"The shaft failed by fatigue." True, perhaps — but useless. Fatigue is a mechanism, not a cause. If you replace the shaft and change nothing else, the same cyclic stress will crack the new one on roughly the same schedule. Root-cause analysis exists to close the gap between naming the mechanism and knowing what to change.
Mechanism, cause, root cause
Three layers, each answering a different question:
- Mechanism — how did it fail? Fatigue, SCC, overload, corrosion. Established from the physical evidence: fractography, metallography, deposits. Example →
- Cause — what drove the mechanism here? The cyclic stress from a misaligned coupling; the chlorides concentrated under failed insulation; the hydrogen from a skipped bake.
- Root cause — why did that condition exist? The alignment procedure that allowed the misalignment; the inspection plan that never looked under insulation; the plating subcontract without a bake requirement. Root causes are usually decisions, procedures and assumptions — which is exactly why fixing them prevents recurrence.
Hypotheses, not hunches
Good investigations run on competing hypotheses. From the first look at the evidence, several credible stories exist; each one predicts something checkable — if it was fatigue from misalignment, the crack should start here and the bearing wear should look like that; if it was a material defect, the origin should show it metallographically. Testing is then targeted at the predictions that separate the stories, and hypotheses are eliminated until the best-supported explanation remains. This is faster and cheaper than "test everything", and far more honest than working backwards from the preferred answer.
What honest conclusions look like
Evidence is not always complete — surfaces get destroyed, histories go unrecorded. An honest report distinguishes a confirmed mechanism (supported by direct evidence) from a most likely mechanism (the best-supported hypothesis, stated with its bounds and with what further evidence would firm it). Beware of investigations that always end in certainty: bounded conclusions are not weakness, they are what evidence-led reasoning looks like when applied by someone who is not overselling.
From root cause to action
The test of a root-cause finding is that it changes something implementable: a material, a design detail, an operating limit, a procedure, an inspection. A finding nobody can act on is a description wearing a root-cause costume. MTIS reports end in recommendations for this reason — the investigation is finished when you know what to do, not when the mechanism has a name. What a good report includes →
When to contact MTIS
When a failure keeps recurring despite fixes — the signature of mechanism-level fixes applied to a cause-level problem; when a failure's consequences justify knowing, not guessing; or when an existing report names a mechanism but leaves you unsure what to change. Our Failure Investigation service → · Start a job request →