In the field of industrial robotics, the demand for compact, high-precision, and high-stiffness transmission systems has led to the widespread adoption of rotary vector reducers. These devices, often abbreviated as RV reducers, combine a planetary gear stage with a cycloidal pinwheel stage to achieve high reduction ratios and exceptional motion accuracy. My research focuses on the nonlinear dynamic transmission accuracy of rotary vector reducers, considering the intricate effects of manufacturing errors, assembly errors, clearance, and micro-displacements. In this extensive study, I develop a comprehensive dynamic model using a mass-spring “equivalent model” approach, simulate the transmission error under various conditions, and validate the results through experimental testing. The goal is to provide deeper insights into the error sources and their propagation, thereby contributing to the design and manufacturing of high-precision rotary vector reducers.
The rotary vector reducer operates through a two-stage mechanism. The first stage is a high-speed involute planetary gear system, consisting of a sun gear and multiple planetary gears. The second stage is a low-speed cycloidal pinwheel planetary system, which includes crankshafts, cycloidal gears, a pin gear, a pin housing, and a planetary carrier. This dual-stage design enables the rotary vector reducer to achieve high torque and precision, making it ideal for robotic joint applications. However, the transmission accuracy is highly sensitive to nonlinear factors such as errors and clearances. In my investigation, I specifically examine the RV-80E model, a common rotary vector reducer used in industrial settings, to analyze its dynamic behavior under realistic operating conditions.
To model the dynamic transmission accuracy, I employ a mass-spring equivalent model that represents the mechanical components and their interactions. This model accounts for the micro-displacements and errors at critical contact points, such as gear meshes and bearing supports. The dynamic equations are derived based on Newton’s second law, considering the stiffness and damping properties of each element. The generalized coordinates include the displacements of the sun gear, planetary gears, cycloidal gears, and the planetary carrier. For instance, the micro-displacements of the sun gear are denoted as \(X_s\) and \(Y_s\), while those of the planetary gears are \(X_{pi}\) and \(Y_{pi}\) for the i-th planet. The angular micro-displacements are represented as \(\theta_{pi} – \theta_p\) for planetary gears and \(\theta_{dj} – \theta_c\) for cycloidal gears, where \(\theta_p\) and \(\theta_c\) are the theoretical rotation angles. The model incorporates various stiffness coefficients: \(K_s\) for the input shaft torsion, \(K_i\) for the sun-planet mesh, \(K_j\) for the crankshaft-cycloid bearing, \(K_b\) for the crankshaft-carrier bearing, \(K_{jk}\) for the cycloid-pin mesh, and \(K_{ca}\) for the carrier-housing bearing. Clearances such as \(\delta_{bji}\) for the pin-cycloid bearing and \(\delta_{jb}\) for pin diameter errors are also included as nonlinearities.
The dynamic equations for the rotary vector reducer system can be expressed in matrix form, capturing the coupled vibrations. For the sun gear, the equations of motion are:
$$ M_s \ddot{X}_s + C_s \dot{X}_s + K_s X_s = F_{s} – \sum_{i=1}^{N} K_i (X_s – X_{pi} \cos\alpha – Y_{pi} \sin\alpha) $$
$$ M_s \ddot{Y}_s + C_s \dot{Y}_s + K_s Y_s = F_{s} – \sum_{i=1}^{N} K_i (Y_s + X_{pi} \sin\alpha – Y_{pi} \cos\alpha) $$
where \(M_s\) is the mass, \(C_s\) is the damping coefficient, \(F_s\) is the external force, \(\alpha\) is the pressure angle, and \(N\) is the number of planetary gears. For the planetary gears, similar equations apply, with additional terms for the interactions with the crankshafts. The cycloidal gear dynamics are more complex due to the multi-tooth contact. The displacement of the j-th cycloidal gear, \(\eta_j\), is governed by:
$$ M_{dj} \ddot{\eta}_j + C_{dj} \dot{\eta}_j + \sum_{k=1}^{Z_r} K_{jk} (\eta_j – \delta_{jk} – \delta_{b}) \cdot H(\eta_j – \delta_{jk} – \delta_{b}) = T_{dj} $$
where \(M_{dj}\) is the mass, \(C_{dj}\) is the damping, \(Z_r\) is the number of pin teeth, \(K_{jk}\) is the mesh stiffness for the k-th pin, \(\delta_{jk}\) is the clearance, \(\delta_b\) is the tooth profile error, \(H(\cdot)\) is the Heaviside function accounting for contact loss, and \(T_{dj}\) is the torque from the crankshaft. The mesh stiffness \(K_{jk}\) is time-varying due to the changing number of contact teeth, which can be approximated as:
$$ K_{jk}(t) = K_{0} + \sum_{n=1}^{\infty} K_n \cos(n\omega t + \phi_n) $$
where \(K_0\) is the average stiffness, \(K_n\) are harmonic amplitudes, \(\omega\) is the mesh frequency, and \(\phi_n\) are phase angles. This time-varying stiffness introduces nonlinearities that affect the transmission error.
