Fan Blade-Off Test: Blowing a Blade Off a Jet Engine at Full Power

What the rule demands: 14 CFR 33.94 and CS-E 810

Every turbine aircraft engine has to show, on a real running engine, that it can survive losing a blade. In the United States the rule is 14 CFR 33.94, “Blade containment and rotor unbalance tests”, added in 1984; in Europe it is EASA CS-E 810, Compressor and Turbine Blade Failure. The engine must contain the damage without catching fire and without failure of its mounting attachments while running for at least 15 s afterwards, unless the damage shuts it down by itself.

• Which blade: the most critical compressor or fan blade. In a high-bypass turbofan that is a fan blade, the largest and most energetic blade in the engine. A second test covers the most critical turbine blade, chosen by weighing blade mass against the strength of the hot turbine case.
• Where it breaks: at the outermost retention groove, just above the dovetail or fir-tree root that locks the blade into the disc, so the whole airfoil goes. A blisk (integrally bladed rotor) has no separate root, so at least 80% of the blade must be released.
• How fast: at maximum permissible rpm (redline), the most energy the blade can ever carry in certified service.

Paragraph (b) lets analysis backed by rig tests, component tests or service experience replace one of the two engine tests, but only the one that produces the less rotor unbalance, and only if the analysis is shown to be equivalent to the test. A fan release leaves far more unbalance, so on a big turbofan it is the one run on a full engine.

The release: why the blade flies along the tangent, not straight out

The instant the charge fires, the pull of the disc through the root, the only force bending the blade's path into a circle, disappears. By Newton's first law the blade carries on in a straight line along the tangent to its circle, at the speed of its centre of mass. It keeps tumbling about that centre at the rotor's rate, because nothing has yet applied a torque to stop it.

The picture of a blade shooting radially outward like a spoke is a rotating-frame view: riding on the disc, you would see it slide straight out and fall behind its neighbours. From the test cell the path is straight, and the centre of mass moves away from the axis only as r√(1 + θ²), where θ is the angle the rotor has turned since release. But the tip starts only millimetres from the wall, so it reaches the case in well under a millisecond, after the rotor has turned just a few degrees.

The blade does not bounce back. Still moving at ~260 m/s in the direction of rotation, it is pressed into the case and skids around the inside of the wall, gouging, bending and breaking up. Travelling on a larger radius and slowed by friction, it falls behind the rotor, and within a few milliseconds the trailing blade slams into it and often snaps off part of its own airfoil. The case therefore has to hold the released blade plus a large piece of its neighbour, all within about one 24 ms revolution, while the remaining blades grind their tips into the damaged wall.

The numbers: ~1 MN of pull, ~0.5 MJ of energy, 24 ms per turn

Take a widebody fan about 3 m across at a redline of ~2,500 rpm, with a ~15 kg blade whose centre of mass sits ~1 m from the axis. These are representative figures; real engines vary.

• Angular speed: ω = 2,500 × 2π ÷ 60 ≈ 262 rad/s, so one revolution takes 60 ÷ 2,500 s = 24 ms.
• Tip speed: at a ~1.53 m tip radius, v = ωr ≈ 262 × 1.53 ≈ ~400 m/s, faster than sound at sea level (~340 m/s).
• Holding force: F = mω²r = 15 × 262² × 1 ≈ 1.03 × 10⁶ N, about ~1 MN (~100 tonnes-force). The centripetal acceleration of ~68,500 m/s² is about 7,000 g, so the root carries some 7,000 times the blade's own weight.
• Released energy: the centre of mass leaves at ωr ≈ 262 m/s, so ½mv² = 0.5 × 15 × 262² ≈ 5.1 × 10⁵ J, or ~0.5 MJ. That is about a 1-tonne car at ~110 km/h (½ × 1,000 × 30.6² ≈ 0.47 MJ), in something you could carry under one arm. The tumble adds roughly a tenth more: for a ~1.1 m blade, I ≈ mL²/12 ≈ 1.5 kg·m² and ½Iω² ≈ 0.05 MJ.

