DOE insights for forged and machined aircraft housings
Power reduction gearboxes for geared aircraft engines place unusual demands on their housings. They must contain high-speed rotating components, carry gear and bearing loads, retain oil, and remain stiff enough to control shaft alignment. Their mass also accumulates across an aircraft, making even a modest saving valuable when fatigue margin, damping, and serviceability are preserved.
A design-of-experiments approach, or DOE, offers a disciplined way to compare forged and machined housings for weight and stiffness without attributing every result to the manufacturing route. The comparison works best when both candidates meet the same functional envelope, material specification, bearing arrangement, surface finish, and acceptance criteria. Otherwise, geometry or material can obscure the real process effect.
The OPTIMIZE Clean Sky project links simulation, tolerance analysis, and physical testing to improve gearbox efficiency, durability, weight, and power density. That combination is particularly useful for housings, where a lighter design can change stress, oil temperature, and bearing behaviour as well as aircraft mass.
For Australian aerospace programmes, the comparison also has practical consequences. Local supplier capability, certification evidence, imported tooling, and transport distances affect whether a laboratory gain can become a repeatable production advantage.
Why housing architecture changes performance
Forging shapes material through controlled plastic flow, allowing integral webs, ribs, bosses, and flanges with less waste than removing those features from plate or block. The process can also align the material flow with principal stresses, improving fatigue resistance. Machining offers fine control of dimensions and can create complex internal passages, but stock removal, setup time, and scrap can erode its weight and cost advantage.
Stiffness is not fixed by the word “forged” or “machined”. It depends on section geometry, material modulus, bearing-seat stiffness, joints, and how the housing deforms under the hyperstatic load path. Forged grain flow may improve durability without increasing bulk stiffness, while a well-designed machined shell can be equally stiff or lighter. DOE and finite-element analysis therefore need to measure both routes against a common duty cycle.
| Criterion | Forged housing | Machined housing | DOE measurement |
|---|---|---|---|
| Material utilisation | Near-net shape can reduce stock removal, subject to flash and allowance | More removal may be needed when machined from plate or block | Mass, scrap mass and process yield |
| Structural form | Integral ribs and bosses depend on forging design | Internal webs and passages can be accurately machined | Geometry, thickness and feature tolerance |
| Fatigue behaviour | Flow lines may follow highly stressed regions | Material history remains important after machining | Fatigue life, crack-growth margin and stress concentration |
| Stiffness | Set by geometry and material modulus, not forging alone | Set by geometry and material modulus, not machining alone | Load deflection, bearing-seat movement and modal frequency |
| Dimensional repeatability | Tooling, temperature and draft affect variation | Setup, tool wear and datum strategy affect variation | Capability indices and tolerance stack-up |
| Production exposure | Die cost and closed-die constraints | Machine time, labour and stock cost | Lead time, rework rate and unit cost |
Defining a fair housing comparison
The first decision is what “machined housing” means. A part machined from a forging blank primarily tests finishing effects, while a part cut from plate or block tests an entire manufacturing strategy. Those are different questions. The OPTIMIZE study should declare the starting stock, heat treatment, final geometry, bearing fits and inspection plan for each route before specimens are made.
The load definition must then match service. Gear-mesh force, shaft bending, bearing preload, gyroscopic effects, thermal distortion and manoeuvring loads should be represented as a load spectrum rather than a single static case. At the shaft interface, the high-speed bearing study shows why friction torque and bearing conditions belong in the housing decision: losses and heat rise can change the mass and stiffness benefit seen on the test stand.
Building the DOE matrix
Process route can be treated as a categorical factor, while web thickness, boss diameter, bearing-seat length, draft, machining allowance and heat-treatment condition can be continuous factors. Responses should include empty-housing mass, load-path deflection, first flexible natural frequency, peak von Mises stress, fatigue margin, critical bearing temperature and manufacturing cost. A useful design measures several responses because the lightest housing is not automatically the best one.
A full factorial screening experiment can expose interactions among forging allowance, local wall thickness and machining allowance. A response-surface design can then refine the promising region and locate a minimum-mass, maximum-stiffness setting. Randomised blocks by material heat, tooling cavity, machine and operator shift protect the comparison from time-based bias, while repeated builds estimate pure error and reproducibility.
