Foundry · Inspection & QC

You know your reject rate. You don’t know your repeat offenders.

Inspection is where a foundry finds out what the previous six stages did. A pass/fail count tells you how much you lost; it does not tell you that four of this month’s rejects are the same defect on the same pattern that appeared last month. Repeat defects are the cheapest thing in a foundry to fix and the hardest to see.

Your ERP records the result: passed, rejected, quantity.

It won’t tell you this is the fourth month running for the same defect on the same pattern.

Why the gap exists

Where the losses hide

The rejects that are the same reject.

A reject rate is a total. A defect family is a cause - and causes are what you can actually act on.

Defect families that repeat

Blow holes, shrinkage, inclusions and dimensional faults each trace to a different stage. Counted together as "rejects", they average into a number that points nowhere.

No route back to the heat

A casting rejected at final inspection came from a specific heat, a specific mould and a specific pour. Without that trace, the finding stops at the inspection bench.

Customer returns that never reach the cause

A return is recorded as a return. Connecting it back to the conditions that produced the part is the difference between a credit note and a fix.

The join

What the record holds, and what Munshi captures beside it.

Same run, same clock, same asset — three streams that only mean something together.

Your ERP records

Inspection outcome, quantity passed and rejected, disposition.

Munshi captures

The defect tagged by family at the bench, against the heat, pattern and batch it came from - and NDT or spectro results tied to the same record.

What that makes possible

Rejects grouped by defect family and traced back to the stage that produced them, instead of a monthly percentage.

What Munshi surfaces

Which defect, how often, and from where.

Teal is normal, maroon is flagged. Illustrative data.

Rejects by defect family

Tagged at the bench by family, not lumped into a single reject count.

One family, month over month

A number that is not improving is a cause that has not been found.

Normal / in-spec Flagged by Munshi -- threshold / average

When an outcome moves and a signal moves with it, Munshi shows both and says they moved together. It does not tell you one caused the other. The operator confirms the cause, and the confirmed cause is written back to that asset’s register — so the pattern is already named the next time it appears.

Related stages

The stages either side of this one.

Heat treatment

Energy per batch, cycle deviation and re-treat loops.

Other assets

Cranes, pumps, compressors, cooling and extraction.

Get started

See this on your own line.

Bring a month you already understand. The useful test is whether Munshi finds what you already know — and then what you didn’t.