Why bearing raceway waviness at specific orders amplifies gearbox structure-borne noise
In a geared aircraft engine, rolling bearings support shafts that rotate at high speed while transmitting substantial torque through compact, highly loaded gear meshes. Their raceways appear smooth at the scale visible during ordinary inspection, yet small periodic deviations can create measurable forces. When those deviations repeat at particular circumferential orders, they can become a powerful source of structure-borne noise.
The problem is governed by an interaction between geometry, speed, contact mechanics, shaft dynamics, housing modes, and acoustic radiation. A waviness component that seems harmless in isolation may align with a bearing-pass excitation or a gearbox structural resonance. The result can be a sharp vibration peak, tonal noise in the cabin or engine nacelle, and increased sensitivity to manufacturing variation.
The OPTIMIZE project’s focus on simulation, tolerance analysis, design of experiments, and physical validation is relevant to this issue because bearing waviness rarely acts as an independent variable. It is one element in a coupled transmission system where small geometric changes can alter efficiency, durability, power density, and noise performance at the same time.
How raceway waviness creates periodic excitation
Raceway waviness is a repeating departure from the ideal circular or profile geometry of an inner or outer bearing race. It may arise from grinding marks, fixture errors, machine-tool harmonics, waviness transfer from a forming process, or local correction operations. Unlike random roughness, which produces broadband disturbances, waviness has a recognizable spatial wavelength around the raceway.
As rolling elements pass over the undulation, the local contact position and elastic deflection vary. The bearing therefore produces a small fluctuation in rolling resistance, contact force, and sometimes axial or radial displacement. Since the raceway rotates relative to the rolling elements, the spatial pattern is converted into a temporal excitation whose frequency depends on shaft speed and the order of the imperfection.
A useful simplified relationship is:
[ f_w = m \frac{n}{60} ]
where (f_w) is the waviness excitation frequency, (m) is the circumferential order, and (n) is rotational speed in revolutions per minute. Real bearings generate additional components because rolling elements distribute the disturbance, contacts are elastic, and the cage introduces its own kinematic frequency. The basic order relationship nevertheless explains why a defect can become prominent at a specific operating speed.
The excitation may be transmitted through the bearing rings, shaft, fits, housing, and gearbox casing. It does not need to be large at the source to become important. A lightly loaded periodic force can produce a strong response when it enters a flexible path with low damping.
Why certain orders are much louder than others
A circumferential order describes how many complete waviness cycles occur around a raceway. Low orders correspond to broad shape errors, while higher orders represent finer repeating patterns. Each order produces a different frequency as the shaft rotates, and each frequency interacts differently with the gearbox structure.
Amplification occurs when an excitation order approaches a natural frequency of the shaft-bearing-housing assembly. This can involve a bending mode, torsional mode, bearing-seat mode, gear blank mode, or local casing panel mode. The response is often described by a frequency-response function: the bearing supplies the force, while the structure determines how much displacement and force transmission result.
The phase relationship between multiple bearings also matters. If two bearings carry related waviness patterns, their forces can reinforce one another along a particular direction. If their phases oppose, partial cancellation may occur. A housing may therefore respond strongly to a raceway order in one assembly configuration and weakly to the same order after a small change in bearing orientation, preload, shaft position, or support stiffness.
Specific orders can also align with gear-mesh harmonics and sidebands. A gear mesh already creates a periodic force at tooth-passing frequency. Bearing waviness can modulate that force, generating sum and difference components around the gear-mesh frequency. These sidebands are especially valuable diagnostically because they can reveal whether a tonal response originates from gear transmission error, bearing geometry, or an interaction between both.
The route from bearing contact to structure-borne noise
Structure-borne noise begins as a mechanical force but ends as radiated sound when vibrating surfaces displace the surrounding air. In a gearbox, the transmission path commonly runs from the raceway contact into the rolling elements, bearing rings, shaft seats, housing walls, mounting interfaces, and finally the engine structure. Each interface has its own stiffness, damping, preload, and frequency dependence.
