How to determine material limits from test data

[twillis]

Hi

I've received uniaxial tension test data for a range of materials. Each material test was repeated either three or five times. I'd like to establish an appropriate UTS limit for each material. At the moment I'm taking the minimum stress at failure from the range of test data -10%. I was wondering if someone could point me in the direction of a suitable standard which details a more robust method.

Thanks in advance.

 

[metengr]

Yes, use engineering statistics to establish a mean and confidence levels. The lower bound should be your minimum UTS.

 

[EdStainless]

You can use statistics, but with so few data points it really becomes an effort in just being reasonable. There are special statistical methods for working with small sample sizes, it basically involves building in extra margin.
In aerospace they require 10 separate tests form each of three different lots (or heats) as a minimum.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube

 

[israelkk]

What is the purpose of the tests? Is to to verify that that the materials are meeting the declared properties by the spec or other reasons?

 

[redpicker]

As israelkk has inferred, the use of statistical methods will only be valid if the specimens were obtained from truly random samples. This is unlikely to be the case. Depending on the material and form, it can difficult to obtain a random sample.

For example, if the tests were performed to verify the properties met the minimum requirements of their respective specifications, it is likely that they were chosen to represent the minimum properties of the lot to help insure the entire lot met the minimum requirements. In such cases, the distribution of test results is more descriptive of process control than the range of properties one could expect.

Of course, this does not help the OP at all. Perhaps the best way for the OP to get the information he wants is to ask the producer of the materials in question. They will have a much better idea than anyone else. They may not want to share all of their information, but they may be able to help answer the question posed by the OP.

 

[twillis]

Hi all

Thanks for the advise, very useful. The test is to determine the properties of the materials. I've used a t-distribution to account for the low sample count and calculated the 99% confidence interval. I've taken the lower limit as my UTS.

 

[israelkk]

You didn't answer my question. If it is a standard material manufactured under a spec such as SAE, ASRM, ASME, AMS, DIN, etc., then your tests can only verify if the actual material you have, meets or doesn't meet the spec. Otherwise, it may be a waste of time. You can not assume nothing for future batches of material if you intend to manufacture more in the future. If you design a new product you have to assume the minimum UTS as it is stated in the spec and design for this value. The only and must benefit from your tests is that you can interpolate what will happen when the UTS will be minimum as declared in the spec, based on the actual tests you perform on the final product made from the material you tested.

 

[twillis]

As mentioned above, the purpose of the tests was to determine the properties of the materials. Specifically, uniaxial stress strain curves and material limits at varying strain rates and temperatures. The materials in question are polyurethane, which do not typically conform to an applicable standard. Therefore there are no minimum UTS values to work to without undertaking some form of testing. The final product will be tested to determine its performance.

 

[Jughandle]

As you may be aware, the mechanical props of polymers are temperature and strain rate dependent, so you would also have account for those variables. (Polyurethane is a thermoset, which may lessen those dependencies some.)

 

[navadado]

can you share some images from the test? suppose u plot the test and will soon figure out where located the YP, UTS and fracture (metal & ceramic), it's different to polymer stress-strain curve .

 

[SWComposites]

Twillis, see CMH-17, volume 1, chapter 8, for statistical methods to calculate design (allowable) values from test data.