Modal Analysis for Mesh-Excited Housing Resonance
Using modal analysis to identify housing resonance modes excited by mesh order connects the rotating load spectrum of a gearbox with the dynamic behaviour of its casing. The aim extends beyond simply listing natural frequencies to determining whether tooth-pair harmonics, sidebands or shaft-related orders can create a repeatable response at an engine operating condition, and where that response concentrates around bearing seats, webs, ribs and bolted interfaces.
A useful workflow combines operating-order definition, finite-element modelling, load-dependent frequency response, test correlation and high-speed measurement. For a power-reduction gearbox, that chain matters because small changes in bearing clearance, support stiffness or assembly preload can move a mode toward a demanding mesh line. Its output should identify what to change, what to measure and how much confidence to place in each decision.
From excitation to resonance
For a fixed-axis gear pair, mesh frequency can be estimated as f_m = Z₁n₁/60 = Z₂n₂/60, where Z is tooth count and n is speed in rpm. The ideal excitation contains the fundamental and integer harmonics at k f_m. Real meshes also create sidebands through transmission error, cyclic load variation and timing modulation, while shaft and bearing orders can occupy adjacent bands. Analysts therefore need harmonic amplitudes and phase alongside the frequency list. The inventory should span the full speed envelope and allow for shifts caused by slip, active ratio changes and gears entering or leaving mesh. For planetary stages, carrier-relative motion must also be included.
A housing mode becomes concerning when an excitation frequency, its sidebands or a running speed produces a repeatable response near that mode. The critical result may be large acceleration at a bearing seat, relative movement across a joint or local strain around an oil-film boundary. Mobility can peak near a lightly damped mode, while a nearby anti-resonance can reduce motion in one region without making the whole housing safe. Mode shape and interface relative motion are therefore as important as frequency in hertz. A short pulse can expose a mode without reaching steady-state resonance, whereas a speed sweep may reveal amplification, phase lag and settling as the crossing is traversed.
Resonance can also alter how mesh loads enter the structure. Cyclic casing deformation may disturb alignment and lubricant-film behaviour when speed, temperature or surface condition is already marginal. Vibration measurements should therefore sit beside surface-finish scuffing research, supported by pressure, film thickness and temperature evidence. Modal analysis identifies a structural vulnerability; it cannot diagnose lubricant failure by itself.
Building a useful housing model
A first finite-element model should represent ribs, webs, bolt groups, bearing interfaces, seals and mass-relevant accessories, avoiding oversimplification as a plain cylindrical shell. Eigenanalysis supplies candidate natural frequencies and mode shapes, but its value depends on credible material properties, joints and constraints. Shaft loads should be applied before extracting operating modes so gravity, torque, centrifugal force and housing deflection can shift the frequencies encountered in service. Meshed components must transfer gear forces into the case: complete bonding can hide interface flexibility, while disconnected components omit the load path entirely.
Boundary assumptions deserve particular care in a hyperstatic arrangement. Multiple bearings and planet-carrier contacts can make reactions dependent on clearance, oil-film stiffness and case deformation, so fixing every support perfectly may create an artificially stiff mode set. Engineers should compare alternative interface models and record whether a bearing seat is grounded, elastically mounted or coupled to another component. Contact pressure and local bearing stiffness deserve the same scrutiny as shaft stiffness. Sensitivity to preload and interface flexibility is often more informative than an extra decimal place in a nominal natural frequency.
Manufacturing variation makes every housing slightly different. Bolt preload, bearing-seat position, rib thickness, casting porosity and material modulus can move response poles and zeros. Australian suppliers often connect precision shops around Melbourne and Geelong with aerospace programmes near Brisbane, Adelaide and Bankstown, making traceable geometry and inspection records especially important across a tiered supply chain. Before another steel article is funded, the team should have a proper look at utilisation and residual value. The valuation logic applied to a pontoon cash-out quote—condition, provenance and buyer demand—illustrates why purchase price is a poor measure of an idle asset’s recoverable value.
