New Gear Design Strategy Cuts Noise in Heavy-Duty Electric Trucks

Electric trucks promise cleaner roads, but they bring an acoustic problem that combustion engines never had: without the broadband rumble of a diesel engine to mask it, the high-pitched whine of meshing gears becomes impossible to hide. In a heavy-duty electric drive axle, where a compact motor delivers enormous torque through multiple gear stages into a flexible housing, that whine can dominate the cabin and the street. A research team led by NianCheng Guo, QiMin Mu, Hai Yang, HaoNan Sun, Yang Gao, and Lu Wang, publishing in Results in Engineering, has now tackled the problem at its source with a system-level gear redesign strategy that treats the entire axle—not a single gear pair—as the design target.

The team’s central insight is that gear noise is not simply a function of how badly the teeth mesh. Sound radiated from the axle housing depends on a long chain of events: elastic tooth deflection and time-varying mesh stiffness generate periodic dynamic mesh forces; those forces travel through shafts, bearings, and bolted joints; the housing vibrates in its structural modes; and only then does the vibrating surface push air into audible pressure waves. A gear parameter that improves transmission error in isolation may, for example, raise axial force, deform the bearings differently, or shift the participation of a housing mode, potentially making noise worse at some speeds. This is precisely why previous optimization studies that focused on single gear pairs, single operating points, or transmission error alone have produced limited real-world gains.

To capture the full source–transfer–radiation chain, the researchers built a rigid–flexible coupled dynamic model of the complete motor–gear–shaft–bearing–housing system, in which the axle housing, shafts, and wheel web were modeled as flexible finite-element components. The housing surface vibration predicted by this model was then mapped onto an acoustic boundary-element model to compute the A-weighted radiated sound pressure. The mesh design targeted a 5 kHz upper analysis frequency, with housing, shaft, and wheel-web elements set at 8 mm to provide numerical margin. Motor excitation was not neglected: torque ripple from the 150 kW rated, 275 kW peak traction motor was prescribed as an input, since electric powertrains introduce distinct electromagnetic orders alongside gear-meshing orders.

Critically, the model was validated against a dedicated e-axle test bench before any optimization was attempted. Measured and predicted mean A-weighted sound pressure levels came in at 79.53 and 78.89 dB(A) respectively—a difference of just 0.64 dB(A)—and the simulated speed–order waterfall reproduced the dominant order ridges seen in the measurements. Those dominant orders, the 11.5th, 17.3rd, and 30th, told an important story: the first two trace back to planetary-stage meshing harmonics and the third to the fixed-axis gear stages, meaning gear excitation, not the motor, drives the radiated noise of this axle.

With a trustworthy model in hand, the team formulated a hierarchical optimization. Operating conditions were not chosen arbitrarily; they were extracted from the road-load duty cycle of a semi-trailer tractor under China’s GB/T 38146.2-2019 heavy-duty commercial vehicle test cycle. Four representative motor torques—120, 240, 360, and 480 Nm—were weighted by how often they occur, with 240 Nm dominating the duty cycle at roughly 52 percent of its 1644-second duration. The first objective functions were the fundamental harmonic and the peak-to-peak value of loaded transmission error, the classic source-side metrics of gear whine, evaluated across all four weighted conditions rather than at one design point.

Macrogeometry came first. An L9 orthogonal array screened the main effects of face width, helix angle, and normal pressure angle on the weighted noise response, ranking helix angle as most influential with a range of 0.44 dB(A), followed by face width at 0.34 dB(A) and pressure angle at 0.29 dB(A). Single-factor analysis revealed why a naive ‘increase the helix angle’ rule fails here: while a larger helix angle raises contact ratio and smooths a single mesh, it also increases axial force, aggravating bearing support deformation and planetary-stage misalignment, which amplifies the 11.5th- and 17.3rd-order responses. A non-dominated sorting genetic algorithm (NSGA-II) then searched the helix angles under hard constraints—center distances fixed, zero-backlash profile shifts maintained, contact ratio and stress limits enforced, and efficiency never below the baseline. The result: helix angles dropped from 18 and 15 degrees to 8.15 and 8.33 degrees, pressure angles moved to 18 degrees, and face widths grew, including a widened planetary sun gear.

Macrogeometry optimization alone reduced mean A-weighted noise by 0.61, 1.69, 0.87, and 2.34 dB(A) at 120, 240, 360, and 480 Nm respectively—but it raised noise peaks at some speeds, a warning that global mesh changes alone cannot tame every resonance. Microgeometry refinement followed. Tooth flank crowning, involute crowning, and lead bias modification—bounded at 0–30 micrometers and 0–10 micrometers respectively—were optimized with the same two transmission-error objectives, compensating for load-induced deformation and misalignment without altering the axle’s layout or ratios. The combined design cut the mean 17.3rd-order vibration acceleration by 23.9, 20.7, 9.2, and 5.7 percent across the four torque conditions, with 11.5th-order reductions of 2.8 to 13.2 percent.

The acoustic payoff was substantial at low and medium loads. Mean A-weighted sound pressure level fell by 1.10, 2.26, 2.32, and 1.43 dB(A) at 120, 240, 360, and 480 Nm relative to the baseline, and at 240 Nm the peak level dropped from 85.33 to 83.39 dB(A)—the most balanced improvement of the four conditions. Microgeometry refinement added further mean-noise reductions of 0.49, 0.57, and 1.45 dB(A) at 120, 240, and 360 Nm compared with macro-only design. The one blemish was the 480 Nm high-load case, where the combined design raised peak noise by 3.80 dB(A) over the baseline, prompting the authors to recommend stronger peak and variability constraints for high torque in future formulations.

Crucially, none of this came at the cost of performance elsewhere. Overall efficiency was maintained at 120 Nm and improved by 0.1, 0.1, and 0.2 percentage points at 240, 360, and 480 Nm, with mesh friction, bearing friction, and oil churning losses computed per ISO/TR 14179-2 and Harris formulations using SAE 75W-90 gear oil. Strength verification at the worst-case 480 Nm condition, per ISO 6336, showed maximum contact stress of 1045.60 MPa and bending stress of 682.34 MPa in the critical first-stage gear—both safely below allowable limits of 1281.8 and 716.3 MPa, though the bending margin of just 4.7 percent is the design’s tightest constraint. The optimized gears remain unmanufactured; the authors stress that the reported gains are model-based predictions pending bench and road validation, and that future work should incorporate tonality and loudness metrics, robustness to manufacturing deviations, and full-vehicle testing. What the study delivers now is a rigorous engineering route: fix the global meshing state with macrogeometry first, then polish the loaded contact with microgeometry—and always judge the result by the sound that actually reaches the microphone.

Subject of Research: Hierarchical macro- and microgeometry optimization of multi-stage gears to reduce vibration and radiated noise in heavy-duty electric drive axles

Article Title: Hierarchical optimization of gear macro- and microgeometry for vibration and radiated noise reduction in a heavy-duty electric drive axle

Article References: Guo, N., Mu, Q., Yang, H., Sun, H., Gao, Y., & Wang, L. (2026). Hierarchical optimization of gear macro- and microgeometry for vibration and radiated noise reduction in a heavy-duty electric drive axle. Results in Engineering, 32, Article 113075. https://doi.org/10.1016/j.rineng.2026.113075

Image Credits: AI Generated

DOI: Not provided

Keywords: electric drive axle, gear whine, transmission error, NSGA-II, microgeometry modification, macrogeometry optimization, radiated noise, NVH, heavy-duty vehicles, planetary gear, acoustic boundary element, transmission efficiency