The risk-control mindset described in an Australian pokies strategy is not a substitute for engineering DOE, but it illustrates a useful local lesson: changing many inputs at once makes outcomes difficult to interpret. For the gearbox programme, that means holding lubricant batch, assembly torque, instrumentation calibration and test sequence under control, then logging every deviation as a factor rather than silently excluding an outlier.
Predicting mass and stiffness
A mass model should separate material that is structurally useful from allowance, flash, chips and non-critical features. For a forged component, compare the finished mass with the predicted net shape and forging envelope; for a machined component, record stock mass, removed mass and recycled swarf. Tolerance analysis then tests whether a nominal lightweight design still satisfies the limit after realistic upper and lower bounds are applied.
Finite-element analysis should model contact, bearing fits, fasteners and the hyperstatic reaction path rather than treating the housing as a free-standing ring. Modal testing and static load-deflection tests provide measured corrections for the model. The same distinction between signal and noise is central to a non-engineering gambling variance analysis: a few unrepeated outcomes can suggest a trend that later observations do not support, just as one apparently light housing can conceal uncontrolled build variation.
Validating high-speed and thermal behaviour
A scale or material coupon cannot reproduce every aspect of an aircraft gearbox, so physical validation should be staged. Begin with dimensional inspection, hardness checks and non-destructive testing, then use low-speed load cases to verify stiffness and instrument response. Progressive spin testing adds centrifugal force, bearing friction and oil-film effects, allowing the team to detect imbalance, looseness, abnormal vibration and early bearing distress before reaching the planned high-speed condition.
Temperature must be treated as a dynamic design variable, especially during start-up and shutdown. The project’s transient thermal analysis is directly relevant because oil-temperature rise changes viscosity, which alters friction torque, bearing load, heat generation and structural deflection. A housing that saves mass but distorts a bearing seat or raises oil temperature may reduce system power density even if its static model looks favourable.
The final acceptance rule should combine several thresholds: minimum natural frequency, maximum bearing-seat deflection, fatigue damage below the agreed limit, stable oil temperature and no unacceptable wear or distress. Replicate housings from both manufacturing routes are essential because average performance plus measured spread is more credible for flight-critical hardware than a single best result.
Managing Australian manufacturing reality
Australian supply geography shapes the test and production plan. Melbourne has a strong advanced-manufacturing and maintenance, repair and overhaul base; Adelaide supports a defence-aerospace cluster; and Brisbane has component and maintenance capability that can shorten turnaround routes. These regional strengths make local forging, machining and inspection partnerships possible, but supplier capacity must be checked early.
Bringing specialist dies, gauges or machine tools into Australia requires correct tariff classification, customs declarations and a clear plan for temporary admission, including an ATA Carnet where appropriate. Imported equipment introduces freight, currency and lead-time exposure that can change the commercial result. A part that wins a laboratory comparison can still lose at scale if local suppliers cannot repeat its tolerance stack-up or if imported replacement tooling takes months to arrive.
The Australian market also requires evidence that can survive certification scrutiny. A CASA-approved engineering and manufacturing organisation will expect traceable material certificates, approved process controls, inspection records and clear change management. Since the domestic supplier base for specialist aerospace forgings is narrower than the general metal-forming market, the pilot should compare both a local source and an established overseas source while applying the same acceptance criteria to each.
Moving from evidence to production
The preferred housing is the candidate that meets the structural and thermal requirements with the lowest risk-adjusted mass, not simply the lightest sample. A ranked score can combine mass, deflection, fatigue margin, oil-temperature rise, variation, lead time and cost, with mandatory pass/fail limits placed ahead of weighted preferences. Confidence intervals and ANOVA effect estimates should show which gains are real and which depend on a particular batch.
Production readiness follows the evidence. For a forged route, verify die-life limits, flash removal and heat-treatment distortion; for a machined route, verify datum strategy, tool wear and machining allowance. Either route needs an incoming inspection plan, tolerance stack-up linked to bearing fit, and a feedback loop from service data. Reviewing the OPTIMIZE project videos can help teams connect the test philosophy with the wider gearbox programme.
The immediate action is to release a controlled pilot matrix with fixed functional requirements, randomised build blocks and shared response definitions. Run the first screening batch, publish the raw measurements with uncertainty, and select the next design point from verified evidence rather than supplier preference or a single successful specimen.