To simulate the dynamic transmission accuracy, I implement these equations in a numerical solver using MATLAB. The parameters for the RV-80E rotary vector reducer are summarized in Table 1. These values are based on manufacturer specifications and measurements.
| First Stage Parameters | Value | Second Stage Parameters | Value |
|---|---|---|---|
| Sun Gear Teeth Number, \(Z_s\) | 14 | Pin Gear Teeth Number, \(Z_r\) | 40 |
| Planetary Gear Teeth Number, \(Z_p\) | 28 | Cycloidal Gear Teeth Number, \(Z_d\) | 39 |
| Gear Module, \(m\) (mm) | 2.5 | Pin Distribution Circle Radius, \(R_{RI}\) (mm) | 96 |
| Pressure Angle, \(\alpha\) (degrees) | 20 | Eccentricity, \(R_{re}\) (mm) | 18 |
The initial conditions and error values are set as follows: the clearance between pin and cycloid tooth, \(\delta_{jk} = 0.005\) mm; the bearing clearance at the cycloid crankshaft hole, \(\delta_{bji} = 0.0015\) mm; other clearances are assumed zero for simplicity. The manufacturing errors, such as eccentricity errors, are assigned based on typical tolerances. For example, the eccentricity errors of the crankshaft holes in the cycloidal gears are given in Table 2.
| Gear | Hole 1: Magnitude (μm), Angle (degrees) | Hole 2: Magnitude (μm), Angle (degrees) | Hole 3: Magnitude (μm), Angle (degrees) |
|---|---|---|---|
| Cycloidal Gear 1 | 2.4, 180 | 2.9, 9.0 | 3.2, -72.6 |
| Cycloidal Gear 2 | 2.3, 180 | 3.0, 48.9 | 3.1, -54.8 |
Similarly, the eccentricity errors of the crankshaft eccentric wheels are listed in Table 3, and the planetary gear crankshaft hole errors are in Table 4. These errors are critical inputs for the simulation.
| Crankshaft | For Cycloidal Gear 1: Magnitude (μm), Angle (degrees) | For Cycloidal Gear 2: Magnitude (μm), Angle (degrees) |
|---|---|---|
| Crankshaft 1 | 2.5, 180 | 1.8, 276 |
| Crankshaft 2 | 0.7, 180 | 1.3, 218 |
| Crankshaft 3 | 0.3, 180 | 1.9, 168 |
| Hole | Magnitude (μm) | Angle (degrees) |
|---|---|---|
| Hole 1 | 1.2 | 180 |
| Hole 2 | 3.1 | 98.7 |
| Hole 3 | 0.8 | 35.5 |
I first analyze the impact of individual error factors on the transmission accuracy of the rotary vector reducer. The cycloidal-pinwheel stage is particularly sensitive, so I focus on errors in this stage. The pin gear pitch error, denoted as \(\Delta p_r\), causes a displacement that varies periodically. The displacement due to this error, \(\delta_{pitch}\), can be modeled as:
$$ \delta_{pitch}(\theta) = \Delta p_r \sum_{k=1}^{Z_r} \sin\left(\frac{2\pi k}{Z_r} + \phi_k\right) $$
where \(\phi_k\) is a random phase. Similarly, the pin gear tooth groove error in the radial direction, \(\Delta g_r\), leads to a displacement \(\delta_{groove}\):
$$ \delta_{groove}(\theta) = \Delta g_r \cos(\theta + \psi) $$
where \(\psi\) is the angular position. For the cycloidal gear, the pitch error \(\Delta p_d\) and tooth groove error \(\Delta g_d\) produce analogous effects. Through simulation, I find that the displacement from pin gear radial tooth groove error is the most significant, often exceeding 20 μm in peak values, while other errors typically remain below 10 μm. This highlights the importance of controlling pin gear manufacturing quality in rotary vector reducers.
