FLAC is the finite difference based software and PLAXIS is the finite element based software. I recommend to you to use PLAXIS 3d tunnel software also to model tunnel related problem.
numerical studies suggest that finite element modelling and finite deference modelling have essentially similar results. However, when it comes to modelling, designers do often prefer to work with FEM software like PLAXIS and ABAQUS etc simply because they are graphical and operator-friendly.
FLAC is more powerful compared to PLAXIS since you can implement more complex constitutive models and more complex geometry through writing your own codes. ABAQUS has similar ability so it is basically a matter of convenience rather than technical issue.
in addition, 2D modelling is considered simplifying method as long as the geometry of tunnel and its surrounding lead to such conclusion. otherwise, for more detailed geometries, 3D is favored.
FLAC is preferred over finite element software to model tunnels and excavations since it can handle huge geometries and large strain problems very well comparatively. Also, generating a null model of any geometry in FLAC to represent excavation or tunnel is much easier compared to other software. The user defined subroutines using FISH is also very simple.
We generally excavate tunnels through rock mass. Rock massess consists of various features such as cracks , fissures ,etc which can be easily Incorporated in modelling using Flac especially Flac3D.Hence Flac has advantage over other FEM softwares.
All researchers explain about FLAC software. I recommend you, determine your goal in the first step. Then, you can decide that use 2D or 3D numerical simulation and also FE or FD or DE method.
Zoran's very reasonable comment leads to a question or two, since he mentions jointed rocks. Is this RS(2)3 similar to Phase 2, or a further development? Is joint behaviour really modelled as in UDEC/UDEC-BB where one can see bolt forces acting with different magnitudes across different joint planes along the same bolt? I have seen - in a court case - the combination of GSI/H-B/ Phase 2 producing significantly exaggerated so-called 'plastic' zones - for a stable tunnel. Have we progressed past this?
To answer the original question - FLAC is a finite difference program, and so approaches the solution by applying velocities. This means that almost any kind of constitutive model can be used, and it follows the stress path to failure. If you are interested in modelling geotechnical failure, or if you are interested in using sophisticated constitutive models, then FLAC is probably a good choice.
Most FE programs only allow you to use their built-in constitutive models, because how they interact with the solution algorithm is important. Some, like ABAQUS and PLAXIS do allow you to write your own code for constitutive models, but this is very difficult even for a PhD student with 3 years to do it. Using the FISH language in FLAC takes a little while to learn, but it is very powerful and gives you a lot of control. It also is easy to debug, as you can easily see where it isn't converging because there will often be massive displacements there. Often with FE programs, when there is a problem you have no results at all to help you find the problem.
PLAXIS has the advantage of being very easy to learn to use. I have had students building 3D models of an advancing tunnel within a couple of hours. FLAC will probably take up to a month to learn how to use, but it is so powerful that this can be worthwhile, particularly for researchers. The manuals are also really really good. ABAQUS is really difficult to learn... compared to FLAC or PLAXIS the manuals are horrible (feel free to disagree if you are an ABAQUS-lover). The one area I have found ABAQUS to be useful is when catastrophic failure of a ductile structure is important, e.g. ship impact into a sheet pile cofferdam where we are looking at displacements of 1-2m. PLAXIS3D has only just introduced elastoplastic shell elements, and I haven't yet had a go at testing them, but you cannot easily replicate the shape of sheet piles or their orthotropic behaviour. There are also real difficulties modelling soil in ABAQUS, particularly beneath water bodies and where there is excavation and/or fill stages, and there are no useful examples in the manuals to help guide you.
If jointed rock masses are involved, and thinking of rock slope or rock tunnel design checks, to verify/modify empirical solutions for rock reinforcement (to be used in different rock mass quality classes), then being 'old-fashioned' I do not see how one can avoid UDEC/UDEC-BB/and eventually 3DEC for important structures. Was numerical modelling supposed to be easy and cheap and basically non-representative of reality, despite all the attractive colour? I am aware of the 'economies' of e.g. continuum modelling with FLAC, or FEM/ Phase 2 or a RocScience update, and black-box H-B linked to GSI picture recognotion. But do not rock masses, and does not the practice of rock mechanics deserve something more than 'corner-cutting'. For instance, a continuum model that shows a so-called 'plastic zone' deeper than the empirical design recommendations for bolts or anchors may be far removed from the reality. Just think of all the assumptions involved in accepting the advanced algebra, which smears out the effect of a clay-filled discontinuity, or a clay-bearing joint set, giving actual anisotropic response.
