The strain wave gear, also known as a harmonic drive, represents a unique class of gearing that relies on the controlled elastic deformation of a thin-walled component—the flexspline—to transmit motion and torque. Its exceptional characteristics, including high reduction ratios, compactness, zero backlash, and coaxial shaft design, have cemented its indispensable role in demanding applications such as space robotics, precision instrumentation, and industrial automation. The core of this transmission’s performance and its primary point of failure is the flexspline. Among various configurations, the cup-shaped flexspline is widely utilized due to its favorable stiffness and torque capacity.
Accurately predicting the stress state within a loaded flexspline is a formidable challenge. Traditional analytical approaches are often based on simplified cylindrical shell theory, which necessitates significant assumptions regarding boundary conditions and the influence of the gear teeth. The interaction between the flexspline and the circular spline involves multiple, simultaneously engaged tooth pairs, leading to a complex, statically indeterminate load distribution. Furthermore, the flexspline undergoes a traveling wave of deformation, making experimental stress measurement difficult and impractical during the design phase. Consequently, conventional analytical methods provide only approximate results, and the industry lacks a unified, reliable strength calculation method. Premature fatigue failure, particularly cracking in the gear rim, remains a persistent and costly problem in strain wave gear applications.
This investigation focuses on a radially engaging, cup-shaped flexspline with a cam-type wave generator. The primary objective is to move beyond simplified analytical models and employ a high-fidelity, three-dimensional finite element analysis (FEA) to comprehensively evaluate the stress and deformation of the flexspline under load. By constructing a realistic simulation model that accounts for elastic contact between multiple tooth pairs, this study aims to reveal the detailed stress distribution in both the gear rim and the cylindrical shell, elucidate the deformation mechanics of the teeth, and ultimately identify the root causes of failure in the design stage. The insights gained will be used to refine key design parameters, such as the tooth influence coefficient, thereby providing a more precise and efficient methodology for the analysis and design of robust strain wave gear flexsplines.
Theoretical Foundation and Analytical Stress Calculation
Before delving into the finite element analysis, it is instructive to review the standard analytical framework for stress calculation in a cup-shaped flexspline. This approach, while approximate, establishes a baseline for comparison. The methodology typically treats the flexspline body as a cylindrical shell and incorporates the effect of the teeth through an empirically or analytically derived “tooth influence coefficient.” The major stress components considered are bending stresses induced by the wave generator’s deformation and the meshing forces.
For a given flexspline geometry and loading condition, the maximum circumferential bending stress at the critical section can be expressed as a summation of several components:
$$ \sigma_{\theta \text{max}} = Y_z [\sigma_{t}(\phi=0^\circ) + \sigma_k] + \sigma_p + \sigma_{3y} $$
where:
– $\sigma_{t}(\phi=0^\circ)$ is the bending stress due to the nominal radial deflection of the wave generator at the major axis ($\phi = 0^\circ$).
– $\sigma_k$ is an additional bending stress component accounting for manufacturing tolerances and assembly gaps.
– $\sigma_p$ is the bending stress resulting from the tangential meshing force (torque transmission).
– $\sigma_{3y}$ is the local bending stress in the tooth space (web) due to the tooth load.
– $Y_z$ is the tooth influence coefficient, a critical factor that amplifies the shell stresses to account for the presence of the teeth. It is defined as:
$$ Y_z = \frac{1}{1 – \xi[1 – (s_1 / h_0)^2]} $$
In this coefficient, $s_1$ is the tooth root thickness, $h_0$ is the nominal wall thickness of the rim, and $\xi$ is a parameter related to the equivalent tooth geometry. The mean circumferential bending stress, $\sigma_{\theta m}$, and the stress amplitude, $\sigma_a$, are also calculated for fatigue life prediction. Similarly, the torsional shear stress, $\tau$, and its amplitude are evaluated.