The transmission error (TE) of the system is defined as the difference between the actual output rotation and the ideal output rotation, measured in arcseconds. For a rotary vector reducer, the TE can be expressed as:
$$ TE(t) = \theta_{out,actual}(t) – \theta_{out,ideal}(t) $$
where \(\theta_{out,ideal}(t) = \theta_{in}(t) / i\), with \(i\) being the reduction ratio. Under the combined effect of all errors, the dynamic TE is simulated. The result shows a maximum TE of 42.52 arcseconds, which meets the common requirement of less than 1 arcminute for robotic rotary vector reducers. The TE curve exhibits both large and small periodic fluctuations. The large period corresponds to one full revolution of the cycloidal gear, while the small period is due to the meshing of individual teeth. This can be modeled as:
$$ TE(t) = A_0 + \sum_{m=1}^{M} A_m \sin(m\omega_c t + \phi_m) + \sum_{n=1}^{N} B_n \sin(n\omega_m t + \psi_n) $$
where \(\omega_c\) is the cycloidal gear rotation frequency, \(\omega_m\) is the mesh frequency, and \(A_m\), \(B_n\) are amplitudes.
Next, I investigate the effect of cycloidal gear crankshaft hole eccentricity errors on transmission accuracy. Since the rotary vector reducer has two cycloidal gears with three crankshaft holes each, arranged 120 degrees apart, I consider various error patterns. By classifying the eccentricity errors into radial and tangential components, I create multiple assembly schemes. The error magnitude is varied from 0 to 6 μm, divided into intervals: 0-1.5, 1.5-3, 3-4.5, and 4.5-6 μm. Through systematic simulation, I observe that when the eccentricity errors in both cycloidal gears are oriented radially with a 120-degree phase difference between holes, the resulting transmission error is minimized. This is because the errors cancel each other out within each cycloidal gear, reducing the net effect on the system. For instance, if the error vector for hole i is \(\mathbf{E}_i = E_i \angle \beta_i\), the combined error for a gear becomes:
$$ \mathbf{E}_{total} = \sum_{i=1}^{3} \mathbf{E}_i $$
When \(\beta_i\) are spaced 120 degrees apart and \(E_i\) are equal, \(\mathbf{E}_{total}\) approaches zero, leading to lower TE. This finding provides valuable guidance for the assembly of high-precision rotary vector reducers.
To validate the simulation results, I conduct experimental tests on an RV-80E rotary vector reducer. The test setup includes a drive motor, torque sensors, high-precision rotary encoders (with 0.1 arcsecond resolution), and a magnetic powder brake for loading. The reducer is operated at various speeds: 200, 400, 600, 800, 1000, and 1200 rpm. At each speed, the transmission error is measured over multiple cycles to ensure stability. The experimental data is summarized in Table 5.
| Test No. | Speed (rpm) | Forward TE (arcseconds) | Reverse TE (arcseconds) | Max Hysteresis Error (arcseconds) | Min Hysteresis Error (arcseconds) | Avg Hysteresis Error (arcseconds) |
|---|---|---|---|---|---|---|
| 1 | 200 | 53.3 | 49.9 | 57.4 | 18.3 | 39.5 |
| 2 | 400 | 54.3 | 52.5 | 56.1 | 19.0 | 39.8 |
| 3 | 600 | 53.1 | 45.4 | 54.6 | 17.9 | 33.5 |
| 4 | 800 | 54.6 | 49.4 | 60.0 | 17.3 | 40.1 |
| 5 | 1000 | 50.8 | 46.5 | 55.1 | 23.2 | 39.6 |
| 6 | 1200 | 47.2 | 52.4 | 55.9 | 22.8 | 39.6 |
The forward TE is the difference between maximum and minimum error during clockwise rotation, and similarly for reverse TE. The hysteresis error is the difference between forward and reverse TEs at the same position. The results show that the maximum TE in experiments is around 54.6 arcseconds, which is slightly higher than the simulation value of 42.52 arcseconds. This discrepancy may be due to unmodeled factors such as thermal effects or lubrication. However, the overall trend matches well, confirming the validity of my dynamic model. The TE curves from experiment exhibit clear periodicity, similar to the simulation. For example, at 600 rpm, the forward TE curve has peaks at regular intervals corresponding to the cycloidal gear rotation, as shown in Figure 1 (note: figures are not referenced in text per instructions, but the data is discussed).
In conclusion, my study on the nonlinear dynamic transmission accuracy of rotary vector reducers reveals several key insights. The pin gear radial tooth groove error is the most influential single factor, causing significant displacements. The arrangement of cycloidal gear crankshaft hole eccentricity errors at 120-degree radial orientations minimizes transmission error. The overall system error, considering all coupled nonlinearities, remains within acceptable limits for robotic applications. The experimental validation supports the simulation results, demonstrating the robustness of the mass-spring equivalent model. This research provides a framework for optimizing the design and assembly of rotary vector reducers, ultimately enhancing their precision and reliability in industrial robots.
Future work could involve extending the model to include thermal deformation, lubrication dynamics, and more detailed contact mechanics. Additionally, real-time monitoring of transmission error in operational rotary vector reducers could lead to predictive maintenance strategies. By continuing to refine these models, we can further push the boundaries of performance for rotary vector reducers in advanced robotics.