New update of the Plaxis, partially, solves the problem mentioned in the post below. Now, if you define a manual tensile strength, this will be reduced during safety analysis. But if you rely on the tensile strength automatically calculated by Plaxis, it will not be reduced and same problem continues for that case, in my opinion. But at least, for the users that are aware of this distinction, there is an option to correctly use Hoek-Brown model.
See the note on Plaxis
This is an investigation of a problem about Hoek-Brown model in FE codes. If you have any addition, test results or correction in my statements, please reach me.
Recently, we have been involved with a lot of FE analyses of excavations in rock with my colleagues in Niras. During the embedment depth and sensitivity analyses, we noticed weird behaviour and started small tests. Our tests showed us a behaviour that worried us - so this article is for people that have dealt with the Hoek-Brown models in Plaxis (or in any FE software) before - especially if tensile behaviour was governing the overall response.
So, the simple test model.
We define two material model - one Hoek-Brown and one Mohr-Coulomb. Stiffness parameters do not matter. For Mohr-Coulomb, enter some strength and for tensile strength, enter 34.5 kPa. For Hoek-Brown, here are the parameters - they result in 34.5 kPa too.
What do we expect if we pull them in extension and make a safety analysis? FS = 34.50 / 30.00 = 1.15. This should be the result. But, in Hoek-Brown, we get 3.40.
There are two questions. Why do we see FS of Hoek Brown much higher than what it is supposed to be and what happens here?
Why?
Plaxis does not reduce the tensile strength during safety calculation. Plaxis follows Benz et. al. (2008) approach and this paper does not talk about tensile strength. Plaxis material manual also does not mention any factorization of tensile strength. If tensile strength is not reduced during safety analysis, what happens?
What Happens?
Our guess here is that Plaxis starts to increase the safety factor to find instability during steps of safety analysis. However, even though it increases it to values around 9, it cannot find a instability, because we are looking for a full tensile failure while Plaxis only reduces the compressive/shear strength. When it reduces other parameters but not tensile strength, with the wording of Plaxis support, the failure surface rotates around the tensile strength while actually it should both rotate and move to the right, because tensile strength is reduced.