Using DOE to Improve Gearbox Housing Joint Reliability
In a geared aircraft engine, a housing joint can look straightforward while carrying a demanding combination of loads. Bolts must maintain clamp force through vibration, thermal cycling, torque reversals and high-speed operation. If preload is too low, the joint may slip or open microscopically. If it is too high, housing distortion, thread damage or excessive local stress can create a different failure path.
Design of experiments (DOE) gives engineers a structured way to study these interactions. Rather than changing one torque setting at a time, a planned test matrix can reveal how torque, surface condition, bolt scatter, joint stiffness and operating temperature influence one another. That evidence supports a setting that is repeatable in production and robust in service.
For the OPTIMIZE Project, this type of investigation fits a wider effort to reduce power losses while improving gearbox efficiency, durability, weight and power density. A housing joint must preserve alignment and oil containment without adding unnecessary material or assembly burden. Its behaviour can therefore affect bearings, gears, lubrication and the whole propulsion system.
The Australian operating context makes robust joint design especially valuable. Components may be developed around Adelaide’s aerospace and defence capability, manufactured through suppliers in Melbourne or Brisbane, and then shipped long distances to a test facility or operator. Heat, dust, salt air and limited maintenance access can turn a small variation in clamp load into a serious reliability concern.
Why Housing Bolt Preload Matters
Housing bolts create the compressive force that holds two mating structures together. That force controls the contact pressure across the joint, the amount of separation under load and the resistance to transverse movement. Joint stiffness is governed by both the bolts and the clamped housing material, so a torque value cannot be assessed in isolation.
A gearbox housing experiences changing forces as gears mesh and shafts rotate. These loads can produce small relative movements at the interface. When the contact surfaces move repeatedly, fretting may remove material, generate debris and create small cracks. Fretting can also damage protective finishes and alter the contact geometry, allowing more movement during later load cycles.
Torque is only an indirect measure of preload. Much of the applied torque is consumed by thread friction and under-head friction, leaving a smaller portion to stretch the bolt. Lubricant, coating, washer design, thread condition and operator technique can therefore produce large preload differences at the same torque setting. A joint that appears correctly assembled may contain bolts with substantially different clamp forces.
The target should be a reliable preload range rather than a single impressive torque number. That range must keep the joint closed under service loads, limit slip, avoid local housing damage and remain achievable with approved tools and processes.
Building A Useful DOE
A sound DOE starts with a clear response definition. Useful responses may include bolt preload, joint separation, interface slip, dynamic stiffness, torque loss after vibration and fretting wear depth. Strain gauges, ultrasonic bolt measurement, load washers, displacement sensors and digital image correlation can each provide evidence, depending on the test article and measurement uncertainty.
Factors should represent realistic sources of variation. Candidate inputs include nominal tightening torque, lubrication state, bolt and nut finish, washer configuration, housing surface roughness, flange thickness, bolt diameter, tightening sequence and temperature. The test plan can also include repeated assembly, because a joint that performs well once may behave differently after maintenance removal and reinstallation.
A screening design can identify the few factors that matter most. A response-surface design can then refine the operating window around promising conditions. For a complex gearbox, a fractional factorial arrangement may reduce test count while still revealing key interactions. Engineers should avoid treating DOE as a statistical exercise detached from mechanics; each factor and response needs a physical explanation.
The test matrix should include centre points and repeated runs. Centre points help detect curvature, while repeats estimate random variation and expose poor measurement repeatability. If the planned torque is near a transition between sticking and slipping, extra observations around that region can be more valuable than a large number of distant test conditions.
Controlling Australian Test Conditions
Australian conditions can make environmental control a practical part of the experiment. A component assembled in humid Brisbane may have a different surface condition from one prepared in dry South Australia. A test conducted in a cool Melbourne laboratory may not reproduce the bolt relaxation and housing expansion experienced during a hot ground run near Perth or in the Pilbara.
Temperature should be recorded at the joint, not merely in the test room. Thermal gradients can change flange contact pressure and create differential expansion between bolts, housing sections and inserts. Cycling between cold soak and elevated temperature can show whether preload is retained after the assembly has expanded and contracted several times.
Measurement setup also deserves care. Optical systems, cameras and laser instruments can be affected by reflections, changing illumination and surface finish. Guidance on ambient light effects is relevant when selecting a measurement arrangement for displacement or surface movement, particularly where a shiny machined housing is being monitored.
Dust and salt exposure may be relevant to fielded hardware or logistics. Assemblies transported from an Adelaide supplier to a coastal facility may encounter different contamination risks from those stored inland. The DOE does not need to simulate every Australian environment, but it should identify which conditions could alter friction, corrosion, lubricant behaviour or contact stability.
Selecting Factors And Responses
A practical DOE separates controllable assembly variables from noise factors. Torque strategy, tightening sequence and lubrication may be controlled by the production process. Ambient temperature, surface contamination and small dimensional deviations may be treated as noise factors. This distinction helps engineers find a setting that remains effective when conditions move away from the laboratory ideal.
Key responses should connect directly to failure prevention. Initial preload indicates whether the assembly process is working. Joint stiffness shows how the housing responds to load. Slip displacement and hysteresis reveal interface movement. Fretting scars, oxide debris and wear depth provide physical confirmation after endurance testing. Torque-angle data can add information about thread engagement and seating behaviour.
