The same casting, two names

Chapter One

A defect is named before it is understood

On the floor a casting is picked up, turned over, and judged in a few seconds. A name is written down, and that name is the only thing that survives.

Everything downstream is built on it. The morning meeting works from it, the corrective action is chosen because of it, and the monthly rejection report is only a count of those names added up. If the name is unreliable, so is everything resting on it.

And the name is often unreliable, for reasons that have nothing to do with carelessness. A blowhole and a pinhole are both gas defects, and on a rough as-cast surface they can look much the same. Sand inclusion and slag inclusion both show as dark, dirty pockets. Shrinkage porosity and gas porosity can be told apart with a lens and some experience, and told apart differently by two people who both have some experience.

The trouble is that these pairs do not share a cause. Blowholes point toward mould and core gas, venting and moisture. Pinholes point toward the melt, toward dissolved gas and a damp ladle. Sand inclusion sends you to the mould's strength and how it was closed; slag inclusion sends you to skimming and to the gating. Each pair carries the same appearance and two different repairs.

There is a second cost, quieter than the first. Where two people disagree, the disagreement is settled by argument rather than by evidence, usually in a meeting, usually by whoever is most senior in the room. Time is spent, and the register slowly becomes a document nobody fully trusts. A plant can carry on like this for years. Most do.

Chapter Two

What the log sheet quietly loses

Between the casting on the floor and the number in the monthly report, the record passes through several pairs of hands. Detail is shed at every handover.

The first loss happens at the moment of inspection. A defect noticed at ten in the morning is often written up near the end of the shift, from memory. Counts drift. The heat gets attached to the wrong casting, or the pattern number is remembered as the one before it. Nothing dishonest has happened; the entry is simply made hours after the thing it describes.

The sheet itself then has to survive a foundry. Paper in that environment picks up oil and sand, gets rained on near the door, gets folded into a pocket and comes back illegible. Sheets go missing between the floor and the office often enough that most plants keep a habit of reconstructing them.

There is also the part nobody puts in writing: a reject recorded is a reject somebody may be answerable for. Where the register is used to apportion blame, the softer number tends to be the one that gets written down.

By the time the figure reaches management it is a total with the reasoning stripped out. It can tell you that rejections were up in August. It cannot tell you which pattern, which impression, which heat or which shift, and it certainly cannot show you the casting.

Chapter Three

A photograph settles what an argument cannot

The cheapest change a foundry can make is to capture the defect where it is found, with a photograph, before the casting moves.

A photograph does something a written name cannot: it keeps the evidence. Six weeks later, when the same defect is being discussed, the casting itself is long gone, but the image is still there and can be looked at again by anyone. Disagreements stop being a contest of recollection.

The second change is a shared list. When every inspector picks the defect from the same short list, matched to what the plant actually casts, the names stop drifting between shifts. It does not make everyone an expert overnight, but it narrows the room for two people to describe one thing in two ways.

The third is a settled way of deciding. An inspector records what they see. A supervisor, or a metallurgist who may be sitting in another plant, can open the same photograph and change the classification if it is wrong. The senior call becomes the record, and the correction is kept alongside the original rather than replacing it silently.

Keeping both readings has a use beyond tidiness. Over a few months the pairs of "what was said" and "what it turned out to be" become the plant's own training material, drawn from its own castings rather than from a textbook. A new inspector learns the difference between a blowhole and a pinhole by looking at fifty of the plant's own examples with the agreed answer attached.

Defect knowledge should not leave the plant when a senior inspector retires.

Chapter Four

Down to the impression that made it

One casting comes from one cavity. Recorded properly, that single number opens up a question the register could never answer before.

A pattern with eight impressions produces eight castings from every mould. They are poured from the same ladle, in the same sand, on the same line, within seconds of each other. When one of them rejects far more often than the other seven, the difference is unlikely to be the metal or the sand. It is far more likely to be that impression: worn, poorly vented, or fed differently from its neighbours.

That question is only askable if the cavity number is captured with the defect. It takes one tap on the floor. Cavity numbers are cast into most patterns already, so the inspector is reading a number that is sitting in front of them.

With the heat number recorded as well, the same register answers a different family of questions. Was this a bad heat, or a bad impression, or a bad shift? Those three have entirely separate remedies, and until the data can separate them, the plant is guessing which one to spend its morning on.

Chapter Five

Sand is often the reason, and the sand lab already knows

A good share of what a green sand foundry rejects traces back to the sand system. Most plants measure the sand carefully and then keep those readings somewhere the defect register never sees.

Moisture, green compression strength, compactability, permeability and AFS grain fineness are checked through the day, written into the sand lab register, and used to steer the muller. Meanwhile the defect register is kept separately, usually by different people, on different paper. Both describe the same shift, and they are rarely read side by side.

The defects that respond to sand are well known to anyone who has run a green sand line. Blowholes and pinholes follow moisture and gas. Sand inclusions, scabs and swells follow mould strength and friability. Metal penetration follows ramming and pouring temperature. When those defects rise, the sand numbers for the same hours are the obvious first place to look.

Placing the two records on one timeline is not a new idea; it is simply awkward to do by hand, so it seldom gets done. Where a plant already runs Versatile sand testing equipment, those readings can be logged automatically, and the defect register can then be read against them rather than beside them. That pairing is where a cause is most likely to become visible.

Chapter Six

From a total to a decision

A register earns its keep at the point where somebody has to choose what to work on this week, and then find out whether it worked.

Ranking the defects is the first step, and it is usually uncomfortable reading. A small number of defect types tend to account for most of what a plant scraps. That concentration is useful: it says where the effort belongs, and it says plainly which problems can wait.

Then find out whether the fix held

The step most often skipped is the last one. An action is agreed in a meeting, someone changes the venting or the moisture target, and the matter is considered closed. Whether the defect actually fell afterwards is rarely checked, because checking it would mean comparing two months of a register that nobody quite believes.

With dated, photographed, consistently named entries, that check becomes a straightforward look at the weeks either side of the change. The loop closes, and the plant starts to accumulate a short list of actions that are known to have worked on its own castings.

The audit takes care of itself

Plants working to IATF 16949 or ISO 9001 already have to show traceable evidence of nonconformance and the action taken. Where the register is photographs, timestamps, the person who classified, the person who changed it and the disposition, most of that evidence is a by-product of ordinary work rather than something assembled the week before the audit.

Chapter Seven

What it actually asks of the floor

None of this is worth much if it slows the inspector down. The test of a defect register is whether it survives a busy shift.

In practice the whole capture is a photograph of the cast-in part number, a tap for the cavity, a photograph of the defect, and a tap on the defect name. It is done standing at the bench with the casting in hand, on a phone, and it takes about as long as writing a line on the sheet. Where there is no network in that corner of the plant, the entries are held on the phone and go up when the signal returns.

A foundry does not adopt a defect register because it wants software. It adopts one because the same argument about the same casting has been had too many times, and because the monthly number has stopped telling anyone anything useful. The change being asked for is small, and it happens at the bench: name it once, in front of the casting, the same way every shift.