I have focused my research on the continuous generating grinding of helical gears that carry longitudinal crowning. In this process, the worm wheel and the helical gear form a crossed-axis meshing pair. Although generating grinding offers high efficiency and high accuracy, it introduces a characteristic geometric error known as tooth surface distortion when the helical gear is modified along the lead direction. This distortion appears because the instantaneous contact lines on the left and right flanks are asymmetrical, so the same worm wheel position removes different amounts of material along the profile. As a result, the pressure angle varies across the face width, and the helical gear loses part of its intended transmission quality. To solve this problem, I established a spatial meshing model, derived the meshing equation, built a distortion model, proposed an arc-shaped worm wheel dressing method, verified the method through digital twin simulation, and finally validated it through grinding experiments and profile measurements.
1. Meshing Analysis of Worm Wheel Grinding for Helical Gears
I treated the worm wheel grinding process as a pair of crossed helical gears in mesh. The worm wheel rotates about its own axis and simultaneously moves along that axis. The helical gear rotates about its axis and also moves along its axis to complete the full face width. The transmission ratio is determined by the number of worm starts and the number of helical gear teeth. The center distance changes when longitudinal crowning is applied, so the meshing condition must be written as a function of the face width position.
I established the fixed coordinate system \(S_0(O_0x_0y_0z_0)\) and the moving coordinate systems \(S_1(O_1x_1y_1z_1)\) and \(S_2(O_2x_2y_2z_2)\) for the worm wheel and the helical gear, respectively. The angle between the two axes is \(\Sigma\). The worm wheel surface is an involute helicoid. Its parametric equation in \(S_1\) is
$$
\begin{aligned}
x_1 &= r_b \cos(\theta+\varphi_1+\sigma_0) + r_b (u+\sigma_0)\sin(\theta+\varphi_1+\sigma_0), \\
y_1 &= r_b \sin(\theta+\varphi_1+\sigma_0) – r_b (u+\sigma_0)\cos(\theta+\varphi_1+\sigma_0), \\
z_1 &= P_1 \theta,
\end{aligned}
$$
where \(r_b\) is the base circle radius, \(u\) and \(\theta\) are the surface parameters, \(\varphi_1\) is the worm wheel rotation angle, \(\sigma_0\) is the involute starting angle, and \(P_1\) is the worm wheel helix parameter. The helix parameter is
$$
P_1 = \frac{m_n n}{2\cos\lambda},
$$
where \(m_n\) is the normal module, \(n\) is the number of worm starts, and \(\lambda\) is the lead angle. For the helical gear, the helix parameter is
$$
P_2 = \frac{m_n z}{2\sin\beta},
$$
where \(z\) is the number of helical gear teeth and \(\beta\) is the helix angle. The shaft angle satisfies
$$
\Sigma = \frac{\pi}{2} \pm \beta \pm \lambda.
$$
I then derived the coordinate transformation matrix from \(S_1\) to \(S_2\):
$$
\mathbf{r}_2 = \mathbf{M}_{2p}(\varphi_2)\mathbf{M}_{p0}\mathbf{M}_{01}(\varphi_1)\mathbf{r}_1 = \mathbf{M}_{21}\mathbf{r}_1.
$$
The relative velocity between the worm wheel and the helical gear at a contact point is
$$
\mathbf{v}^{(12)} = \mathbf{v}^{(1)} – \mathbf{v}^{(2)}.
$$
The meshing equation for the generating contact is
$$
\mathbf{v}^{(12)} \cdot \mathbf{n} = 0,
$$
where \(\mathbf{n}\) is the normal vector of the worm wheel surface at the contact point. Because the process is a double-parameter envelope, the meshing equation can be separated into two independent conditions:
$$
\begin{aligned}
U_1 \cos\varphi_1 – V_1 \sin\varphi_1 &= W_1, \\
U_2 \cos\varphi_1 – V_2 \sin\varphi_1 &= W_2,
\end{aligned}
$$
where \(U_1, V_1, W_1\) and \(U_2, V_2, W_2\) are functions of the worm wheel surface parameters and the helical gear parameters. Solving these equations yields a unique pair \((u,\theta)\) for a given rotation angle \(\varphi_1\). By traversing the end profile from the root to the tip, I obtained the instantaneous contact lines on the left and right flanks of the helical gear.
