In the field of precision motion transmission, harmonic drive gears represent a cornerstone technology due to their compact design, high reduction ratios, and excellent positional accuracy. However, a persistent challenge lies in the meshing performance, often compromised by backlash and interference stemming from the inherent elastic deformation of the flexspline. Traditional design approaches, which rely on planar tooth profiles, fail to account for the three-dimensional, tapered deformation of the cup-shaped flexspline under the wave generator’s influence. This oversight leads to suboptimal contact conditions, increased wear, and reduced load capacity. In this work, I present a comprehensive methodology for designing a spatial tooth profile for the circular spline in a harmonic drive gear, explicitly considering the flexspline’s conical deformation. The core of this approach involves discretizing the spatial design problem into a series of planar conjugate tooth profile designs across multiple cross-sections perpendicular to the gear axis. By solving the conjugate equations in each section, fitting the results with modified involute profiles, and employing a lofting algorithm to construct the final spatial surface, a significant improvement in meshing characteristics is achieved. This article details the theoretical foundation, the step-by-step design and analysis process, and presents a rigorous verification through three-dimensional modeling and backlash calculation, demonstrating that the proposed spatial profile enlarges the meshing zone, increases the number of contacting tooth pairs, and ultimately enhances the gear’s durability and power transmission capability.
The fundamental operation of a harmonic drive gear relies on the controlled elastic deformation of the flexspline, typically induced by an elliptical wave generator. This deformation forces the teeth of the flexspline to engage and disengage with those of the rigid circular spline in a wave-like pattern. The common assumption of in-extensibility for the flexspline’s neutral surface during this deformation is central to kinematic modeling. For a cup-type flexspline, the radial deformation is not uniform along the tooth width; it exhibits a tapered characteristic, being maximal at the front end (near the diaphragm) and diminishing towards the rear. This spatial variation necessitates a three-dimensional approach to tooth profile design. My analysis begins by establishing the coordinate systems and the mathematical description of this deformation.
I define a fixed coordinate system $S(O-XYZ)$ attached to the circular spline, with the Y-axis aligned with its major axis. A moving coordinate system $S_1(o_1-x_1y_1z_1)$ is attached to the flexspline tooth, and $S_2(o_2-x_2y_2z_2)$ is attached to the wave generator. The angular positions are related by the wave generator rotation angle $\phi_2$, the flexspline rotation angle $\phi_1$, and the angle $\mu$ which denotes the tilt of the tooth’s symmetry line relative to the radial vector due to deformation. The key relationship is $\Phi = \mu + \phi_1$, where $\Phi$ is the absolute orientation of the flexspline tooth in the fixed frame.
The radial displacement $w$ of a point on the flexspline’s neutral surface, located at angular position $\phi$ (relative to the wave generator’s major axis) and axial coordinate $z$, is modeled as:
$$ w(\phi, z) = w(\phi) \cdot \frac{l – z}{l – b/2 – f} $$
where $l$ is the total length of the flexspline cup, $b$ is the width of the tooth rim, $f$ is the transition length at the front, and $w(\phi)$ is the radial displacement function in the central cross-section. At the major axis ($\phi=0$), the maximum displacement is:
$$ w(0, z) = \frac{l – z}{l – b/2 – f} w_0 $$
with $w_0$ being the maximum radial displacement at the tooth rim, typically set to the gear module $m$. The angle $\mu$ is derived from the deformation geometry:
$$ \mu(\phi, z) = -\arctan\left( \frac{dw/d\phi(\phi)}{r_m + w(\phi, z)} \right) $$
where $r_m$ is the radius of the neutral surface before deformation. The condition of in-extensibility links the angular parameter $\varphi$ on the deformed curve to the initial angle $\phi$ through an integral:
$$ \varphi = \int_{0}^{\phi} \sqrt{ \left(1 + \frac{w}{r_m}\right)^2 + \left( \frac{1}{r_m} \frac{dw}{d\phi} \right)^2 } d\phi $$
This forms the basis for precisely locating the position of each flexspline tooth after deformation.
