Hi dear Researchers,
In terms of uncertainity and reliability assessment, apart from geotechnical justifications, I need to have a statistical support for the estimated parameters in a series of triaxial tests. As you know, in triaxail tests the of interest item is estimation of the p-q line (the failure line), the line then could be used to derive the behavioural parameters, the line equation could be something like "y=ax+b", and "a" and "b" are the fitted coefficient and constant.
I'm aware that there are some tables in pure statistics which could relate the confidence levels to the degree of freedom (for known and unknow populations) but sometimes the required number of samples are not practical in engineering (e.g. 50 triaxial tests for just one site). we normally use three or four points to form/fit the line, but in reliability assessments, as both "a" and "b" values (as line gradient and y-intercept) are subjected to uncertainities, I'm not sure how to find the sample size to remain in 80% reliablity level.
It would be highly appreciated if anyone could advise me with him/her similar experience.
thanks,
I have the same case with you, four samples to find out the behavior of the soil I think is enough but if you are in doubt you can use eight samples (in this statistical theory already represent) ... because of problems in Civil Engineering, if we need a large sample it is necessary a lot of money
Thanks Faisal estu Yulianto for your answer, highly appreciated.
As far as I realised, (sometimes) the financial issues could govern the number of samplings/tests, and the number of samples/test should be (very) optimom.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size