offset error $\Delta E$, the resulting flank deviation is distributed relatively stably along the tooth width direction and increases in an approximately linear manner with the error amplitude. The flank deviation caused by the radial error $\Delta r$ is significantly lower than that caused by the other machining errors, which shows that its sensitivity in the flank forming process is weak and that its influence on the flank error is limited. The displacement error $\Delta\sigma$ has the most significant influence on the flank deviation. The highest point of the flank deviation occurs at the discrete point located at the outer radius of the gear, and the maximum flank deviation can reach 0.46 mm. This demonstrates that the displacement error $\Delta\sigma$ is the dominant factor affecting the machining accuracy of face gears, and special attention should therefore be paid to the possible influence of the displacement error $\Delta\sigma$ on the flank during machining.
| Error type | Symbol | Range examined | Distribution along tooth width | Sensitivity |
|---|---|---|---|---|
| Angular error | $\Delta\alpha$ | 0.02–0.10 rad | Accelerating growth, strongest at the outer radius | High |
| Offset error | $\Delta E$ | 0.02–0.10 mm | Stable, approximately linear growth | Moderate |
| Radial error | $\Delta r$ | 0.02–0.10 mm | Low and weakly varying | Low |
| Displacement error | $\Delta\sigma$ | 0.02–0.10 rad | Maximum at the outer radius | Dominant |
The numerical magnitudes deserve closer inspection. For the offset error, the flank deviation remains confined to a relatively narrow band across the entire set of amplitudes examined, and the corresponding maximum values stay in the range of approximately $1.3\times10^{-4}$ mm to $1.3\times10^{-3}$ mm, depending on the amplitude and on the position of the reference point. For the radial error, the flank deviation occupies an intermediate range that reaches about 0.60 mm at the largest amplitude considered, but the growth with amplitude is gradual and the spatial distribution is comparatively uniform. For the angular error, the deviation band extends from roughly 0.080 mm to 0.160 mm over the same set of amplitudes, and the spread between the lowest and the highest value at a fixed amplitude reflects the strong dependence on tooth width position. For the displacement error, the deviation reaches its maximum of about 0.46 mm at the outermost reference point, which places this error in a category of its own.
| Reference point index along tooth width | Trend for $\Delta E$ | Trend for $\Delta r$ | Trend for $\Delta\alpha$ | Trend for $\Delta\sigma$ |
|---|---|---|---|---|
| 1 (inner region) | Low, nearly constant | Low | Moderate | Moderate |
| 3 | Low, slowly rising | Low | Moderate | Steadily rising |
| 5 (mid region) | Intermediate | Intermediate | Rising | Rising faster |
| 7 | Intermediate | Intermediate | Rising faster | High |
| 9 | Higher | Higher | High | Very high |
| 10 (outer region) | Highest of the $\Delta E$ series | Highest of the $\Delta r$ series | Highest of the $\Delta\alpha$ series | Maximum of all series |
The physical interpretation of these trends follows from the geometry of the generating motion. The displacement error $\Delta\sigma$ acts as a rotational perturbation of the virtual production gear relative to the blank, and its effect is amplified by the distance from the axis of rotation. Points at the outer radius therefore experience the largest normal deviation, which explains why the maximum deviation occurs at the outermost reference point and why the value can reach 0.46 mm. The angular error $\Delta\alpha$ tilts the tool axis relative to its nominal orientation, and the resulting deviation grows with the local radius in a manner similar to that of the displacement error but with a smaller coefficient. The offset error $\Delta E$ translates the tool in a fixed direction, so its effect is nearly uniform along the tooth width and varies almost linearly with the error amplitude. The radial error $\Delta r$ displaces the tool along the radial direction of the gear, a direction that is nearly tangent to the generated flank in the region of interest, and this geometric alignment is the reason why the radial error produces the smallest normal deviation of all the errors examined.
| Ranking | Error | Principal consequence for the face gear flank | Control priority |
|---|---|---|---|
| 1 | $\Delta\sigma$ | Largest normal deviation, concentrated at the outer radius | Highest |
| 2 | $\Delta\alpha$ | Strong dependence on tooth width position | High |
| 3 | $\Delta E$ | Nearly uniform offset of the whole flank | Moderate |
| 4 | $\Delta r$ | Weak influence on the normal deviation | Lower |
It is worth noting that the ranking above is not simply a ranking of the amplitude of the error but a ranking of the coupling between each error and the direction normal to the generated face gear flank. An error that displaces the tool predominantly in a direction tangent to the flank is largely absorbed by the generating motion and leaves comparatively little trace in the final geometry, whereas an error that displaces the tool along the normal direction is imprinted almost directly onto the flank. This observation explains the surprising result that the radial error, which is comparable in magnitude to the other translational errors, produces the smallest deviation. It also explains why the displacement error, which is rotational rather than translational, produces by far the largest deviation, because its effect grows with radius and therefore accumulates rather than cancels across the tooth width.
7. Verification of the Generated Geometry
To confirm the validity of the entire modeling chain, I compared the flank obtained with the modified hob against the ideal flank over the full working region. The comparison used the normal deviation defined earlier, evaluated on a uniform grid with interval $\Delta d$ in both the $X_2$ and the $Z_2$ directions. The result confirmed that the hobbed face gear flank possesses high geometric precision and that the forming method is geometrically sound. The comparison also confirmed that the modification of the hob along the limit line removes the singular points that would otherwise appear near the inner radius, so that the transition between the working flank and the transition flank becomes smooth rather than distorted.
