Application of artificial neural networks to fracture analysis at the AspoHRL, Sweden: fracture sets classification

This study investigates the potential of artificial neural networks (ANNs) to recognize, classify and predict patterns of different fracture sets in t he top 450m in crystalline rocks at the Aspo Hard Rock Laboratory (HRL), So utheastern Sweden. ANNs are computer systems composed of a number of proces sing elements that are interconnected in a particular topology which is pro blem dependent. ANNs have the ability to learn from examples using differen t learning algorithms; these involve incremental adjustment of a set of par ameters to minimize the error between the desired output and the actual net work output. Six fracture-sets with particular ranges of strike and dip hav e been distinguished. A series of trials were carried out using backpropaga tion (BP) neural networks for supervised classification. and the BP network s recognized different fracture sets accurately. Self-organizing neural net works have been used for data clustering analysis with supervised learning algorithms; (competitive learning and learning vector quantization), and un supervised learning algorithms (self-organizing maps). The self-organizing networks adapted successfully to different fracture clusters (sets). A set of trials has been carried out to investigate the effect of changing the ne twork's topologies on the performance of the BP networks. Using two hidden layers with tan-sigmoid and linear transfer functions was beneficial for th e performance of BP classification. ANNs improved fracture sets classificat ion that was based on Kamb contouring method with constraint on areas betwe en fracture clusters. (C) 2001 Published by Elsevier Science Ltd.