In the field of precision mechanical transmission, the harmonic drive gear stands out as a revolutionary technology that relies on the elastic deformation of a thin-shell component, known as the flexspline, to transmit motion and power. As an engineer deeply involved in this area, I have encountered numerous challenges in accurately predicting the stress and deformation behaviors of the flexspline, particularly the column-shaped variant. Traditional methods, based on simplified cylindrical shell theories, often yield approximate results that fail to capture the complex interactions in real-world applications, leading to persistent issues such as cracking in the ring gear of the flexspline. This article presents a comprehensive study using three-dimensional elastic contact finite element analysis to simulate and analyze the stress distribution, deformation patterns, and strength of the column-shaped flexspline in a harmonic drive gear system. By developing a realistic entity model and employing advanced contact mechanics, I aim to provide a more precise and efficient approach to understanding and mitigating flexspline failures, ultimately enhancing the reliability and performance of harmonic drive gear systems.
The harmonic drive gear operates on the principle of controlled elastic deformation, where a wave generator induces radial displacement in the flexspline, enabling meshing with a rigid circular spline. This mechanism offers advantages like high reduction ratios, compact design, and minimal backlash, making it invaluable in aerospace, robotics, and industrial automation. However, the flexspline, being a critical load-bearing element, undergoes complex stress states due to simultaneous multi-tooth engagement and cyclic deformation. In my experience, the column-shaped flexspline, commonly used in cup-style harmonic drive gear reducers, is particularly susceptible to fatigue cracking at the ring gear section, often stemming from inadequate stress predictions during design. Theoretical calculations, while useful, rely on assumptions such as simplified geometry and empirical coefficients—like the tooth influence factor—that may not reflect actual conditions. For instance, the traditional approach approximates the maximum circumferential bending stress using formulas derived from cylindrical shell theory, but it neglects detailed tooth deformation effects and boundary interactions. To address this, I turned to finite element analysis (FEA), which allows for a more detailed simulation of the harmonic drive gear’s behavior under load.
In this study, I focus on a harmonic drive gear reducer with a radial meshing, internal four-force cam-type wave generator, where the circular spline is fixed, and the flexspline serves as the output. The reducer has a transmission ratio of 10, an output torque of 800 N·m, and uses a column-shaped flexspline made of 30CrMnSiNi2A steel, subjected to quenching, tempering, and shot peening. Key parameters include a module of 0.8 mm, 202 teeth, a pressure angle of 20°, and a radial deformation coefficient of 1.1. The wave generator operates at 1,450 rpm with an action angle of 25°, rotating clockwise. Starting from the long-axis meshing zone at a position angle φ = 0°, I analyze the flexspline’s response. Theoretical stress calculations, based on simplified methods, yield the maximum circumferential bending stress as follows, incorporating the tooth influence factor to account for gear effects:
$$ \sigma_{hmax} = K_z \left[ \sigma_{t}(\phi=0^\circ) + \sigma_k \right] + \sigma_p + \sigma_{3s} = 261.489 \, \text{MPa} $$
where \( K_z \) is the tooth influence factor, calculated as:
$$ K_z = \frac{1}{1 – \xi \left[1 – (s_1 / s_h)^2\right]} = 1.23046 $$
Here, \( \xi \) is the tooth thickness influence coefficient, \( s_1 \) is the root thickness of the flexspline tooth, \( s_h \) is the wall thickness at the ring gear, and other terms represent stresses from wave generator forces. The average bending stress and shear stress are derived similarly, but these values often underestimate actual stresses due to simplifications. For example, the bending stress amplitude is computed as:
$$ \sigma_a = \frac{1}{2} \left[ \sigma_{t}(\phi=0^\circ) – \sigma_{t}(\phi=90^\circ) \right] + \sigma_k = 187.088 \, \text{MPa} $$
and the shear stress amplitude as:
$$ \tau_a = \frac{T}{2 \pi r^2 \delta} = 9.7 \, \text{MPa} $$
However, these calculations do not fully capture the three-dimensional contact and deformation effects inherent in the harmonic drive gear, prompting the need for a finite element approach.
