Failure Autopsy

Weight 25% · Capability: Diagnose · Due 14 May 2027, end of Week 11

Reconstruct the most defensible failure sequence one supplied case will support, and say clearly where the evidence stops.

Task at a glance#

• Mode — individual.
• Format — a forensic report built on one case from the supplied failure-case pack: evidence inventory, event sequence, annotated failure diagram, causal chain, competing explanation, uncertainty record.
• Scale — one case, worked to the limit of its evidence rather than several cases surveyed.
• Evidence — the supplied pack. You may not substitute a specimen of your own for the case, though you may reference your own earlier tests as comparison material.
• Route — one file, submitted to the SLOP2895 submission point before the deadline.

The brief#

Each case in the pack is a structure that has already come apart, documented after the fact by someone who was not watching when it happened. You get photographs taken from two or three angles, a handful of measurements, some context about handling and transport, and gaps. The gaps are not an oversight in the pack. They are the condition every real failure investigation works under, and they are the reason this task exists.

The declared problem is that a failure’s appearance and a failure’s cause are different things, and the first is much easier to obtain than the second. A separated interface looks the same whether it separated because the interface layer lost grip, because the layer above it was displaced by a shear load during transport, or because moisture had already compromised the boundary hours earlier and the handling merely finished it. Visual resemblance to a failure mode you have seen before is a hypothesis. It is never, on its own, a confirmed cause, and a report that treats it as one loses marks under evidence handling however confident it sounds.

What is assessed is the reconstruction: an ordered sequence of events in which each step is supported by a named item of evidence, at least one rival sequence that the same evidence also permits, and a clear account of what would distinguish them if you could go back and measure it. A report that reaches a confident single answer the evidence cannot support marks below one that reaches two candidate sequences and says precisely why it cannot choose.

Required submission#

1. An evidence inventory: every item in the case, numbered, with what it shows and what it does not, and an explicit list of the evidence the case lacks.
2. A proposed event sequence, ordered in time, with each step keyed to the inventory items that support it.
3. An annotated failure diagram locating the initiation point, the propagation path and the final separation on the specimen itself.
4. A probable causal chain, distinguishing the initiating condition, the contributing conditions and the immediate trigger.
5. At least one competing explanation, tested against the same inventory rather than dismissed in a sentence.
6. A record of unresolved uncertainty: what remains undetermined, and the specific observation or measurement that would settle it.

Recommended process#

1. Observe the case in full and inventory it before forming any account of what happened.
2. Document what each item can and cannot establish, including the angles and regions the photographs never show.
3. Classify the failure by where it occurred — within a layer, at an interface, or across the assembly — before proposing why.
4. Diagnose an initiation point, and state the evidence that places it there rather than one layer over.
5. Compare at least two sequences that the inventory permits, and identify the evidence each one predicts.
6. Test both sequences against the items that ought to distinguish them, and report the result even when it separates nothing.
7. Revise the causal chain to the strongest claim the evidence still supports after that comparison.
8. Defend the surviving reconstruction by naming the single observation that would overturn it.

Evidence standard#

This report stands or falls on four registers staying visibly apart.

• Observed

  Drawn solid: present in the evidence.

• Inferred

  Drawn dashed: reasoning laid over the evidence.

• Uncertain

  Drawn as an open span: the evidence fixes no one value.

• Limitation

  Drawn as far as the method reached, and no further.

Observed is what the pack shows, cited by inventory number. If a claim cannot be traced to a numbered item, it is not an observation.

Inferred is the step from evidence to event. Each inference states the items it rests on and the assumption it adds to them — that a mark was made in one direction, that two items document the same moment, that the specimen was upright when photographed.

Uncertain covers what the evidence underdetermines: sequence where only outcome is visible, elapsed time, conditions before documentation began, and any point where two explanations remain equally supported. Name the measurement that would resolve each one.

Limitation is what the pack itself cannot deliver — the angle not photographed, the interface not sectioned, the handling not recorded — as distinct from what you chose not to pursue.

An uncertainty closed by assertion is a heavier error here than an uncertainty left open, because the report’s value is the boundary it draws around what is actually known.

Equivalent participation#

The case comes from the supplied pack for every student, so the evidence base is identical across the cohort. The three routes apply to any supporting work you choose to bring alongside it:

• Physical specimen — reproduce a candidate failure mode to compare against the case, with no requirement to consume anything.
• Non-food analogue — reproduce the same mode in foam, sponge, cardboard, clay or textile.
• Supplied evidence — work from the pack alone, using the comparison cases provided with it.

All three assess the same outcome and no route loses marks. Supporting work is optional in every route; a report resting entirely on the supplied pack can receive full marks. See the participation policies.

Safety, cost and waste#

The core of this task requires no materials at all. Any comparison specimen you choose to make must be the smallest that serves the comparison, and a non-food analogue is preferred wherever the comparison is destructive. Nothing tested may be presented to anyone as safe to eat. Cases in the pack are cleared for reproduction within the course; do not republish them elsewhere. Full detail is on the policies page.

Preparation#

Weeks 7 to 11 build the diagnostic capability, on the evidence discipline established in the Field Log.

• Week 7: Thermal Conflict — one assembly whose behaviour changes over time.
• Week 8: The Order of Layers — why the same material set fails differently when reordered.
• Week 9: Cutting Geometry — how a cut changes support, grip and exposed boundaries.
• Week 10: Packaging Is Part of the Structure — transport damage, and distinguishing protection from imposed load.
• Week 11: Failure Autopsy — reconstructing a sequence from limited evidence, worked through in the teaching week this task falls due.