In my work on large straight spur gear metrology, one of the most critical challenges is accurately locating the geometrical center of the gear before any tooth profile evaluation can be performed. Large straight spur gears, typically defined as those with a diameter exceeding 0.8 m, are widely used in heavy machinery such as mining equipment, wind turbines, marine propulsion systems, and large-scale power generation units. The quality and service life of these gears directly influence the overall performance of the entire mechanical system. Therefore, precise measurement of tooth profile deviations, including the total profile error, is essential for quality assurance. However, due to the immense size and weight, traditional gear measuring centers or coordinate measuring machines often cannot accommodate these components, and in-situ measurement techniques must be employed. The first step in such in-situ measurement is the reliable determination of the gear’s geometrical center, which serves as the reference for all subsequent error calculations.
In this paper, I present a novel method for calculating the geometrical center of a large straight spur gear based on discrete sampled data from a subset of its teeth. The approach builds upon an involute gear discrete data model and uses an iterative optimization scheme to find the center that minimizes the distance between measured tooth flank points and theoretical involutes. I then use the obtained center to compute the total tooth profile deviation and compare the results with those computed using the known theoretical center. The simulation results demonstrate that the proposed method is robust even under significant measurement noise, with the center error remaining below 1 µm and the profile deviation error below 10 µm. This work provides a practical foundation for the on-site measurement of large straight spur gears.
1. Involute Gear Discrete Data Model
To simulate the measurement process, I first establish a discrete data model of an involute straight spur gear. The parametric equations of an involute curve in Cartesian coordinates are:
$$
\begin{cases}
x_k = r_b (\cos \varphi_k + \varphi_k \sin \varphi_k) \\[4pt]
y_k = r_b (\sin \varphi_k – \varphi_k \cos \varphi_k)
\end{cases}
$$
where \(r_b\) is the base circle radius and \(\varphi_k\) is the roll angle at point \(k\). The limits of \(\varphi_k\) are determined by the root and tip circles:
$$
\varphi_f = \tan \alpha_f = \tan\left(\arccos\frac{r_b}{r_f}\right), \quad
\varphi_a = \tan \alpha_a = \tan\left(\arccos\frac{r_b}{r_a}\right)
$$
with \(r_f\) the root radius, \(r_a\) the addendum radius, and \(\alpha_f, \alpha_a\) the corresponding pressure angles. For a symmetrical tooth centered on the \(x\)-axis, the left and right flanks must be rotated by initial angles:
$$
\theta_r = -(\tan\alpha – \alpha + \frac{\pi}{2z}), \quad
\theta_l = \tan\alpha – \alpha + \frac{\pi}{2z}
$$
where \(\alpha\) is the nominal pressure angle and \(z\) the number of teeth. For the \(i\)-th tooth, the entire tooth profile is rotated by:
$$
\omega_i = i \times \frac{2\pi}{z}
$$
Thus, the coordinates of the \(k\)-th point on the \(j\)-th flank of the \(i\)-th tooth are given by:
$$
\begin{bmatrix} x_{ijk} \\ y_{ijk} \end{bmatrix} =
\begin{bmatrix} \cos\omega_i & -\sin\omega_i \\ \sin\omega_i & \cos\omega_i \end{bmatrix}
\begin{bmatrix} \cos\theta_j & -\sin\theta_j \\ \sin\theta_j & \cos\theta_j \end{bmatrix}
\begin{bmatrix} r_b(\cos\varphi_k + \varphi_k\sin\varphi_k) \\ r_b(\sin\varphi_k – \varphi_k\cos\varphi_k) \end{bmatrix}
$$
Using this model, I generate the complete set of standard involute tooth flank points for any specified straight spur gear geometry.