To overcome the limitations of theoretical models, I developed a detailed finite element model of the harmonic drive gear system using CAD software and imported it into ANSYS for analysis. The model includes the column-shaped flexspline and circular spline with realistic tooth profiles, meshed with SOLID95 elements for the ring gear to accurately represent tooth boundaries. For the contact simulation, I employed surface-to-surface contact elements, where the flexspline teeth are defined as flexible contact surfaces (CONTA174) and the circular spline teeth as rigid target surfaces (TARGE170). Using APDL language, I created contact pairs across the entire tooth profile from tip to root, covering 25 tooth pairs in the meshing zone defined by \( -5\pi/24 < \phi < \pi/24 \). This setup allows for a precise simulation of the elastic contact problem in the harmonic drive gear. The wave generator’s reaction forces, derived from experimental data to include effects like clearance and ball distribution, are applied to the flexspline’s inner surface using surface effect elements (SURF154). A distributed load function is implemented to represent these forces circumferentially and axially. The circular spline is subjected to a driving torque via a pilot node, while the flexspline’s cylindrical end is constrained in all degrees of freedom. Given the geometric nonlinearity (radial deformation-to-thickness ratio > 0.2) and contact nonlinearity, I used a full Newton-Raphson method with Lagrangian multipliers for contact resolution, ensuring convergence through iterative checks. The finite element equation in incremental form is expressed as:
$$ \begin{bmatrix} \mathbf{K} + \mathbf{K}_c \end{bmatrix} \begin{bmatrix} \Delta \mathbf{u} \\ \Delta \lambda \end{bmatrix} = \begin{bmatrix} \mathbf{F}_{t+\Delta t} \\ -\mathbf{g}_t \end{bmatrix} $$
where \( \mathbf{K} \) is the stiffness matrix, \( \mathbf{K}_c \) is the contact stiffness matrix, \( \Delta \mathbf{u} \) is the displacement increment, \( \Delta \lambda \) is the Lagrange multiplier (representing contact forces), \( \mathbf{F}_{t+\Delta t} \) is the load vector, and \( \mathbf{g}_t \) is the gap at time t. Contact states (stick, slip, or separation) are determined based on criteria such as normal force and friction, with updates in each increment. This approach enables a comprehensive analysis of the harmonic drive gear’s behavior under static loading, providing insights into stress distribution and deformation that are unattainable through theoretical methods alone.
The finite element results reveal critical aspects of the harmonic drive gear’s performance, particularly regarding tooth deformation and stress patterns. In the column-shaped flexspline, tooth deformation primarily occurs at the tooth slots due to their lower stiffness relative to the tooth bodies. I analyzed radial and circumferential deformations at key positions, such as the meshing zone near φ = 39° and φ = -5.4°, using nodes along the tooth width from the front to the back face. The results, summarized in the table below, show that deformation varies along the tooth width, indicating uneven loading and tooth tilting during engagement.
| Position Along Tooth Width (mm) | Radial Deformation at φ=39° (mm) | Circumferential Deformation at φ=39° (mm) | Radial Deformation at φ=-5.4° (mm) | Circumferential Deformation at φ=-5.4° (mm) |
|---|---|---|---|---|
| 1.00 | -0.032991 | 0.89470 | -0.01838 | 0.03499 |
| 5.00 | -0.02926 | 0.89292 | -0.01828 | 0.03468 |
| 10.00 | -0.02690 | 0.86452 | -0.01824 | 0.03332 |
| 15.00 | -0.02498 | 0.83530 | -0.01807 | 0.03190 |
| 20.00 | -0.02302 | 0.78601 | -0.01797 | 0.02990 |
| 25.00 | -0.02191 | 0.74964 | -0.01753 | 0.02839 |
From this data, it is evident that at φ = 39°, the circumferential deformation causes the back tooth flank to become the working surface, while at φ = -5.4°, the front flank dominates due to higher deformation. This alternating load pattern during meshing increases the risk of fatigue cracking in the ring gear, particularly near the front face tooth slots. Moreover, the radial deformation shows that the front of the ring gear experiences greater contraction or expansion than the back, leading to contact with the flexible bearing edges—a phenomenon observed in practice but difficult to capture theoretically. The overall deformation of the harmonic drive gear’s flexspline, visualized in the model, indicates that the tooth slot generatrix remains largely straight, with no significant curvature at the junction between the ring gear and cylinder, suggesting that stress concentrations may arise elsewhere.
Stress analysis from the finite element simulation provides detailed insights into the column-shaped flexspline’s behavior in the harmonic drive gear. The circumferential bending stress in the cylinder, at different axial positions, shows a decay with distance from the ring gear. For instance, at the junction of the ring gear back face and cylinder, the maximum stress is 217.99 MPa, while 20 mm away, it drops to 82.227 MPa—a reduction of 62.3%. This trend aligns with experimental observations and highlights the localized effect of tooth engagement. The stress distribution exhibits dual peaks near the long-axis meshing zone due to tooth influence, as illustrated in the plot below. When teeth are removed from the model and equivalent forces are applied at the tooth roots, the stress pattern remains similar but with reduced magnitude, confirming that teeth primarily amplify stress levels rather than alter distribution. Specifically, the circumferential bending stress in a toothless flexspline under the same deformation is 163.67 MPa, compared to 217.99 MPa in the toothed version—a factor of 1.33 increase. This underscores the importance of accurately accounting for tooth effects in harmonic drive gear design.
