In the field of high-power transmission systems, the herringbone gear is widely recognized for its exceptional load-carrying capacity, smooth operation, and inherent self-balancing of axial forces. These characteristics make it an indispensable component in aerospace, marine, and heavy industrial applications. However, the performance of a herringbone gear is highly sensitive to manufacturing and assembly errors, particularly when the gear pair operates with a high contact ratio. To mitigate issues such as stress concentration, non-uniform load distribution along the tooth width, and meshing impact, tooth modification becomes a critical design step. Through my research, I have developed a comprehensive optimization methodology for the parabolic modification of herringbone gear teeth, integrating tooth contact analysis with a multi-objective genetic algorithm to achieve superior transmission performance.
The study begins with the establishment of a mathematical model for the herringbone gear tooth surface. A herringbone gear can be considered as two helical gears with opposite helix directions but identical helix angles. For economic and practical reasons, I focused the modification efforts on the pinion gear only, assuming that all deformation and error compensation would be concentrated on the smaller member. The modification is implemented through parabolic profiles in both the tooth profile direction and the tooth trace direction. For the tooth profile modification, I replaced the standard straight-line rack-cutter profile with a parabolic curve. The left-side parabolic tooth profile of the rack-cutter in the coordinate system Sa is expressed as:
$$ r_a(u_i) = \left[ a_i u_i^2, \; -u_i, \; 0, \; 1 \right]^T $$
Here, a_i is the parabola coefficient for the tooth profile, and u_i is the coordinate parameter along the tooth profile direction. The rack-cutter surface is then transformed into the pinion tooth surface through a series of coordinate transformations, yielding the final equation:
$$ r_c(u_i, l_i, \phi_i) = M_{cb} \cdot M_{ba} \cdot r_a(u_i, l_i) $$
$$ f(u_i, l_i, \phi_i) = 0 $$
This approach allows for precise control of the modification amount along the involute profile, which is particularly beneficial for herringbone gears operating under high loads.
In addition to profile modification, tooth trace modification is essential for addressing load distribution along the tooth width. I adopted a symmetrical parabolic modification method, taking the midpoint of the tooth width as the vertex and applying a parabolic curve along the helix line. The normal modification amount is given by:
$$ y = a^* x^2 $$
where a* is the parabola coefficient for the tooth trace modification. This method effectively compensates for the torsional and bending deformations that occur during the meshing process of a herringbone gear.
To evaluate the meshing performance of the modified herringbone gear pair, I conducted a comprehensive tooth contact analysis (TCA). The TCA model accounts for typical installation errors such as center distance error ΔE, axial offset error ΔL, and axis parallelism error Δγ. The coordinate system for the analysis includes a fixed coordinate system Sg, two rotating coordinate systems S1 and S2 fixed to the pinion and gear housings, and several auxiliary coordinate systems Se, Sf, and Sh. The contact conditions require that at any instant of meshing, the two tooth surfaces have a common contact point with a common normal vector. The governing equations are:
$$ r_g^{(1)}(u_1, l_1, \phi_1) = r_g^{(2)}(u_2, l_2, \phi_2) $$
$$ n_g^{(1)}(u_1, l_1, \phi_1) = n_g^{(2)}(u_2, l_2, \phi_2) $$
By solving these equations with φ1 as the input parameter, I obtained the instantaneous contact points. The transmission error of the gear pair is defined as:
$$ \Delta \phi_2 = \phi_2 – \phi_2^{(0)} – \left( \phi_1 – \phi_1^{(0)} \right) \frac{z_1}{z_2} $$
For the herringbone gear, the analysis is performed separately for the left and right helical sections. Due to manufacturing and assembly errors, the two sections may not engage simultaneously. I handled this by introducing a phase difference Δφ between the rotations of the left and right pinion sections:
$$ \phi_L = \phi_R – \Delta \phi $$
where Δφ = Σe / (rb cos βb). This formulation allows the TCA to accurately simulate the load-sharing behavior of the herringbone gear under realistic error conditions.
The core of this research is the multi-objective optimization of the modification parameters for the herringbone gear. I defined two conflicting objective functions. The first objective is to minimize the fluctuation of the geometric transmission error to ensure smooth power transmission:
$$ \text{min} \; f_1 = \text{min} \left( \max(\Delta \phi_2) – \min(\Delta \phi_2) \right) $$
The second objective is to maximize the number of contact points on the left helical tooth surface of the herringbone gear, which directly relates to reducing edge loading and improving load distribution:
$$ \text{min} \; f_2 = \text{min} \left( -PN \right) $$
Here, PN represents the number of contact points on the left tooth surface. To simplify the optimization, I assumed identical modification coefficients for both the left and right helical sections of the pinion, reducing the design variables to three: the profile parabola coefficient α1, the profile modification vertex position u01, and the trace parabola coefficient α*. The optimization was performed using the fast non-dominated sorting genetic algorithm (NSGA-II), which effectively avoids local convergence and yields a Pareto-optimal set of solutions.
