In the field of precision engineering, particularly for industrial robotics, the RV reducer plays a critical role in ensuring high torque transmission and positional accuracy. As a key component within the RV reducer, the main bearing significantly influences its mechanical performance, including load capacity, fatigue life, and operational stability. However, the design and theoretical analysis of the main bearing in RV reducers are still in nascent stages, often relying on imitation rather than innovative optimization. To advance the development process and enhance performance, a multi-objective optimization approach grounded in thorough theoretical analysis is essential. This paper focuses on establishing a quasi-static model for the RV reducer main bearing and proposes an efficient iterative solution method to address the computational challenges posed by its multi-ball characteristics.
The RV reducer main bearing features a multi-ball design, which, while beneficial for distributing loads and improving precision, results in a quasi-static model with a large-scale system of nonlinear equations. Compared to traditional bearings, the RV reducer main bearing’s model involves multiple degrees of freedom and numerous contact points, leading to convergence difficulties under extreme working conditions. Specifically, the model must account for geometric relationships, motion parameters, and force equilibria across all balls, making it computationally intensive. In optimization routines like NSGA-II, where the quasi-static model is evaluated repeatedly for each candidate design, computational efficiency becomes paramount. Thus, developing a rapid and robust iterative method is crucial for practical applications.
To address these challenges, we first derive a 5-degree-of-freedom (5-DOF) quasi-static model for the RV reducer main bearing. This model incorporates external loads, centrifugal forces, gyroscopic moments, and contact deformations between balls and raceways. The geometry of the bearing is simplified by considering the angular positions of the balls, denoted as $\psi_j = 2\pi(j-1)/z$ for the $j$-th ball, where $z$ is the number of balls. For an RV reducer main bearing, $z$ is typically around 51, leading to a system of equations with $4j + 5$ unknowns per ball set, totaling over 200 variables for the full bearing. The key equations include the distances between raceway curvature centers, contact deformations, and force balances. For instance, the axial and radial distances between the inner and outer raceway curvature centers at position $j$ are given by:
$$A_{1j} = l_{BD} \sin \alpha + \delta_x + R_i \theta_z \sin \psi_j – R_i \theta_y \cos \psi_j$$
$$A_{2j} = l_{BD} \sin \alpha + \delta_z \cos \psi_j + \delta_y \sin \psi_j$$
where $\delta_x$, $\delta_y$, $\delta_z$, $\theta_y$, and $\theta_z$ represent displacements and rotations, $R_i$ is the inner race radius, and $\alpha$ is the initial contact angle. The contact deformations $\delta_{ij}$ and $\delta_{oj}$ for inner and outer raceways relate to the ball loads via the Hertzian contact theory:
$$Q_{ij} = K_{ij} \delta_{ij}^{1.5}, \quad Q_{oj} = K_{oj} \delta_{oj}^{1.5}$$
Here, $K_{ij}$ and $K_{oj}$ are load-displacement coefficients, and $Q_{ij}$ and $Q_{oj}$ are the contact loads. The force equilibrium for each ball involves centrifugal forces $F_{Cj}$ and gyroscopic moments, expressed as:
$$F_{Cj} = 0.5 m D_m \omega^2 (\omega_m / \omega)^2$$
where $m$ is the ball mass, $D_m$ is the pitch diameter, $\omega$ is the rotational speed, and $\omega_m$ is the orbital speed. The balance equations for the entire bearing in 5-DOF form are:
$$F_x – \sum_{j=1}^z (Q_{ij} \sin \alpha_{ij} – F_{ij} \cos \alpha_{ij}) = 0$$
$$F_y – \sum_{j=1}^z (Q_{ij} \cos \alpha_{ij} + F_{ij} \sin \alpha_{ij}) \sin \psi_j = 0$$
$$F_z – \sum_{j=1}^z (Q_{ij} \cos \alpha_{ij} + F_{ij} \sin \alpha_{ij}) \cos \psi_j = 0$$
$$M_y + \sum_{j=1}^z [(Q_{ij} \sin \alpha_{ij} – F_{ij} \cos \alpha_{ij}) R_i + F_{ij} f_i] \cos \psi_j = 0$$
$$M_z – \sum_{j=1}^z [(Q_{ij} \sin \alpha_{ij} – F_{ij} \cos \alpha_{ij}) R_i + F_{ij} f_i] \sin \psi_j = 0$$
These equations, combined with geometric constraints, form a nonlinear system that is challenging to solve, especially for the RV reducer main bearing with its many balls.