Comparing analysis evidence
Four forms of evidence answer different engineering questions about the same housing:
| Method | Question answered | Strongest use | Important limitation |
|---|---|---|---|
| Linear eigenvalue analysis | Where are the undamped natural frequencies? | Rapid geometry and load-case screening | Idealised joints and damping |
| Prestressed modal analysis | How do service loads shift poles and mode shapes? | Hyperstatic housings and installed configurations | Contact and oil-film nonlinearities are linearised |
| Calculated frequency response | How strongly does each excitation order excite a point or interface? | Harmonic response, phase and anti-resonance estimates | Sensitive to damping, force distribution and load case |
| Measured response function or operational shape | Does the real structure show the predicted motion? | Model correction and validation under controlled excitation | Requires suitable access, instrumentation and safe test conditions |
The most defensible workflow uses all four without treating them as equivalent. Predicted poles should align with measured peaks or mobility minima, and modal assurance criteria should be checked after scaling, sign conventions and sensor placement are consistent. Anti-resonances can constrain joint mobility even when an apparent mode shape differs. Residue fits, damping estimates and frequency uncertainty then show whether a mismatch is meaningful. Evidence governance is equally important: traceable instrument settings, calibration records and raw responses matter, while a polished unrelated test-site page is no substitute for an instrumented test report.
Validating modes under operating load
Experimental modal testing usually begins with the rig stopped. Impact-hammer strikes or a controlled shaker sweep can excite several modes, while accelerometers or laser vibrometry measure the response. The team should fit frequency-response functions, refine poles and residues, and compare mode shapes with the finite-element prediction using frequency error, damping estimates and the modal assurance criterion. Point measurements can miss an anti-resonance, so sensor placement should sample both bearing supports and comparatively stiff attachment regions. Measurements should also preserve the force-input reference needed to distinguish a genuine mobility change from inconsistent hammer contact.
The high-speed test then establishes whether the assembled gearbox responds at the calculated orders. A once-per-revolution tachometer, torque and temperature channels, strain gauges and housing accelerometers should be synchronised with a sampling rate high enough for the highest mesh harmonic of interest. Order tracking can separate changing excitation from fixed resonances, while an operational deflection shape reveals the real pattern under torque and oil pressure. Repeat runs should check repeatability as bearings heat, clearance changes and torque moves through its schedule. Sampling rate, anti-alias filters and sensor bandwidth need enough margin to prevent a high-order mesh response from disappearing into the instrumentation noise floor.
Testing must also fit Australian delivery and assurance constraints. NATA-accredited laboratories, calibrated reference equipment and AS9100-controlled records may be required by the programme, while CASA approvals apply to defined activities where a certificated civil design is involved. Moving a delicate rig or instrumentation between Australian facilities adds lead time, so calibration and spare-sensor plans should be settled before shutdown. Hot ambient conditions can also change damping and lubricant properties, making a warm-condition repeat more representative than a cool baseline alone.
Turning resonance findings into design gains
Once a mode is confirmed, evaluate remedies in the same load path that created it. Local ribs, bearing-seat stiffness, joint preload, support isolation and damping treatments can alter response, as can gear macrogeometry changes that reduce a troublesome harmonic. A thicker rib may lower motion but add mass and churning loss; a coating may improve amplitude yet reduce fatigue strength or lose damping as it heats. Weight, efficiency, durability and power density must be assessed together rather than optimising acceleration at any cost.
Nominal separation provides an incomplete robust margin. Run tolerance and sensitivity studies on bearing centres, joint stiffness, material properties and rib thickness, then track how uncertainty moves both resonance and anti-resonance frequencies. Monte Carlo sampling can expose rare combinations, while targeted experiments can verify the variables with the greatest influence. The design target should keep important responses away from strong mesh orders across the credible speed, load and temperature range, including manufacturing spread. This prevents a small tolerance excursion from returning the gearbox to the same troublesome operating condition.
Physical validation closes the loop. Test the revised housing first at stationary modal frequencies, then through the intended speed sweep using the same order definitions and mounting interfaces as the original correlation. Retain rejected concepts as documented trade studies so procurement and certification teams can see why a lighter or simpler solution was not accepted. That record is especially valuable when international suppliers, local machining partners and Australian test laboratories divide responsibility for the final evidence package.
Use the next design review to nominate the mesh harmonics, speed bands and housing response points that matter, assign ownership of correlation evidence and tolerance sensitivity, and set a measurable acceptance criterion. Review current project updates for recorded methods and related test material, then carry one representative housing mode into the next correlation and redesign cycle with amplitudes, uncertainty and design assumptions attached.