How a batch of acrylic tube is actually inspected
Somewhere between your purchase order and the container, a person looked at a sample of your tube and decided the batch was acceptable. This is what that decision is made of — the sampling standard behind it, how many pieces it really covers, and the parts of it no inspection can reach.
Ask an extruder how they inspect and you will often get the phrase "100 % inspection". For extruded tube, in almost every case, that is not true — and it is worth knowing why before you take it as reassurance.
Extrusion does not produce units. It produces a continuous section that runs off the line for hours and gets cut into lengths at the end. There is no natural point at which one piece stops and the next begins, and nothing to count until after the saw. So a factory running tube has to decide what a batch is, decide when to look at it, and decide how many pieces out of it to measure. Those three decisions are the whole of quality control on an extrusion line. A supplier who answers with a slogan has not told you any of them.
Here are ours, in the order they happen to your goods.
Decision one: a batch is two hours long
On our lines, each two-hour window of output is closed out as its own batch, sampled and recorded on its own. Not each shift, not each order, not each day's production.
That single choice does more for the goods you receive than any piece of measuring equipment in the building, for a reason that has nothing to do with effort and everything to do with arithmetic. Every sampling plan works by inspecting a small number of pieces and inferring the condition of the rest. The inference is only as good as the assumption that the lot is reasonably uniform. And extrusion drifts: a cooling water temperature creeps up, a haul-off belt wears, a new box of resin behaves slightly differently, the afternoon gets hot.
Drift is gradual, and gradual is exactly what a large lot hides. Sample a whole shift's output as one lot and a drift that began after lunch is a minority of the pieces — statistically outvoted by the good ones from the morning. Sample the afternoon in two-hour slices and the slice where it started fails on its own.
This is the first question worth asking any extruder, and it is a much better question than asking about AQL numbers: what do you call a batch? If the answer is a shift, a day, or an order, their sampling plan is weaker than its numbers suggest, no matter what the numbers say.
Decision two: when someone looks
Inspection on our lines is not one event at the end. It is eight, and the first four all happen before volume production is allowed to continue.
| When | What happens |
|---|---|
| After every setup | First-article inspection. Nothing runs in volume until the first piece has been measured and passed. |
| At every shift handover | First-article inspection again. The incoming shift does not inherit the outgoing shift's result — it establishes its own. |
| After any adjustment | First-article inspection again, every time the process is touched or the output starts looking unstable. |
| Every 2 hours | Self-check by the production crew against the running standard, recorded. |
| Every 2 hours | Patrol inspection by the line leader — the batch above, sampled to plan. |
| Before stock | Final inspection by QC to the sampling plan, including the weight of at least one piece per batch. Bundled product is unwrapped, not judged through the film. |
| After the annealing oven | Clear acrylic is re-checked for out-of-roundness, residual stress and oil mist. It does not enter stock until that check passes. |
| Before shipment | Outgoing inspection after packing, against your order. |
Two of those rows carry more weight than the others.
First-article inspection at shift handover sounds like bureaucracy and is not. A line that was running correctly at 3 p.m. is not evidence that it is running correctly at 11 p.m., and the handover is the moment when responsibility for that assumption changes hands. Making the incoming shift measure a piece for itself is how a drift that started at the end of one shift gets caught at the beginning of the next rather than at the end of the week.
The re-check after the oven exists because annealing is a process that can go wrong on its own. Twenty hours at temperature can leave a tube out of round, or with oil mist on the surface, or having moved less than it should have. A tube that went into the oven good and came out wrong must not quietly rejoin the batch it left with — so it is inspected again as if it were new. If you have not read what annealing is for, that article covers the oven itself; this one only covers the fact that we check its work.
Decision three: how many pieces — and what AQL really promises
We sample to ISO 2859-1, normal single sampling — the international standard, adopted in China as GB/T 2828.1, so the plan reads the same in your incoming inspection department as it does in ours. What that gives us is a sampling scheme, not a single number: a sample size and an accept/reject rule derived from a general inspection level and a quality limit, chosen to suit what the product is and what happens if it is wrong. Here is how those choices are made on our lines:
| Extruded product, entry to stock | A reduced special inspection level — fewer pieces measured per batch, because the failure we are looking for is a trend across the batch rather than a single bad piece |
|---|---|
| Marking and labelling | The tightest quality limit we apply to anything we ship |
| Machined and fabricated product | A general inspection level, and a larger sample than the extruded product, because each piece has had more operations done to it |
| Any batch re-submitted after a failure | A tighter quality limit than the original inspection — and the inspection level steps up one grade for every re-submission, until the batch is inspected 100 % |
The specific levels, quality limits and sample sizes we apply to a given section are set out in our inspection plan and confirmed in writing against your order — ask at enquiry stage and we will put the numbers for your part next to it. The reasoning below matters more than the figures, because it holds whatever numbers a supplier is working to.
