Measure first. Explain second.#

This manual holds the methods, the vocabulary and the evidence standards used across all twelve weeks and all four assessments, so that a week’s brief can name a method instead of restating it. It is a reference, not a reading. Every week page and every assessment brief links to the section it uses; where a brief and this page disagree on method, tell a member of staff so one of the two can be corrected. The conditions under which any test may be run are set by the participation policies, which govern this page.

System vocabulary#

These nine terms are used in one sense throughout the course. A record that uses them in another sense is not wrong so much as unreadable by a second person.

The nine terms in full, with the sense each carries in this course:

The four failure families#

The course sorts recurring failures into four families. Each is anchored in one week, and every later week reuses the vocabulary rather than inventing its own.

These four families are this course’s own simplified analytical model. They were designed for SLOP2895 so that failures recorded in twelve different investigations can be compared in one vocabulary. They are not a recognised professional classification, they appear in no standard, and they must not be cited as one — not in an assessment, and not outside the course.

Treat a family name as a comparison label and nothing more. A claim about what a material or an interface did rests on the observation behind it; if the observation is removed, the family name carries no evidence on its own. A specimen may also sit in two families at once, and often does: a front that reaches an interface produces Soak and then Slip, and saying so is more defensible than choosing one.

Observation protocol#

Seven steps, in this order, for every test in the course. The order is the method: steps one and two are what make step six possible.

All seven steps, in order:

1. Identify. Give the specimen an identifier and record what it is made of, in stack order, before anything is applied. An unidentified specimen cannot be compared with a second one.
2. Condition. State the condition the specimen is held in, and list what you are holding constant — dimensions, stack order, ambient temperature, support positions, span, and who is reading the instrument. A constant you did not name is a variable you did not control.
3. Load. Apply one declared load, at a stated magnitude, for a stated duration, in the same way each time. One load per test. If two conditions changed, the comparison has nothing to attribute the difference to.
4. Observe. Watch the specimen while the load is on it. Record the moment behaviour changes, not only the state you find at the end.
5. Record. Write each measurement with its unit, the location on the specimen it was taken from, and the elapsed time. A measurement without its time and place is not evidence in this course.
6. Compare. Set the reading against a control, or against another declared condition. A single reading describes a specimen; it establishes nothing.
7. Interpret. State what the comparison supports, then state what it does not. Separate observed, inferred and unknown before writing a cause.

Test methods#

Six methods. Each one states what it varies, what it measures, and what it cannot establish — the third of those is the part that gets left out, and the part that is marked.

Compression#

Varies the applied load, in steps, at a fixed span with fixed support positions. Measures central deflection at each load step, and the geometry recovered on unloading. Cannot establish the strength of a material: the reading belongs to that specimen at that span with those supports, and a shorter span on the same material gives a different answer. It cannot say when a layer will hinge under a load it was not taken to. Used at Week 2: Bread Is a Beam, reused wherever a stack is loaded at Week 5: Load-Bearing Fillings and Week 8: The Order of Layers, and the readiest variable for the Component Test.

Slippage#

Varies the surface pairing at one interface, and the normal load pressing that pairing together. Measures the angle of first movement on a tilted support, and the displacement travelled from it. Cannot establish a coefficient belonging to one material: resistance is a property of the pair, under that normal load, at that moment, and a pairing that exchanges moisture across its interface will not give the same figure twenty minutes later. Nor does it locate which of the two surfaces is responsible without a residue reading. Taught at Week 3: Grip, Friction and Slip, diagnosed at the film scale in Week 6: Sauce at the Interface, and accepted as a declared variable in the Component Test.

Moisture migration#

Varies elapsed time before observation, together with one declared boundary condition — the position of the moist component in the stack, or the presence of a barrier at an interface. Measures penetration depth of the visible front from the boundary, read at stated elapsed times, against a dry control held alongside.

Cannot establish why the front advanced, whether it reached an interface at which it is not visible, or which structural effect accounts for a failure. One front can soften a layer, swell it, or lubricate an interface, and a depth reading does not separate the three.

On the model: the standard capillary relation for a porous medium (Lucas–Washburn) gives penetration proportional to the square root of elapsed time, under idealising assumptions — a rigid, uniform, non-swelling pore network and a constant contact angle. A real outer layer satisfies none of them. Read the shape of the advance, which is the transferable part; do not fit a curve to it, and do not draw the intervals between readings you did not take. Taught at Week 4: Moisture Is a Load with the full method in the Week 4 lecture, met again from inside the assembly at Week 7: Thermal Conflict, and carried into the Component Test.