To expedite the solution, we optimize the initial guess for the iterative process. Since the RV reducer operates at relatively low speeds, static analysis provides a reasonable approximation for displacements and contact angles. The static model solves for axial displacement $\delta_a$, radial displacement $\delta_r$, and misalignment angle $\theta$ under external loads. Using these, we estimate the quasi-static variables such as $X_{1j}$, $X_{2j}$, $\delta_{ij}$, and $\delta_{oj}$ through derived relations. For example:
$$X_{1j} = [(f_o – 0.5)D + \delta_{oj}] \sin \alpha_{oj}$$
$$X_{2j} = \sqrt{[(f_o – 0.5)D + \delta_{oj}]^2 – X_{1j}^2}$$
where $D$ is the ball diameter, $f_o$ is the outer race curvature coefficient, and $\alpha_{oj}$ is approximated from static results. This optimized initial guess reduces the sensitivity of the iterative solver and enhances convergence, which is vital for the RV reducer main bearing’s complex model.
For solving the nonlinear system, we propose a combined Newton-Broyden iterative algorithm. The Newton-Raphson method offers quadratic convergence but requires computing and inverting the Jacobian matrix at each step, which is computationally expensive for large systems like that of the RV reducer main bearing. The Broyden method, a quasi-Newton approach, approximates the Jacobian and achieves superlinear convergence with less computational overhead. However, it is sensitive to initial guesses and may diverge if poorly chosen. Our combined algorithm leverages the strengths of both: we use a damped Newton-Raphson method initially to ensure convergence from the optimized guess, then switch to Broyden’s method for faster refinement. The steps are as follows:
- Start with the optimized initial vector $\mathbf{x}_0$ from static analysis.
- Apply a damped Newton-Raphson iteration: $\mathbf{x}_{i+1} = \mathbf{x}_i – \lambda_G [\mathbf{F}'(\mathbf{x}_i)]^{-1} \mathbf{F}(\mathbf{x}_i)$, where $\lambda_G < 1$ is a global damping factor (e.g., 0.2) to avoid singularities.
- Introduce an acceleration factor $\beta > 1$ (e.g., 2) to speed up convergence once stable.
- Incorporate a local damping factor $\lambda_0$ for backtracking if the residual increases, ensuring descent direction: if $||\mathbf{F}(\mathbf{x}_{i+1})|| > ||\mathbf{F}(\mathbf{x}_i)||$, reduce $\lambda_0$ and recompute.
- When the residual falls below a tolerance $\tau$, switch to Broyden’s method: $\mathbf{x}_{i+1} = \mathbf{x}_i – \mathbf{A}_i^{-1} \mathbf{F}(\mathbf{x}_i)$, where $\mathbf{A}_i$ updates via the formula $\mathbf{A}_{i+1} = \mathbf{A}_i + \frac{(\mathbf{s} – \mathbf{A}_i \mathbf{r}) \mathbf{s}^T}{\mathbf{s}^T \mathbf{s}}$ with $\mathbf{s} = \mathbf{x}_{i+1} – \mathbf{x}_i$ and $\mathbf{r} = \mathbf{F}(\mathbf{x}_{i+1}) – \mathbf{F}(\mathbf{x}_i)$.
- Continue until the residual meets a final tolerance $\epsilon$.
This hybrid approach balances robustness and efficiency, making it suitable for the RV reducer main bearing’s quasi-static model. The algorithm includes a backtracking mechanism to handle non-convergence, which is essential for extreme load conditions where traditional methods may fail.
To validate the method, we conduct numerical experiments on an RV reducer main bearing model with parameters derived from a typical design. The bearing has 51 balls, ball diameter $D = 10.319$ mm, inner race diameter $d_i = 187.533$ mm, outer race diameter $d_o = 208.426$ mm, and curvature radii $r_i = 5.26$ mm and $r_o = 5.37$ mm. We apply a constant axial load $F_x = 13,720$ N (the maximum specified for this RV reducer type) and a rotational speed $n = 40$ rpm, with varying radial loads $F_z$ from 0 to 4.5 kN. The quasi-static model is solved using three methods: pure Newton-Raphson (NR), pure Broyden (B), and our combined Newton-Broyden (NB) approach. Computational efficiency is measured by iteration steps and time, with results averaged over 10 runs. The following table summarizes the performance under different radial loads:
| Radial Load $F_z$ (kN) | Method | Iteration Steps | Time (s) | Residual | Complex Solution? | Converged? |
|---|---|---|---|---|---|---|
| 0 | NR | 1 | 21.97 | 7.2e-7 | No | Yes |
| NB | 2 | 14.24 | 6.3e-8 | No | Yes | |
| B | 1 | 14.40 | 6.6e-8 | No | Yes | |
| 1 | NR | 5 | 50.40 | 6.4e-7 | No | Yes |
| NB | 10 | 44.50 | 3.8e-7 | No | Yes | |
| B | 20 | 24.60 | 6.5e-6 | No | Yes | |
| 3 | NR | 18 | 410.20 | 4.9e-9 | Yes | Yes |
| NB | 11 | 68.30 | 2.2e-6 | No | Yes | |
| B | 53 | 186.60 | 5.2e-6 | Yes | Yes | |
| 4 | NR | – | – | – | – | No |
| NB | 13 | 74.10 | 4.2e-8 | No | Yes | |
| B | – | – | – | – | No |
As shown, for low radial loads (0 and 1 kN), all methods converge, with Broyden’s method being fastest. However, for higher loads (3 and 4 kN), the pure Newton-Raphson and Broyden methods either produce complex solutions (indicating numerical instability) or fail to converge, while our combined algorithm succeeds with reasonable time. This demonstrates the robustness of the NB method for the RV reducer main bearing under extreme conditions. The computational savings are significant: at $F_z = 3$ kN, NB reduces time by over 80% compared to NR and avoids the pitfalls of B. The RV reducer’s performance is thus analyzable even near theoretical load limits, which is crucial for optimization.
