The design and manufacture of noncircular gears present unique challenges not encountered in standard cylindrical gear systems. One of the most critical defects that can compromise the functionality and structural integrity of a noncircular gear is tooth undercutting. Undercutting weakens the tooth root, significantly reducing its bending strength and load-bearing capacity. Furthermore, it shortens the active profile of the tooth, which can lead to a decrease in the contact ratio, potentially causing uneven motion transmission, increased noise, vibration, and even intermittent loss of kinematic control. Therefore, predicting and evaluating undercutting during the design phase is paramount for ensuring the reliability and performance of noncircular gear drives.
Traditional methods for assessing the risk of undercutting in noncircular gears often rely on simplified criteria. A common approach involves evaluating the minimum radius of curvature, $\rho_{min}$, of the pitch curve. The gear is conceptually replaced by an equivalent spur gear with a pitch radius equal to $\rho_{min}$. The condition for avoiding undercutting when cut by a rack tool is then applied, leading to a formula for the maximum allowable module, $m_{max}$. A typical form of this simplified criterion is:
$$ m_{max} \le 0.117 \rho_{min} $$
While this method provides a quick, conservative check, it has significant limitations. Firstly, it assumes generation with a rack cutter. In practice, gear shaping is a far more common manufacturing process for complex gear profiles. A shaper cutter, being a generating gear itself, generally produces less severe undercutting than a rack tool for the same gear parameters. Consequently, a gear deemed to have undercutting via the rack-based criterion might be perfectly acceptable when produced via gear shaping. Secondly, and more importantly, this simplified method offers no insight into the degree or location of undercutting. It provides a binary “yes/no” answer for the entire gear, ignoring the fact that undercutting is a local phenomenon that varies from tooth to tooth around the non-circular pitch curve. A designer cannot tell which specific teeth are affected or how severe the undercut is, making it impossible to make informed trade-offs between geometric design, strength, and manufacturability.
To overcome these limitations, a precise, tooth-by-tooth analysis method is required. This article develops a comprehensive methodology for analyzing undercutting in noncircular gears based on a numerical model of the gear shaping process. The core of this approach is the simulation of the tooth generation process itself, allowing for the exact computation of the tooth profile and the direct investigation of the conditions that lead to undercutting.
1. Numerical Model for Tooth Generation in Gear Shaping
The fundamental principle of gear shaping is one of envelope generation. The shaper cutter (tool) and the workpiece (noncircular gear blank) are maintained in a strict kinematic relationship simulating meshing: their pitch curves roll without slip. The cutting edges of the shaper cutter successively remove material, and their collective swept volume defines the generated tooth space of the noncircular gear. Mathematically, the final tooth flanks of the noncircular gear are the envelope of the family of surfaces representing the successive positions of the shaper cutter’s tooth profiles.
To construct a numerical model, we fix the coordinate system $Oxy$ to the noncircular gear. The shaper cutter, with its own coordinate system $O_s x_s y_s$, rotates as its pitch circle rolls along the noncircular pitch curve $r = r(\phi)$, where $\phi$ is the polar angle of the noncircular gear. For a given instant of the gear shaping process defined by a point $K$ on the pitch curve with coordinates $r_K = r(\phi_K)$, the shaper cutter has rotated through an angle $\eta_K$. The center of the shaper cutter, $O_1$, and the line $O_1K$ (which is the normal to the pitch curve at $K$) can be calculated precisely.
Instead of solving complex analytical envelope equations, a discrete computational approach is adopted. The key idea is to calculate points on the generated tooth profile as intersections between the noncircular gear’s tooth boundary and a series of curves that are normal offsets from its pitch curve. The addendum and dedendum curves of a gear are, by definition, normal offset curves from the pitch curve. We can define a series of $k$ such offset curves, spaced evenly from the addendum towards the root, effectively slicing the tooth space.
The distance from the pitch curve to the $t$-th offset curve is given by:
$$ h_t = \pm \left( h_a – t \frac{h_a + h_{a0}}{k} \right) $$
where $h_a$ is the addendum of the noncircular gear, $h_{a0}$ is the addendum of the shaper cutter, $t = 0, 1, 2, …, k$, and the sign is positive for external gears and negative for internal gears. For $t=0$, $h_t = \pm h_a$, which is the addendum curve. As $t$ increases, we move towards the tooth root.
