Parallel distributed processing in rock engineering systems

摘要

Rock Engineering Systems are a collection of ideas, mathematical tools and computerudtechnology all of which are designed to solve problems in rock engineering withudinteracting components. The interactions between components can be complex and theudrock engineering problems themselves contain a high degree of uncertainty. Theudresearch described in this thesis investigates the incorporation of computationaludtechniques known as parallel distributed processing methods into the disciplines ofudrock mechanics and rock engineering. Two main applications of parallel distributedudprocessing methods in rock engineering are investigated in this thesis.ud1) Multilayered perceptron artificial neural networks are used successfully toudencapsulate the laboratory behaviour of rocks under triaxial compression. Trainedudartificial neural networks are then used to replace conventional constitutive modelsudwithin finite difference geomechanical numerical modelling codes.ud2) Two multilayered perceptron artificial neural networks are developed to assistudin the task of discrimination of rock fracture presence within digital imagery of rockudexposures. The first is trained using samples of the image that contain fracture imageudcontent and samples that do not, and provides a probability-like measure of fractureudpresence. It was sufficiently successful to permit estimation of fracture intensityudparameter , . The second was developed specifically to identify fracture terminationudcondition by matching samples to a set of fracture termination condition templates.udSeven original contributions to the rock mechanics and rock engineeringuddisciplines have resulted across the three application areas. These contributions areuditemised, with details, at the beginning of the final Chapter of the thesis.