This paper reviews the feasibility studies conducted for a new research theme that establishes a methodology for the production of adequate dynamic empirical models of the deformation response of rocks and soils to engineering. The key to the new methodology is an automated capability for back analysis of rock engineering problems. The proposed methodology couples genetic algorithms and the ideas of evolutionary computing to existing numerical modelling methods. The possible area of application of the proposed methodology within rock engineering is wide, ranging from the analysis of laboratory testing, to in-situ measurement of rock mass properties, to automated routine condition monitoring of rock engineering structures during the post-construction phase.


Ce papier revait les etudes de faisabilites conduites pour un nouveau theme de recherches qui etablit une methodologie pour la creation de "modeles empiriques dynamiquement adequat" de la deformation des roches et sols pour l'ingenirie. La clede la nouvelle methodologie proposee est unw capacite automatisee pour I'analyse retrospective de problems d'ingenirie. La methodologie proposee couple des algorythmes genetiques avec les idees d'amelioration evolutionaire des ordinateurs concernant les methodes pour modeller numeriquement qui existent. L'itendue possible d'application de la methodologie proposee avec I'ingenirie des roches est vastes, allont de I'analyse de tests en laboratoire, jusqu' à la measure des proprietes de masses rocheases, et la condition routine automatisee surveillance de structures d'ingenirie des roches pendant la phase de post-construction.


Diese Arbeit bietet einen Rueckblick auf eine Studie, die ueberdie Anwendbarkeit eines neuen Forschungsthemas durchgefuehrt wurde, welches eine Methodologie fuer die Produktion von adaquaten, dynamischen, empirischen Modellen ueber die Deformations-Reaktionen von Fels und Erde entwickelt. Der Schluessel zu dieser neuen Methodologie ist die Fahigkeit fuer eine automatisierte Rueckanalyse Von Problemen im Felsbau. Die angewandte Methodologie fuegt genetische Algorithmen und die ldeen von evolutionarer Berechnung zusammen zu existierenden numerischen Modellmethoden. Die Anwendungsmöglichkeiten dieser Methodologie im Felsingenieurswesen ist weit: Von der Analyse in Laborversuchen ueber Messungen von Gebirgseigenschaften in situ, zur automatisierten Übersicht von Routinekonditionen in Felsbau Strukturen wahrend der Postkonstruktionsphase.


In rock engineering the basic design objectives are to understand how any given rock mass will behave when an engineering structure is placed on, in or within it. Having identified the likely response of the in situ rock mass when the structure is introduced, the next phase of the rock engineering process is to eliminate those aspects of the predicted response that are considered unfavourable, i.e. to engineer the rock mass so that its performance is satisfactory throughout construction, commissioning, its service life, and decommissioning. This paper discusses how a technique termed evolutionary computing coupled with numerical methods leads to a new tool capable of automating back-analysis. This tool enables the in situ response of a rock mass to be decalibrated to produce high quality models that can be used in design. Examples of the automated back-analysis are given and extensions of the basic ideas are presented.


Back-analysis procedures are familiar to most practising rock engineers. The essence of back analysis is best explained through contrast with engineering design. In engineering design the performance of an engineered structure in rock is predicted on the basis of a design model with known values for the parameters of this model and any initial or boundary conditions. In back analysis, a performance history is known and the values for the design parameters, or boundary or initial conditions of the same model are unknown and are sought. The values for these sought after quantities can be collectively termed the solution set.

This content is only available via PDF.
You can access this article if you purchase or spend a download.