Skip to main content
D
Dumitru1-Visitor
1-Visitor
July 7, 2010
Question

Stress Strain Statistical Analysis

  • July 7, 2010
  • 2 replies
  • 6367 views

Hi, there! I have the attached sheet with 30 stress-strain diagrams (each out of 50 points). I need to do a statistical analysis to establish a mean curve. By the looks of it, there are about 2 groups, but again how can I be sure (grouping the curves which are most "look alike") ?.

Please HELP!

Kind regards, D.

2 replies

J
jeanGiraud1-Visitor
1-Visitor
July 7, 2010

"By the looks of it, there are about 2 groups, but again how can I be sure (grouping the curves which are most "look alike") ?."

_____________________________________

Not so sure about two groups, looking at the hay pile !

It should interest some experts in that kind of analysis.

The 30 experiment do correlate but like the girl ..........

"slightly pregnant".

jmG

D
Dumitru1-VisitorAuthor
1-Visitor
July 8, 2010

"By the looks of it, there are about 2 groups, but again how can I be sure (grouping the curves which are most "look alike") ?."

Well, there seem to be 2 extreme values for the stress (one accuring at a strain of about 1 mm/ 1000 mm and the other at about 1,75 mm/1000 mm) , therefore the two forementioned groups. The idea is, based on similar strain values (or stress, which ever is best - still at figuring out) the diagrams most "looking alike" would form a group for which I need to determine a mean curve. For that mean I'll then try to find a best fit...

F
FDS13-Aquamarine
13-Aquamarine
July 8, 2010

Dimitru, there is a very nice package called Kornucopia which works in combination with MC 14 or higher. With this package you should be capable to make an in-depth analysis. You can find more information: http://2kornucopia.com/index.html

T
tubar1-Visitor
1-Visitor
July 11, 2010

Is not established by now an analytic relation of strain vs stress.

It depend in what region you are:

-in elastic region of material stress=relative displ * Young modulus

- in plastic region the relation isnt linear.

Your groups of data have to calculate correlation with some official test data. You'll get a coeficient of correlation

Which group correlate more, that one is more "look like".

But result will still remain relative anyway, because still isnt a analytic formula in plastic region.

J
jeanGiraud1-Visitor
1-Visitor
July 11, 2010

By just reading, you are correct on all counts !

This was a long thread too, to discover that the data or the experiments were wrongly carried. Same thing happened in another collaboration. I had implored and implored about the data, that they were incorrect... finally the data were wrong . For that particular "Stress strain" I have implored to get a "certified" data set so to avoid ramping in the clouds. If those experiments have or may have value for the "concrete community" they should be on the design board of some regulatory something, not just home experiments. I watched the long video [about 1 hr] from the previous suggestion, not to avail. Can't even figure what the guy is talking about [onion soup, choux à la crême] ?

jmG

Terms & ConditionsAccessibility statement