[Brian_K_White]

You can't really test. The tests you can do don't actually tell you what you really need to know.

You can't prove the material is good, you can only trust that the material is good, and 50 years later observe how it held up.

You can't find out the distribution of the alloy ingredients, or detect voids, or crystal structures, or traces of other elements, except by sawing the part in half and looking at the cut surface.

You can't find out the critical properties by looking at it. All you can do is be sure you know the full truth of the history of the material and the part. You only know that if a certain recipe is followed, then the material will be good. You have to trust that the supplier did do the recipe exactly as specified. You can't look at the part after the fact and tell that. Even stress testing to failure doesn't tell you that because the material may pass the test today but fail from fatigue over time.

The only empirical test is actual use in actual conditions for the full actual time.

You can accelerate some tests, and failing an accelerated test obviously proves the material was bad, but it doesn't go the other way. Passing an accelerated test does not prove that the material is good for actual use in actual conditions for the full normal time.

The end of the article has it right, if the parts seem ok from what testing is possible, then they are probably ok for this minute, and it's probably good enough to just replace them at the first opportunity during routine maintenance.

[schlauerfox]

Reminds me of when a favorite restaurant is bought and changes just enough to not be a favorite anymore, despite seemingly having the same menu. That feels like a similar analogy. Engineering has important details in the subtlety.

[Arnt]

I'm curious:

I assume that the documentation asserts something acceptable about the manufacturer testing (accelerated, destructive, what have you). In theory it could assert that the production process was such and such without any information about resulting quality assurance, but that seems improbable.

Why can't those tests be repeated (on samples, obviously)?

[HeyLaughingBoy]

Because it's not just about testing. Like in high-quality software, testing is only the final step. The primary determinant of quality is the source material and how it's processed, and testing can't completely prove whether or not it was processed correctly.

[Brian_K_White]

What tests? I just told you there are no tests which can actually tell you what you need to know. Even destructive tests.

Even if there was a destructive test that actually predicted lifetime performance, the total sample sizes are probably too small for statistics to be valid unless you destroy something like 10% or more. If you only have say 100 of something, a random samling of 1% is too few. 2 or 3 is no better. Maybe 10, IF all 10 gave perfectly consistent results. But there is no such test anyway.

[Arnt]

There's a bottle of water on the table next to me. The label says it's... basically official wording for "high quality" and that the source, result and production process was thoroughly tested on a named date, and with a less thorough using continuous testing regime.

I find it difficult to believe that aircraft metal production has less testing. And if not less, then surely that which the manufacturer does can be repeated on a sample by the aircraft company?

[Brian_K_White]

There are all kinds of tests you can do, and they do do all those tests of course, and they can tell you a lot, but they can not tell you what kolnowledge of the raw material source and production process tells you. Tests can prove that a part is bad, but can not prove that a part is not bad.

All the xrays and ultrasound and strain guages and spark chromatography in the world don't tell you how a part will perform and develop over time. But prior observation of a parts full lifetime and knowledge of it's production does.

Even destructive examination of random samples aren't sufficient for high stakes items when the total quantity is small.

Find it difficult to believe all you want. Or look into it and then not rely on uneducated lack of credulity to decide if something is bs or not.

When it comes to a chunk of alloy, the only way to trust the end product is to know that you created it according to a known protocol that previously has been shown to produce a certain performance result.

That protocol starts right with where the raw materials were sourced from, and every process they've been subject to along the way.

The only way to really know is if you did it all yourself.

Next best is to have documentation that you have reason to trust, ie, the supplier has a valuable reputation that they wouldn't dare risk all future jobs for the small short term gain from lying about any one job.

In this case, the supplier was a nameless supplier several subcontractor levels deep away from Boeing, and had no such reputation to worry about. The small immediate gain from a single sale was all they were ever after and they got it. Tomorrow they can do the same thing again just fine under a new random name to a new customer. And most customers won't even care because they are making bike parts and camping equipment and gimmik wallets and phone bodies not jet parts.