2. Geometrical Center Calculation Method
To determine the geometrical center of a large straight spur gear, I select two pairs of teeth that are symmetrically opposite when the tooth count is even, or a pair of teeth and their opposing spaces when the tooth count is odd. For each selected tooth (e.g., tooth A), I obtain measured points on both the left and right flanks: \((x_{Alk}, y_{Alk})\) and \((x_{Ark}, y_{Ark})\) for \(k = 1,\dots,n\). An initial estimate of the center \(C_0(x_0,y_0)\) is obtained by averaging all point coordinates from the eight flanks (two teeth × two flanks × two symmetrical positions):
$$
x_0 = \frac{1}{8n}\sum_{k=1}^{n}\left( x_{Alk}+x_{Ark}+x_{A’lk}+x_{A’rk}+x_{Blk}+x_{Brk}+x_{B’lk}+x_{B’rk} \right)
$$
$$
y_0 = \frac{1}{8n}\sum_{k=1}^{n}\left( y_{Alk}+y_{Ark}+y_{A’lk}+y_{A’rk}+y_{Blk}+y_{Brk}+y_{B’lk}+y_{B’rk} \right)
$$
Then, starting from \(C_0\), I generate four additional candidate centers at a step size \(p\) (initially \(p=1\)): \(C_1(x_0+p, y_0+p)\), \(C_2(x_0-p, y_0+p)\), \(C_3(x_0-p, y_0-p)\), \(C_4(x_0+p, y_0-p)\). For each candidate center, I construct theoretical involute curves on the same four tooth flanks using the same base circle radius \(r_b\). For a given measured point \(Q_i(x_i,y_i)\) on a flank, I find the corresponding theoretical point \(Q’_i(x’_i,y’_i)\) on the same radial line (i.e., on a circle of the same radius centered at the candidate center). The squared Euclidean distance between them is:
$$
w_i = (x_i – x’_i)^2 + (y_i – y’_i)^2
$$
Summing over all \(N\) sampled points on all eight flanks yields the total squared distances for candidate center \(C_l\):
$$
\Psi_l = \sum_{k=1}^{8}\sum_{i=1}^{N} \left[ (x_{k,i} – x’_{k,i})^2 + (y_{k,i} – y’_{k,i})^2 \right]
$$
I compare the five \(\Psi_l\) values (\(l=0,\dots,4\)) and select the center that gives the smallest \(\Psi_l\) as the new best estimate. The step size \(p\) is then reduced to \(0.1\) times the coordinate values of the selected center, and the process is repeated iteratively until convergence. The final center coordinates \(C_{\text{final}}(x_{\text{final}}, y_{\text{final}})\) are taken as the geometrical center of the large straight spur gear.
3. Tooth Profile Deviation Calculation
Once the geometrical center is determined, I compute the total tooth profile deviation according to the standard GB/T 10095.1-2008. The total profile deviation is defined as the distance between two design involute curves that just enclose the actual tooth profile over the evaluation range. For a measured point \(K(x_K, y_K)\) on the tooth flank, I first find the intersection point \(P’\) of the line \(OK\) with the base circle (center \(O\) is the gear center):
$$
x_{P’} = \frac{r_b}{\sqrt{(x_K – x_0)^2 + (y_K – y_0)^2}}\,(x_K – x_0) + x_0
$$
$$
y_{P’} = \frac{r_b}{\sqrt{(x_K – x_0)^2 + (y_K – y_0)^2}}\,(y_K – y_0) + y_0
$$
where \((x_0, y_0)\) is the gear center. The pressure angle at point \(K\) is:
$$
\alpha_k = \arctan\frac{\sqrt{(x_K – x_0)^2 + (y_K – y_0)^2 – r_b^2}}{r_b}
$$
The roll angle \(\varphi_k\) is:
$$
\varphi_k = \frac{\sqrt{(x_K – x_0)^2 + (y_K – y_0)^2 – r_b^2}}{r_b}
$$
The rotation angle \(\beta\) from point \(P’\) to the theoretical involute start point \(P\) is:
$$
\beta = \varphi_k – \alpha_k
$$
Then the theoretical start point \(P\) is obtained by rotating \(P’\) by angle \(\beta\) (clockwise for right flank, counterclockwise for left flank). For a left flank, the rotation matrix is:
$$
\mathbf{A} = \begin{bmatrix} \cos\beta & -\sin\beta \\ \sin\beta & \cos\beta \end{bmatrix}
$$
and for a right flank:
$$
\mathbf{A} = \begin{bmatrix} -\cos\beta & -\sin\beta \\ \sin\beta & -\cos\beta \end{bmatrix}
$$
Applying this to all measured points on a flank yields a set of start points \(P_i\) on the base circle. The angle between any two such points with respect to the gear center gives the angular spread of the actual involute. Specifically, for a pair of start points \(P_k\) and \(P_i\), the angle \(\theta_i\) is:
$$
\theta_i = \arcsin\frac{y_i x_k – x_i y_k}{\sqrt{x_k^2 + y_k^2}\sqrt{x_i^2 + y_i^2}}, \quad \theta_i \in (-\frac{\pi}{2}, \frac{\pi}{2})
$$
The total profile deviation \(\Delta f\) for that flank is then:
$$
\Delta f = r_b \times (\theta_{\max} – \theta_{\min})
$$
where \(\theta_{\max}\) and \(\theta_{\min}\) are the maximum and minimum angles among all \(\theta_i\). This procedure is repeated for every tooth flank of interest.
