Tube expanding calculator

[clarnibass]

Hi

I just tried Chat GPT for this and I'm not sure if it's right...

Is there a decent and simple calculator for calculating change in wall thickness when expanding a tube? I don't think the expanding method makes a difference? If it does, then I mostly use a rolling method (same as how jewelers flatten rods) but sometimes an inner expander that pushes it larger from the inside (more risky).

I'd like a general calculator/formula so I can always able to use it, but in this specific case I'm trying to calculate it for a 27.5mm OD tube, 0.7mm wall thickness, to see if it's possible to expand it to 28.5mm OD and what the wall thickness would be if it is. GPT claims my wall thickness would be -0.3mm and since it's impossible to have a negative thickness, it's not possible to expand the tube to 28.5mm. Is this correct...?

Thanks

 

[Garwood]

It's just basic displacement.

Figure the crossectional area of the stock tube. Figure the new id based on the desired od with the same crossectional area.

 

[Corn]

Yes, the material will not disappear. So in your case if you expand it to 28.5mm the wall thickness would be ~0.68mm. (Rounded from: 0.67657). So only negligible differences.

 

[Clive603]

In a practical world it is perfectly adequate for that sort of small change to assume everything changes pro rata.

So the ratio between old and new wall thickness is can be taken as the same as the ratio of diameters.

In any seriously engineered application the bigger worry will be what the stretch is doing to material properties. The dimension change is getting towards 5% which is about where I'd definitely be running some numbers to make sure nothing hinky is going on due to the induced stresses when the material stretches.

I found it quite easy to burst thin wall tube by careless expansion! That particular bit of tube wasn't annealed like the stock list label said.

Clive

 

[clarnibass]

Thank you. Calculating the area works great for this.

I tried it "manually" and now I'd like to find a good online calculator to make much faster. No need for a full calculator of "wall thickness".
I found a good calculator for annular ring area (and also found it is called an annular ring in English). What I'm looking for now is one for calculating e.g. the ID from the OD and area (the ones I found for area don't have this in reverse).

Corn, did you just manually calculate it with a calculator? Or else maybe you know of a good online one for this specific calculation?

As far as the part/metal properties and any issues from expanding, that doesn't matter. For now I only need to know the theoretical dimensions, excluding any of those potential issues.

 

[WizardOfBoz]

So for those interested in AI, I would recommend "The Book of Why" by Judea Pearl. He points out some interesting facts, all of which I've been trying to figure out and explain for 20 years. Roughly

Step 1: If you take a dataset (say "yearly cheeseburger intake in a sample of people in Nebraska, and cardiovascular disease") and analyze it, you can say something about the actual population you tested. That is "The extent to which cheeseburger intake in a specific set of Nebraskans affects cardiovascular events is 'x'". Its a correlation, not a causal relationship. From a stats standpoint, you can't say ANYTHING about a causal relationship. And you can't extrapoloate to, say, Kansans.

Step 2: If you do a randomized controlled trial in which you divvy up a population Nebraskans into two groups with approximately equal characteristics (sex, age, health, weight, etc) and apply (for example) a drug treatment, you may be able to say that any observed difference is causal (an effect is due to the drug) but you still can't extrapolote to other populations. And remember, it's unethical to to a trial where you assign a control arm to do something unhealthy (go without cancer care, eat only Big Macs for a month, start smoking...).

Step 3: A lot of times (in science, and in the questions that people ask AI bots) we want to know what will happen if we do an experiment that's never been done. "How will novel drug candidate A affect blood glucose?" This question is called a counterfactual (as in the experiment not yet done is not a fact yet). Pearl's point is that to allow any hope of prediciting a counterfactual, the underlying model (made by humans or AI) MUST be causal. You just can't predict stuff not yet done on the basis of statistical fits of experiments that have been done.

Another key point is that in step 2 (the randomized trial) if we try to account for "confounders" (like age, which might affect both disease severity but also receptiveness to the treatment) without accounting for causality, then we can actually make the conclusions reached LESS likely to be true
It gets pretty technogeeky (but I'm a technogeek). If your interested in this stuff I think that the book is fanstastic.

Why am I posting this? Well if you use AI (ChatGPT or anything else) to figure out the physical process of tube expansion, unless there's a causal model that it discerns in all the data, the ChatGPT predictions will likely fail.

Anyway, recommended: "The Book of Why".