Question of Concrete Statistics

[BigH]

Dear All:

Just wanted to see if anyone has any opinions on the topic. ACI/AASHTO, etc. typically use 2 cylinder breaks at 28-d in order to show compliance with the specified strengths subject to three moving averages being greater than the specified strength and no single cyliner less than 3.5 MPa (500 psi) below the specified strength. This is based on no more than 1 in 20 being below the specified strength and is a function of the standard deviation of the test results.

In Asia, it is customary to break three cylinders (or cubes - but now generally cylinders) rather than 2 cylinders. It is my contention that the average of the three cylinders is "closer" to the true value of the set than the average of 2 cylinders. Would this change the definition of compliance? Say, that the moving average be no less than 0.5 MPa (100 psi) less than the specified average and no single test value less than, say, 4 MPa below the average?

On our job we are breaking three but still using the ACI definition of compliance.

Any thoughts? Ron??

Cheers to all - hoping that we are all still "above water".

 

[JAE]

From a picky-technical perspective, the applicable building code usually will reference the concrete spec (such as ACI 318) which in turn states definitively that the acceptance is the average of 2 cylinders.

But the building code also allows for other methods and means when "acceptable to the building official".

So I think you would have to find particular statistical evidence or research that would convince your local building official...which might be difficult even with the research since it has not been codified.

 

[BigH]

Thanks JAE - I appreciate your thoughts.

I am posing this question only from an interest's sake. I don't have to worry about a local building official . Just wanting to see if anyone had any thoughts on using 3 or even 4 cylinders for determining the average and how ACI compliance statistics might be adjusted - perhaps, in the future, ACI might go to 3 cylinders (which I think is the case now if you use 4x8 inch moulds - yet, as far as I know, they have not changed the statistics.

 

[JAE]

BigH,
I'm sure that a statistica case could be made for three cylinders vs. 2 and possibly a change in the acceptance threshold.

But the variability of one sample of concrete (making 3 cylinders vs. 2) may not be directly linked to the overall variability of the concrete strength as it varies from load to load. Each represents a different sampling and measures a different form of the variability.

So if that's true, there may be a change in the acceptance threshold for strength...but it might not be as much as you suggest (100 psi) in your OP.

 

[woofar]

G'day Howie,

First up I must stress that this is only my opinion and although I have been putting this theory to colleagues for many years I have never really recieved much support.

No matter how many cylinders you take to form a set, I think that the only one that should be reported is the highest result achieved by any single cylinder.

My take on this that there are many different reasons for getting variability in a set of cylinders.
1)The truck driver may not have spun the bowl properly so he may not have achieved a uniform mix.
2) The batcher may not have introduced the ingredients in the correct sequence.
3)The sampling procedure may have been incorrect.
4)The moulding procedure may have been incorrect.
5)The curing procedure may have been incorrect.
6)The cracking procedure may have been incorrect.

My contention for accepting only the highest result is that despite all the possible variables you have managed to achieve a high result from that particular batch at least in one instance therefore there is nothing really wrong with the brew, the rest is just to do with sloppy workmanship and therefore can only be addressed by constant surveillance. I don't believe that you can address these problems by statistical analysis.

Just to qualify myself again I must reiterate that this is only my personal opinion, I have nothing concrete(if you will pardon the pun) to back me up on this theory.

Cheers
Michael

 

[Ron]

BigH...you and I probably come from the same school of thought...3 is better than 2. I get asked occasionally by younger engineers and even in trial testimony, why I select a minimum of 3 samples for my evaluation. My answer is simply that one sample shows no trend, two samples can confuse you (one good, one bad...how do I interpret?), so at least 3 samples are necessary to provide direction (to myself!).

As for ACI, I surmise that the reason for the two specimens is a compromise of sorts in the early committee activity and it stuck. As you know, most of our standards and codes are "committee compiled". In that process, there is usually a range of interests represented and in the process someone gets stubborn and wants this or that, so rather than hold up a standard or code item, a compromise is struck.

Also, with sampling by one technician and testing by another with multiple processes in between, are we chasing a false accuracy by increasing the number of samples? Somewhat like measuring with a micrometer, marking with a crayon, then cutting with an axe.