Until very recently, rock mechanics engineers made very little use of 3D simulations. Designers preferred so-called equivalent 2D runs to rough out problems before using their engineering judgement. The availability of a new FEM code called VIPLEF 3D gave the authors an opportunity to call in question several widespread approaches. Several classic layouts of leached salt caverns are presented and for each case, the authors discuss the contribution of 3D simulation.
Jusqu'à un passe très recent, les mecaniciens des roches utilisaient peu les simulations 3D. Les concepteurs preferaient donc effectuer des simulations 2D jugees intuitivement "equivalentes" pour degrossir Ies problèmes avant d'utiliser leur jugement d'ingenieur. La mise à disposition d'un nouveau code de calcul, baptise VIPLEF-3D, a permis aux auteurs de remettre en cause certaines approximations couramment utilis∼es jusqu'alors. C'est ainsi que plusieurs configurations "classiques" de cavites salines sont presentees. Dans chacun de ces cas, les auteurs examinent l'apport des simulations 3D par rapport aux approches les plus repandues.
Bis vor kurzem benutzten die Felsmechaniker die 3D-Simulation kaum, gaben die Konzepteure deshalb zweidimensionalen und intuitiv als "gleichwertig" gewerteten Simulationen den Vorzug. Die Bereitstellung eines neuen Berechnungscodes, VIPLEF-3D genannt, hat es den Autoren ermöglicht, gewisse und bislang allgemein eingesetzte Approximationen in Frage zu stellen. Aus diesem Grund stellen wir mehrere "klassiche" Salzhohlraum-Konfigurationen. Im Vergleich zu den verbreitesten Methoden pruefen die Autoren fuer jedes Beispiel den Beitrag der 3D-Simulation.
FEM (Finite Element Method) appeared in the Sixties, started to be taught to young engineering students in the seventies, and had a continuous development since then in parallel with the drastic evolution of computer performance. Its use in rock mechanics did follow that trend with the limitation that so-called exact modelisation has never been expected due to the difficulties in obtaining realistic input data. The word numerical simulation is then preferred to modelling. At the same time, description of the rockmass also reflected constant progress using both empirical methods based on hundreds of case histories and more rational approaches based mainly on laboratory or in-situ testing. To the question "has rockmass description made enough progress to justify fully computerized design for underground construction ?" our answer would be "no", but we are trying in this study to give some answers to the question: "has numerical simulation techniques made enough progress to justify one step further toward accuracy in using 3D capacities ?".
Numerical simulation progress in stress and strain distribution started in fact with 1D solution (Kirsch, Lame, etc.). Analytical solutions were given long ago for cylindrical tunnel and shaft or for spherical isolated storage caverns. During many years, rock engineers tried to solve their 2D problem by comparing them to known 1D conditions. Early developments of convergence-confinement method give typical examples. Nowadays we have moved one step further toward geometrical accuracy, most design firms use regularly 2D numerical simulation techniques but still try to avoid 3D codes. Some typical cases are presented in this paper, they are all dealing with rock salt behaviour. Its almost isotropical behaviour and the isotropical initial stress conditions linked to its high plasticity make leached caverns more susceptible of axisymmetrical simulations. Some limitations of this current practice are given below. Case I: Elliptical shape. In a project where rock salt was vertically layered with non soluble beds we had the feeling that the leached shape could result rather elliptical than cylindrical. Classical approach using 2D codes would analyse this dissymmetry with various runs on vertical (axisymmetry) and horizontal (plane strain) section. Case 2: Field of cavities. When several, systematically set, parallel leached caverns are considered, the usual assumption states that each cavern and pillar behaviour is to be considered in between two extreme cases. A middle field case of a cavern surrounded by an infinite number of others and a border field case where the cavern is supposed to be isolated from others in an infinite medium(l).