Applying this theoretical model to a specific case study (a flexspline with module m=1.5 mm, 202 teeth, transmitting 800 Nm torque), the key calculated stresses are summarized below. The material is 30CrMnSiA steel.
| Stress Component | Symbol | Calculated Value (MPa) |
|---|---|---|
| Max. Circumferential Bending Stress | $\sigma_{\theta \text{max}}$ | 261.5 |
| Mean Circumferential Bending Stress | $\sigma_{\theta m}$ | 76.5 |
| Bending Stress Amplitude | $\sigma_a$ | 187.1 |
| Tooth Influence Coefficient | $Y_z$ | 1.230 |
| Max. Shear Stress (Torsion) | $\tau_{\text{max}}$ | 66.1 |
| Shear Stress Amplitude | $\tau_a$ | 66.1 |
Based on these stresses and using the distortion energy theory for static strength and a Goodman-type criterion for fatigue, the theoretical safety factors were found to be adequate for static load ($n \approx 1.9$) but potentially borderline or insufficient for high-cycle fatigue depending on the assumed fatigue limits. This approximate calculation, however, masks the complex three-dimensional stress state and the true nature of stress concentrations, prompting the need for a more detailed analysis.
High-Fidelity Finite Element Modeling of the Strain Wave Gear
To overcome the limitations of analytical methods, a detailed three-dimensional nonlinear finite element model of the engaged flexspline and circular spline was developed. The primary goal was to simulate the actual operating conditions as closely as possible, minimizing simplifying assumptions.
1. Geometry and Meshing: A precise, parametrically generated 3D solid model of the cup-shaped flexspline with a wide tooth space design was created. The model included the full gear rim, the transitional fillet, and the cylindrical shell. A segment of the circular spline corresponding to the engagement zone was also modeled. The region of engagement was defined based on the nominal wave generator deflection ($\omega_{0r}=1.1 \text{mm}$), covering an angle of $\pm52.5^\circ$ from the major axis, encompassing approximately 25 tooth pairs. The finite element mesh was refined in critical areas, particularly at the tooth roots, the rim-shell junction, and the contact surfaces. The gear teeth were discretized using SOLID95 higher-order 3D elements, which can accurately model curved boundaries.
2. Contact Modeling: This is the most crucial aspect of the simulation. The interaction between the flexspline and circular spline teeth was modeled as a surface-to-surface elastic contact problem. The flexspline tooth surfaces were defined as “contact” surfaces (CONTA174 elements), and the circular spline tooth surfaces as “target” surfaces (TARGE170 elements). A Lagrange multiplier algorithm was employed to enforce the contact constraints, which is well-suited for problems where the contact area is unknown beforehand. The contact formulation accounted for the possibility of sticking and sliding friction, though a frictionless condition was initially assumed to isolate the stress effects. The governing variational principle for the contact problem incorporating Lagrange multipliers $\lambda$ is:
$$ \Pi = \Pi_0 + \Pi_c = \Pi_0 + \int_{\Gamma_c} \lambda^T (g) d\Gamma $$
where $\Pi_0$ is the total potential energy without contact, $\Pi_c$ is the contact constraint functional, $g$ is the gap function, and $\Gamma_c$ is the contact boundary. The resulting discrete finite element equations take the incremental form:
$$
\begin{bmatrix}
\mathbf{K} & \mathbf{K}_c^T \\
\mathbf{K}_c & 0
\end{bmatrix}
\begin{Bmatrix}
\Delta \mathbf{u} \\
\Delta \boldsymbol{\lambda}
\end{Bmatrix}
=
\begin{Bmatrix}
^{t+\Delta t}\mathbf{R} – ^{t}\mathbf{F} \\
-^{t}\mathbf{g}
\end{Bmatrix}
$$
Here, $\mathbf{K}$ is the standard stiffness matrix, $\mathbf{K}_c$ is the constraint matrix related to contact, $\Delta \mathbf{u}$ is the displacement increment, $\Delta \boldsymbol{\lambda}$ is the increment of contact forces (Lagrange multipliers), $\mathbf{R}$ is the external load vector, and $\mathbf{F}$ is the internal force vector.
3. Loads and Boundary Conditions: The circular spline was fixed in space. A torque of 800 Nm was applied to the flexspline’s output end (the shell’s open end) to simulate the output loading condition. The action of the cam-type wave generator was modeled not by explicitly modeling the cam, but by applying a distributed radial pressure load to the inner surface of the flexspline’s cylindrical shell. This pressure distribution was derived from the theoretical radial force function of the wave generator, which varies sinusoidally in the circumferential direction and is uniform axially over the contact length. This approach directly incorporates the effects of the wave generator’s preload and deformation.
