Janbu vs Morgenstern-Price and Spencer in block stability 1

[sgsibob]

We have compared stability analysis methods using SLOPE/W for a block failure geometry in a large rock slope. The problem is a tall slope at roughly 45 degrees which is backed by a fault zone which dips steeper than the slope, and various hypothetical inclined slip surfaces exiting the slope through a broad zone of weaker material that intersects the fault and is also exposed on the slope. The mode of failure is therefore down the fault and out toward the slope through the weak material. What we have found is that Janbu consistently results in lower FOS than Morgenstern-Price or Spencer even if the geology, failure geometry and material properties are identical.

Is there a theoretical basis for this or is the answer found in some nuance of the application?

 

[mconti]

Janbu it is semplified method? if yes this is difference

 

[sgsibob]

Others performed the analyses...I believe the Janbu simplified method was used because runs attempted using the Janbu generalized option in the software return "no solution" so it seems that was not the Janbu model that was used. In one case the FOS using Janbu in the block failure mode was a little more than unity but the exact same geometry using Morgenstern-Price was almost 1.9.

Some feel Janbu is preferable for a noncircular geometry with a low-angle release. Why?

 

[ItsOnlyDirt]

I suspect the issues you are observing are to do with how Janbu deals with, or rather doesnt deal with, interslice shear forces.

Janbu doesnt deal with interslice shear, only with interslice normal forces, SO: the resultant is always horizontal.

With methods such as Spencer, MP and GLE which consider interslice shear force there will be an increased contribution to the base normal force and thus an increase in resiting forces, and a therefore a higher FoS.

M-P or GLE is a much better model as it considers both force and moment equilibrium. The spencer method is similar to both MP and GLE however is has a constant interslice force function i.e. not defined by the modeller, e.g. MP typically uses the half sine function.

If it was me I would use MP, but perhaps using a hoek brown soil model.

Hope this helps

 

[BigH]

Maybe you should be using a rock programme as you are not dealing with soils - SWEDGE, if I am not mistaken, is one. (just a little Sulawesi aftershock - cool!)

 

[ItsOnlyDirt]

SLOPE/W can cope more than adequately with rock slopes, though it requires some understanding of the available strength models.

Using the hoek brown model would seem appropriate possibly with some anisotropy to account for your weaker materail or a even seperate unit.

I would also do some sensivity analysis/probalistic analysis on the water table an possibly rock parameters. This is all available in SLOPE/W.

Might be worth running some FE on this too, deformation may give you a better idea of behaviour than limit state. I dont know how you intend to deal with the instability but it may also give you pointers as to when the slope is about to fail. i.e for a slope inspection manual say.

SWEDGE is more for large wedge failures. If it were me I wouldnt use it for this application.

The choice is yours mon ami.

Cheers

 

[sgsibob]

Thank you for the replies. Not having gone through hand calculations of the Janbu vs M-P methods I will take your word for it. If Janbu generally results in lower safety factors because it leaves out resistive forces that are components of interslice shear, then that is good to know. On the other hand, from my common sense assessment of the results given the rock conditions I was not surprised at some of the low safety factors.

As for the material properties, the rock mass was modeled as irregular geotechnical domains each corresponding to an isotropic Hoek-Brown rock mass strength, with the exception of the major faults which were modeled as tabular bodies with strengths properties as measured in large-scale laboratory direct shear testing. The rock fabric anisotropy question was identified early on as an unknown and in future modeling we will address it.

The slopes are too large for programs like SWEDGE to have any validity. We have run SWEDGE on smaller slope components and the nice thing is that a probabilistic analysis can be run that addresses rock structural variability, whereas when you do a slices model you pretty much define the rock structural arrangement and that's it.

We were directed to assume that all the slopes are fully dewatered. Future efforts will explore the impact of that assumption more fully.

We plan in future efforts to run probabilistic analyses. In that instance we would characterize the rock mass strengths in terms of a probability density function and let the program routines give us failure probabilities. However that will not address the greatest uncertainties, which are the actual arrangement of the subsurface geotechnical units and the representation of strength anisotropy.

We have run finite element analyses in instances where limit equilibrium showed safety factors lower than we would like and obtained more favorable stability factors. However, the strength reduction method used for giving a "safety factor" is not accepted in all circles, although it is widely used.