Two design recommendations follow directly from the verification. First, selecting a smaller pressure angle for the hob narrows the deviation band between the machined flank and the theoretical flank, and it therefore improves the achievable machining accuracy of the face gear. Second, restricting the tooth width to a reasonable range avoids the outermost region where the sensitivity to the displacement error is greatest, and it therefore reduces the maximum deviation that must be tolerated.
| Assessment criterion | Observation with the modified hob | Observation with the standard hob |
|---|---|---|
| Presence of singular points near the inner radius | Suppressed along the limit line | Present, causing severe distortion |
| Addendum region | Regular, close to the theoretical profile | Warped |
| Dedendum region | Regular, close to the theoretical profile | Warped |
| Consistency with the ideal flank | High | Low |
| Sensitivity to the displacement error | Reduced but still dominant | Aggravated |
8. Discussion
The model I have presented treats the face gear hobbing process as a two-parameter enveloping problem embedded in the six-axis kinematics of a modern machine tool. This treatment has several consequences for the analysis of face gears. First, it makes the meshing conditions explicit as a pair of scalar equations that must hold simultaneously, which is a necessary condition for any analysis that seeks to identify singular points on the generated flank. Second, it supplies a computable criterion, expressed through the determinant condition, for locating the limit line on the hob. Third, it connects the limit line directly to a modification design procedure, so that the theoretical criterion is translated into a manufacturable tool geometry.
The comparison model complements the generating model by quantifying how closely the machined face gear flank approaches the ideal flank and how that closeness degrades in the presence of installation errors. By separating the four error types and by projecting the resulting surface mismatch onto the normal direction, the model produces a ranking that can be used to allocate manufacturing tolerances rationally. Because the displacement error dominates the ranking by a wide margin, the practical implication for the assembly and operation of face gears is that the rotational alignment of the tool and of the workpiece must be controlled far more tightly than the translational alignment.
The pressure angle also emerges as a design variable that affects accuracy. A smaller pressure angle narrows the deviation band, which indicates that the choice of the pressure angle should not be made solely on the basis of load capacity and contact stress considerations but should also account for the attainable geometric accuracy of the generated flank. In the example studied here, the pressure angle of 25.0 degrees produced the narrowest deviation band, the pressure angle of 27.5 degrees produced an intermediate band, and the pressure angle of 30.0 degrees produced the widest band.
| Modeling element | Formulation | Analytical benefit |
|---|---|---|
| Surface family | $\mathbf{r}_2=\mathbf{M}_{s2}\mathbf{r}_s$ | Complete description of the generated flank |
| Meshing conditions | $f^{(1)}=0,\;f^{(2)}=0$ | Simultaneous treatment of the A and X axis motions |
| Singularity criterion | $\mathbf{v}_m^{(2)}=\mathbf{0}$ | Detection of undercutting and distortion |
| Limit line | $\Delta_1^2+\Delta_2^2+\Delta_3^2=0$ with $f=0$ | Basis for the hob modification |
| Error surface | $\mathbf{r}_2’=\mathbf{M}_{s2}^{err}\mathbf{r}_s$ | Isolation of individual error contributions |
| Deviation measure | $\sigma=\left|(\mathbf{r}_2-\mathbf{r}_2′)\cdot\mathbf{n}_2\right|$ | Quantitative accuracy assessment |
9. Conclusions
I established a hobbing model for face gears on a six-axis CNC machine tool, derived the coordinate transformation relations that govern the generating motion of the face gear, and obtained the corresponding coordinate transformation matrices. On the basis of the requirement to restrict the appearance of singular points on the face gear flank, I proposed a modification design method for the hob. The method proceeds from the velocity condition at a contact point, through the determinant criterion for a non-trivial solution of the differentiated meshing system, to the limit line that defines the modification constraint, and it produces a tool geometry that suppresses the undercutting that a standard involute hob would introduce near the inner radius of the face gear.
I further proposed a tooth surface comparison analysis method and applied it to compare the flank produced by the modified hobbing process with the ideal flank. The comparison showed that the face gear flank obtained by the hobbing method possesses high geometric precision and that the forming method is geometrically reasonable. The choice of a smaller pressure angle together with a reasonable tooth width range effectively improves the machining accuracy of the flank.
With respect to error sensitivity, the analysis of the four installation error types demonstrated that the displacement error $\Delta\sigma$ is the principal factor affecting the machining accuracy of the face gear, and it should therefore be the focus of attention during face gear hobbing. The angular error $\Delta\alpha$ is the second most important factor, and its influence grows with the radius and is therefore most pronounced at the outer region of the tooth width. The offset error $\Delta E$ produces a nearly uniform and approximately linear deviation along the tooth width, while the radial error $\Delta r$ is the least influential of the four because its direction is largely tangent to the generated flank. Taken together, these findings provide a rational basis for assigning tolerances and for selecting process parameters when face gears are manufactured by hobbing on a six-axis numerical control machine tool.
| Conclusion | Content | Practical implication |
|---|---|---|
| Kinematic model | A six-axis hobbing model with a two-parameter envelope was established | Provides the transformation matrices for the generating motion |
| Hob modification | A modification method based on the limit line was proposed | Restricts singular points and improves the machined flank |
| Comparison analysis | A machined-versus-ideal surface comparison method was developed | Quantifies consistency and supports feasibility judgments |
| Pressure angle | Smaller pressure angles yield narrower deviation bands | Suggests selecting a smaller pressure angle for accuracy |
| Error control | The displacement error $\Delta\sigma$ dominates the deviation | Requires the strictest control during assembly and operation |