Student Number: 0420801P Master of Science in Engineering. Faculty of Engineering and the Built Environment / Underground observations at a coal mine indicated failure of the immediate roof above the bords while pillars were observed to be intact. To determine the underlying causes of roof failures, careful observations and photographic recording of occurrences of roof failures have been made. Rock samples of the immediate shale roof were collected for laboratory testing to determine the rock strength and deformation properties. Numerical simulations were carried out to analyse stress and strain distributions and also to attempt to explain the guttering process. Mapping of roof failures showed that these took place mainly towards the centre of the roadways. The roof failures, termed “roof guttering”, were observed to occur violently and with little warning. Occurrence of roof guttering had a negative impact on production. Some panels are abandoned, production times have increased and safety of workers is compromised. The mine authorities initially thought that roof guttering was caused by shear failure of the rock mass. Roof bolts are therefore used as a means of primary support. No improvements have been observed. Increasing the size of pillars has not solved the problem either. It has only increased the amount of coal left in the pillars without any improvements in reducing roof failures. Stress measurement results carried out in 2001 showed that high horizontal stresses exist at the mine. The immediate shale roof was observed to be weak. Laboratory testing showed that the shale rock is transversely isotropic. Numerical modelling results indicated that there are insignificant stress concentrations towards the centre of the roadway using the elastic and transversely isotropic elastic models. Stress concentrations were predicted at the roof-pillar contact area. It is therefore expected that failure should initiate and occur at the roof-pillar contact area. The Mohr-Coulomb and Mohr-Coulomb strain softening models predicted shear failure at the roof-pillar contact area. The two models over predicted the depth and under predicted the width of failures. The extension strain criterion predicted correctly the depth and width of failures although the failures were predicted at the roof-pillar contact area while the observations indicated failure mainly towards the centre of the roads. Initiation of failure was predicted ahead of the coal face at the centre of the road position using the extension strain criterion. Although none of the constitutive behaviours predicted correctly the observed underground failures the extension strain criterion has shown the best agreement. Guttering that occurred at the roof-pillar contact was modelled successfully using the extension strain criterion. The extension strain criterion predicted initiation of failure ahead of the coal face at the road centre position. It is possible that fracture initiation could be taking place in this location ahead of the coal face, and, on blasting the rock that has been fractured falls forming a gutter at the centre of the road.