The concrete cube test, and the four weeks you wait

The cube test is how a claim about concrete is turned into evidence.

Concrete arrives, somebody fills a set of steel moulds from it, those cubes are cured in water, and at a stated age they are crushed in a machine that records the load at which they fail. Divide that load by the area of the face and you have a compressive strength. Compare it with what was specified and you have an answer.

The procedure is simple enough to describe in a paragraph. Almost everything that goes wrong with it happens in the ten minutes at the beginning, when nobody is watching.

What the test is measuring#

A cube is not the structure. It is a sample of the same material, treated as well as anybody knows how to treat it, so that its result represents the best the concrete could do.

That is a deliberate choice and it has a consequence people find counter-intuitive: a cube result tells you about the concrete as supplied, not about the concrete as placed. Poor compaction, poor curing and bad weather in the structure are real problems that a perfect cube result will not reveal. The cube answers one question — was the material we were given capable of the specified strength — and it answers it well. It answers nothing else.

The first ten minutes#

The sample must be representative. Taken from the discharge of the load, not from the first splash and not from the scrapings at the end. A sample taken from the wrong part of a load is a test of something that was never in the structure.

The moulds must be right. Clean, correctly sized, properly oiled, and assembled so the faces are square. A distorted mould gives a cube with a face that is not flat, which changes how the load is applied and therefore what number comes out.

Compaction has to be done properly and consistently. Air voids left in a cube reduce its strength substantially. A cube that was rodded half-heartedly because the person was in a hurry will fail, and the failure will be blamed on the supplier.

They must be marked. This is the one that ruins whole months of testing. A cube with no identification, or with an identification that has washed off in the curing tank, is a result belonging to nothing. It cannot be attached to a pour, a location, a grade or a supplier. It is a number with no owner, and the only honest thing to do with it is discard it — which means the pour it came from has no test at all.

Curing is where the argument starts#

Cubes are cured in water at a controlled temperature until they are tested. That is the standard condition, and it exists so that results from different sites are comparable.

It also means the cubes are treated better than the structure almost always is. Somebody will point this out, usually when a result is disappointing, usually in the form "the cubes were not cured the same way as the slab". They are correct about the fact and wrong about what it implies. The cube is a control sample. Its job is to hold everything constant except the concrete. If you cure cubes the way the slab was cured, you have measured the curing, not the concrete, and you have lost the only comparable number you had.

If the concern is the structure rather than the supply, that is a different test — cores taken from the actual member, or non-destructive methods — and those come with their own corrections and their own arguments.

Seven days and twenty-eight days#

Twenty-eight days is the age at which concrete is judged. Seven-day cubes are taken as an early warning.

The temptation is to convert one into the other with a fixed ratio. Resist it. The relationship depends on the cement, the mix, the admixtures and the temperature, and the range across ordinary construction is wide enough that a fixed factor will cause both false alarms and false comfort.

What a seven-day result is genuinely good for: telling you that something is badly wrong, four weeks earlier than you would otherwise know. A seven-day result far below what that mix has historically produced is a reason to stop pouring from that source and find out why. It is not, by itself, a rejection.

What a failing result actually means#

Not much, on its own. Compliance is assessed on groups of results against criteria set out in the code, not on individual cubes, precisely because the material is variable and the test has its own scatter.

The sequence when results are unsatisfactory is investigative rather than punitive: check the test itself, check the records for that batch, then if necessary move to tests on the structure. Skipping straight from one low cube to demolition, or from one low cube to a shrug, are the two commonest errors and they are made by different people for the same reason — neither has understood that the specification is a statement about a distribution.

The grade specification itself, and why it is a statistical claim rather than a label, is explained in what M25 actually means.

The record is the deliverable#

A cube test produces a number. The value of that number is entirely a function of what it can be attached to.

Each result needs, at minimum: the date the cubes were cast, the date tested, the identification mark, the grade, the pour and its location in the structure, the supplier or batching source, the challan or batch reference, the slump recorded on arrival, and the load at failure.

Without the pour location, you cannot answer the only question that ever gets asked later: is the concrete in this particular column all right? Without the challan reference you cannot connect the result to a load, and therefore cannot connect it to a supplier or to a payment. That connection is what makes cube results into something more than a filing exercise, and it is the same connection that makes a delivery challan worth capturing properly in the first place.

When we imported eight months of a concrete register into our own system, the fields that turned out to carry the value were exactly these — the receiving officer, the challan number, the rate, the slump, and the seven- and twenty-eight-day results where they had been taken. The rows without them were not wrong. They were just no longer evidence of anything. They are also among the few unrepeatable items in what belongs in a handover file, because nobody can cast a cube for a pour made three years ago.

The short version#

The crushing machine is the least interesting part of the cube test.

Sample properly, compact properly, mark the cube so the mark survives a month under water, cure it to the standard condition even when somebody objects, and record the result against the pour it came from.

Do those five things and the number means something. Miss any one of them and you have spent four weeks producing a figure you cannot use.