A rigid-looking housing is not dynamically rigid across the full operating range. Thin sections, ribs, inspection covers, flanges, bearing pedestals, and oil passages can produce local modes. A bearing support that is stiff in a static load case may be compliant near a mode shape. This is why static stiffness alone cannot predict noise behavior reliably.
The force reaching the housing is influenced by bearing internal clearance, contact angle, roller or ball geometry, preload, lubricant film thickness, and operating temperature. Lubrication changes contact damping and friction, while temperature changes fits and clearance. At high speed, centrifugal effects alter rolling-element loads and cage behavior. These effects can move the excitation frequency, change its amplitude, or shift the structural resonance enough to produce a noticeable order-tracking peak.
The audible result is often a narrow tonal component rather than a general increase in broadband sound. Passengers or maintenance personnel may perceive it as a whine, buzz, or high-frequency growl. Even when overall sound pressure remains within a limit, a strong discrete tone can be undesirable because human hearing is sensitive to tonal prominence.
How simulation and testing reveal the amplification mechanism
A reliable investigation combines bearing kinematics, contact modeling, multibody dynamics, finite-element analysis, and test measurements. A bearing model can convert measured raceway waviness into time-varying contact forces. A gearbox dynamic model can then propagate those forces through shafts, gears, bearings, and housing interfaces.
Finite-element modal analysis identifies the natural frequencies and mode shapes most likely to amplify an order. Component mode synthesis or reduced-order models can make this analysis practical across many operating conditions. Frequency-response calculations then show whether the bearing excitation enters a highly responsive part of the structure or a region with substantial damping.
Order tracking is essential because rotational speed changes continuously during a run-up or run-down. A fixed-frequency spectrum can smear an order across several bins, while an order map shows whether the response follows shaft speed, cage speed, gear mesh, or a structural resonance. Accelerometers near bearing seats, casing panels, and mounting points can distinguish local excitation from global housing response.
Physical tests should measure both source and path. Bearing vibration, shaft displacement, housing acceleration, sound pressure, oil temperature, speed, torque, and preload provide a connected data set. Tests with controlled raceway specimens or deliberately varied waviness orders can establish causality. The comparison between predicted and measured order amplitudes is more informative than a single overall noise value because it reveals whether the model has the correct excitation mechanism and transfer path.
Comparing excitation patterns and diagnostic clues
The table below summarizes how different geometric or dynamic sources tend to appear in gearbox measurements. Actual signatures depend on bearing architecture, load, speed, sensor location, and structural damping.
| Source | Primary frequency relationship | Typical signature | Useful diagnostic check |
|---|---|---|---|
| Raceway waviness | Integer order of shaft speed, with bearing kinematic combinations | Discrete order peak that may rise sharply near a housing mode | Compare order maps with measured raceway order and rotate or replace the bearing |
| Rolling-element diameter variation | Ball-pass or roller-pass related frequencies | Repeating modulation and sidebands around shaft-related components | Check consistency across load and temperature changes |
| Cage irregularity | Cage rotational frequency and harmonics | Low-frequency modulation, sometimes with intermittent amplitude changes | Compare with calculated cage speed and inspect cage condition |
| Gear transmission error | Gear-mesh frequency and tooth-related harmonics | Strong mesh tones with shaft-order sidebands | Measure gear microgeometry and compare mesh-order predictions |
| Housing resonance | Natural frequency of casing or support structure | Large response at a narrow frequency, often across several source orders | Perform modal testing and compare sensor phase patterns |
| Lubrication or preload variation | Speed-, load-, and temperature-dependent components | Amplitude drift, broadband growth, or shifted tonal peaks | Repeat tests after thermal stabilization and controlled preload changes |
The important distinction is between an excitation source and a transfer-path amplifier. Raceway waviness may create the periodic force, but the housing mode determines whether that force becomes a dominant noise feature. This distinction prevents engineers from removing material or stiffening a casing without first confirming the forcing order.