Further analysis reveals insights into the load distribution and contact angles of the RV reducer main bearing. Under combined axial and radial loads, the bearing exhibits symmetry about the line normal to the external radial force contact point. Let $\gamma$ denote the angle between this normal and the position of a ball. The contact load $Q_j$ and contact angle $\alpha_j$ vary with $\gamma$. For instance, when $\gamma$ is acute (less than 90°), the load decreases as $\gamma$ increases, whereas for obtuse $\gamma$, the load increases with $\gamma$. Conversely, the contact angle shows an opposite trend: it increases with $\gamma$ for acute angles and decreases for obtuse angles under constant radial force. This behavior is summarized by the following relations derived from the quasi-static model:
$$Q_j \propto \frac{1}{\gamma} \text{ for } \gamma < 90^\circ, \quad Q_j \propto \gamma \text{ for } \gamma > 90^\circ$$
$$\alpha_j \propto \gamma \text{ for all } \gamma$$
These patterns are critical for designing the RV reducer main bearing to avoid ball-raceway separation under high loads. For example, at $F_z = 4.5$ kN, the minimum contact load drops to about 49 N, nearing the detachment limit. The symmetry ensures that loads are balanced across the bearing, enhancing the RV reducer’s durability. The following table illustrates the contact loads and angles for selected balls under $F_z = 3$ kN, highlighting the $\gamma$-dependence:
| Ball Position $\psi_j$ (degrees) | Angle $\gamma$ (degrees) | Inner Load $Q_{ij}$ (N) | Outer Load $Q_{oj}$ (N) | Contact Angle $\alpha_j$ (degrees) |
|---|---|---|---|---|
| 0 | 0 | 1250.3 | 1248.9 | 40.2 |
| 45 | 45 | 980.5 | 979.1 | 41.5 |
| 90 | 90 | 750.8 | 749.6 | 42.8 |
| 135 | 135 | 1020.7 | 1019.3 | 41.0 |
| 180 | 180 | 1300.1 | 1298.7 | 40.1 |
The data confirms that loads are minimized near $\gamma = 90^\circ$ and maximized at $\gamma = 0^\circ$ and $180^\circ$, while contact angles increase monotonically with $\gamma$. This understanding aids in optimizing the RV reducer main bearing for uniform stress distribution and extended service life.
In conclusion, the quasi-static model for the RV reducer main bearing, with its multi-ball architecture, presents a formidable computational challenge due to the large-scale nonlinear equations. Our proposed combined Newton-Broyden iterative algorithm, coupled with an optimized initial guess from static analysis, effectively addresses this issue. The method demonstrates superior convergence and efficiency, particularly under extreme radial loads where traditional Newton-Raphson and Broyden methods fail. For the RV reducer application, this enables accurate prediction of load distributions and contact angles, facilitating multi-objective optimization for enhanced performance. Key findings include the symmetric load patterns dependent on the angle $\gamma$, which inform design choices to prevent raceway detachment. Future work could extend the model to account for flexible raceways or dynamic effects, but the current approach provides a robust foundation for analyzing the RV reducer main bearing in practical scenarios. This advancement supports the ongoing development of high-performance RV reducers for industrial robotics, ensuring reliability and precision in demanding operations.
The RV reducer’s significance in robotics cannot be overstated, and improving its main bearing through methods like this contributes to overall system efficiency. By leveraging computational techniques, we can push the boundaries of what’s achievable in RV reducer design, ultimately fostering innovation in automation and manufacturing. The iterative strategy outlined here is generalizable to other bearing types with complex geometries, making it a valuable tool in mechanical engineering. As RV reducers evolve, continued refinement of such models will be essential for meeting the growing demands for speed, accuracy, and durability in industrial applications.