4. Solution Strategy: The analysis accounted for both material and geometric nonlinearities (due to large deformation relative to wall thickness). A full Newton-Raphson iterative procedure was used to achieve convergence. The problem was solved as a static analysis, representing a snapshot of the stress state at a specific angular position of the wave generator (major axis at a defined location).
Finite Element Results: Stress, Deformation, and Mechanism
The results from the nonlinear FEA provide a detailed and revealing picture of the flexspline’s behavior under load, offering significant insights not available from analytical calculations.
1. Tooth Deformation and Load Distribution: A key finding is the non-uniform deformation of the individual gear teeth along their face width. The following table shows the radial ($\delta_r$) and circumferential ($\delta_\theta$) deformation at the tooth root for nodes along the face width at two different engagement positions: near the entry zone ($\phi = 39^\circ$) and near the major axis ($\phi = -5.4^\circ$).
| Position Along Face Width (Normalized) | $\phi = 39^\circ$ (Entry) | $\phi = -5.4^\circ$ (Near Major Axis) | ||
|---|---|---|---|---|
| $\delta_r$ (mm) | $\delta_\theta$ (mm) | $\delta_r$ (mm) | $\delta_\theta$ (mm) | |
| Front (0.0) | -0.03299 | -0.03499 | 0.08947 | -0.01838 |
| 0.25 | -0.02926 | -0.03468 | 0.08929 | -0.01828 |
| 0.50 | -0.02690 | -0.03332 | 0.08645 | -0.01824 |
| 0.75 | -0.02498 | -0.03190 | 0.08353 | -0.01807 |
| Rear (1.0) | -0.02302 | -0.02990 | 0.07860 | -0.01797 |
The data indicates that the tooth deformation is not symmetric. At the entry zone, the radial contraction is larger at the front of the tooth, while near the major axis, the radial expansion is larger at the front. More importantly, the circumferential deformation shows that the active tooth flank shifts from the rear flank to the front flank as the tooth moves from entry into the main engagement zone. This leads to a cyclic reversal of the bending load on the tooth root and a non-uniform load distribution across the face width, contributing significantly to stress concentration and fatigue initiation. This phenomenon also explains the observed “run-in” wear patterns near the edges of the flexible bearing raceway in physical assemblies.
2. Stress Distribution in the Cylindrical Shell: The FEA clearly maps the decay of stress away from the high-stress gear rim. The maximum circumferential bending stress in the shell occurs at its junction with the gear rim. At a location just 20 mm axially away from this junction, the stress magnitude drops by over 60%.
$$ \frac{\sigma_{\theta}(z=20\text{mm})}{\sigma_{\theta \text{max}}(z=0\text{mm})} \approx 0.38 $$
This rapid decay confirms that the rim is the critical region. Furthermore, the presence of the teeth modifies the stress field in the adjacent shell. A comparative analysis between the model with teeth and a simplified model where tooth loads were applied as nodal forces at the theoretical root circle showed that the teeth primarily amplify the stress magnitude rather than drastically alter its distribution pattern. The “tooth influence” is largely confined to a region within approximately 20 mm of the rim.
3. Stress Concentration in the Gear Rim: The most critical results concern the stress in the gear rim itself. The FEA predicts the maximum circumferential bending stress location to be in the tooth space (web) between the 3rd and 4th tooth from the major axis, near the front face of the gear rim. The stress distribution around the circumference at this critical section is highly irregular due to the discrete nature of the teeth. The predicted maximum stress value from FEA is significantly higher than the theoretical calculation:
$$ \sigma_{\theta \text{max}}^{FEA} = 448.3 \text{ MPa}, \quad \sigma_{\theta \text{max}}^{Theory} = 261.5 \text{ MPa} $$
The stress amplitude for fatigue calculation is also correspondingly higher: $\sigma_a^{FEA} = 355.8 \text{ MPa}$.
