A pressure cabin, three openings, one origin
The Hole Beside the Window
The usual telling points to square passenger windows. For G-ALYP, Withey reconstructed the fatal crack at a 10 mm countersunk bolt hole beside the rear ADF cutout, while retaining roles for cutout stress, local structure and fixing method. Run his conditional 1,286-cycle inverse estimate, normalized because its tabulated weight function is not printed, and see why about 100 micrometres is a model result, not an observed flaw.
For G-ALYP, the familiar passenger-window-corner origin story points to the wrong opening. Withey reconstructed the crack at a bolt hole, but estimated its roughly 100 µm starting size under a uniaxial Paris-law model whose tabulated weight function is not printed. It was not an observed manufacturing flaw.
Click the opening
Actual origin: reconstructed from G-ALYP wreckage
10 mm countersunk bolt hole, more than 50 mm from the rear ADF cutout edge.
Calculated stress in the vicinity: about 70 MPa. The final fatigue crack grew mainly toward the cutout and was 25 mm long.
This correction does not acquit geometry. Withey reports a field above three times remote stress around cutouts and about 315 MPa at the ADF edge. He judged fastener fixing probably more damaging than cutout shape, while noting that non-square windows might have permitted Redux bonding and avoided failure. The fatal origin was not the edge, but geometry, local structure and the geometry-dependent fixing method remain in the causal chain.
1,286 cycles
Matches the G-ALYP calibration target to within one cycle.
The baseline is normalized to Withey's published inverse result. Turning on the 315 MPa edge field while keeping the hole factor stacks two different physical locations. That combination is a sensitivity experiment, not a claim about G-ALYP. At extreme settings the model can predict only a few cycles; this is not a validated short-crack forecast.
The cause is a chain
The sophisticated objection is right: moving the origin does not make the cutout shape innocent. The record supports four linked roles, not one mascot.
Field
Cutout geometry shaped the surrounding stress. The measured edge maximum was about 315 MPa.
Origin
A countersunk fastener hole and an inferred initial defect supplied the local crack.
Growth
Cabin pressure returned near zero each flight, then rose again. Repetition accelerated the crack.
Test history
Prior proof loading may have plastically altered the separate section test at its high-stress sites.
Why 18,000 was not the fleet life
The forward cabin section survived this many cycles before fatigue failure near a window corner. Its fitted end bulkheads made it nonrepresentative of a whole aircraft.
The section had first been proof tested at twice operating pressure. Withey says local plastic deformation may have changed the high-stress regions. Fleet proof tests were 11 psi, not 16.5 psi.
Leave the overload toggle off and the crack model can print a normalized life. Turn it on and the model refuses: the cited record supplies no retardation law or uncertainty band that would convert the 2P proof load into extra cycles.
The check
The page computes the normalized life live from Withey's Paris constants. It also says exactly what that arithmetic can and cannot establish.
baseline: a₀ = 0.100 mm, aᶠ = 25 mm, Δσ = 70 MPa, F = 2.4155
Closed-form integral: 1,286.000 cycles. Separate log-space quadrature: 1,286.000 cycles.
Withey reports about 100 µm to 25 mm in 1,286 cycles, plus the distinct 70 MPa origin and 315 MPa edge fields. The cycle pair calibrates F; it does not validate F.
The checker rejects the 77 MPa poison: 878.36 cycles, outside the one-cycle anchor tolerance.
Prior 2P proof load returns: no sourced overload-retardation model. It never prints a made-up life.
What would fail: a changed Paris constant, unit conversion, factor, stress, crack bound or integration routine moves the independent routes or the 1,286-cycle target. A checker that accepted the poisoned 77 MPa case fails the control. The offline verifier also string-matches these rendered figures and the page metadata.
Every uncertainty and free choice
- Normalized factor. Withey used a tabulated crack-from-hole weight function but did not print its values. This page computes the constant effective factor 2.4155376321540145 required to reproduce his reported 100 µm, 25 mm and 1,286-cycle inverse result. It is a disclosed calibration, not a recovered table.
- Loading. The page uses Withey's uniaxial assumption. He says biaxial loading would slightly increase the inferred initial flaw.
- Environment. The Paris data came from G-ALYR in a water tank. Withey could not isolate whether its fast growth reflected alloy, application or environment.
- Residual stress. Compressive stress left by forming the hole could slow early growth and imply a larger initial defect.
- Short cracks. Paris-law extrapolation to 100 µm is a model. The initial flaw was not observed directly.
- G-ALYY. Its similar cause was inferred from circumstances. Insufficient wreckage was recovered to prove an origin, so this page does not draw one.
- Prior overload. The page refuses to quantify retardation from the separate section's 2P proof load.