The conjugate tooth profile for the circular spline is derived using the envelope theory. In each cross-section at a fixed $z$, the process is treated as a planar meshing problem. Given the flexspline’s tooth profile in $S_1$, denoted by $R(x_1, y_1)$, the conjugate profile $G(x_2, y_2)$ on the circular spline in $S$ must satisfy the system of equations derived from coordinate transformation and the envelope condition:
$$
\begin{aligned}
x_2(u, \phi) &= x_1(u) \cos \Phi + y_1(u) \sin \Phi + \rho \sin \gamma, \\
y_2(u, \phi) &= -x_1(u) \sin \Phi + y_1(u) \cos \Phi + \rho \cos \gamma, \\
\frac{\partial x_2}{\partial u} \cdot \frac{\partial y_2}{\partial \phi} – \frac{\partial x_2}{\partial \phi} \cdot \frac{\partial y_2}{\partial u} &= 0.
\end{aligned}
$$
Here, $u$ is the tooth profile parameter, $\Phi = \Phi(\phi, z) = \varphi_1 + \mu(\phi, z)$, $\rho$ is the distance from the flexspline center to the tooth, and $\gamma$ is an associated angle. Solving the third equation (the meshing equation) yields the relation between $u$ and $\phi$ for conjugate points, which are then substituted into the first two equations to generate discrete points $(x_{2K}, y_{2K})$ for the theoretical circular spline tooth profile in that cross-section.
To make the design practical for manufacturing, I fit these discrete points with a standard involute profile. The involute curve $G_w$ for the circular spline is expressed as:
$$
\begin{aligned}
x_{2wK} &= r_2 [ -\sin(u_{2K} – \theta_2) + u_{2K} \cos \alpha_0 \cos(u_{2K} – \theta_2 + \alpha_0) ], \\
y_{2wK} &= r_2 [ \cos(u_{2K} – \theta_2) + u_{2K} \cos \alpha_0 \sin(u_{2K} – \theta_2 + \alpha_0) ],
\end{aligned}
$$
where $r_2$ is the pitch radius of the circular spline, $u_{2K}$ is the involute roll angle, $\alpha_0$ is the standard pressure angle (e.g., 20°), and $\theta_2$ is half of the angular tooth thickness on the pitch circle, which depends on the profile shift coefficient $x_2$. The fitting process aims to minimize the average deviation $\epsilon_m = \sum d_K / n$ while ensuring no interference, i.e., $\Delta x_K = x_{2wK} – x_{2K} \ge 0$ for all points. This yields an optimal profile shift coefficient $x_2$ for each cross-section. By repeating this process for multiple sections along the tooth width (e.g., front, middle, rear), I obtain a set of planar involute profiles with varying parameters. The spatial tooth surface for the circular spline is then constructed by lofting a Non-Uniform Rational B-Spline (NURBS) surface through these sectional curves.
A critical performance metric for any harmonic drive gear is its backlash. To analyze the meshing characteristics of the designed spatial profile, I developed a precise backlash calculation method based on the exact positioning of the deformed flexspline teeth. The position of the $i$-th tooth on the undeformed flexspline is given by $\phi_i = 2\pi i / z_1$. After deformation, its actual angular position $\varphi_i$ and orientation $\Phi_i$ are calculated using the integral and $\mu$ formulas mentioned earlier. For a given meshing position (defined by the wave generator angle), the circumferential backlash $j_t$ between a point on the flexspline tooth and the adjacent circular spline tooth profile is approximated as the distance between corresponding points:
$$ j_t \approx \sqrt{ (x_{K2} – x_{K1})^2 + (y_{K1} – y_{K2})^2 }. $$
The coordinates $(x_{K1}, y_{K1})$ for a point on the flexspline involute (with profile shift $x_1$) in the fixed coordinate system are:
$$
\begin{aligned}
x_{K1} &= r_1 \{ \sin[\Phi – (u_{K1} – \theta_1)] + u_{K1}\cos\alpha_0 \cos[\Phi – (u_{K1} – \theta_1 + \alpha_0)] \} + \rho\sin\varphi_1 – r_m \sin\Phi, \\
y_{K1} &= r_1 \{ \cos[\Phi – (u_{K1} – \theta_1)] – u_{K1}\cos\alpha_0 \sin[\Phi – (u_{K1} – \theta_1 + \alpha_0)] \} + \rho\cos\varphi_1 – r_m \cos\Phi,
\end{aligned}
$$
where $r_1$ is the flexspline pitch radius and $\theta_1$ is related to its profile shift $x_1$. The corresponding point $(x_{K2}, y_{K2})$ on the circular spline involute (with its sectional $x_2$) is found by intersecting a circle of radius $r_K = \sqrt{x_{K1}^2 + y_{K1}^2}$ with its profile. By calculating the minimum $j_t$ across the tooth height for multiple engagement positions, I obtain the backlash distribution for each cross-section. Typically, the minimum backlash occurs either near the tip of the flexspline tooth or the tip of the circular spline tooth, depending on the engagement zone.
