A component installed correctly, torqued to specification and inspected without finding is not, on that basis, a component that was sound. Hydrogen-assisted cracking needs time under load: hydrogen has to diffuse to the region of highest triaxial stress, accumulate there, and reduce the local cohesive strength enough for a crack to start. That takes hours, days or weeks. The interval between a clean inspection and a brittle fracture is not evidence that something happened in between.
Incubation is part of the mechanism
Most mechanisms familiar to a claims file leave a trail: fatigue accumulates cycles, corrosion removes measurable metal, creep produces measurable strain. Hydrogen damage does none of that. Between hydrogen entering the metal and a crack initiating, nothing changes that a field measurement would detect.
What is happening is diffusion. Atomic hydrogen migrates toward regions of high triaxial tensile stress and concentrates there. When local concentration and local stress are jointly sufficient a crack initiates, and in a material with little ductility to blunt it, propagation follows quickly.
Sustained load, not an event load
The load that fails a hydrogen-embrittled part is usually the ordinary one: preload from assembly, dead weight, a residual stress left by fabrication or forced fit-up. There is often no event to point to, which is why these failures arrive described as the part having simply broken.
An investigation that begins by hunting for an overload can spend a long time finding nothing, because the applied stress at fracture may have been well below yield and entirely within design intent. The absence of a triggering event is itself a signal.
What a passed inspection actually establishes
Receiving inspection, dimensional checks, hardness verification, magnetic particle or dye penetrant examination and torque verification all sample properties that hydrogen embrittlement does not change until it fractures the part. A component carrying enough hydrogen to fail next month passes every one of them this month.
The argument is narrower than it is usually pitched. A clean inspection establishes that no crack was present at that moment. It does not establish that the component was fit to remain in service, and where the mechanism is hydrogen those are different propositions.
Proof testing can mislead
A sustained-load test and a proof load are not the same instrument. Short-duration proof and wedge tensile testing of the kind set out in ASTM F606 for threaded fasteners confirms strength, but applies load for far too little time for hydrogen to redistribute. A part can pass a proof load comfortably and fracture under a lower sustained load days later. The tests that do reveal susceptibility either hold load for a long period or ramp it very slowly, which is the design of the ASTM F519 sustained-load specimen and the incremental step loading of ASTM F1624.
Baking, and the clock attached to it
Post-plating baking gives absorbed hydrogen a route out: heat the part, hold it, let hydrogen diffuse to the surface and escape. ASTM B850 is the guide for these post-coating treatments. Time at temperature depends on section thickness and strength level, but the variable that most often decides a dispute is the delay before the bake started.
Some coatings act as a barrier that slows escape, so specifications call for baking as soon as practicable after plating rather than at a convenient point in the production run. A bake performed days later, or after the part was assembled and preloaded, is a different act from the one the specification contemplates.
The documents that carry the case
Where the mechanism is invisible and the timeline is the argument, process records become the primary evidence rather than the supporting evidence: travellers and route cards with time stamps, plating bath logs, oven charts for the specific load, lot traceability from the component back to the heat and the finishing batch, and certificates of conformance read for what they actually certify.
Retention is governed by contract and quality system requirements rather than by anyone's anticipation of a dispute, so requesting these records early matters more here than in most mechanisms.
Assembly sits inside the timeline
The clock that matters runs from hydrogen introduction to fracture, and assembly sits in the middle of it. When the part was plated, when it was baked, when it shipped, when it was installed and when load was first applied together determine whether the observed interval is consistent with hydrogen.
Installation records also carry load history: torque values, whether a component was tightened, loosened and re-tightened, whether it was reused. Re-torquing raises stress on a part that may already be charged.
One component or the whole lot
Because hydrogen is usually introduced by a process applied to a batch, a single delayed fracture raises an immediate question about everything else that went through the same process, which is a risk-management question before it is a liability one.
Quarantining the remainder of the lot serves both purposes. It removes exposure, and it supplies the exemplar material susceptibility testing requires, with the same processing history as the part that failed.
Where these opinions are challenged
Predictably: that the interval between assembly and fracture was asserted rather than documented; that the bake record produced was for a different lot; that the inspection relied upon tested for a property the mechanism does not change; that a proof-load result was offered as evidence of freedom from embrittlement.
The answer is a timeline built from dated records rather than recollection, with an explicit statement of which tests would and would not have detected the condition when they were performed.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.