Concrete statistical analysis

[Robert168]

A concrete test requires at least two cylinders being broken, of which the average is used.

During a concrete testing on a project, I got a low strength result from a pair of cylinders on the 7 day break. As it was very low, I just broke one cylinder on the 28 day break and save the last one for the 56 day break.

So I had an one cylinder test result for the 28 day break and it was, as expected, low in strength.

We finally took cores and the core test result was satisfactory.

My question is whether I shall include that one cylinder test in the concrete statistical analysis.

If I include it, it is an one cylinder test. If not, the whole analysis would lose the most obvious change in the construction process.

Any advice will be appreciated.

 

[BigH]

Why did you break two at 7-days - usually, you take three cylinders, break 1 at 7 and 2 at 28-d for your "test result"? AASHTO Bridge code permits you, even with two at 28-day to toss out obvious outlier result if attributable to poor casting or other such problems. I would presume the single cylinder test result would be included in the concrete statistics.

 

[iandig]

In my opinion the statistical analysis is carried out on the quality of the concrete used, not on the quality of samples taken. From the way you describe the situation, checks made on the concrete as placed have shown that the strength of the concrete was OK, but that the original cylinder samples were not as the measured strength was low. In this instance there seems to be a good argument to say that the original samples were not representitive of the concrete as placed (poorly made/compacted/cured etc...), and as such should not be included in the statistical analysis of the concrete batch.
If however, you 'feel' that the original samples were a true reflection of the concrete as used, then they should definatley be used in the assessment. Judgment is required as to what data should be included, and this is the same as if you had a series of 'abnormally' high results, you need to be confident that the results are relevant. If you indiscriminately use every result, your statistical analysis will be as much to do with the quality of sampling as it is on the quality of the concrete used so the question is "is this what you want it to show?".
In order to aid in the assessment of the validity of the original samples, it is always worth looking at other data apart from just the strength, was the density of the concrete 'typical' for the mix and how does it compare to the cores. Is there any evidence that the way in which the original samples were made and cured was different to the previous tests, did the same technician make all the samples etc... all this should help in deciding if the original results should be included.

 

[Robert168]

Thank you again for your help.

The specification of the project requires two cylinders be broken, so I just follow it. The "trouble" cylinders were not expected.

The low strength cylinders are more representive of the concrete at the spot where the lowest strength core was taken. That core was thoroughly checked and no defects were found.

So I will include the one cylinder test in the statistical analysis, which, by doing so, will tell the true and complete story happening on the site.

 

[HughL]

Robert168,

<<the core test result was satisfactory>>
Why bother with the cylinders at all if you have cores. Just enter the core result for that point in your analysis.

regards Hugh