A Curvic Coupling is a face-toothed coupling with circular-arc teeth ground simultaneously on both mating faces, so each tooth meshes with its partner along a curved profile rather than a straight one. Rolls-Royce Trent jet engines use them to clamp compressor discs into a rotor stack. The coupling self-centres under axial preload, transmits high torque, and indexes precisely on every reassembly. That gives you concentric repeatability under 5 µm and torque capacity into the tens of thousands of Nm in a part the size of a dinner plate.
Curvic Coupling Interactive Calculator
Vary preload, flank angle, tooth count, and starting offset to see how Curvic teeth convert axial clamp load into radial self-centering force.
Equation Used
This calculator models the Curvic Coupling self-centering action described in the worked mechanism section: axial preload squeezes the curved face teeth together, and the flank wedge angle converts part of that clamp load into radial centering force. Larger preload or flank angle increases centering force; more engaged teeth reduces the force carried by each tooth.
- Curved tooth flanks are approximated as symmetric wedges.
- Axial preload is shared evenly by the engaged teeth.
- Friction, local elastic deformation, and manufacturing pitch error are not included.
- Flank angle alpha is measured as the effective wedge angle converting axial preload into radial centering force.
Operating Principle of the Curvic Coupling
A Curvic Coupling works by squeezing two face-ground discs together with a tie-bolt or central drawbar. The teeth on both faces are cut by the same grinding wheel in a single setup, so when you press the two halves together they self-locate — any radial offset gets pushed back to centre as the curved tooth flanks slide into full contact. That's the trick. You don't need dowel pins, you don't need a pilot diameter. Preload alone aligns the joint, and the curved tooth geometry means contact is a true line contact along the full tooth length, not a point.
The Gleason Curvic process — invented at Gleason Works in Rochester in the 1940s — grinds all teeth in one revolution of the workpiece against a cup-shaped grinding wheel. That single-setup grind is what gives the coupling its self-centering coupling behaviour and its sub-5 µm concentricity. If the tooth flanks are off by even 10 µm in lead or pitch, the rotor stack runs eccentric, you get 1× vibration at running speed, and on a gas turbine disc coupling that shows up as bearing distress within hours.
Tolerances matter ruthlessly here. Tooth-to-tooth pitch error must stay under 3 µm on aerospace-grade parts. Surface finish on the flanks needs to be Ra 0.4 µm or better — anything rougher and the contact pattern goes patchy under axial preload, fretting wear starts on the high spots, and the joint loses its self-centering ability after 50-100 thermal cycles. The classic failure mode is fretting fatigue at the tooth roots when preload drops below the design value, usually because the tie-bolt has stretched or the disc faces have crept under cyclic thermal load.
Key Components
- Male Curvic Half: The disc face carrying the convex circular-arc teeth, ground simultaneously with the female half on the same machine. Tooth count typically runs 24 to 180 depending on diameter — a Trent 1000 HP compressor disc uses around 60 teeth at roughly 250 mm pitch diameter.
- Female Curvic Half: The mating disc face with concave teeth of identical curvature. Pitch error between male and female must stay under 3 µm for aerospace work, 10 µm for industrial gas turbines and CNC indexing applications.
- Tie-Bolt or Central Drawbar: Provides the axial preload — typically 60-80% of bolt yield — that pulls the faces into full tooth contact. On rotor-stack coupling assemblies the tie-bolt runs through multiple discs and clamps the entire stack as one rigid shaft.
- Tooth Flank: The working face of each tooth, ground to a true circular arc. Surface finish must be Ra 0.4 µm or better. Flank angle is usually 30° per side, giving a 60° included angle that balances radial centering force against axial preload demand.
- Tooth Root Fillet: The rounded transition between tooth flank and disc face. Root radius is the highest-stress region and the typical fatigue initiation site — keep it generous, usually 0.5-1.0 mm depending on tooth size, and shot-peened on rotating aerospace parts.
Where the Curvic Coupling Is Used
Curvic Couplings show up wherever you need to stack rotating discs into a single shaft assembly with sub-micron concentricity and the ability to disassemble for inspection without losing alignment. Jet engines and industrial gas turbines are the headline use case, but precision CNC rotary indexing tables run them too — a face spline coupling that locks at every 1° or 5° position with repeatability better than 1 arc-second.
- Aerospace propulsion: Rolls-Royce Trent 1000 and Trent XWB compressor and turbine disc stacks — each rotor disc joins to its neighbour through a Curvic Coupling clamped by a central tie-shaft.
- Industrial power generation: Siemens SGT-800 industrial gas turbine rotor — multiple compressor discs and turbine wheels stacked through Curvic joints for field-serviceable rotor reassembly.