For longitudinal crowning, the center distance is no longer constant. I expressed the modification amount as
$$
x_i = L(z_i),
$$
where \(z_i\) is the face width position and \(x_i\) is the corresponding radial modification. Substituting this relation into the meshing equation gives the crowning-modified meshing condition:
$$
\begin{aligned}
& \left[ -\frac{v_{01}}{\omega_1} + z_1 \right] n_x + \left[ -\sin\Sigma – a + L(z_i) \right] n_y + \left[ \cos\Sigma \right] n_z = 0, \\
& \left[ -\frac{v_{01}}{\omega_1} + z_1 \right] n_y + \left[ \sin\Sigma – a + L(z_i) \right] n_x + \left[ \cos\Sigma \right] n_z = 0.
\end{aligned}
$$
The contact line axial distance from the pitch line to the start and end of the instantaneous contact line is
$$
L_f = \left( \sqrt{r^2 – r_b^2} – \sqrt{r_f^2 – r_b^2} \right) \tan\beta_b,
$$
$$
L_a = \left( \sqrt{r_a^2 – r_b^2} – \sqrt{r^2 – r_b^2} \right) \tan\beta_b,
$$
where \(r\) is the pitch circle radius, \(r_b\) is the base circle radius, \(r_f\) is the involute start radius, \(r_a\) is the involute end radius, and \(\beta_b\) is the base helix angle. These lengths directly influence the distortion magnitude. A longer contact line produces a larger distortion because the modification amount varies more strongly along the line.
| Parameter | Symbol | Value |
|---|---|---|
| Number of helical gear teeth | \(z\) | 50 |
| Normal module | \(m_n\) | 3 mm |
| Normal pressure angle | \(\alpha_n\) | 20° |
| Transverse pressure angle | \(\alpha_t\) | 22.8° |
| Helix angle | \(\beta\) | 30° |
| Face width | \(b\) | 30 mm |
| Crowning amount | \(C_r\) | 27.5 µm |
2. Tooth Surface Distortion Mechanism
I analyzed the distortion mechanism from the viewpoint of the instantaneous contact line. During generating grinding, the worm wheel contacts the helical gear along a spatial curve. All points on the same contact line receive the same radial modification because the worm wheel position is fixed for that instant. However, the helical gear end profile is generated by a sequence of different contact lines. Each contact line corresponds to a different face width position and therefore to a different crowning amount. This contradiction produces the distortion. The left and right flanks are affected in opposite directions, so the helical gear shows a symmetric but opposite pressure angle deviation on its two flanks.
I defined the height difference between the actual modification curve and the theoretical modification curve as
$$
S_v = \frac{\pi m_n}{4} \sin\beta_b.
$$
The base helix angle is related to the pitch helix angle by
$$
\cos\beta_b = \frac{\cos\beta \cos\alpha_n}{\cos\alpha_t}.
$$
For a crowned helical gear, the theoretical modification curve is
$$
f(z) = \frac{C_r}{(0.4b)^2} z^2,
$$
where \(z\) is measured from the center of the face width. The actual modification on the left and right flanks becomes
$$
f(z_k) = \frac{C_r}{(0.4b)^2} z_k^2 \pm S_v,
$$
where the plus sign applies to one flank and the minus sign to the other. The profile deviation at face width position \(k\) is then
$$
f_{H\alpha k} = f(z_k) – f(z_1),
$$
and the corresponding pressure angle deviation is
$$
\Delta\alpha_k = -\frac{f_{H\alpha k}}{L_{ea} \tan\alpha_t},
$$
where \(L_{ea}\) is the involute length and \(\alpha_t\) is the transverse pressure angle. This equation shows that the pressure angle deviation changes continuously along the face width. It also shows why the left and right flanks of the same helical gear have opposite signs.
I studied the influence of the base helix angle and the normal module on the height difference. The results are summarized in the following table. The height difference increases with both the base helix angle and the normal module, which means that the distortion becomes more severe for helical gears with large helix angles and large modules.
| Base helix angle \(\beta_b\) | Normal module \(m_n\) | Height difference \(S_v\) |
|---|---|---|
| 20° | 2 mm | 0.54 µm |
| 20° | 3 mm | 0.81 µm |
| 20° | 4 mm | 1.07 µm |
| 30° | 2 mm | 0.79 µm |
| 30° | 3 mm | 1.18 µm |
| 30° | 4 mm | 1.57 µm |
| 40° | 2 mm | 1.01 µm |
| 40° | 3 mm | 1.52 µm |
| 40° | 4 mm | 2.02 µm |
The profile deviation also depends on the crowning amount and the helix angle. For a fixed face width position, a larger crowning amount or a larger helix angle produces a larger profile deviation. The sign of the deviation reverses from the top end to the bottom end of the face width. On the left flank, the profile deviation changes from positive to negative, while on the right flank it changes from negative to positive. This opposite behavior is the core difficulty of distortion compensation for a helical gear.