4. Simulation and Validation
To validate the proposed method, I performed numerical simulations on a large straight spur gear with the following parameters: module \(m = 40\ \text{mm}\), number of teeth \(z = 100\), pressure angle \(\alpha = 20^\circ\), helix angle \(\beta = 0^\circ\). The base circle radius is \(r_b = \frac{mz\cos\alpha}{2} = 1879.385\ \text{mm}\). Using the discrete data model, I generated a set of standard involute points. A portion of the simulated measured data is listed in Table 1.
| \(x_i\) | 1949.4 | 1950.4 | 1951.3 | 1952.2 | 1953.1 |
|---|---|---|---|---|---|
| \(y_i\) | 4652.1 | 4629.0 | 4605.7 | 4582.3 | 4558.7 |
| \(x_i\) | 1954.0 | 1954.9 | 1955.8 | 1956.8 | 1957.7 |
| \(y_i\) | 4535.0 | 4511.0 | 4486.9 | 4462.7 | 4438.2 |
I then added random noise with amplitudes \(\delta = 0, 10, 50, 100, 150, 200, 250, 300, 318\ \mu\text{m}\) to the standard involute points to simulate realistic measurement errors. The noise was generated using MATLAB’s uniform random number function multiplied by half the amplitude. For each noise level, I applied the center-finding algorithm and computed the geometrical center coordinates \((x_l, y_l)\). Using both the theoretical center (0,0) and the computed center, I calculated the total profile deviation \(F_\alpha\) and \(F’_\alpha\), respectively. The results are summarized in Table 2.
| \(\delta\) (μm) | 0 | 10 | 50 | 100 | 150 | 200 | 250 | 300 | 318 |
|---|---|---|---|---|---|---|---|---|---|
| \(x_l\) (μm) | -0.0383 | -0.0456 | -0.0302 | -0.0289 | -0.0325 | -0.0565 | -0.0578 | -0.0649 | -0.0520 |
| \(y_l\) (μm) | 0.0321 | 0.0586 | 0.0741 | 0.0753 | 0.0718 | 0.0478 | 0.0464 | 0.0394 | 0.0523 |
| \(F_\alpha\) (μm) | 0 | 10 | 50 | 99.8 | 149.5 | 199.4 | 249.6 | 299.5 | 317.2 |
| \(F’_\alpha\) (μm) | 0 | 10 | 59.9 | 102.2 | 159.1 | 208.6 | 258.6 | 306.8 | 318.5 |
| \(F_\alpha – F’_\alpha\) (μm) | 0 | 0 | -9.9 | -2.4 | -9.6 | -9.2 | -9.0 | -7.3 | -1.3 |
From Table 2, it is evident that the maximum error in the computed geometrical center is less than 0.1 μm, which is negligible for practical purposes. The difference between the profile deviations calculated using the theoretical center and those using the computed center is within 10 μm for all noise levels up to 318 μm. This demonstrates the robustness and accuracy of the proposed method even under significant measurement noise, making it suitable for the on-site inspection of large straight spur gears.
The image above illustrates a large straight spur gear typical of those considered in this study. The method I have developed can be directly applied to such gears during in-situ measurement, where the gear remains mounted on its shaft or within the machine.
5. Conclusion
I have presented a systematic approach for determining the geometrical center of a large straight spur gear from discrete tooth flank measurements, and for computing the total tooth profile deviation based on that center. The method employs an iterative optimization of candidate centers using the sum of squared distances between measured points and theoretical involutes. Simulation results on a gear with 3.2 m diameter show that the center can be located with sub‑micrometer accuracy even when random measurement noise up to 318 μm is present. The resulting profile deviation error remains below 10 μm, which meets the typical tolerance requirements for large gears. This technique eliminates the need for an external reference and is well suited for integration into portable measuring systems, thereby enabling practical quality control of large straight spur gears in industrial settings.
Future work will extend the method to helical gears and to gears with non‑standard tooth modifications. Additionally, experimental validation using a large straight spur gear on a dedicated test rig is planned to confirm the simulation findings and to assess real‑world measurement uncertainties.