Testing of shear pins

[acrockett]

Hi,

The design i am working on uses a shear pin to avoid damging the equipment (in fault scenario). In order to commission the equipment we are testing a singe hsear pin to see what the failure load is. However, we are unsure if a single test is sufficient for testing shear pins. Can anyone provide some guidance on the testing of shear pins (i.e. what is best practice?). I have had a look and it seems we should expect variations of upto 15%, but for our shear pin do we nee to test a certain number of times over a full range of tolerances?

Thanks

Andrew

 

[MintJulep]

Testing a single anything is generally useless.

You need to test a population sufficient to capture the variation in both dimensions and material properties.

 

[acrockett]

That was our feeling. Is there any guidance on the number of tests required to get reliable data. We will be manufacturing all shear pins from the same billet and they should all be made to the correct size (with pretty tight tolerances).

Thanks

 

[MintJulep]

http://stattrek.com/lesson6/samplesize.aspx

 

[TVP]

Page 4 of this document has some very general guidance values that are typically used in the aerospace and automotive industries:

http://asqbuffalo.com/wp-content/uploads/2011/10/3a-SampleSize.pdf

 

[zekeman]

Here's how how I would approach the problem.
First since it is a shear pin safety problem, you have to decide the range of failure with an associated probabity that any shear pin would fail within that range .

Then you go to the mfr and ask for his capability of meeting that requirement

If it can be done your testing is rather staightforward using the techniques of sampling suggested above.

For instance, let's say that you want the shear pin to fail between 80000psi and 100,000 psi with a probability of 0.95. Assuming a normal distribution, you would like the mean to be about 90000 and since p=0.95, you need 2 sigma to be 10000 or less.
Now when the lot comes in for checking, you use the sampling methods to get sigma and the mean , verifying that

mean-2sigma > 80000
mean+2sigma < 100000

 

[chicopee]

From a Probability and Statistics for Engineers course, I seem to remember that a population of 50 pins should be tested to establish a distribution.

 

[GregLocock]

Good grief, testing 50 seems a bit OTT. Wouldn't they all be from one batch?

First up you need a testable requirement, something like "99% of pins must fail at X kN or less, but none must fail before Y kN".

I'd create a model that predicts the load for the relevant factors. Then I'd verify the model against real tests for the outliers and the centre. Then I'd use the model, and the known tolerances and distribution for each parameter to predict the entire population of loads.











Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[Cockroach]

I do this all the time with shear pins for use in Pump Out plugs for oilfield downhole tool applications.

You make your shear pins from rod representative of the piece you will use. Use a random number table and use the Student T curve to get a sample size representing 95% confidence level. Those are tested under a control setting in the laboratory. You're looking for the material properties and loading as a result of failure by shear.

For example, I apply a fluid hydraulic pressure to the pump out plug for which the shear pin properties are desired. The test is preformed with good repeatability, therefore you get a confident and tightly grouped bunch of numbers for load and shear strength of the material in question.

Done. Good luck with it.

Regards,
Cockroach

 

[GregLocock]

"Assuming a normal distribution".... could be a bit ass bitey.

Plot your results on normal distribution paper or a weibull chart to make sure that you are getting a normal distribution.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[MintJulep]

Assuming that all of the shear pins that might ever be made over the life of the equipment will be made from the same billet could be ass bitey too.

 

[chicopee]

Just opened TVP's link and page 4 summarizes sampling sizes.

 

[acrockett]

Thanks guys. Think I have a way forward based on your responses and have found some work that was done on a previous project.