- CNC machining: DMG MORI NTX 2500 turn-mill centre B-axis indexing head — Curvic-style face gear teeth lock the head at programmed positions with arc-second repeatability.
- Helicopter drivetrains: Sikorsky CH-53K main gearbox planetary stages — Curvic Couplings transmit torque between gear stages while accommodating disassembly for overhaul.
- Heavy machine tools: Schiess vertical turret lathes — large-diameter Curvic indexing rings position the turret to within 2 arc-seconds across 24 stations.
- Marine propulsion: MAN Energy Solutions marine gas turbine packages — rotor stacks assembled with Curvic Couplings for shipboard maintenance access.
The Formula Behind the Curvic Coupling
The torque capacity of a Curvic Coupling depends on the axial preload from the tie-bolt, the friction coefficient at the tooth flanks, and the mean tooth radius. At the low end of typical preload — around 40% of bolt yield — the joint will transmit torque but may slip under shock load and lose its self-centering coupling action. At nominal preload of roughly 70% yield, you get full tooth contact, predictable torque capacity, and stable centering across thermal cycles. Push preload past 85% and you risk yielding the tie-bolt during transient overspeed or thermal spike events. The sweet spot for most rotor stack coupling designs is 65-75% bolt yield.
Variables
| Symbol | Meaning | Unit (SI) | Unit (Imperial) |
|---|---|---|---|
| T | Torque capacity of the coupling before tooth slip | N·m | lbf·ft |
| Fpreload | Axial clamp force from the tie-bolt | N | lbf |
| μ | Coefficient of friction at tooth flank contact (typically 0.10-0.15 for ground steel) | dimensionless | dimensionless |
| rm | Mean tooth radius (pitch radius of the Curvic teeth) | m | in |
| α | Tooth flank pressure angle, typically 30° | degrees | degrees |
Worked Example: Curvic Coupling in an industrial gas turbine rotor stack
Sizing the Curvic Coupling between the third and fourth compressor discs of a 25 MW industrial gas turbine. Mean tooth radius is 140 mm, tie-bolt provides nominal axial preload of 800 kN, flank friction coefficient is 0.12, flank angle is 30°. We need to know the torque capacity at low, nominal, and high preload to verify the joint won't slip under the 95 kN·m design torque.
Given
- rm = 0.140 m
- Fpreload,nom = 800,000 N
- μ = 0.12 dimensionless
- α = 30 degrees
- Design torque = 95,000 N·m
Solution
Step 1 — compute 1 / tan(α) for the 30° flank angle, which captures how flank geometry converts axial clamp force into tangential grip:
Step 2 — compute torque capacity at nominal preload of 800 kN, which corresponds to roughly 70% yield on a typical M36 tie-bolt in 4340 steel:
For a full Curvic with 60 teeth all engaged simultaneously, the effective torque capacity scales with the integrated radius — the practical rule used in Gleason design data gives total coupling torque as roughly:
where neff ≈ 5.5 captures the effective load-sharing factor across the tooth ring at 70% preload. This sits comfortably above the 95 kN·m design torque — about 35% margin.
Step 3 — at the low end of typical preload, 460 kN (around 40% bolt yield, what you'd see if the bolt has stretched 0.05 mm under thermal cycling):
That's below the 95 kN·m design torque — the joint will start slipping under any transient, the flanks micro-slide, fretting wear begins, and within a few hundred hours you'll see vibration climb on the 1× line. At the high end, 920 kN preload (about 80% yield, near the practical ceiling), capacity rises to roughly 147,000 N·m, but the tie-bolt is now operating with very little reserve for thermal transients and you risk bolt yielding on an overspeed event.
Result
Nominal torque capacity comes out at about 128,000 N·m — a healthy 35% margin over the 95 kN·m design torque. At low preload (460 kN) capacity drops to 73,600 N·m, which means the joint slips under design load and you get the classic fretting-wear signature on the flanks. At high preload (920 kN) capacity rises to 147,000 N·m but the tie-bolt is running close to yield. If your measured torque-to-slip is below predicted, look first for: (1) tie-bolt relaxation from thermal cycling — re-torque procedures should specify a hot re-check after first heat run, (2) flank surface finish degraded above Ra 0.8 µm from light corrosion, which drops effective μ from 0.12 to under 0.08, or (3) tooth pitch error above 5 µm causing only partial tooth engagement so neff falls from 5.5 to closer to 3.