| Face width position | Left flank deviation | Right flank deviation |
|---|---|---|
| Top 1/10 position | Positive | Negative |
| Center position | Near zero | Near zero |
| Bottom 1/10 position | Negative | Positive |
3. Anti-Distortion Compensation by Arc Worm Wheel Dressing
I proposed a compensation method that dresses the worm wheel into an arc shape along its axial direction. The arc shape changes the effective pressure angle of the worm wheel in a controlled manner. When the worm wheel moves along the helical gear axis, different arc sections engage the helical gear. Because the arc radius varies along the worm wheel axis, the pressure angle correction also varies. This variation is designed to cancel the natural distortion of the helical gear.
The conventional method uses a diamond roller that is tilted by an additional rotary axis. That approach requires high dynamic accuracy and can introduce vibration during dressing. In contrast, my method keeps the diamond roller in a more stable orientation and introduces the compensation through the axial profile of the worm wheel. The arc radius is calculated from the required pressure angle deviation at each face width position. The relationship between the compensation length along the worm wheel and the face width position is
$$
\Delta W_k = W \frac{\Delta Z_k}{B},
$$
where \(\Delta W_k\) is the compensation length on the worm wheel, \(W\) is the total usable compensation length, \(\Delta Z_k\) is the measurement position along the face width, and \(B\) is the measurement face width. The pressure angle deviation and the compensation length are related by
$$
\Delta W_k = R_k \cos\alpha_k – r \cos\lambda,
$$
where \(R_k\) is the arc radius in the normal plane of the helical gear, \(\alpha_k\) is the local pressure angle, \(r\) is the pitch radius, and \(\lambda\) is the lead angle. Combining these equations gives the arc radius:
$$
R_k = r + \frac{W \Delta Z_k \cos\lambda}{B \Delta\alpha_k \cos\alpha_k}.
$$
Because the diamond roller dresses the worm wheel in the axial plane, I projected the normal plane radius onto the axial plane. Using the Baxter curvature relation, I obtained
$$
R’_k = \frac{R_k R_w \cos\lambda}{R_k + R_w \cos(90^\circ – \lambda)} = \frac{R_k R_w \cos\lambda}{R_k + R_w \sin\lambda},
$$
where \(R_w\) is the nominal worm wheel radius. This axial radius is the value used by the NC dressing program. The worm wheel is dressed with a spline curve that follows these calculated radii. The diamond roller moves along the worm wheel axis while the radial infeed axis follows the spline data. The result is an arc-shaped worm wheel whose pressure angle changes smoothly along its usable length.
| Face width position \(\Delta Z_k\) (mm) | Pressure angle deviation \(\Delta\alpha_k\) (rad) | Compensation length \(\Delta W_k\) (mm) | Normal arc radius \(R_k\) (mm) | Axial arc radius \(R’_k\) (mm) |
|---|---|---|---|---|
| 12 | -0.00628 | 30 | 4386.8 | 3921.4 |
| 10 | -0.00532 | 25 | 4322.9 | 3872.6 |
| 8 | -0.00435 | 20 | 4230.6 | 3798.5 |
| 6 | -0.00338 | 15 | 4085.3 | 3682.1 |
| 4 | -0.00242 | 10 | 3823.4 | 3468.9 |
| 2 | -0.00145 | 5 | 3209.7 | 2954.3 |
| 0 | -0.00048 | 0 | 86.5 | 86.5 |
| -2 | 0.00049 | -5 | 9190.4 | 7815.2 |
| -4 | 0.00145 | -10 | 6243.0 | 5521.7 |
| -6 | 0.00242 | -15 | 5643.4 | 5041.8 |
| -8 | 0.00338 | -20 | 5385.3 | 4832.5 |
| -10 | 0.00435 | -25 | 5241.7 | 4715.9 |
| -12 | 0.00532 | -30 | 5150.2 | 4642.7 |
I simulated the grinding process with the arc-shaped worm wheel. The simulation compared the tooth flank deviations before and after compensation. The results showed that the compensation reduced the lead deviation at the tip and root of the helical gear by more than 94%. For a measurement position at 12 mm from the center, the left flank profile deviation dropped from 47.5 µm to 2.3 µm, a reduction of 95.2%. The right flank profile deviation dropped from -40.2 µm to -1.5 µm, a reduction of 96.3%. These values confirm that the proposed arc dressing method can effectively cancel the distortion of the helical gear.
