Mill test certificate verification in fewer hours

A lorry arrives with bar stock, plates and a bundle of pipe. The stores in-charge counts the pieces and books the receipt. Then the paperwork goes to QC: a folder of mill test certificates, some printed by the mill, some re-issued by the stockist, one a photograph forwarded on WhatsApp. Nothing can be issued to the shop floor until someone has checked them.

If you make valves, fittings, flanges, castings, forgings or pressure equipment for oil and gas and process plants, mill test certificate verification is part of every week. Search for how to do it and you will find good checklists for reading one certificate, mostly written by steel mills and traders, and software pages written by vendors. What you will not find is the part that costs you time: checking hundreds of certificates, in dozens of formats, against orders that each ask for something slightly different.

This article is about that part.

What does mill test certificate verification involve?

The individual checks are well known. For each certificate, someone confirms that:

• the certificate type is the one the order asked for
• the material standard and grade match the purchase order exactly
• the heat number on the certificate matches the heat number marked on the material
• the chemical composition falls within the limits of the material standard and of the order, which is sometimes tighter
• the mechanical results (tensile, yield, elongation, hardness, impact where required) meet the minimums
• heat treatment, size and quantity agree with what was ordered and what arrived
• any supplementary requirement in the order appears on the certificate or on an attached report: impact testing at a stated temperature, sour-service hardness limits, positive material identification, extra testing

The certificate types themselves come from EN 10204, the European standard that defines the types of inspection document a purchaser can ask for with metallic products. The two you see most often in this work are 3.1, where the manufacturer reports test results for the material supplied, or the heat or lot it belongs to, and the document is validated by an inspection representative independent of manufacturing, and 3.2, where the purchaser's representative or an inspector named by regulation also validates it. If the order says 3.2, a 3.1 is not good enough, however clean it looks.

None of this is difficult for an experienced QC engineer. The difficulty is the volume, and the fact that the information needed to judge each certificate lives in two or three other documents.

Where do the hours actually go?

Ask how long it takes to check a certificate and most people will say "five minutes". Watch it being done and the reading is the shortest part.

Finding what the order requires. The grade is on the PO. The tighter carbon limit is in the client's material specification, attached three emails back. The impact test temperature is in the datasheet. Before checking a certificate, someone has to assemble the acceptance criteria for this order, and on the next order, assemble them again.

Certificates that cover more than you received. A mill certificate often lists several heats or a whole lot. Someone has to find the lines that apply to this delivery and ignore the rest.

Stockist and trader certificates. Material bought from a stockist may arrive with the mill's certificate, the stockist's own document, or both. Working out which one is traceable to the bar in your store takes time and sometimes a phone call.

Physical matching. The heat number has to be found on the material itself, which may be faint, painted over, or on the end of a bar now at the bottom of a stack.

Chasing. A certificate is missing, illegible, or for the wrong heat. Each gap means an email to the supplier, a wait, and material held in the stores.

Recording. Once checked, the result has to go somewhere useful: an incoming inspection register, a heat-number log, a file that will later feed the manufacturing record book. If this is skipped, the same certificate gets checked again months later, from memory.

Notice that almost none of this is typing. It is finding, matching and comparing, and it is done by the one or two people who know the orders well enough.

How do you measure the hours in your own receiving?

Before changing anything, find out what verification costs you now. Your goods receipt register already holds most of the answer.

Take the last month and note:

1. Receipts that needed certificates, and the number of certificates received.
2. Heats per certificate, roughly. A one-heat certificate and a twelve-heat certificate are different jobs.
3. Certificates that failed or were queried, and why: missing, wrong type, wrong grade, value out of limits, heat number not found.
4. Days material waited in stores between receipt and release to production.
5. Who did the checking, and what they were pulled away from.

Then time one week as it happens: a simple log of start and stop times for each receipt, split into finding the requirements, checking, chasing and recording. A log kept at the time is worth far more than an estimate made afterwards.

The arithmetic is simple enough to check yourself. For example, if you receive 300 certificates a month and each takes 8 minutes to check against its order, that is 40 hours. If one in ten needs a query that takes 20 minutes of emails and follow-up, add another 10 hours. That is more than a working week of a senior QC person every month, before anyone has recorded a single heat number. Your numbers may be much smaller or much larger. The point is to know them.

What can you fix without software?

Several of the slowest hours come from how certificates are asked for and received, not from the checking itself.

• Put the certificate requirement in the purchase order. State the certificate type, the material standard and grade, any supplementary tests, and that material will not be accepted without a legible certificate traceable by heat number.
• Make the certificate a condition of receipt. Material that arrives without one waits at the gate, not in the stores. The supplier then chases it, not you.
• Write an acceptance sheet per grade and per client. One page with the limits that apply, drawn from the standard and the client's specification. Built once, used for every receipt.
• Ask suppliers for digital certificates. A text-based PDF from the mill is far easier to check, and to search later, than a scan of a photocopy.
• Keep one heat-number register. Record each heat when it is received and checked, and record where it was issued. This is the same register that later saves days on the record book.

If you receive a few dozen certificates a month from a handful of regular mills, these steps may be all you need.

When is AI the wrong tool for checking MTCs?

AI is not always the answer, whatever a vendor says. Be sceptical if:

• Volume is low. A few certificates a week rarely justify more than an acceptance sheet and a register.
• The certificate is missing. No system can check a document a supplier never sent. That is a purchasing problem.
• Your mills already send structured data. If certificates arrive as consistent digital files, a simple comparison against a limits table may be enough. No reading is needed.
• The decision is a judgement. Whether to accept a value slightly outside the limit, ask for a retest, or raise a concession with the client is a decision for your QC head and the client. A system can show the deviation; it should not decide.

What does a working system for MTC verification look like?

Where the volume is real and the hours are in reading and comparing, a well-built system does the reading and the comparing, not the judging. In practice it should:

• Read each certificate, whatever its layout, including scans and phone photographs, and pull out the heat numbers, grade, certificate type, chemistry, mechanical values and test results as named fields.
• Keep the source. Every extracted value points back to the place on the page it came from, so the QC engineer can confirm it in seconds.
• Build the acceptance criteria from the order. Combine the limits from the material standard, the purchase order and the client's specification for that order, and show where each limit came from.
• Compare and show the result per value: pass, fail, or missing, with the reason.
• Flag doubts to a person. An unreadable figure, a heat number that matches nothing on the receipt, a 3.1 where the order asks for 3.2: these go to QC, not into the register.
• Write the result into the register you already keep, so the heat number is traceable when the record book is put together.

The QC engineer still signs the release. The difference is that they review a checked draft instead of building each check from scratch. Our document intelligence page shows an example of what we build for documents like these.

What should you ask before buying anything?

Whether it is software, a service or a custom build, ask:

1. How many hours does our verification take today, measured, and how many will this remove?
2. Does it read the certificates we actually receive, including stockist documents, scans and photographs?
3. Where does it get the acceptance limits for each order, and can we see them?
4. Can every value it reports be traced back to the original page?
5. What happens when it is unsure? Does it guess, or does it flag?
6. Does it write into our existing register or system, or start another one?
7. If our volume is low, will you tell us this is not worth doing?

If the answer to the last question is anything but yes, keep looking.

Where to start

Take one month of your goods receipt register and count three things: certificates received, certificates queried, and days material waited for release. That page will tell you whether you need a better purchase order clause, a register, or something that reads and checks certificates for you.

If you would like a second opinion on it, book a 30-minute call about this one process at prakreon.com/contact. We will ask who checks your certificates, how many arrive and how long each takes, and we will tell you plainly if the answer is a better register rather than new software.