To validate the entire design methodology and visualize the spatial interaction, I constructed a three-dimensional assembly model using SolidWorks. The process involved creating sketch profiles for the flexspline teeth in multiple cross-sections based on the calculated deformed positions, then using the loft feature to generate the deformed flexspline body with its spatial tooth surfaces. Similarly, the circular spline’s spatial tooth surface was created by lofting the different sectional involute curves corresponding to the fitted $x_2$ values. The assembly model clearly shows the full engagement at the major axis region and complete disengagement at the minor axis. Interference checks and direct measurement tools within the software were used to verify the absence of geometric interference and to corroborate the calculated backlash values, providing a strong visual and quantitative confirmation of the design’s efficacy.
For a concrete demonstration, I applied this methodology to two sets of harmonic drive gear parameters, differing primarily in flexspline cup length. The common base parameters are: number of teeth $z_1=200$ (flexspline), $z_2=202$ (circular spline), module $m=0.5$ mm, standard pressure angle $\alpha_0=20^\circ$, undeformed flexspline neutral radius $r_m=50.375$ mm, wave generator type is a four-roller with installation angle $\beta=30^\circ$, and flexspline profile shift coefficient $x_1=3.0$. The radial displacement function $w(\phi)$ for the central section corresponds to a pure radial deformation with $w_0 = 1.0m = 0.5$ mm. I analyzed two cases: a longer cup with $l=80$ mm and a shorter cup with $l=50$ mm. The tooth rim width $b$ and transition length $f$ are derived from standard design practices.
For each case, I divided the tooth width into three sections: Front, Middle, and Rear. The maximum radial displacement $w(0,z)$ varies per section due to the taper factor $(l-z)/(l-b/2-f)$. The conjugate theory and involute fitting process were executed for each section to determine the optimal circular spline profile shift coefficient $x_2$ and other derived geometric parameters. The results are summarized in the following tables.
| Parameter | Flexspline | Circular Spline (Section) | ||
|---|---|---|---|---|
| Front | Middle | Rear | ||
| Profile Shift Coefficient, $x$ | 3.0000 | 2.7170 | 2.6676 | 2.6254 |
| Half Angular Tooth Thickness on Pitch Circle, $\theta$ (deg) | 1.0756 | 1.0786 | 1.0659 | 1.0551 |
| Pitch Radius, $r$ (mm) | 50.000 | 50.500 | 50.500 | 50.500 |
| Tip Radius, $r_a$ (mm) | 51.8740 | 51.7325 | 51.7076 | 51.6863 |
| Root Radius, $r_f$ (mm) | 50.8250 | 52.5335 | 52.5088 | 52.4877 |
| Tooth Thickness on Pitch Circle (mm) | 1.8773 | 1.9013 | 1.8790 | 1.8600 |
| Parameter | Flexspline | Circular Spline (Section) | ||
|---|---|---|---|---|
| Front | Middle | Rear | ||
| Profile Shift Coefficient, $x$ | 3.0000 | 2.7511 | 2.6676 | 2.6300 |
| Half Angular Tooth Thickness on Pitch Circle, $\theta$ (deg) | 1.0756 | 1.0873 | 1.0659 | 1.0563 |
| Pitch Radius, $r$ (mm) | 50.000 | 50.500 | 50.500 | 50.500 |
| Tip Radius, $r_a$ (mm) | 51.8740 | 51.7497 | 51.7076 | 51.6689 |
| Root Radius, $r_f$ (mm) | 50.8250 | 52.5505 | 52.5088 | 52.4900 |
| Tooth Thickness on Pitch Circle (mm) | 1.8773 | 1.9167 | 1.8790 | 1.8620 |
The conjugate existence intervals, calculated using the precise algorithm for each section, further illustrate the benefit of the spatial design. For the longer cup harmonic drive gear, the intervals (in terms of flexspline polar angle) are approximately $[-1.4589^\circ, 3.4776^\circ]$ for the Front section, $[0.9143^\circ, 5.8981^\circ]$ for the Middle, and $[3.9252^\circ, 8.9848^\circ]$ for the Rear section. This sequential engagement across sections effectively widens the overall active meshing zone for the harmonic drive gear assembly.