Thermal comparison#

Varies nothing in the specimen. One condition is observed at three declared time points against an ambient control, and nothing is observed between them. Measures a temperature and the location it was taken from, the condition of the inner face of the outer layer, and where liquid is visible, at each time point. Cannot establish a temperature for the assembly — a gradient has no single value, only a value and a place — and cannot separate a change caused by heat from one caused by twenty minutes elapsing, unless the ambient control was run. The course treats the assembly as three lumped thermal zones, which is a teaching simplification and not a claim about how heat distributes. Taught at Week 7: Thermal Conflict, and the first artefact with times on it for the Failure Autopsy.

Cutting geometry#

This method issues no cutting instruction and holds no blades for general use. It works from pre-cut stock supplied by the laboratory, from the station’s cut-path diagrams, or with equipment operated by a demonstrator on request, and the same three routes are available at every station under the participation policies.

Varies cut angle across a fixed specimen width. Measures exposed-face length at each angle, and the span and support positions before and after the cut, marked together on one section drawing. Cannot establish anything about how a cut was made or what a blade did on its way through: the method treats the cut as geometrically clean, with no crushing and no drag, which no real cut is. That simplification is what allows geometry to be studied by itself, and it is a limitation to declare rather than to leave implied. Taught at Week 9: Cutting Geometry, and it sets which sequences were available to a specimen in the Failure Autopsy.

Transport disturbance#

Varies the route: a tilt held at a recorded angle, a compression load applied through the packaging rather than directly to a layer, and a recorded number of pick-ups, in fixed order. Measures the specimen’s condition at each station, photographed twice — in its packaging and out of it — against its own pre-route record. Cannot establish which station produced a displacement when only the end state was photographed, and establishes nothing at all about the interior from a packaged photograph alone. Taught at Week 10: Packaging Is Part of the Structure, and applied to a declared route in The Sandwich Must Hold and Week 12: The Final Stress Test.

Evidence records#

Six record types are accepted as evidence. Each has a threshold: a record missing what its row names is a description, and is marked as one.

Observed, inferred, unknown#

One packaged specimen, photographed before a route trace and again on arrival. The two lengths below are illustrative course values, not measurements.

Observed is what a second person could read off your record without your help. Inferred is what you conclude from it, and it belongs in its own column because it is arguable — a reader who disagrees needs somewhere to disagree. Unknown is the evidence you do not have, and naming it is what stops an inference being read as a measurement.

The failure this separation prevents is not a wrong answer. It is a confident one: an inference written into the observation column, which removes the reader’s ability to test it. Assessment work is marked on the separation as much as on the finding, from the Sandwich Field Log onward.

Failure notation#

One symbol set, used on every diagram, table and sheet in the course. Plain text throughout, so a record can be written by hand, typed, or copied into a message without losing its meaning.

Recording sheets#

Three templates, in field labels only. Copy one, fill it in, and submit it as part of the artefact the week asks for. A field you cannot fill is answered with ?, which is a recorded result.

Method limits#

Every method on this page supports comparative reasoning inside a declared condition. That is the whole of what it supports.

On this evidence a student may claim that one condition behaved differently from another under the declared load, in the direction the record shows; that a named layer, interface or edge is where a change first became visible; that a proposed sequence is consistent with the evidence, or that the evidence does not decide between two sequences; and that a reading is uncertain, together with what would settle it.

On this evidence a student may not claim a material property, a coefficient, a strength or a keeping time as a general value; that a result holds outside the condition declared for it; that a specimen is safe, unsafe, hygienic, or fit to eat; that a failure family is an established classification; or that the curve between two readings is data.

None of this is professional engineering analysis. The rigs are teaching instruments, the quoted figures across the site are illustrative course values, and no result here transfers to a load-bearing, structural or regulated application. Neither is any of it food-safety certification: no method on this page measures contamination, microbial growth or fitness for consumption, and nothing in the course produces a finding about them. Food safety is handled as participation policy in the participation policies, where it governs what may be run at all — a structural result never overrides a food-safety requirement.

Where each method is used#

Each week page links back to the method section it uses, and each assessment brief names the records it expects. The Sandwich Field Log runs no test method: it draws on the observation protocol and the evidence records instead, and Week 1: What Counts as a Sandwich? draws on the vocabulary alone. Week 11: Failure Autopsy reads the records of all six methods backwards, which is why it adds none of its own.

The full schedule sits on the weeks index, the weightings and due dates on the assessment overview, and the course-level framing on the course guide.