What an AQL does not mean
The most common misreading in purchasing is that a quality limit means up to that share of your tube is allowed to be defective. It does not, and the real meaning matters in both directions.
An AQL is a property of the process, not a licence for the shipment. It is the quality level at which a plan is designed to accept lots with high probability. Take a hypothetical process running at exactly that defective rate: the plan will let most of its output through. Run the same plan against a process that is clearly better and it passes comfortably; against one that is clearly worse and it starts rejecting. So it is a filter tuned to a threshold, not a quota someone is allowed to fill. Nobody in our building is permitted to send you a batch they know contains bad pieces because "it is within AQL".
But the honest half of that sentence is the other half. A sampling plan is a filter, not a guarantee. A special inspection level deliberately uses a small sample — tens of pieces from a lot of thousands, not hundreds. It is built to catch the thing that is systematically wrong: a drifting diameter, a wall going thin, a surface deteriorating, a length cut short. It is not built to find one stray piece in a thousand, and no plan at any AQL will reliably do that. If a single non-conforming piece in your delivery would stop your line, sampling is not the mechanism that protects you — the conversation to have is about the process, the packing, and what we do when it happens, not about which AQL number to write on the order.
Why the label is inspected harder than the tube
Look again at that table: marking and labelling carries the tightest limit we apply to anything we ship, far tighter than the tube itself. That is not a typo and it is not a priority error.
A marginal tube costs you one tube. A mislabelled bundle costs you the delivery. If a bundle of Ø30 × 3 mm is labelled Ø30 × 2 mm, nothing in the bundle is defective and every piece of it is in the wrong place — in your rack, in your kit, in your customer's assembly, and the error is found by the person who needs it least. Identification errors propagate; dimensional ones stay where they are. So they are inspected at nearly thirty times the tightness.
The row worth reading twice
The last line of that table is the one we would check if we were buying rather than selling: when a batch fails here, the next inspection of it is stricter than the first one. It drops to a tighter quality limit, and the inspection level steps up a grade for every re-submission, until the batch is being inspected 100 %.
The reason this matters is that the default human behaviour is the opposite. A rejected batch gets sorted, and the sorted batch gets re-checked at the same level as before — which is the one circumstance where the original plan is no longer valid, because the lot is no longer the lot the plan was chosen for and everyone involved now wants it to pass. A supplier who re-inspects a failed batch at the original level is not really re-inspecting it. Ask what happens on the second submission; the answer tells you more than the AQL on the first.
What it gets measured with
- Profile projector
- Vernier callipers — one per operator, not shared from a tool crib
- Micrometers
- Steel tape for length
- Colorimeter for colour matching
- Integrating sphere for light transmittance and haze
- Accelerated weathering tester
- Melt flow indexer for incoming resin
The callipers line is a real answer to a real problem: when measuring tools live in a crib and have to be fetched, measurements get skipped, and the skipped ones are always the inconvenient ones. One per operator removes the excuse.
Two of the instruments change what we can tell you rather than what we can catch. The integrating sphere means a transmittance and haze figure for your material can be measured here instead of copied off the resin maker's datasheet. The melt flow indexer means incoming resin is checked before it reaches a die — a batch of material that is off-spec is found at the silo, not diagnosed afterwards from a pile of scrap.
What no inspection catches
This is the part of the subject that gets left out of quality pages, and it is the part that decides whether your parts survive.
Every check described above happens on finished tube, in a factory, on the day it was made. Three of the ways acrylic tube fails are invisible to all of them.
1. Residual stress
It does not show on a dimensional report, in a photograph, or in any incoming inspection you can run on arrival. It shows up weeks later as silvering along a solvent bond, as a split running out from a drilled hole, as crazing after cleaning. No sampling plan at any AQL will find it, because there is nothing to find until the part is worked on. The only answer is process — the anneal, and the check after it — and annealing lowers the stress, it does not remove it.
2. How it behaves against your medium
Whether a tube survives contact with a fuel, a cleaner or an adhesive is not a property the tube can be inspected for. It depends on the medium, the temperature, the pressure and the stress state of the part — a combination that only exists on your bench. We test what we can and publish the results, including the cases where we tell you to use something else. What we cannot do is inspect a batch and pronounce it fit for a medium nobody has told us about.
3. What it will look like in five years
Outdoor life is a property of formulation, not of the batch in front of the inspector. It is decided by the resin grade and by whether UV and anti-ageing additives went in during production — and it is evidenced by accelerated weathering, which is why there is a weathering tester on the equipment list rather than a line in the inspection plan.