| Flank | Tip lead deviation before (µm) | Tip lead deviation after (µm) | Root lead deviation before (µm) | Root lead deviation after (µm) |
|---|---|---|---|---|
| Left | 34.7 | 1.9 | -52.9 | -1.8 |
| Right | -34.7 | -1.9 | 52.9 | 1.8 |
4. Machine Tool Kinematics and Digital Twin Simulation
I built a digital twin of the worm wheel grinding machine to verify the arc dressing motion. The machine has fourteen axes, including thirteen motion axes and one virtual axis. The main axes are the radial infeed axis \(X\), the tangential infeed axis \(Y\), the axial stroke axis \(Z\), the worm wheel rotation axis \(B\), and the helical gear rotation axes \(C1\) and \(C2\). Additional axes include the wheelhead swivel axis \(A\), the workpiece spindle \(C3\), the diamond roller swivel axis \(C4\), the diamond roller rotation axis \(B2\), the tailstock axes \(W1\) and \(W2\), the oil nozzle axis \(U\), and the virtual axis \(YM\).
| Axis | Function | Type |
|---|---|---|
| \(X\) | Radial infeed of the worm wheel | Linear |
| \(Y\) | Tangential shift along the worm wheel axis | Linear |
| \(Z\) | Axial stroke along the helical gear axis | Linear |
| \(A\) | Swivel of the worm wheel head | Rotary |
| \(B\) | Rotation of the worm wheel | Rotary |
| \(C1\) | Rotation of workpiece 1 | Rotary |
| \(C2\) | Rotation of workpiece 2 | Rotary |
| \(C3\) | Workpiece spindle indexing | Rotary |
| \(C4\) | Diamond roller swivel | Rotary |
| \(B2\) | Diamond roller rotation | Rotary |
| \(W1\) | Tailstock of workpiece 1 | Linear |
| \(W2\) | Tailstock of workpiece 2 | Linear |
| \(U\) | Oil nozzle positioning | Linear |
| \(YM\) | Virtual compensation axis | Virtual |
I used a three-dimensional model of the machine tool and imported it into a machine builder environment. Each moving axis was represented by a coordinate transformation node and a motion node. The kinematic chain was built with \(C1\) as the reference axis. The worm wheel chain was connected through \(X\), \(Z\), \(A\), \(Y\), and \(B\). The workpiece chain was connected through \(C3\), \(C1\), \(C2\), \(C4\), \(W1\), and \(W2\). The diamond roller was attached to \(B2\), which is a child of \(C4\). The virtual axis \(YM\) was superimposed on the \(Y\) axis to provide additional compensation.
I wrote NC subprograms to call the arc dressing data. For a right-handed helical gear, the negative-direction spline subprogram was used, and for a left-handed helical gear, the positive-direction spline subprogram was used. The spline motion was executed with the \(X\) axis and the \(Y\) axis in a synchronized group. The following is a representative structure of the dressing subprogram:
N100 proc t_dm_spline(real drepos,real tdinf) N102 def real xpos[800],ypos[800],offset N104 def real trlen,twlen,tulen,minusy N106 def real rlen,wlen,startposy N108 def int cnt,tpnts N110 minusy=gn_tdPar8 N114 tpnts=gh_clPar1 N118 trlen=gh_ctPar1 N120 twlen=gh_ctPar2+2*trlen N122 tulen=twlen/tpnts N124 rlen=gh_cwPar13 N126 wlen=gh_tPar3 N128 offsety=g_par2-gh_tPar21 N130 startposy=-wlen/2-offsety+(rlen-trlen)+twlen N132 for cnt=0 to tpnts N134 ypos[cnt]=startposy-cnt*tulen N136 xpos[cnt]=gh_ctPar4[cnt] N138 endfor N142 fgroup(shi_axis) N144 g90 g01 g64 N146 ax[inf_axis]=drepos ax[shi_axis]=ypos[0] f=tdinf N148 aspline bauto eauto N150 for cnt=0 to tpnts N152 ax[inf_axis]=drepos+xpos[cnt] ax[shi_axis]=ypos[cnt] f=tdinf N154 endfor N156 fgroup(inf_axis,shi_axis,str_axis) N158 fnorm N160 g90 g01 ax[shi_axis]=minusy f=tdinf N164 ret
The digital twin simulation showed that the diamond roller follows the calculated spline path without collision. The arc-shaped profile of the worm wheel was generated correctly. The simulation also confirmed that the dressing direction must be opposite to the grinding direction for a right-handed helical gear and identical to the grinding direction for a left-handed helical gear. This directional relationship ensures that the compensation is applied to the correct face width position.