Choosing the Curvic Coupling: Pros and Cons
Curvic Couplings compete with Hirth joints and conventional spline couplings for high-torque rotor-stack jobs. Each has a sweet spot, and the choice usually comes down to torque density, repeatability after disassembly, and manufacturing cost.
| Property | Curvic Coupling | Hirth Joint | Splined Shaft Coupling |
|---|---|---|---|
| Concentricity after reassembly | < 5 µm | < 3 µm | 20-50 µm (depends on pilot fit) |
| Torque capacity (typical 250 mm OD) | 100-150 kN·m | 80-130 kN·m | 40-80 kN·m |
| Maximum operating speed | 20,000+ RPM | 20,000+ RPM | 8,000-15,000 RPM |
| Manufacturing cost (relative) | High — Gleason curvic grinder required | Very high — straight teeth, slow generation grind | Low — standard hobbing and broaching |
| Disassembly / reassembly count before measurable wear | 100-200 cycles | 200-500 cycles | 10-50 cycles before fretting |
| Self-centering under preload | Yes — inherent to curved flanks | Yes — straight radial teeth | No — requires pilot diameter |
| Best application fit | Aero engine and gas turbine rotor stacks | Highest-precision indexing heads | General industrial drivetrains |
Frequently Asked Questions About Curvic Coupling
Static concentricity on a bench check doesn't predict dynamic concentricity under preload and centrifugal load. The most common cause is uneven tooth-flank contact pattern — under the bench check the joint sits in one stable position, but under tie-bolt preload it shifts to a different equilibrium where 2-3 teeth carry most of the load.
Blue the flanks with engineer's blue, assemble at full preload, disassemble, and look at the contact pattern. You want at least 80% of teeth showing contact across at least 70% of flank length. Anything less and you have either a pitch error in grinding or a face that's not flat to within 2 µm.
The design target is 65-75% of tie-bolt yield. That gives you full tooth-flank contact, stable centering across thermal cycles, and reserve for transient overspeed events. Below 50% the joint can slip under shock load and lose its self-centering action. Above 85% you eat into your reserve for thermal expansion — when the rotor heats up on startup, the bolt sees additional thermal stress on top of the mechanical preload, and you can yield the bolt on the first hot start.
Rule of thumb: torque the bolt cold to 70% yield, run a heat soak, then re-check torque. If preload has dropped more than 10%, the disc faces are still bedding in and you need another heat cycle.
Hirth wins on pure repeatability — the straight radial teeth give you marginally better concentricity (around 3 µm vs 5 µm) and slightly better angular indexing accuracy. But Hirth is harder to grind and costs significantly more for the same diameter.
For a CNC B-axis on a turn-mill centre where you need 1 arc-second repeatability across 360 index positions, Curvic is the practical choice — the cost premium of Hirth isn't justified unless you're chasing the absolute last 0.5 arc-second. For a precision rotary table on a jig borer, Hirth is worth the money.
That's a classic symptom of face flatness error or non-parallel mating faces. The teeth are circular-arc-ground perfectly, but if the disc face isn't flat to within 2-3 µm, preload concentrates the contact at the OD where the gap is smallest.
Check both mating disc faces on a surface plate with a 0.5 µm indicator. Faces should be flat across the full annulus carrying the teeth. If you find a dish or crown over 3 µm, the face needs to go back to the grinder before the teeth see any more service — running it as-is will fret the OD teeth and eventually crack the tooth roots.
You can, but it's a worse joint. A central tie-bolt applies preload symmetrically about the axis and the resulting clamping pressure distributes evenly across the tooth ring. Peripheral bolts apply preload at discrete points, so the clamping pressure varies around the ring — teeth nearest each bolt see higher contact, teeth between bolts see less.
The result is uneven flank wear, reduced effective tooth count, and worse centering repeatability. Peripheral-bolt Curvics work for low-speed industrial joints where the torque is well below capacity, but for any high-speed rotor application, use a central tie-bolt or a tie-shaft running the full length of the stack.
Industry standard is 30° flank pressure angle and a tooth count that gives you 3-5 mm circular pitch at the mean radius. For 200 mm pitch diameter that's roughly 48-60 teeth. Lower tooth counts give higher per-tooth torque but worse load sharing under any pitch error. Higher tooth counts share load better but become more sensitive to grinding errors.
For aerospace and high-speed gas turbine work, 60 teeth at 30° flank angle is the well-trodden path — Gleason has published design data for this geometry going back to the 1950s, and most curvic grinders are tooled for it as standard.
References & Further Reading
- Wikipedia contributors. Curvic coupling. Wikipedia
Building or designing a mechanism like this?
Explore the precision-engineered motion control hardware used by mechanical engineers, makers, and product designers.