The backlash distributions for both gear sets across the three sections were calculated and are depicted conceptually in the analysis. For the longer cup harmonic drive gear, the minimum backlash curve shows that in the polar angle range $[-20^\circ, 2.1521^\circ]$, the Front section has the smallest backlash (thus is the active meshing section). In the range $[2.1521^\circ, 5.6708^\circ]$, the Middle section takes over, and for $[5.6708^\circ, 60^\circ]$, the Rear section becomes active. A similar phased engagement is observed for the shorter cup harmonic drive gear, with transition points at $0.8540^\circ$ and $5.8392^\circ$. This demonstrates that the spatial tooth profile ensures a smooth handover of load sharing along the tooth width, increasing the total number of tooth pairs in contact at any instant compared to a single planar profile design.
To quantify the advantage, I compared the backlash of the spatial profile against a conventional planar design. For the same gear parameters, a planar circular spline profile with a constant $x_2 = 2.7$ is required to avoid interference across all sections. The backlash values for this planar profile are consistently larger than those for any of the three spatial sections across a significant portion of the meshing cycle (e.g., from $-20^\circ$ to about $16^\circ$). This directly implies that the spatial profile achieves tighter meshing and greater contact area. The geometric consequence is evident in the lofted models: the circular spline tooth is not a straight extrusion but a surface where the tooth slot is wider and the tooth height is slightly smaller at the front section compared to the rear. This tailored geometry accommodates the tapered deformation perfectly. For the longer cup harmonic drive gear, the front section pitch tooth thickness is $0.0413$ mm larger than the rear, and the front tooth height is $0.0462$ mm smaller. The differences are more pronounced for the shorter cup harmonic drive gear ($0.0547$ mm and $0.0611$ mm, respectively), underscoring that shorter, stiffer flexsplines induce a more severe taper, making the spatial design even more critical.
Furthermore, to confirm that the design approach is not overly sensitive to the number of cross-sections used for discretization, I analyzed the longer cup harmonic drive gear using five equally spaced sections. The resulting backlash curves showed that the intermediate sections’ backlash values fell within the envelope defined by the front and rear sections. The transition points for the minimum backlash shifted slightly, but the overall phased engagement behavior remained unchanged, confirming the robustness of the lofting-based spatial construction method.
The three-dimensional models built in SolidWorks provided the final validation. No interference was detected during the virtual assembly and rotation simulation. The ability to section the model and directly measure clearances confirmed the calculated backlash trends. The visual representation clearly shows the gradual engagement from the front to the rear of the tooth width, eliminating the risk of edge contact or localized stress concentration that plagues planar profile designs.
In conclusion, the spatial tooth profile design methodology for harmonic drive gears, which I have developed and detailed here, addresses a fundamental limitation of traditional planar approaches. By explicitly modeling the tapered deformation of the cup-type flexspline and decomposing the spatial conjugate problem into manageable planar sections, I derived optimized involute profiles for the circular spline that vary systematically along the axis. The lofted spatial surface constructed from these profiles ensures a continuous and favorable meshing condition. The backlash analysis and three-dimensional modeling verification conclusively demonstrate that this design significantly expands the functional meshing domain, increases the contact area between teeth, and promotes load sharing among more tooth pairs. For harmonic drive gear applications demanding high precision, high torque capacity, and long service life—such as robotics, aerospace actuators, and precision instrumentation—adopting this spatial design philosophy can yield substantial improvements in performance and reliability. The mathematical framework, combining precise deformation analysis, envelope theory, and geometric optimization, along with modern CAD verification, provides a comprehensive and practical toolkit for advancing the state-of-the-art in harmonic drive gear design.