5. Experimental Validation
I validated the proposed method on a double-station worm wheel grinding machine equipped with a Siemens SINUMERIK ONE controller. The compensation module was developed as a custom interface. It allows the operator to input the gear parameters, the crowning curve, and the three-section profile deviation values. The module calculates the arc dressing data and writes an initialization file that is called by the NC program. The machine then dresses the worm wheel and grinds the helical gear.
The helical gear used in the experiment had the following parameters:
| Parameter | Symbol | Value |
|---|---|---|
| Number of teeth | \(z\) | 75 |
| Normal module | \(m_n\) | 2.425 mm |
| Normal pressure angle | \(\alpha_n\) | 22.5° |
| Transverse pressure angle | \(\alpha_t\) | 25.652° |
| Helix angle | \(\beta\) | 30.4° |
| Face width | \(b\) | 34 mm |
| Crowning amount | \(C_r\) | 8 µm |
Before enabling the distortion compensation, I ground the helical gear and measured the profile deviations at three sections along the face width. The top section was \(c\), the middle section was \(b\), and the bottom section was \(a\). The measured differences between the sections were used as input for the compensation module. On the left flank, the difference between \(c\) and \(b\) was 4.9 µm, and the difference between \(a\) and \(b\) was 3.1 µm. On the right flank, the difference between \(c\) and \(b\) was 4.7 µm, and the difference between \(a\) and \(b\) was 4.2 µm.
After entering these values, the module generated the arc dressing data. The worm wheel was dressed with the arc profile, and the helical gear was ground again. I performed two compensation trials and measured the results with a Klingelnberg gear measuring center. The results are shown in the following table.
| Measurement trial | Flank | Section difference | Before compensation (µm) | After compensation (µm) | Reduction |
|---|---|---|---|---|---|
| First trial | Left | \(c-b\) | 4.9 | 0.6 | 87.8% |
| First trial | Left | \(a-b\) | 3.1 | 0.3 | 90.3% |
| First trial | Right | \(c-b\) | 4.7 | 0.4 | 91.5% |
| First trial | Right | \(a-b\) | 4.2 | 0.9 | 78.6% |
| Second trial | Left | \(c-b\) | 4.9 | 1.0 | 79.6% |
| Second trial | Left | \(a-b\) | 3.1 | 0.0 | 100% |
| Second trial | Right | \(c-b\) | 4.7 | 0.4 | 91.5% |
| Second trial | Right | \(a-b\) | 4.2 | 1.0 | 76.2% |
The experimental results show that the overall distortion of the helical gear was reduced by approximately 85%. The left and right flanks both improved significantly. The remaining deviations are within a few tenths of a micrometer, which is close to the measurement uncertainty of the gear measuring center. These results confirm that the arc-shaped worm wheel dressing method is feasible and reliable for the anti-distortion grinding of helical gears.
6. Conclusions
I established a spatial meshing model for the worm wheel grinding of helical gears with longitudinal crowning. I derived the meshing equation using the double-parameter envelope principle and obtained the instantaneous contact lines on the left and right flanks. I showed that the contact line length increases with the base helix angle and the normal module, which explains why distortion becomes more severe for certain helical gear designs.
I built a tooth surface distortion model and derived the pressure angle deviation as a function of face width position. The model shows that the left and right flanks have opposite deviations and that the distortion changes continuously along the face width. I proposed an anti-distortion compensation method that dresses the worm wheel into an arc shape. The arc radius is calculated from the measured pressure angle deviation at each face width position. I derived the relationship between the normal arc radius and the axial arc radius and used a spline-based NC dressing program to generate the arc profile.
I created a digital twin of the grinding machine, built the kinematic chain, and simulated the arc dressing motion. The simulation verified the correct direction of dressing and the collision-free motion of the diamond roller. Finally, I conducted grinding experiments on a helical gear with a crowning amount of 8 µm. The three-section profile measurements showed that the distortion was reduced by about 85% on average. The left flank reduction reached 87.8% and 90.3% in the first trial, and the right flank reduction reached 91.5% and 78.6%. The second trial produced similar results. These outcomes confirm that the proposed method can effectively control the tooth surface distortion of helical gears in continuous generating grinding.
For future work, I will study the influence of thermal deformation and axis motion errors on the stability of the compensation. I will also optimize the compensation module interface and extend the method to different machine configurations. The goal is to make the anti-distortion grinding of helical gears more robust and more widely applicable in high-precision gear manufacturing.