The general point is worth stating plainly, because it cuts against the way quality is usually sold: measuring catches the tube that is the wrong size. It does not catch the tube that is the right size and fails six months later. Inspection is necessary and it is not sufficient, and a supplier whose entire quality story is an inspection plan has told you about the easier half of the problem.
The rule behind the rules
Our written standard contains one sentence that matters more than the tables it sits in: a part which meets the standard but still has a defect clearly affecting its appearance or its use is to be judged non-conforming.
That is in there deliberately. No specification anticipates everything, and an inspector who is only permitted to read the table will pass parts that everyone in the room can see are wrong. Giving the inspector that authority costs us some batches. It is cheaper than the alternative.
Three more rules govern the standards themselves:
- A product's own agreed standard wins. Where you and we have agreed a standard for your part, that governs; our general standard applies only where the agreed one is silent.
- Standards are reviewed every two years and tightened as the process improves — the limits move in one direction.
- No single department can change one. A revision needs Technical, Production and QC to agree, and the General Manager to approve it. Nobody can quietly relax a limit to make a difficult month look better.
One thing we do not publish is the grading table itself — the defect classifications and dimensional bands our inspectors judge against. That is the document our competitors would find most useful and our customers least, because the numbers that matter to you are the ones for your part, and those get agreed with you before production rather than read off a general table. Ask at enquiry stage and we will put the criteria for your section in writing as part of the order.
When a batch fails
The first thing that happens is physical: the batch is separated and tagged. Our shop floor runs on status tags, so a batch is always visibly awaiting inspection, passed, isolated or on concession. Nothing sits untagged, because an untagged pallet is how a rejected batch ends up in a container.
From there it is sorted, reworked, scrapped, or released on concession — and the concession is the one worth understanding, because it is the route by which out-of-specification goods legally reach a customer anywhere in this industry. In our system it is not an inspector's call and not a production manager's call. It needs the General Manager or sales to sign, against a confirmed understanding of what you actually need the part to do. Whatever route the batch takes, it is re-inspected before it moves, at the tightened level described above, with the re-submission counted on the record.
Records are kept in three copies — quality keeps the original, the warehouse releases against its copy, production keeps the third. The point of three copies is not paperwork. It is that no one department can close out a problem on its own.
Why the system looks like this
None of the above was designed in the abstract. It is what we were left with after getting a shipment wrong, years ago, for a customer in Canada.
We had their surface requirement in writing, and we met the requirement we had been given. What we had not done was ask what the parts were actually going to be used for — so the standard we set ourselves and the standard the application needed were not the same standard, and the first batch was not fit for the job. We paid for it, and then we rebuilt the way an order runs from end to end:
- Order accepted
- Technical review
- Small-batch trial
- Written standard agreed
- First-article inspection
- In-process inspection
- Final inspection
- Sampling inspection
- Post-packing inspection before shipment
The important part of that list is where step four sits. The standard gets written and agreed before production, not inferred from the enquiry — and because we now know what the failure mode of an inferred standard looks like, we will usually come back to you at technical review with our reading of your specification and a suggestion or two. That can feel like friction on a simple order. It is the cheapest hour in the whole project.
Five questions for any extruder
You will get more out of these than out of asking for a certificate. They work on us too.
- What do you call a batch? A shift or a day means their sampling has poor resolution whatever AQL they quote. Ask what closes a batch out.
- What happens on the second submission of a failed batch? If it is re-checked at the same level, the re-check is decoration.
- Who can authorise shipping something out of specification, and against what? If the answer is quality or production alone, commercial pressure has a short path to your container.
- What do you check that is not a dimension? Dimensions are the easy half. Ask what catches the failures that appear after the goods arrive.
- What in your quality documentation is yours, and what belongs to your resin supplier? This one separates the honest answers from the rest. Ours: we hold ISO 9001, continuously since October 2008. RoHS and REACH declarations come from the resin manufacturers whose grades we run, and we pass the current declaration for your grade through with your order — we do not hold them in our own name, and we would rather tell you now than have you find out during an audit.
What we will and will not claim
We will not tell you we inspect every piece, because on an extrusion line that claim is usually false and always unverifiable. What we will tell you is what a batch is, when it gets looked at, which standard the sample size comes from, what happens when it fails, and which failure modes inspection cannot reach at all.
If your incoming plan is stricter than ours, send it before you order rather than after the goods arrive. Tell us what the part has to survive and which documents your auditor will ask for, and we will tell you in advance whether we can meet it — including when the answer is no. The whole inspection plan is set out on our quality control page, and the stock sizes it applies to are published: 211 round tube sizes from Ø4 to Ø400 mm, plus square, rectangular and hexagonal sections.
Send us your inspection standard
Tell us the section, the quantity and the standard your incoming inspection actually applies — or your customer's. We will tell you before you order whether we can meet it, and what we would want written into the order so that both sides are judging the goods the same way.