Force and energy both scale with ω², so 10% more speed means 21% more of each, which is why the test is run at redline. The same arithmetic explains the rule's other half, rotor unbalance. Before release, the ~1 MN pulls of opposite blades cancel. Remove one and they no longer do: the rotor is left with an unbalance of mr ≈ 15 kg·m, which at 262 rad/s is a ~1 MN rotating force sweeping around about 42 times a second. That is typically a thousand times or more the residual unbalance a fan is balanced to in service, and it goes straight into the bearings, frames and mounts.

Catching it: hardwall, softwall and composite cases, and the bearing fuse

The containment zone is a band of the fan case in the plane of the blades. It must turn ~0.5 MJ, plus the trailing-blade fragments, into deformation and heat without letting anything through. Three approaches are used:

• Hardwall: a thick ring of steel, titanium or aluminium alloy stops the blade on its own through plastic bending, denting and local shearing. It is simple and stiff, but it must be thick all the way around a ~10 m circumference, so it is heavy.
• Softwall: a thin aluminium case that the blade is allowed to punch through, wrapped in many plies of Kevlar aramid fabric, which has several times steel's strength-to-weight ratio. The belt stretches and bulges outward like a catcher's mitt, absorbing energy as the fibres strain, and the nacelle must leave room for that bulge.
• Composite: the GEnx introduced a braided carbon-fibre composite fan case, and the CFM LEAP uses a 3D-woven, resin-transfer-moulded one. They work as a lighter hard wall, absorbing energy through fibre breakage, delamination and matrix cracking.

The ~1 MN rotating unbalance then hammers the fan's front bearing, the No. 1 bearing, about 42 times a second, and a rigid support would pass most of it into the frame, mounts and pylon. Many large turbofans therefore build a frangible fuse (also called a decoupler or load-reduction device) into the No. 1 bearing support. Above a set load it breaks, the support's radial stiffness collapses, and the fan is free to orbit.

This is Jeffcott rotor behaviour. A rotor spinning well above the natural frequency of its support whirls about its centre of mass: the shaft centre orbits with a radius close to the mass offset e, and the force passed to the structure is roughly the now-soft stiffness times e rather than the full mω²r. For a fan rotor of several hundred kilograms, 15 kg·m of unbalance moves the centre of mass only a few centimetres. The cost comes later. Running down, the fan passes back through the lowered critical speed, and the aircraft must then tolerate it windmilling unbalanced for the rest of the flight. Airframe rules such as 14 CFR 25.362 (engine failure loads) cover that separately.

Running the test

A blade-off is one of the most expensive single events in an engine certification programme, and it normally wrecks the test engine. The chosen blade's root is machined to take a small explosive charge, placed so that it severs the blade at the outermost retention groove. The engine, with its production fan case and mounts, sits on an instrumented thrust stand. It is run up to redline and held there, and the charge is fired remotely.

• High-speed cameras capture the blade's path, the trailing-blade hit and the case bulge. At 10,000 frames per second, one 24 ms revolution spans 240 frames.
• Load cells, strain gauges and accelerometers on the mounts, case and bearing supports measure the unbalance loads, which later feed the aircraft maker's engine-failure models.
• Speed and vibration records show when the bearing fuse let go and how fast the engine ran down. Teardown then confirms that nothing penetrated the case.

The run is the final exam after years of explicit finite-element simulation (codes such as LS-DYNA) of blade impact, fibre failure and rotor whirl. Its fire, mount and 15 s criteria catch what models can miss: oil and fuel lines torn by the whirling rotor, rub fires, and bolts failing under the shaking.

What this test is not: bird strikes and disc bursts

Bird ingestion (14 CFR 33.76, CS-E 800) is a separate test. Real birds, up to ~3.65 kg (8 lb) for a single large bird on the biggest engines, are fired into a running engine from a gas cannon. Blades may bend or break, but the threat comes from outside, not from the engine's own most critical blade at redline.