Definitely, yes, one needs to carefully consider how the ground mass will behave and whether continuum analysis is an acceptable approximation. Modelling the most important/relevant aspects of ground behaviour sometimes requires the use of more sophisticated/less convenient tools. The original question was about choosing between FD or FE - but you are right that there are other choices such as DEM or even empirical methods that may be more appropriate than continuum modelling in many situations, particularly in rock.
I have some practical experience in continuum modeling of different engineering structures (tunnels, high cuttings etc.) in various rock masses.
From my (South-eastern Europe) point of view FEM modeling is still state of the art concerning practical applications, regardless of the nature of rock material involved. In combination with certain experience (read design failures) it can give reasonable prediction of actual rock-structure response.
I am wondering if we have reached an era for routine practical application of DEM (with basic knowledge in discrete element modeling, to be honest)?
What is the cost of investigations (both field and laboratory) that should be performed in order to come to reliable conclusions regarding rock parameters, fracture network, etc.?
Time needed to perform DEM analysis, software cost?
Can DEM provide "A class" prediction of rock-structure behavior (better than continuum-experience based approach with all related shortcomings) in forward analysis?
Zoran! Thanks for the stimulating discussion. I have some specific points of view due to 'advanced age' in relation to you, and therefore a chance to start early with DEM - namely UDEC-BB with the benefit of having at that time (1985 onwards) an active DEM team at NGI. So with hundreds of such models over the years you could say we have been 'biased' in favour of DEM as an automatic preference. And we have cheap ways of measuring or estimating input data: JRC, JCS, E-mass (stress-dependent) from Q, scale effects Ln/L0 etc. As for comparison with FEM (probably?) continuum analyses despite rock MASSES I invite you to take a look at our Gjøvik cavern modelling (NB et al. 1994, RG journal paper). Besides the challenge of the big cavern, you might be interested to see the results of the introductory physical models, with 20,000 blocks created by double-bladed guillotine (more detail in NB and HH, 1979, conference, RG). With the same boundary conditions: medium or high horizontal stress, the jointing orientations and continuity contra stepped (with real cohesion) jointing gave quite different results. So how often are actual rock masses behaving differently to FEM continuum? Often I would expect, especially if GSI H-B modelled. As I have asked before: where is the effect of a clay-filled discontinuity or one clay-filled joint set - in such continuum modelling?
Professor, having no arguments any more I thank you for thorough and stimulating answer. I will read mentioned references for sure and hopefully try to catch up at least small part of state of the art DEM modeling.
Every time I had some doubts needing explanations Professor M. Maksimovic used to ask "What is the first word of Quran?"
I just want to draw attention to the fact that there is also the code DISROC developed by FRACSIMA which is a Finite Element code but specially designed for fractured rocks (DIScontinuous ROCks). It makes it possible to extend the power of the FEM to the fractured media and therefore to save a lot of computation time compared to the DEM codes (Zoran’s remark on the computation time). Of course, the problem of determining the parameters of fractures and rockjoints remains intact whatever the numerical method, FEM or DEM, used. Also I agree with Benoit Jones that a numerical method, proposed in Finite Differences like FLAC, but above all applying velocities, allows to introduce any type of constitutive law. But, first, I think applying dynamic evolution concepts to quasistatic deformation can be time consuming, and, the, I don't think this is essential for modeling softening materials: DISROC has also softening constitutive models for bulk materials and rockjoints that work well and have been applied to many examples at least in research projects. In my opinion, many discussions of modeling in fractured rocks and rock masses would take a different turn if the experience of modeling with DISROC was also included. But it is true that this code is still little known.