The algorithm parameters were set as follows: population size of 108, maximum generation of 100, crossover probability of 0.8, mutation probability of 0.3, crossover distribution index of 10, and mutation distribution index of 30. I applied this method to a specific herringbone gear pair used in a high-power transmission system. The basic parameters of the herringbone gear are summarized in the table below.
| Parameter | Pinion / Gear |
|---|---|
| Number of Teeth | 23 / 231 |
| Module (mm) | 4.051 |
| Pressure Angle (°) | 20 |
| Helix Angle (°) | 34.096 |
| Face Width (mm) | 112.51 |
| Center Distance Error (mm) | 0.04 |
| Axis Parallelism Error (‘) | 0.01 |
The evolution of the feasible solutions during the optimization process clearly demonstrated the effectiveness of the algorithm. In the 10th generation, the number of contact points on the left tooth surface ranged from 0 to 2, indicating severe edge loading where the left section had no contact when the right section was meshing. By the 40th generation, the left surface contact points increased to 2, and by the 60th generation, they stabilized at 3. In the final 100th generation, two solutions even achieved 4 contact points. Simultaneously, the transmission error fluctuation was reduced to a range of 0 to 1.25 arc-seconds. The optimization results showed that axis parallelism errors have a more pronounced impact on the contact pattern of the herringbone gear than center distance errors. A set of optimized modification coefficients selected from the Pareto front is presented below.
| Modification Parameter | Before Optimization / After Optimization |
|---|---|
| Profile Parabola Coefficient α1 | 0.005 / 0.00637 |
| Profile Modification Vertex u01 (mm) | -0.01 / -0.0329 |
| Trace Parabola Coefficient α* | 2E-11 / 4E-6 |
| Transmission Error Fluctuation (arc-sec) | 1.27 / 0.332 |
| Number of Left Tooth Contact Points | 0 / 3 |
The comparison of the tooth contact patterns and transmission error curves before and after optimization revealed significant improvements. Prior to optimization, the herringbone gear exhibited a severe edge loading condition. When the right helical tooth surface was in contact, the left surface had no contact points at all. This asymmetric contact led to a discontinuous transmission error curve, indicating high levels of vibration and impact during operation. After the application of the optimized parabolic modification, the contact pattern on the right tooth surface became more uniformly distributed along the tooth trace, and the left tooth surface now featured three distinct contact points. The transmission error curve transformed into a continuous, parabolic shape, which is characteristic of smooth gear meshing. Although the peak value of the transmission error increased slightly, it remained well within the acceptable tolerance range. This demonstrates that the tooth trace modification effectively compensates for alignment errors, reducing the sensitivity of the herringbone gear to manufacturing imperfections.
To validate the practical benefits of this optimization, I conducted an experimental noise test on the original and optimized herringbone gear pairs. The gears were mounted on a high-speed gear vibration and noise reduction test rig. The noise measurement system included six microphones placed 1 meter away from the gearbox surface at the right side and rear of the box. The test was performed under loaded conditions to simulate real operating scenarios. The noise levels in decibels (dB) measured at the six points are summarized in the following table.
| Measurement Point | Before Optimization (dB) | After Optimization (dB) |
|---|---|---|
| Point 1 | 114.34 | 108.72 |
| Point 2 | 117.13 | 111.11 |
| Point 3 | 113.56 | 108.24 |
| Point 4 | 115.27 | 109.56 |
| Point 5 | 116.79 | 111.49 |
| Point 6 | 115.76 | 108.91 |
The experimental results show a consistent reduction in noise levels across all six measurement points. The optimized herringbone gear pair exhibited a noise reduction of 5 to 7 dB compared to the unmodified pair. This significant decrease in noise is directly attributable to the improved contact pattern and reduced transmission error fluctuation achieved through the parabolic modification. The smoother meshing process, with better load distribution and reduced edge contact, lowers the excitation forces that generate vibration and airborne noise.
In summary, this research has successfully developed a systematic approach for the optimal modification design of high-contact-ratio herringbone gears. The key contributions include: first, a parabolic modification method for both tooth profile and tooth trace, which is simple to implement in manufacturing; second, a robust tooth contact analysis model that incorporates typical installation errors and phase differences between the left and right helical sections; third, a multi-objective optimization framework using the NSGA-II algorithm that simultaneously minimizes transmission error fluctuation and maximizes contact point number, avoiding local convergence; and fourth, experimental validation demonstrating a 5-7 dB noise reduction. The findings confirm that this optimization methodology is effective in improving the load distribution, reducing meshing impact, and lowering the vibration and noise levels of herringbone gear transmissions. This work provides a valuable reference for the design of high-performance herringbone gear systems in demanding applications such as helicopter main gearboxes and marine propulsion drives.