The same logic applies to gear transmission error. A gearbox can exhibit a gear-mesh tone that appears to be the primary problem, while a bearing waviness order is actually modulating the mesh force. Order-related sidebands and phase measurements can separate these mechanisms. This is particularly valuable when several components share harmonically related frequencies.
Design and manufacturing variables that control the response
Reducing waviness-related noise begins with controlling the geometry of the raceway and the process that creates it. Specification limits should address amplitude by circumferential order, rather than relying only on a single total waviness value. Two bearings with the same overall deviation can produce very different vibration if their energy is concentrated at different orders.
Manufacturing process capability studies can identify grinding wheel condition, spindle runout, fixturing, heat treatment distortion, and measurement filtering as influential variables. Metrology must preserve order information; excessive filtering can conceal a problematic harmonic, while poor datum control can introduce a false one. Measurements from multiple circumferential positions also help distinguish a true raceway pattern from instrument or setup artifacts.
Designers can reduce sensitivity through bearing selection, preload control, support stiffness, damping treatments, and changes to shaft or housing geometry. Moving a structural mode away from a dominant excitation order is often more robust than attempting to eliminate every source of waviness. A tuned rib, altered bearing span, compliant isolator, or local damping treatment may reduce force transmission without imposing a major mass penalty.
The OPTIMIZE initiative examines this kind of multidisciplinary trade-off through integrated gearbox research, linking high-speed operation, lubrication, manufacturing variation, durability, and power density. That systems perspective is important because a noise fix that increases friction, reduces bearing life, or adds excessive weight may be unsuitable for an aircraft propulsion gearbox.
Building a robust investigation and control strategy
Because many parameters interact, screening experiments can be more efficient than changing one variable at a time. Factors may include waviness order and amplitude, bearing preload, radial load, shaft speed, housing stiffness, lubricant viscosity, gear microgeometry, and temperature. A structured design-of-experiments approach can identify the variables that govern vibration amplification and expose interactions between them.
The Taguchi screening methods described for gearbox transmission error provide a useful example of how orthogonal arrays can reduce the number of simulation or test combinations. Similar methods can screen bearing and structural variables before a higher-resolution response-surface study. The objective is not to replace detailed physics, but to focus it on the combinations most likely to affect noise and reliability.
A practical control strategy should connect specification, simulation, and verification:
- Measure raceway profiles with order-resolved metrology and retain raw circumferential data.
- Use order-tracked vibration and acoustic measurements during speed, load, and temperature sweeps.
- Correlate bearing excitation models with housing modal tests and transfer-path measurements.
- Set order-specific manufacturing limits instead of relying solely on overall roundness or waviness values.
- Validate the final design with representative lubrication, preload, thermal growth, and mounting conditions.
This workflow supports robust optimization because it treats variation as part of the design problem. It can reveal whether a gearbox is intrinsically sensitive to a particular bearing order or whether the response appears only under a narrow combination of speed, temperature, and assembly conditions.
A strong validation program should include repeated builds and multiple bearings from production-representative processes. The goal is to establish a distribution of noise and vibration responses, not simply to demonstrate one quiet prototype. Statistical tolerance analysis can then identify which variation sources deserve tighter limits and which have little practical influence.
The interaction between raceway waviness and gearbox structure-borne noise is ultimately an order-matching problem. A periodic geometric feature supplies energy at a predictable kinematic relationship, while the mechanical structure filters and amplifies that energy according to its modes, damping, and interfaces. Understanding both sides makes it possible to distinguish a bearing manufacturing issue from a housing resonance or a gear-mesh interaction.
Apply order-based metrology, dynamic simulation, and controlled testing early in the gearbox development cycle. Explore the OPTIMIZE Project resources and use the linked methodology to connect bearing geometry, structural response, and aircraft propulsion performance before a tonal noise issue reaches full-scale integration.