Statistical Compaction Acceptance 3

[Grouser]

With all this discussion of 95% compaction going on in the other forums, I would like to grind a personal axe and suggest that acceptance of compaction should be done using statistical methods rather than the pass/fail technique commonly used. If the mean of the test values met or exceeded the "specified" value, and contained no outliers more than x deviations, then the compaction would be acceptable. X can be defined as standard deviation units or as deviation from the "specified" value.

With the advent of nuclear testing devices, an adequate number of density results can be quickly obtained thus establishing sufficient population for the statistical procedure to apply.

Comments and alternative viewpoints welomed.

 

[Focht3]

In general, your approach is quite reasonable. I can see some circumstances where this might not be a good idea, but it would be a good way to approach the "big" jobs. It would generally require full time inspection -

Small sites and irregular sites might still warrant the current system. And, of course, you have to have enough test samples...



Please see FAQ731-376 for great suggestions on how to make the best use of Eng-Tips Fora. See faq158-922 for recommendations regarding the question, "How Do You Evaluate Fill Settlement Beneath Structures?"

 

[cdh61]

Grouser,

Something else to considered, what if you have a zone (say northeast 1/4 section)with lower values, but the mean and x deviation is still OK based on all readings. You could end up with more settlement in one zone than the others.

regards,

 

[dirtsqueezer]

I respectfully disagree. Even with the nuclear densiometer, inspectors don't test every square inch of a building pad, especially those with the larger square areas. A test represents a miniscule portion of what's known about compaction, and is used to represent the vast majority of the rest of it. 95% only tells you compaction, "To the best of our knowledge".

I think if you are approving compacted structural fill designedmeant to behave as a competent stuctural unit, you should be at least meeting the minimm standard in all areas. If you were to get a 93 and average it to a passing score, I believe there's a chance some areas will be lower. There is a certain level of competency required for a soil inspector. Leave the interpretation to them- that's what they get paid for. If you find them incompetent, it's time to hire someone else.

 

[Focht3]

... A test represents a miniscule portion of what's known about compaction, and is used to represent the vast majority of the rest of it. ...

Yes, that's where the full time inspection comes in. It's important to remember that the idea of percent compaction is relatively new; I know some engineers that are still practicing today that have monitored fill placement using <specified number of passes> each way with <particular compaction equipment>. It worked then (1930's / 1940's - and before), and should work now.


I think [blue]Grouser[/blue]'s point is a good one - on some, but not all - jobs. Let's face it - statistical 'low spots' are a geotechnical engineer's stock in trade. Insisting on a minimum percent compaction for all tests won't prevent weak zones. In my experience, a first-rate field tech can do a much better job. And at some point we have to recognize that some failing areas must occur, regardless of the test results...

Put [blue]Grouser[/blue]'s argument a different way: we're relying on test results that have 'no failures' to give us a false sense of security about the overall quality and likely performance of a completed fill. We need to put the human element back in - and treat the test statistics honestly.

I continue to think it's a reasonable idea.



Please see FAQ731-376 for great suggestions on how to make the best use of Eng-Tips Fora. See faq158-922 for recommendations regarding the question, "How Do You Evaluate Fill Settlement Beneath Structures?"

 

[cdh61]

Grouser,

I do agree with the statistical approach, but like Focht3 said, on some but not all jobs. We have worked on a few jobs which used statistics to pay out contractor bonuses, and these jobs do require full time inspection. Statistics is not used more (I think) due to cost. It does require time and energy, and some people will just end up saying you have results that do not meet the compaction criterion. I have encountered this on a number of occasions, even on jobs which we are providing full time inspection services, and the cost of the project is in the tens of millions. One result under compaction criterion, then comments like -- who is going to guarantee end product? I have had to write letters over 0.5%. So - my response, go compact again and then do a retest. End result, compaction criterion met, not my problem any more, you accept responsibility. Not the best of attitude, but many people want you to ensure things are OK without doing the required work, do not care about statistics, and do not what anyone to tell them that there could be areas which will not meet the criterion.