The ring gear itself experiences significant stress variations. The maximum circumferential bending stress on the ring gear’s circumference is 448.324 MPa, with a minimum of -263.356 MPa, indicating high cyclic loading. The shear stress extremes are 66.074 MPa and -64.352 MPa, nearly symmetric about zero. These values, derived from finite element analysis, exceed theoretical predictions and help identify critical failure locations. Based on these results, I performed strength evaluations for the harmonic drive gear’s flexspline. Using the third strength theory, the equivalent stress is computed as:
$$ \sigma_{eq} = \sqrt{\sigma_{max}^2 + 4\tau_{max}^2} = 448.324 \, \text{MPa} $$
with a static safety factor of:
$$ n = \frac{\sigma_s}{\sigma_{eq}} = 1.89 > [n] \ge 1.4 $$
where \( \sigma_s = 850 \, \text{MPa} \) is the yield strength. This satisfies static strength requirements. However, fatigue strength is more critical due to cyclic stresses. For bidirectional stable varying stress, the safety factor is given by:
$$ n_r = \frac{1}{\sqrt{\left( \frac{\sigma_a}{\sigma_{-1}} + \frac{\psi_\sigma \sigma_m}{\sigma_b} \right)^2 + \left( \frac{\tau_a}{\tau_{-1}} + \frac{\psi_\tau \tau_m}{\tau_b} \right)^2}} $$
where \( \sigma_a = 355.84 \, \text{MPa} \) and \( \tau_a = 66.074 \, \text{MPa} \) are stress amplitudes, \( \sigma_m = 92.454 \, \text{MPa} \) and \( \tau_m = 0 \, \text{MPa} \) are mean stresses, \( \sigma_{-1} = 480 \, \text{MPa} \) and \( \tau_{-1} = 275 \, \text{MPa} \) are fatigue limits, and \( \psi_\sigma, \psi_\tau \) are influence coefficients. With stress concentration factors \( k_\sigma = 1.9 \) and \( k_\tau = 1.425 \), the calculated fatigue safety factor is \( n_r = 0.968 < [n_r] \ge 1.3 \), indicating insufficient fatigue resistance. This correlates with observed cracking in harmonic drive gear flexsplines, typically initiating at tooth slots near the front face and propagating axially and circumferentially.
To address the discrepancy between theoretical and finite element results, I investigated the tooth influence factor in the harmonic drive gear. Theoretical calculations yielded \( K_z = 1.23046 \), but finite element analysis of a flexspline under wave generator forces alone showed a maximum circumferential bending stress of 334.157 MPa for the toothed version versus 206.431 MPa for a toothless version with identical deformation. This implies an actual tooth influence factor of:
$$ K_{z,actual} = \frac{334.157}{206.431} = 1.618 $$
which is 23.99% higher than the theoretical value. This underestimation explains why theoretical stress predictions are lower, leading to premature failures in harmonic drive gear applications. Revised stress calculations using the corrected factor yield:
$$ \sigma_{hmax,revised} = K_{z,actual} \left[ \sigma_{t}(\phi=0^\circ) + \sigma_k \right] + \sigma_p + \sigma_{3s} = 310.011 \, \text{MPa} $$
and a revised bending stress amplitude of 229.113 MPa. Re-evaluating strength with these values still results in a fatigue safety factor below the required threshold, confirming that the harmonic drive gear’s flexspline is prone to cracking under design loads. This analysis highlights the need for refined coefficients in traditional methods, but more importantly, it advocates for the use of finite element analysis as a standard tool in harmonic drive gear design.
In conclusion, this study demonstrates the efficacy of three-dimensional finite element analysis in understanding the complex behavior of the column-shaped flexspline in harmonic drive gear systems. Through detailed modeling of contact mechanics and nonlinear deformations, I have identified key factors contributing to stress concentrations and fatigue failure, such as tooth deformation patterns and the influence of the ring gear on cylinder stresses. The harmonic drive gear’s performance is intricately linked to the flexspline’s design, and traditional theoretical approaches, while useful, often fall short due to simplifications like the tooth influence factor. By integrating finite element results, I have provided a more accurate assessment of stress distributions and strength limits, enabling better prediction of failure sites—typically at tooth slots near the front face of the ring gear. This method offers a robust framework for optimizing harmonic drive gear components, potentially extending their lifespan and reliability in demanding applications. Future work could explore dynamic analyses or material variations to further enhance the harmonic drive gear’s capabilities, but for now, this finite element approach stands as a vital step toward more reliable and efficient harmonic drive gear systems.