The parametric equations for the $t$-th normal offset curve in the $Oxy$ coordinate system are:
$$
\begin{aligned}
x_t(\phi) &= r(\phi)\cos\phi + h_t \frac{r'(\phi)\sin\phi + r(\phi)\cos\phi}{\sqrt{r^2(\phi) + r’^2(\phi)}} \\
y_t(\phi) &= r(\phi)\sin\phi – h_t \frac{r'(\phi)\cos\phi – r(\phi)\sin\phi}{\sqrt{r^2(\phi) + r’^2(\phi)}}
\end{aligned}
$$
where $r'(\phi)$ is the first derivative of the pitch curve function.
The numerical algorithm for gear shaping simulation proceeds as follows: For a given tooth space and a specific offset curve index $t$, the algorithm searches for the point $m_t(x_{2t}, y_{2t})$ where this offset curve intersects the envelope (the tooth flank). Simultaneously, it identifies the corresponding point on the shaper cutter profile that generated $m_t$. This shaper point is conveniently characterized by its radial distance from the cutter center, denoted as $R_t$. By performing this calculation for all $t$ from 0 to $k$, we obtain a discrete but accurate representation of the tooth profile, along with the vital associated data of the generating shaper points $\{R_t\}$.
| Symbol | Description |
|---|---|
| $r(\phi)$ | Polar equation of the noncircular gear pitch curve |
| $\phi_K$ | Polar angle defining a specific meshing position $K$ |
| $h_a$ | Addendum of the noncircular gear |
| $h_{a0}$ | Addendum of the shaper cutter |
| $R_a$ | Tip radius of the shaper cutter ($R_a = r_{0} + h_{a0}$, where $r_0$ is cutter pitch radius) |
| $k$ | Total number of normal offset curves used in the simulation |
| $h_t$ | Normal offset distance for the $t$-th curve |
| $m_t(x_{2t}, y_{2t})$ | Point on the generated noncircular gear tooth profile (on offset curve $t$) |
| $R_t$ | Radial distance from shaper cutter center to the generating point on its profile for $m_t$ |
2. Mechanism and Quantitative Detection of Undercutting
To understand undercutting in gear shaping, we must consider the two distinct parts of a generated tooth profile: the active (or functional) flank and the fillet (or transition) curve. The active flank is generated by the involute portion of the shaper cutter tooth. The fillet curve is the trochoid-like path traced by the tip of the shaper cutter tooth (or its corner). In a properly generated tooth without undercutting, these two curves meet smoothly at a point of tangency, denoted as point $e$.
Undercutting occurs when the envelope generation process reaches a limit. Geometrically, this corresponds to the appearance of a cusp (a sharp point, $J$) on the theoretical generated profile. Beyond this cusp, the theoretical profile loops back into the material of the gear tooth. In the physical gear shaping process, the shaper cutter continues its motion and will cut away this looped-back material, including part of the previously generated active flank. The final manufactured profile is therefore shorter than the theoretical one, and the junction point $e$ between the active flank and the fillet is no longer a point of tangency but an intersection point where the fillet curve cuts into the active profile. This is the root cut, or undercut.
Detecting this intersection point $e$ directly from discrete numerical data is challenging. However, the phenomenon manifests a clear signature in the sequence of shaper cutter radii $\{R_t\}$ computed during the simulation.
- Normal Generation (No Undercut): On the active flank (from tip towards root), the points $m_t$ are generated by different points on the shaper’s involute profile. Therefore, $R_t$ increases gradually and continuously as $t$ increases (moving root-wards). When the simulation reaches the fillet region, the points $m_t$ are generated by the shaper’s tip corner. Consequently, $R_t$ becomes approximately constant and equal to the shaper’s tip radius $R_a$. The transition at point $e$ is smooth, resulting in a continuous, monotonic increase in $R_t$ that asymptotically approaches $R_a$.
