The purpose of this paper is to introduce the idea of representing all the mechanisms operating in a rock engineering project by a parallel processing model with appropriate supporting modules. The paper implements the Rock Engineering Systems approach with a hybrid neural network and expert system, which consists of a dada base and associated processing, backpropagation neural network, expert system and Hopfield neural network. An application to rock cavern performance auditing is illustrated
Le but de cet article est d'introduir l'idee qui represente tout les mecanisms fonctionnant dans Ie projet de genie de roche, en utilisant une modèle parallele avec des modules supportant convernables. L'article montre une approche de systèm de genie de roche avec un hybride neurale et un systèm d'expert qui comprend une base des donnees avec un processus concerne, une reseau neurale de backpropagation, un systèm d'expert et la reseau neurale de Hopfield. Une application de verification du fonctionnement cavernous a ete inllustree.
Der Zweck dieser Arbeit ist die Vorstellung der Idee fuer diè Darstellung allen laufenden Mechanismen in Felsingenieurprojekt durch Parallelvorgangs modell mit passenden Unterstuetzungsmoduln. Die Arbeit Iiefert das Felsingenieursystem eine Methode mit Hybridnervalen Netzwerk und Expertensystem, welche aus Datenbase, verbundenen Vorgange, Rueckverbreitung bzw. Hopfield nervalen Netzwerk und Expertensystem bestehen. Eine Anwendung zur Pruefung der Höhlenleistung wurde erlautert.
Rock engineering activities, such as cavern excavation and slope cutting, generally disturb the natural state of a rock mass and generate interactions between interrelated parameters of the rock mass and the boundary conditions. Such interaction is a dynamic and cyclic process. Initial changes in any rock mass parameter or boundary condition caused by external disturbances induce changes of other parameters, and then these induced changes feed back to induce further changes of the others via a multi-pass loop, until a new equilibrium state is reached (Hudson, 1992). The interaction mechanisms and the dynamic interaction process between the altered parameters are the keys in auditing the performance of a rock engineering project. Because rock masses are complex discontinuous geo-systems involving many geological factors, it is difficult to model the interaction mechanisms and all the interaction process with conventional approaches. Responding to this problem, a systematic view and a total systems approach, termed Rock Engineering Systems (RES), was proposed by Hudson (1991, 1992). This approach studies rock engineering projects within a systems perspective, and has the potential to solve the rock engineering problems as a completely integrated system. The RES approach is an overview methodology. It needs to be fully implemented and computerised with appropriate computational techniques. This paper utilises a neural network and knowledge based expert system for this purpose, with specific example application to rock cavern performance auditing.
The RES approach treats the rock mass, boundary conditions and engineering activities as a complete, interactive and dynamic system. It firstly establishes the engineering objective, then considers all the potentially relevant parameters and their relations, and subsequently develops an appropriate model. Within this procedure, the parameters considered must be objective-orientated because any rock engineering system is developed for a specific engineering purpose. Within the rock engineering system, the constituent parameters of rock mass, boundary conditions and engineering activities interact, through interaction mechanisms, to produce a variety of complex characteristics and dynamic behavioural modes of the rock mass. The basic device for representing the interaction is the interaction mechanism matrix. It represents the key parameters or variables as leading diagonal terms and their interaction mechanisms as off-diagonal terms.