Disc bursts are a different and far worse category. A burst disc throws out thick, compact chunks, conventionally modelled as one-third-disc fragments, whose energy sits in a blunt mass that no case of flyable weight could stop. Certification does not try. Instead, discs get overspeed margins (14 CFR 33.27), strict cycle limits and inspections as life-limited parts (33.70), and airframes are laid out to limit what a fragment can hit (FAA AC 20-128A).

• United Airlines 232 (19 July 1989): the stage 1 fan disc in the DC-10's tail-mounted GE CF6-6 broke apart along a fatigue crack that grew from a metallurgical defect in the titanium. Fragments cut all three hydraulic systems. Steering with engine thrust alone, the crew reached Sioux City, and 184 of the 296 people aboard survived.
• Qantas QF32 (4 November 2010): an oil fire inside a Rolls-Royce Trent 900 on an Airbus A380, fed by a fatigue-cracked oil feed stub pipe, let the intermediate-pressure turbine disc overspeed and burst. Fragments pierced the wing and damaged fuel tanks and systems, but everyone survived.

When contained is not enough: Southwest 1380 and common misconceptions

Passing 33.94 proves the engine holds its own debris, not that the nacelle around it stays on. On Southwest Airlines Flight 1380 (17 April 2018), a fan blade on a Boeing 737-700's CFM56-7B broke away at its dovetail root from a low-cycle fatigue crack. The case contained the blade, but it hit the case at a spot critical to the fan cowl's structure. The inlet and fan cowl broke up, a piece of cowl struck the fuselage beside a window, the window came out, and a passenger was fatally injured. The NTSB found that this impact location and the cowl's response had not been fully evaluated in certification, and it called for blade-out assessments to cover critical impact locations and the nacelle. Fatigue-cracked hollow titanium blades on Pratt & Whitney PW4077 engines caused similar inlet and cowl separations on United 1175 (2018) and United 328 (2021).

• “Passing means the engine survives.” The test engine is usually scrap. The pass is about the aircraft surviving.
• “Contained means nothing happens outside the engine.” Impact loads can still shed cowls, and the airframe must ride out the unbalance and the windmilling vibration.
• “So blade failures are acceptable.” Blades are still life-managed and inspected: after Flight 1380 the FAA ordered ultrasonic inspections of CFM56-7B fan blades.

Fan blade-off versus the tests and failures it is confused with: only a blade release is something the engine case is built to contain.

Event	Governing rule	What happens	What must be shown
Fan blade-off	14 CFR 33.94(a)(1) / CS-E 810	Most critical fan blade cut free by a charge at maximum permissible rpm (redline), at the outermost retention groove; at least 80% of the blade for a blisk	Blade and fragments contained, no fire, mounts intact, 15 s of running unless the damage shuts the engine down
Turbine blade-off	14 CFR 33.94(a)(2) / CS-E 810	Most critical turbine blade released at redline, chosen by blade weight against the strength of the hot turbine case	Same criteria; whichever of the two blade tests produces the less rotor unbalance may be replaced by analysis shown to be equivalent
Bird ingestion	14 CFR 33.76 / CS-E 800	Real birds fired into a running engine: a single large bird (up to ~3.65 kg for the biggest inlets) plus flocks of smaller birds	No fire, no hazardous fragments through the case, engine can still be shut down; flock tests also demand continued thrust
Rotor overspeed	14 CFR 33.27 / CS-E 840	Rotor discs spun beyond redline, ~115–120% of maximum permissible rpm for 5 minutes	No burst; permanent growth and damage within limits
Disc burst (in service)	Not contained by design: life limits (33.70) and airframe layout (AC 20-128A)	The disc itself cracks and releases large, blunt fragments, as on United 232 (1989) and Qantas QF32 (2010)	Prevented by overspeed margins, life limits and inspection; the aircraft is laid out to tolerate fragments