Focht3, put the human element back in - I agree, but the court room or potential legal action can change people’s opinions. Even simple change orders are not been reviewed and accepted by some firms (just reviewed but never officially accepted) because of responsibility. Legal action or the potential of, I feel has removed some sound engineering, which includes honestly.

regards,

 

[SirAl]

Subtle variations in the laboratory testing procedures can produce variations in the moisture density relationship used as compaction reference. Further, interpretation of the test points (how the moisture-density curve is drawn) can be subjective. Finally, soil conditions in the field are never perfectly uniform. As such, compaction testing is subjective and requires interpretation by an experienced individual. On this basis, the statistical approach by Grouser makes good sense to me. Good points brought forward by all respondants.

 

[jheidt2543]

The statistical criteria for the acceptance of concrete test cylinder breaks has a method for handling low breaks. I should think the geotechnical community could develope a similar proceedure for compaction testing. It could address the problem put so well by Focht3: "we're relying on test results that have 'no failures' to give us a false sense of security about the overall quality".

A statistical method could provide a framework for a reliability basis for foundation settlement design. The new codes promote that in steel design, concrete design, wind loading and seismic loading, along with a number of other areas. I do realize that soil properties are not as homogeneous as steel or "concrete", but soil is generally more predictable than a seismic event!

An additional feature of such an approach, I think, would be the advantage a local geotechnical firm might have in knowing the soil conditions in their own area, not just compaction, but the overall approach to their geotechnical work.

Just a few thoughts on a snowy morning waiting for SPRING!

 

[BigH]

Stastical "testing" has been around for quite a while. Ontario has an "acceptance" chart they use (and I am sure other jurisdictions do too). Here in India, it is big - it is required on road construction jobs:

average compaction >= specified compaction + (1.65 -
(1.65/sq.rt N)) x standard deviation.

N is number of tests - 6 minimum if sand cone; 12 minimum if nuclear density machine) - this for each 500 m3 of fill ( I may be off on the later - it is usually per stretch and on our project each stretch is less than the normal m3).

I don't like it that much - I understand why they do it but it does require 95% compaction to have an average compaction of about 96% or slightly more.

 

[dirtsqueezer]

Just another thought on statistical methods.

Nuclear densiometers have a range of 0-12 inches. The sphere of influence of your builing is likely to have orders of magnitude larger effect at depth on a subgrade. Statistical methods are one way to do it, but remember 88 + 95 + 99 = 95% compaction. That 88 could represent an area of up to 2500 ft2. This is not good.

I would also argue that because it's basically an easier standard, it makes the inspector's job 99% more difficult. Say he fails the first test- no problem. He's got four more- no need to change anything, could be just the gods. By the time the second test rolls around, and he fails- right there you're losing money on the budget, when half the work done so far could have been done correctly.

Basically, I think statistical methods are a bad idea. They lower standards for construction, and make it harder for honest contractors/inspectors to stay honest. Amen.

 

[Focht3]

I think you missed the point.

We're not talking about "small" jobs - we're talking about the big ones: resident geotechnical (or materials) engineer, full time inspection, on site lab - the works. And lots of fill to be placed. We're not talking about a Wal-Mart SuperCenter, or an office/warehouse park. We're talking large highway projects and other mass regrading jobs. For example, a "typical" residential subdivision regrading job in southern California is on the order of 30 million cubic yards. That's a lot of fill to excavate, place, moisture adjust and compact!

Frankly, full time inspection will do much more for improving the quality of the compacted fill body than simply insisting that every test meet or exceed 95 percent. What hasn't been said so far is why we see the very rigid requirement for 95 percent relative compaction. The simple historical fact is that owners have been unwilling to pay for inspection, and the requirement for 95 percent or better has been the geotechnical engineer of record's response. It's a sorry state of affairs. But it's where we are right now - on most jobs. We have gotten to the point of taking the judgment out of this aspect of construction. And that's a serious mistake.

In my view, [blue]Grouser[/blue] simply wants the engineer's judgment put back into the process, aided by the use of statistical tools to improve our ability to assess when to permit low tests, and when to yell STOP!



Please see FAQ731-376 for great suggestions on how to make the best use of Eng-Tips Fora. See faq158-922 for recommendations regarding the question, "How Do You Evaluate Fill Settlement Beneath Structures?"

 

[jheidt2543]

dirtsqueezer,

If we were to use soil testing methods based on statistical analysis of the compaction tests similar to the way concrete cylinder tests are handled, there is a proceedure for low tests. There is still an absolute minimum test and also a range within which all tests must fall. There are also proceedures regarding how to handle the materials that test low.

