Failure mechanisms, evidence and systematic failure analysis

Give the statements different jobs

A failure mode states how an item fails its required function: open circuit, unwanted leakage or insufficient output, for example. A mechanism describes the process producing the mode, such as wear or corrosion. An effect describes what the failure changes at a chosen system boundary. The same event can occupy different roles when the boundary changes.

Start by describing what happened, where and when, and what should have happened. Keep original condition records, operating history, instrument information and the sequence of events. Distinguish direct observations from interpretations and missing information. Altering a damaged item before recording its state can remove evidence that would discriminate between explanations.

Use competing explanations to direct evidence

A total measurement can locate a system-level problem while leaving its origin ambiguous. Subsystem observations help localise it, provided the measurement conditions are comparable and the measurement itself does not create the apparent difference. The figure is a supplied resistance account, not instructions to test live equipment.

A hypothesis becomes useful when it predicts an observation that differs from a competing hypothesis. Contact contamination, changed contact force and a damaged conductor can all affect an electrical path. Evidence about one junction may narrow the search but does not identify the chemical or mechanical process by itself. A comparison unit also needs a relevant history, rather than simply looking similar.

Connect physical evidence to the wider conditions

Immediate events and underlying conditions answer different questions. A component defect might interact with exposure, a missed inspection, a misleading interface or an unsuitable requirement. Several causes can contribute without there being one universal “root cause.” State which links are established and which remain hypotheses.

A proposed correction should address the supported mechanism and its enabling conditions. Replacing a part may restore function without preventing recurrence. Assign an owner for a change, define the evidence that will show it worked, and retain contrary findings. A persuasive story is weaker than an explanation that survives tests designed to disprove it.

Make the rival explanations predict different records

Continue the joint B example with a separate, invented evidence packet. Two candidate explanations are H-surface: a poorly conducting surface layer is the main source of the excess resistance; and H-force: reduced contact normal force alone accounts for it. Normal force is the compressive contact load. Real metallic contact occurs at small surface regions, so both their mechanical contact and intervening films can matter. These mechanisms can coexist; calling one the sole explanation creates a stronger claim to test.

For this classroom comparison, supply records from four matched contact specimens. Material, nominal geometry, temperature and electrical measurement conditions are held comparable. Each specimen has either the comparison surface state or the returned surface state, at one of two supplied normal forces. The returned surface category includes any associated surface changes; it is not a chemically identified contaminant. These are completed low-energy laboratory records to interpret, not directions or recommended loads for testing a connector.

H-force alone predicts that matching the normal force should remove the large returned-versus-comparison difference. H-surface predicts that a substantial difference can remain at matched force. The table supplies the observations rather than asking the reader to accept either prediction as a fact. Separate supplied examination finds a layer on the returned contact faces but not the comparison faces; its composition and origin remain unidentified. A corresponding bulk-conductor record shows no resolved returned-versus-comparison difference, further weakening conductor damage alone as an account of the large excess.

Localise a difference in an ideal series account

Supply matching low-energy laboratory records for a returned unit: lead 0.2 Ω, joint A 0.1 Ω and joint B 4.7 Ω. The ideal series sum is 5.0 Ω. A comparison unit has 0.2, 0.1 and 0.2 Ω at the corresponding locations, summing to 0.5 Ω.

The total difference is 4.5 Ω, and all of it appears at joint B in these supplied measurements. That supports localisation to this part of the account. It does not prove contamination, explain when a change arose or establish why it escaped earlier checks. Those questions require discriminating observations, measurement uncertainty and a reliable history.

At either matched force, returned minus comparison resistance is 4.50 Ω. Using the stated absolute bounds conservatively, that difference lies between 4.46 and 4.54 Ω. Changing the supplied force from 2 to 4 N changes resistance by only 0.05 Ω in either surface category, with a conservative difference range 0.01–0.09 Ω. The large surface-associated contrast remains when force is matched. This contradicts reduced force alone as the explanation within this packet and favours a surface-associated resistive mechanism consistent with the observed layer.

That is a positive but bounded conclusion. It does not identify the layer chemistry, separate film resistance from every associated surface change, establish when the condition arose or prove an organisational cause. Nor does it say force never matters outside these two loads. Preserve both hypotheses and the records; investigate the layer and exposure/history before selecting a recurrence-prevention change. Merely replacing or cleaning the contact would change several conditions and would not, by itself, establish that causal chain.

Try the reasoning

1. Is “high resistance” a complete failure mechanism?

   Show the answer and reasoning

   No. It describes measured behaviour or a mode relative to a requirement; several processes can produce it.

2. Does replacing joint B prove the original causal explanation?

   Show the answer and reasoning

   No. Restoration is useful evidence about the affected path, but replacement changes several possible causes at once.

3. Why retain an observation that contradicts the favoured hypothesis?

   Show the answer and reasoning

   It can reveal a wrong explanation or an additional mechanism. Discarding it makes the investigation less capable of preventing recurrence.

4. In a different matched packet, both surface categories give 4.70 Ω at 2 N and 0.20 Ω at 4 N, with the same stated uncertainty bounds. Which explanation gains support, and what should happen to the earlier conclusion?

   Show the answer and reasoning

   Here the matched-force surface contrast is zero within the bounds, while the force contrast is 4.50 Ω. This favours a force-dependent contact mechanism in the new packet and does not reproduce the earlier surface-associated contrast. A visible layer alone would not justify carrying the earlier explanation across. Check that the specimens, measurement conditions and histories really are comparable; neither packet establishes a universal law or a field prevention measure.

5. The original packet favours a surface-associated mechanism. Can it establish that a particular worker introduced the layer or that one cleaning change prevents recurrence?

   Show the answer and reasoning

   No. Layer composition, formation/transfer history and the wider task and exposure conditions remain unresolved. An action needs an evidence-supported causal link and a defined effectiveness check. Identifying a likely physical contribution is different from identifying who or what created it.

What to carry forward

Localise the behaviour, test the mechanism and trace supported causal links before deciding what prevents recurrence.