- Generation with Undercut: The process begins similarly on the active flank, with $R_t$ increasing gradually. However, at the theoretical cusp $J$, the generation by the shaper’s involute flank ceases. The remaining active flank (from $e$ to $J$) is lost. The fillet curve, still generated by the shaper tip, now intersects the truncated active flank at point $e$. In the numerical data, this appears as a sudden jump or discontinuity in the $R_t$ values. We will observe a point $m_k$ on the active flank with $R_k < R_a$, and the very next computed point $m_{k+1}$ on the fillet will have $R_{k+1} \approx R_a$. The difference $R_{k+1} – R_k$ will be significantly larger than the incremental changes in $R_t$ seen on the active flank.
This leads to a robust quantitative criterion for detecting undercutting. Let $k$ be the index of the last computed point on the active flank before the suspected jump. We can identify $k$ by finding the point where $R_t$ first approaches $R_a$:
$$
\begin{cases}
R_{k+1} \approx R_a \\
R_k \ll R_a
\end{cases}
$$
We then define an Undercut Judgment Factor, $\xi$, as:
$$ \xi = \frac{R_{k+1} – R_k}{R_k – R_{k-1}} $$
The interpretation is straightforward:
- If $\xi \approx 1$ (or $\xi < 1$), the transition in $R_t$ is smooth, indicating no undercutting. Point $e$ is a point of tangency.
- If $\xi \gg 1$, a significant jump in $R_t$ exists, indicating that point $e$ is an intersection point and undercutting has occurred.
Furthermore, the severity of the undercutting can be quantified. In a standard, uncut gear, the start of the fillet (point $e$) occurs at a fixed clearance height $h_c$ from the pitch line. When undercutting occurs, this point moves up towards the tooth tip. We define the Undercut Start Height, $H_C$, as the radial distance from the gear center to the undercut intersection point $e$. Since $e$ lies between offset curves $k$ and $k+1$, we can approximate it as:
$$ H_C = h_{a0} + 0.5(h_k + h_{k+1}) $$
where $h_k$ and $h_{k+1}$ are the normal offset distances for curves $k$ and $k+1$, calculated from Equation for $h_t$. The larger the value of $H_C$ is compared to the nominal clearance height $h_c$, the more severe the undercutting, as a greater portion of the active profile has been lost.
| Step | Action | Purpose / Outcome |
|---|---|---|
| 1 | Run the numerical gear shaping model for all teeth. | Obtain datasets $\{m_t, R_t\}$ for left and right flanks of each tooth. |
| 2 | For each tooth flank, analyze the sequence $\{R_t\}$. | Identify the index $k$ where $R_{k+1} \approx R_a$ and $R_k$ is significantly smaller. |
| 3 | Calculate the Undercut Judgment Factor $\xi$ using $R_{k-1}, R_k, R_{k+1}$. | Quantify the discontinuity in the shaper point radius. |
| 4 | Judge: If $\xi \gg 1$, undercutting is present. If $\xi \approx 1$, no undercutting. | Make a definitive, tooth-specific judgment on undercut occurrence. |
| 5 | If undercutting is present, calculate the Undercut Start Height $H_C$. | Quantify the severity of the undercutting for that specific tooth flank. |
| 6 | Compile results for all teeth. | Provide a complete map of undercutting across the entire noncircular gear. |
3. Illustrative Example and Discussion
To demonstrate the efficacy of this method, we analyze an elliptical gear with the following parameters:
- Ellipse semi-major axis, $A = 100$ mm
- Eccentricity, $k_1 = 0.92531$
- Module, $m = 5$ mm
- Pressure angle, $\alpha = 20^\circ$
- Number of teeth on gear, $Z_2 = 29$
- Number of teeth on shaper cutter, $z_0 = 20$
- Shaper cutter addendum, $h_{a0} = 6$ mm (Tip radius $R_a = 56$ mm)
The minimum radius of curvature $\rho_{min}$ for this ellipse is approximately 14.36 mm, occurring at the ends of the major axis.
Applying the traditional simplified criterion (Equation 1):
$$ m_{max} \le 0.117 \times 14.36 \approx 1.68 \text{ mm} $$
Since the design module ($5$ mm) is much greater than $1.68$ mm, the traditional method predicts severe undercutting for the entire gear. However, it gives no further detail.