4. Reevaluation of the Tooth Influence Coefficient: The substantial discrepancy between FEA and theoretical stress values points directly to an underestimation in the analytical model. The primary source of error is identified as the tooth influence coefficient $Y_z$. Using the FEA results from a model with only wave generator loading (no output torque) and a comparable model without teeth, the actual influence of the teeth can be isolated. The FEA-derived tooth influence coefficient is calculated as:
$$ Y_z^{FEA} = \frac{\sigma_{\theta \text{max}}^{(\text{With Teeth, No Torque})}}{\sigma_{\theta \text{max}}^{(\text{No Teeth, Same Deflection})}} = \frac{334.2 \text{ MPa}}{206.4 \text{ MPa}} \approx 1.619 $$
Comparing this to the theoretically calculated value of 1.230 reveals an underestimation of approximately:
$$ \frac{1.619 – 1.230}{1.619} \times 100\% \approx 24\% $$
Strength Prediction and Failure Analysis
Using the more accurate stress data from the finite element analysis, a revised strength assessment is performed. The safety factor for static strength, based on the distortion energy theory, remains acceptable. However, the fatigue safety factor, calculated for combined alternating bending and mean stress, falls below the typical required minimum of 1.3.
$$ n_{\sigma} = \frac{\sigma_{-1}}{k_{\sigma} \sigma_a + \psi_{\sigma} \sigma_m} \quad \text{and} \quad n_{\tau} = \frac{\tau_{-1}}{k_{\tau} \tau_a + \psi_{\tau} \tau_m} $$
$$ n_r = \frac{n_{\sigma} n_{\tau}}{\sqrt{n_{\sigma}^2 + n_{\tau}^2}} \approx 1.07 < [n_r] $$
Where $\sigma_{-1}, \tau_{-1}$ are the material’s fatigue limits, $k_{\sigma}, k_{\tau}$ are fatigue stress concentration factors, and $\psi_{\sigma}, \psi_{\tau}$ are sensitivity coefficients. The low predicted fatigue safety factor correlates with the observed failure mode in practical applications: fatigue cracks initiating in the tooth root fillet region near the front face of the gear rim, subsequently propagating through the rim. The FEA successfully pinpoints this exact location as the point of maximum stress.
If the theoretical model is corrected by using the FEA-derived tooth influence coefficient $Y_z^{FEA} \approx 1.62$, the recalculated maximum theoretical stress becomes:
$$ \sigma_{\theta \text{max}}^{Corrected} = 1.62 \times [\sigma_t + \sigma_k] + \sigma_p + \sigma_{3y} \approx 310.0 \text{ MPa} $$
While still lower than the FEA maximum (due to other assumptions in the shell theory), this corrected value leads to a fatigue safety factor prediction that is also below the required threshold, aligning the theoretical assessment with the FEA prediction and observed failure trends.
Conclusion
This detailed three-dimensional nonlinear finite element analysis of a cup-shaped flexspline in a strain wave gear has provided critical insights that are unattainable through conventional analytical methods. The study demonstrates that the maximum stress is located in the gear rim, specifically at the tooth root fillet near the front face. The influence of the teeth is primarily to amplify the stress magnitude in the adjacent shell and rim, with this effect concentrated within a region roughly 20 mm from the rim. The complex, non-uniform deformation of the teeth along their face width leads to cyclic load reversal and localized stress concentrations, which are the primary drivers of fatigue crack initiation.
The key finding is that standard analytical calculations, reliant on simplified cylindrical shell theory and an empirical tooth influence coefficient, systematically underestimate the peak stress levels. In the case studied, the theoretical tooth influence coefficient was under-predicted by about 24%, leading to non-conservative fatigue life estimates. The FEA methodology presented—incorporating realistic geometry, multi-tooth elastic contact, and accurate loading—serves as a powerful tool to correct these coefficients, validate designs, and predict failure locations with high precision.
For designers of strain wave gears, this work underscores the necessity of using advanced simulation techniques to supplement traditional calculations, especially for high-reliability applications. Future work could involve dynamic FEA to simulate the traveling stress wave, thermal-structural coupling for high-speed applications, and probabilistic analysis to account for manufacturing variances, further enhancing the predictive power and reliability of strain wave gear design.