Secondly, while the compaction test only monitors the 12" depth you note, the fills are placed and tested in lifts and hopefully the exact same spot is not picked for each lift test.

It would be interesting to analyze what the overall costs are of a strict 95% minimum compaction value vs. full time inspection with some lower value based on statistical methods, say 90% (pick a number or analyze a range of numbers).

Just some additional things to think about on the topic.

 

[Focht3]

Secondly, while the compaction test only monitors the 12" depth you note, the fills are placed and tested in lifts and hopefully the exact same spot is not picked for each lift test.

YES! This is a classical problem with a lot of construction projects. Some bean counter (or young engineer with little or no field experience) decides that a density test should be performed every 50 feet in each direction in order to get enough tests per lift to meet the contract requirements. The contractor is instructed to maintain a site grid - laid out in 50 foot spacings - so that the testing lab can report the test locations. All is fine, right?

[red]Heck, NO![/red] Where do you think the contractor tells his operators to concentrate the compactive effort? Along the grid lines. A lazy tech - or one who did not see what the contractor has done because s/he was not present - will simply follow the grid pattern as s/he conducts the tests. The fill pad passes, all the test results exceed the requirement, and the fill body is poorly and unevenly placed. (I had quite a few confrontations with a Pakistani construction supervisor in Saudi Arabia (1976) over this. I wouldn't test along his nice, neat little compacted strips. Boy, would he get mad!)

This happens all the time. And it's why I say that we need full-time inspection on "important" fill jobs (including the small ones.) And the use of statistical sampling would permit a wider number of tests spread out over the site. So if a certain lift did not meet the statistical requirements for compaction, the entire lift would have to be re-worked - not simply one little area.



Please see FAQ731-376 for great suggestions on how to make the best use of Eng-Tips Fora. See faq158-922 for recommendations regarding the question, "How Do You Evaluate Fill Settlement Beneath Structures?"

 

[iandig]

Just a couple of points to add to the discussion, based on someone who has been the 'good site technician' and the 'young engineer'.
There are lies, damn lies and statistics.
If you are concerened with unscrupulos parties, then statistics can cause just as many problems as deciding that you require 95% compaction, without appreciating what you are trying to achieve. Compaction is the removal of voids from a material by the application of some form of compactive effort. It is important beacause if it is carried out succesfully, it will leave a material which should be stable and will undergo limited long-term settlement. The percentage compaction is a comparison between the density achieved on site, compared to a maximum dry density [MDD] from a laboratory test, but dependant upon the compactive effort used, you can/will end up with different MDD. Air voids are also very important to control, it is possible to achieve over 95% compaction and still have high/unacceptable air voids (in excess of 10% dependant upon material type and compactive effort used).
My suggestion would be that if the assessment of the compaction achieved was to involve a better use of statistics, then you should be looking at more than just the density.The materials moisture content, air voids and 'strength' are all important properties, and if the arguement is that a simple pass/fail criteria is insufficient for a particular project, then more than one property should be considered when doing the 'bean count'.
I have just read through this and it sounds like a bit of a rant and its not meant to be.
I guess the point I am really trying to make is that I am all for a better assessment of the results, statistics should play a part in this and from my own experience a site with zero failures usually means something has been hidden.
But PLEASE PLEASE PLEASE do not make come to the conclusion that one party in the UK has, that the highest risk to statistical analyis are those sites that undertake a high degree of testing. The arguement may be sound, but the way it was published was that the less testing you do, the less failures you get. If you don't test it it won't fail.

 

[BigH]

Focht3 - on the large highway jobs it is usually specked (and a good idea) to use randomly generated test locations. I do this for my cores, especially in asphalt. Still, even with random checking, we make sure that a certain percentage represents "edge" conditions or, in the case of using a paver doing half-widths, in the joint area. Also, even with statistical methods, the inspector should also pick audit test locations - these he can pick "good" or bad areas. Again, as we all will agree (I think) tests are only as good as the will of the contractor to do the job right. And, we all have problems in the non-flexibility - but as KRS said in another thread, the test results are nearly the only way to ensure compliance by the contractor to his contractual obligations - always the technical vs contractual dilemma.

