Impact of initial pressure on OCR in Oedometer tests

[Francis20167]

For incremental loading Oedometer tests, BS1377 - Part 5 - 1990 suggested that for firm to stiff clay, the initial pressure can be the next lower level from the in-situ vertical stress, i.e. 0.5 sig'v. Therefore, at for example 30m depth (assuming effective unit weight of 9kN/m3), the initial pressure to be assigned is 0.5 * 30 * 9 = 135kPa.

However, we know that in many laboratories, the initial pressure usually starts from 12 or 25 kPa (provided this pressure is higher than the swelling pressure), which is recommended in ASTM D-2435.

We have a project where the clay can extend to 50-70m depth. The previous lab tests carried out by another lab adopted 12/25kPa as the initial pressure, and OCR derived by Casagrande method is less than 1. However, we find that if BS1377 - part 5 can be followed, for the firm/stiff clay at this depth, the initial pressure should be 225-315kPa, which is much greater than 25kPa assigned by another lab. The results indicate OCR slightly higher than 1. I am not clear whether the difference in OCR is due to the difference in the assigned initial pressure or due to the sample disturbance because of the different GI contractors.

As we know, in Casagrande method, the point associated with the maximum curvature should be determined, which is frankly speaking quite arbitrary. I believe that with an initial pressure of 25kPa, the point with the maximum curvature occurs earlier than that with an initial pressure of 225-315kPa, which probably leads to the underestimate of the OCR. But unfortunately, I could not find any references to demonstrate that.

Anybody knows about the rationale behind this issue? Many thanks.

 

[GeoPaveTraffic]

I would always recommend starting with a lower pressure instead of a higher pressure. Once you place the sample under load, you can't go back.

As for the results you discuss in the post, I would say there is no difference in result between slightly less than 1 and slightly greater than 1. We don't know anything that closely in this business.

Mike Lambert

 

[Francis20167]

Mike,

Thanks for your comments. In our project, the OCR is important, because we get significant settlement in the area of interest, because of the assumption of OCR = 1 by the previous designer. The reason that the previous designer assumed OCR=1 is because the previous oedometer test results indicated OCR < 1 (some are even below 0.5) whereas the empirical relations from CPT cone resistance showed OCR of 1.2 to 3. Therefore, they conservatively assumed OCR = 1.

By conducting new GI, we would like to improve the interpretation of Oedometer tests, as historically the OCR should be larger than 1 as the soil has been deposited for a long time. However, once we follow the British Standard with the initial pressures from 0.5 sig'v, we don't find much curvature in the e-log sigv curve. I am not sure whether BS standard deemed that the loading pressure below 0.5 sig'v can't contribute to the estimate of preconsolidation stress (that's why BS standard suggests 0.5 sig'v as initial pressure for firm/stiff clay), or BS standard simply ignored the derivation of OCR as it does not include the corresponding details.

But I feel that if we start from a lower pressure than 0.5 sig'v, we possibly would get a lower preconsolidation stress, as the maximum curvature may occur earlier. But whether this earlier point should really be taken into account needs to be discussed, because the stress below 0.5 sig'v is very low and sample disturbance may be more severe if we let the sample be exposed under low pressures for a long time.

 

[GeoPaveTraffic]

I understand why you want to sharpen your pencil as much as possible to reduce the predicted settlement, but how of a difference is there in calculated settlement between an OCR of 1.0 and 1.2? Shouldn't be much and should be well within the range of settlement that you are going to report or plan for.

So many things effect our settlement calculations and for most of those things we, at best, know a range of possible values for them. Run the settlement calcs using a various combinations of parameters and develop a range that likely contains the settlement that will happen with the project is constructed. In the end, this is the best you can do.

Mike Lambert

 

[BigH]

No comment on the starting pressure other than you want to make sure you get enough points on both sides of the p'c value. As for deterning the p'c value, have you ever given thought to plotting log strain vs log pressure? A number of references, and I have seen it too, see that you get two straight lines and p'c is the intersection. Someone wrote a paper in EJG? some time back comparing a number of methods like this

 

[Francis20167]

Mike,

Yes, one of the reasons is to minimize the settlement, as we found the settlement is very sensitive to the OCR value because we only have very limited surcharge but a very thick clay layer. That's why the client pays much attention to the quality of the oedometer tests.

We think the oedometer test should be conducted carefully in order to derive the OCR of the slightly over-consolidated clay. I did not quite understand why British standard has such a guide that for firm/stiff clay, the initial pressure should be greater than 0.5 sig'v. We followed the British Standard in the end anyway, but have now to explain this difficulty in deriving OCR.

Thanks for your suggestion that we perform a series of sensitivity study to obtain the possible ranges of settlements. I completely agree with this approach.

 

[Francis20167]

BigH,

Thanks for your comments. We are aware that there are many methods available apart from the well-known Casagrande method, such as Janbu (1969), Pacheco Silva (1970), Becker et al. (1987) and Karlsrud (1991). All these methods have been compared in Grozic et al. (2003) or Paniagua et al.(2016).

For the method you mentioned is from Oikawa (1987) or Onitsuka et al. (1995), which can provide better bi-linear curves. But again it requires sufficient points around pc.

 

[LRJ]

I speculate that the reason the British Standard specifies what it does for stiff clay is because there is relatively little strain between 0.5σ'[sub]v[/sub] and σ'[sub]v[/sub] when compared to a softer clay. Moreover, it is probably just a pragmatic recommendation. Nonetheless, starting at a lower pressure is generally better as it can give an idea of sample disturbance (comparison of initial void ratio to estimated in situ void ratio). It also usually helps in subsequent interpretation to have a longer line at the start.

Regardless of the above, starting at a lower pressure will not influence the value of σ'[sub]p[/sub]; it may influence a subjective interpretation, however, which is why several different methods should always be employed when interpreting the data. Personally I find the log-log and Pacheco Silva methods to be provide consistent results even if performed by different engineers. If I think there is some sample disturbance (which sounds likely if you're getting OCR < 1 in soil where there is no geological reason for underconsolidation) then an energy method like Becker et al. (or the DSEM update) can be very useful.

 

[Francis20167]

Thanks LRJ. I agree that different methods should be adopted for a comparison. There are many papers which have compared different methods, and many papers were published since 2000s. This means that after 70 years since Casagrande (1936), there are still queries on the correctness of each method to estimate OCR, which also demonstrates how difficult it is for a reasonable prediction.

 

[jcm1890]

I realise that this thread is a few weeks old now, but from my experience:

Agree with those above that initial starting pressure should have little impact upon OCR derivation providing this is suitably below the predicted p'c value.

In terms of using Casagrande to determine p'c and hence OCR I am always reluctant when dealing with incrementally loaded oedometers due to the fact that the max curvature will always coincide with the point of a stress increment. I would agree with BigH and say to compare with also the log-log method, and I have also found the Pacheco Silva construction method very useful for removing subjectivity and is therefore replicable and unambiguous.

I would also make the case that given OCR is seen to be a significant issue, and that prior investigations had already been performed, that it would have been worthwhile performing constant rate of strain (CRS) oedometers to provide a more useful data curve (though it must be noted that strain effects may need to be accounted for in OC clays).

In order to determine sample disturbance/quality, there was a method outlined by Lunne et al. (1997 possibly) based on the initial change in void ratio. As a guide to sample disturbance, the gradient of an unload/reload loop, if performed, can be compared to the initial loading gradient - if a significant difference is noted then it may indicate sampling disturbance.