- Preload retention: the percentage of clamp force remaining after thermal and vibration exposure
- Interface movement: relative displacement during representative gearbox loading
- Dynamic stiffness: resistance of the joint to vibration-induced deformation
- Fretting severity: wear, debris, surface damage and crack indications
- Assembly variation: spread caused by tools, operators, coatings and repeated installation
The factor ranges must be realistic enough for manufacturing. Testing an ideal bolt with a perfectly clean flange may produce a beautiful result that cannot be repeated on a production line. Include the approved lubricant, actual coatings, representative housing materials and the torque tools available to Australian assembly partners.
Connecting DOE With Simulation
Finite element analysis can predict bolt stretch, flange contact pressure and housing deformation before hardware is built. A detailed model can identify areas where the joint is likely to separate or where contact pressure falls during a gear-mesh load cycle. It can also show whether increasing torque improves stiffness or simply concentrates stress around a boss, counterbore or thin wall.
The physical DOE should be used to calibrate and challenge the model. Preload measurements can check bolt strain assumptions, while displacement data can reveal whether the contact formulation is realistic. If the model predicts a stiff, stable joint but testing shows slip, the difference may come from surface roughness, embedment, friction or local geometric flexibility.
Tolerance analysis extends the approach to production variation. Bolt diameter, hole position, flange flatness, surface waviness and material strength all influence the final joint. A Monte Carlo model can combine these distributions with the DOE response surface to estimate the probability of separation or inadequate stiffness. That is more useful than reporting a nominal result from one perfect assembly.
The same workflow can support decisions about washers, inserts and housing thickness. If a small washer change reduces preload scatter, it may deliver more reliability than a higher torque value. If extra housing material increases mass without improving contact stability, the design team can redirect that weight to a location with greater structural value.
Managing Fretting In Service
Fretting prevention begins with eliminating harmful relative motion. Higher clamp force can increase frictional resistance, but it is not a universal remedy. Excessive preload may deform the flange, close an oil passage, distort a bearing seat or reduce fatigue life around a threaded hole. The best solution may combine appropriate torque with improved flange geometry, surface treatment, joint alignment or a more suitable fastener arrangement.
Bolt tightening sequence can be especially important in a long gearbox housing. Tightening from one end can trap gaps or create uneven contact, while a controlled cross-pattern or staged sequence may produce a more uniform flange pressure. The DOE should test the sequence used by technicians, including tool access limitations and realistic pauses between stages.
Lubrication and oil retention also influence the evidence collected from a test. A nearby bearing or seal may distribute oil across the housing interface, changing friction and debris transport. Research on bearing shield oil retention offers a useful reminder that lubrication behaviour under acceleration can affect surrounding component interfaces and test interpretation.
Inspection should combine visual examination with measurement. Fretting may begin as discolouration or fine powder before it becomes an obvious groove. Surface profilometry, microscopy, dye penetrant inspection and crack-focused non-destructive testing can help distinguish harmless marking from progressive damage. Record the location, length and depth of any damage so results can be compared across DOE runs.
Turning Results Into A Production Window
DOE results should be translated into an assembly specification that people can apply consistently. The output might define a torque range, lubricant type, tightening sequence, tool calibration interval and acceptance checks. Where torque alone creates too much scatter, a torque-angle method, direct tension indicator or calibrated electronic tool may provide better control.
The production window should account for uncertainty. Measurement error, bolt friction variation and environmental differences all widen the true range of preload. A setting that works only at the centre of a narrow laboratory range is a weak control strategy. Capability analysis can show whether the process can hold the required preload distribution over time.
For Australian aerospace supply chains, documentation and traceability matter. A component may pass through several specialist businesses before final integration, with transport between states and long lead times for replacement parts. Clear records of bolt batch, coating, lubricant, tool identification and assembly temperature make it easier to investigate a deviation without scrapping an otherwise valuable assembly.
Training should use plain, practical language. A technician in regional Queensland may need a process that is just as unambiguous as one used in an Adelaide integration workshop. The instructions should state what to do when a bolt is removed, a thread is damaged, a tool calibration is overdue or the flange shows contamination. Calling out these situations prevents informal workarounds that quietly change the DOE-approved process.
Linking Joint Design To The OPTIMIZE Mission
A gearbox housing joint sits within a larger efficiency problem. Poor stiffness can alter shaft alignment, increase gear contact losses and transmit unwanted vibration into bearings. Fretting debris can contaminate lubricant and accelerate wear. Conversely, a well-controlled joint may allow a lighter housing while maintaining the dimensional stability required for high-speed operation.
The OPTIMIZE Project’s combination of simulation, tolerance analysis, physical testing and design-of-experiments methods is well suited to this trade-off. DOE identifies influential variables, simulation explains the mechanics, and testing confirms whether the predicted operating window survives real variation. Together, these methods reduce the risk of optimising one metric while damaging another.
The project’s work can also help researchers communicate results across disciplines. A gearbox designer may focus on stiffness, a manufacturing engineer on torque scatter, a tribologist on fretting debris and a test engineer on sensor uncertainty. A common response framework gives each specialist a way to connect local observations with propulsion-level performance.
Teams developing a relevant test case or seeking project information can use the OPTIMIZE contact page to follow the initiative and identify appropriate points of connection. Sharing well-documented results helps build a stronger evidence base for efficient, durable geared aircraft engines.
A robust torque strategy is therefore more than an assembly instruction. It is a controlled design decision supported by preload measurement, joint mechanics, tolerance analysis and service-representative testing. When the process is planned carefully, the result is a gearbox housing that resists slip and fretting while preserving the low mass and high power density demanded by modern aircraft propulsion.
Use the DOE findings to define the preload range, validate it with thermal and vibration testing, and lock the approved method into manufacturing documentation. That disciplined path gives engineers and suppliers a defensible basis for reliable housing joints across Australian facilities and international aerospace programmes.