We now apply the proposed numerical gear shaping-based analysis. The simulation is run with $k=30$ offset curves. Let’s examine the first tooth of the gear, specifically its left flank (oriented near the major axis, where curvature is minimum). A portion of the generated data is shown below:
| Offset Index $t$ | Shaper Radius $R_t$ (mm) | Gear Profile Point ($x_{2t}$, $y_{2t}$) (mm) |
|---|---|---|
| 13 | 49.50 | (192.48, 4.25) |
| 14 | 49.79 | (192.19, 4.02) |
| 15 | 50.12 | (191.90, 3.80) |
| 16 | 50.54 | (191.59, 3.62) |
| 17 | 55.99 | (191.25, 3.57) |
| 18 | 55.99 | (190.89, 3.60) |
The pattern is clear. From $t=13$ to $t=16$, $R_t$ increases gradually from 49.50 to 50.54 mm. At $t=17$, there is a dramatic jump to $R_{17} = 55.99 \approx R_a = 56$ mm. This identifies $k=16$. We calculate the Undercut Judgment Factor:
$$ \xi = \frac{R_{17} – R_{16}}{R_{16} – R_{15}} = \frac{55.99 – 50.54}{50.54 – 50.12} = \frac{5.45}{0.42} \approx 12.98 $$
Since $\xi \approx 13 \gg 1$, we conclusively diagnose undercutting on this flank. The Undercut Start Height is calculated. First, find $h_{16}$ and $h_{17}$ using Equation for $h_t$ with $h_a = 5$ mm (standard addendum for $m=5$ mm), $h_{a0}=6$ mm, and $k=30$:
$$ h_{16} \approx -1.45 \text{ mm}, \quad h_{17} \approx -1.07 \text{ mm} $$
$$ H_C = 6 + 0.5(-1.45 – 1.07) = 4.74 \text{ mm} $$
This value is significantly greater than a typical clearance of $1.0$ mm, confirming that the undercutting is severe and has removed a substantial part of the active profile. This tooth would be very weak.
In contrast, analyzing the right flank of the same tooth (oriented near the minor axis, where curvature is larger) yields a different result. The $R_t$ values increase smoothly and approach $R_a$ asymptotically. For the transition point identified at, say, $k=22$, we might find $R_{21}=53.96$, $R_{22}=54.97$, $R_{23}=55.99$. The judgment factor is:
$$ \xi = \frac{55.99 – 54.97}{54.97 – 53.96} = \frac{1.02}{1.01} \approx 1.01 $$
Since $\xi \approx 1$, we conclude that this flank is free from undercutting. The point $e$ is a proper point of tangency between the active flank and the fillet.
This example vividly illustrates the power of the gear shaping numerical analysis. It not only confirms the presence of undercutting where the traditional method predicted it (at the high-curvature region) but also provides a quantitative measure of its severity ($H_C=4.74$ mm) and identifies specific teeth/flanks that are safe from undercutting. A designer can use this detailed map to decide if the gear is acceptable as-is, if certain teeth need localized strengthening, or if the pitch curve/module must be modified.
4. Conclusion
The analysis of tooth undercutting is a critical step in the design of functional and durable noncircular gears. Relying on oversimplified criteria based on minimum curvature can lead to overly conservative designs or, conversely, a failure to identify localized severe undercutting. The methodology presented herein, grounded in a precise numerical simulation of the gear shaping manufacturing process, offers a powerful and accurate alternative.
By constructing the tooth profile through the discrete calculation of points along normal offset curves and tracking the corresponding generating points on the shaper cutter, we gain deep insight into the generation mechanics. The core of the undercut detection algorithm lies in monitoring the radial distance $R_t$ of these shaper points. A discontinuity or jump in this sequence, quantified by the Undercut Judgment Factor $\xi$, is the definitive signature of undercutting. Furthermore, the method introduces a quantitative metric, the Undercut Start Height $H_C$, to assess the severity of the defect, moving beyond mere detection to evaluation.
This gear shaping-based numerical approach provides designers with a precise tool to perform tooth-by-tooth validation of noncircular gear designs. It allows for informed optimization, enabling trade-offs between geometric constraints, strength requirements, and manufacturability. Future work may involve integrating this analysis directly into CAD/CAM systems for noncircular gears, automating the undercut check, and extending the model to account for other generation processes like hobbing or to investigate profile modifications aimed at mitigating undercutting while preserving the desired kinematic function.