[Arnt]

I hear what you're saying, but I don't think you understand my question. No doubt my phrasing is bad.

The manufacturer had to produce the material in a certain way, right? Mix specific amounts of other things into the titanium, use specific heat, specific cooling. (I don't know anything about metal, really, just assuming that these things are like how high-performance concrete is made.) Now, the choice of additives, amounts, temperatures, pressure etc. is based on testing, right? Someone chose a particular pressure after doing many tests using a range of pressures. The manufacturer isn't allowed to just set up a production process that matches that spec and just assume that the result will match the results elsewhere. But the manufacturer can't take decades to check the product at the normal passage of time either. So the manufacturer has to do some sort of accelerated test to check that the production process works as intended.

That testing is naturally not perfect. I understand that. And whatever testing Airbus/Boeing can do after taking delivery is also not perfect.

My question was rather: Why can't Airbus/Boeing reach the same standard of testing as the manufacturer? If the manufacturer can do some tests and document them (or just fake the documentation) and assert that its production process matches the spec, then I don't see why Airbus/Boeing can't. I do realise that it isn't sure to match reality, the thing I don't see is why Airbus/Boeing can't get as close to testing the spec as a (proper) manufacturer can. I'd like to understand that.

If the answer is that some significant aspect is unobservable afterwards, then my next question is how that was chosen to begin with.

Does this make more sense?

[Brian_K_White]

The testing that produced the recipe is the full eventual observed lifetime performance in the past.

[kragen]

voids and crystal structures can be detected by x-rays and routinely are. you're right that the precise composition of the interior can't be, but the precise composition of the surface can be (spark spectroscopy or xrf, also both routine), and the suspicion is not that spirit made fake parts and thinly plated them with the correct metal; it's that they got fake metal. so i don't think any sawing will be needed

[Brian_K_White]

You can only detect gross structures with xrays. It cannot tell you that a material will not be more likely to fail early due to included contaminants, or lack of, or grain structure etc.

By interior composition and distribution I'm not talking about anything as comically stupid as plating like the inside is aluminum.

The surface of a finished part is routinely intentionally quite different from the interior, ie spin casting and case hardening etc. Frequently the performance of the part actually requires that the interior be different from the surface, ie hard shell resilient interior.

You can observe a lot about a finished part in various ways, like just tapping it and observing the sound can be more useful than an xray. But there's a lot you can not know after the fact through observation, except by observation of the eventual failure or not.

For one example, dissimilar materials, either within a casting or even just 2 parts in contact with each other, or a part and a brazing material, can migrate and diffuse into each other over time. Small differences in the initial conditions change how that develops over time, and can result in big changes in the performance of a part later.

You can't examine a finished part to determine that it was fabricated according to the recipe. You can only detect gross problems. You must trust that the supplier and their suppliers all followed the various recipes.

Here's another angle:

They first detected the forged paperwork because the guys on the factory floor observed that the material looked wrong.

So, it's the opposite of "you can't detect the difference". They detected a difference just plain visually.

The counterfeit parts might actually be perfectly sound. We don't know they will fail early, we only know that we can't trust the paperworks claims about how they were produced, where the materials were sourced from, how they were processed etc. Whatever the source and processes actually were, the end result might be inferior, or might be equivalent or even superior. (although detecting pitting they didn't expect does not lean towards the parts being superior)

They are able to observe that there is something different about these parts. They visually looked different enough to raise the question. Yet so far, they haven't been able to say that the parts are actually unsound through any testing or that initial visual observation.

It's not only that a part that looks perfect might not be, it's also true that even when you do detect a difference, it doesn't mean the part is bad.

You can observe a lot, but there is no amount of after-the-fact observation or testing that can replace knowledge of how a thing was produced.

[kragen]

i see, thanks! this has been very educational