 

[dirtsqueezer]

Here's the long and short of it in my mind. A few clarifications:

Yes, full-time inspection. Otherwise, how do you know about lift they installed below?

Large project? Multiple inspectors, each having brought their judgement. You should also have multiple proctors and even more gradations for these individuals to view, thus ensuring compaction in a reasonable and efficient manner.

Low concrete tests: The difference is that soil can be improved right now! There is nothing you can do to improve the concrete. The presence of a competent inspector on site can almost always get the job done cheaply and immediately with water, vibratory rollers, dump trucks, ect.

Basically, I think inspectors are armed with enough tools these days to be able to help with the minor adjustments that can be made while they're still on site: water, rolling pattern, fill variability, to ensure that compaction is met. When I fail a contractor on a test, I am communicating a thing to the engineer: they have a soft area. If an inspector cannot find a way to meet compaction on site where it could be met, and fails you with a 93%, he is not worth his salt. If he cannot find a way to meet compaction on site where there's something terribly wrong with the fill, then the problem is documented, and when your foundation cracks, we know why. If I leave site with a failing report saying the problem still isn't solved, I am communicating a concern, not a number.

Now, about the statistical methods. I see two scenarios, both of which are flexible: a set standard for 95% compaction coupled with the judgement of the inspector for those 94s, and a statistical average where low tests don't need the judgement of anyone, due to possibly higher tests in a completely different area. The first system requires judgement by the inspector. The second does not. The second system allows for soft areas on record, the first does not. I would personally rather play a role where my observations and judgements could do some good for the project, instead of the prospect of being just a button-pusher.

Now, having said all that, I think BigH's point about a 93% in a large unit being de minimus is a good one. That is where judgement comes in. If they think it is an issue, it shouldn't take more than 1 or 2 more passes on a unit like that to achieve their 95%- that is why they're there. It's when they don't affect a fix of the minor problems when it goes to the engineer's desk, and they have to deal with it after the fact that things get convoluted, leading I think partly, to this very discussion.

Good topic!

 

[Focht3]

Yes. Good thread, all!



Please see FAQ731-376 for great suggestions on how to make the best use of Eng-Tips Fora. See faq158-922 for recommendations regarding the question, "How Do You Evaluate Fill Settlement Beneath Structures?"

 

[dirtsqueezer]

One last comment from me, with respect to method-based inspection:

...some engineers that are still practicing today that have monitored fill placement using [method-based inspection] It worked then (1930's / 1940's - and before), and should work now...

I believe the reason I am employed in the industry as a soils inspector is to ensure that a building does work. That's why you spend the extra money to have me there. That way, you have method monitoring as well as empirical results. Otherwise, I'm tending a fry machine at your local Mcdonald's until I get a better degree. :] Just trying to point out that there is a sentient being at the other end of those reports- not just a gauge.

Again, good topic.

 

[Focht3]

And your presence is appreciated. I find it hard to keep my temper when others say that field inspection is "a waste." It clearly is not.

Tests alone are not sufficient.



Please see FAQ731-376 for great suggestions on how to make the best use of Eng-Tips Fora. See faq158-922 for recommendations regarding the question, "How Do You Evaluate Fill Settlement Beneath Structures?"

 

[Grouser]

Thank you to all contributers. Many thoughtful and thought provoking responses. My basic thrust was to insert rational decision making into the compaction acceptance process rather than using a codified value. I see there are aspects I had not considered. While I believe that most of the acceptance concerns expressed above can be accounted for by further statistical criteria, who is going to pay for this analysis time and how can it be communicated to the contractor in a timely manner? I was hoping that the contractor attempting to properly compact a fill lift could get it accepted if his average met the requirement and thus would NOT have to perform additional compaction beyond what is necessary to provide for uniform settlement but it looks like we would be getting back to the 95% discussion I was hoping to avoid.
I don't think anyone can inform me as to whether a two foot fill of one foot lifts compacted with the bottom lift at 94% and the top lift at 96% will exhibit total settlement greater than the two foot fill with both lifts at 95%.
I think a failing test may actually be at 95% but the gradation is slightly different at that location thus bringing the density down. Insufficient densification is not the only reason for a "low" reading.
Despite these considerations, it looks like the statistical acceptance method may be before its time now. Thanks again for participating.