Search

Geological and geophysical interpretation is characterised by large and localised datasets that are extremely expensive to acquire. There are clear advantages in applying deep learning techniques to such datasets, but this requires a large amount of suitable data for effective training. Creation of training data can be time-consuming, but novel countermeasures such as Generative Adversarial Networks (GANs) are an enticing alternative with the potential to alleviate this issue by providing synthetic data that is representative of the real data. This paper details an investigation of the potential of GANs for the augmentation of labelled seismic facies from a mass transport complex in training a convolutional neural network (CNN) for facies classification. The study adopts a specific GAN approach, known as the Conditional Style-Based Logo GAN (LoGANv2), because of its capability to generate conditioned output with improved training stability. By using LoGANv2 to synthesise examples that mimic the behaviour of the real data, based on a 3D seismic dataset from the North Carnarvon Basin in Australia, the accuracy of the traditional CNN for facies classification improved from 38% (the benchmark result where no augmentation was applied) to 52%, 61%, 94% and 74% for four trials with different degrees of augmentation. The results show that increasing training size, either through manual annotation (trial 2) or standard manipulations such as image rotation (trial 3 and trial 4), improves the performance of LoGANv2 and hence the CNN classification. The experiments indicate that although standard manipulation can increase the diversity of the training dataset, the balance between diversity and consistency within the datasets is also important, and this should be optimized for LoGANv2 to achieve better performance. In addition, we explored the adaptability of the proposed LoGANv2-CNN approach to unlabelled samples from another survey to demonstrate the robustness and flexibility of the model. • A LoGANv2-assisted CNN model was developed for augmentation and classification of seismic facies. • With LoGANv2 proving its capacity to augment different types of seismic facies with minimal manual labelling, the synthetic facies generated and the real ones are passed down to a CNN for further classification. • The streamlined process demonstrates that, with little manual annotation, the generation of synthetic seismic facies could lead to a better understanding of the underlying geological structures. [ABSTRACT FROM AUTHOR]

Destress blasting is regarded as an active mitigation measure for preventing rockbursts in deep underground works, whose mechanism and effectiveness, however, have still been in debate. A self-developed combined finite-discrete element method software is implemented in this study to first model the rock fracture process induced by a single-hole destress blasting based on a practical blast in a deep gold mine with high in-situ stresses. The effects of accelerometer holes and bedding planes are discussed. Then, a two-round tunnel face excavation in deep underground with high in-situ stresses is modelled, which reproduces the occurrence of the excavation-induced rockbursts. Finally, destress blast-holes are introduced into the same model prior to the two-round tunnel face excavation to investigate the effect of the destress blasting on the excavation-induced rockbursts. It is found that the fractures induced by the destress blasting shift the high abutment pressure induced by the excavation away from the tunnel face and then protect the tunnel wall from fracturing. The numerical modellings reveal that the success of the destress blasting in controlling excavation-induced rockbursts depends on not only the site condition but also the design of the destress blasting such as the arrangement of the blast-holes and the explosive charges. [ABSTRACT FROM AUTHOR]

Rockburst is one of the severe geotechnical problems that may occur while tunnelling under high in-situ stresses. Despite numerous studies in the relevant literature, the evaluation and prediction of the rockburst still remain a common challenge worldwide. In this paper, a self-developed combined finite-discrete element method (FDEM) is used to numerically investigate the rockbursts in deep tunnelling under high in-situ stresses. The rockbursts in the drainage tunnel at the Jinping II Hydropower Station are modelled and the effects of the geological and geotechnical characteristics of the site on the occurrence of rockbursts are investigated. The FDEM numerical modelling vividly simulates the fracture initiation and propagation, as well as the fragment expulsion, ejection and flyout resulting in the rockburst process that could be difficult to capture on the site or via the conventional continuum modellings. The effect of the pre-existing fault on the site is studied, and the critical roles of the location and the dip angle of the fault in determining the rockburst development around the tunnel are highlighted. Then, the effects of in-situ stresses on rockburst development are investigated by applying different lateral pressure coefficients in the model. It is found that based on the Jinping site condition, a low lateral pressure coefficient could contribute to the alleviation of the rockburst around the tunnel but the alleviation effect is relatively limited, while a high lateral pressure coefficient could result in the rockburst that breaks more rock masses on the tunnel surface. Moreover, the influences of tunnel shapes on the rockburst incidents are also studied by further modelling rockbursts around tunnels in square shape and horseshoe shape. It is found that the square shape is inadequate in controlling rockbursts at the roof and floor areas, and the most suitable tunnel shape to resist rockbursts in such field condition is the circular shape. By vividly reproducing the rockburst processes in deep tunnelling under high in-situ stresses and elaborating the effects of several critical geological and geotechnical characteristics on the rockburst development around the tunnel, this study contributes to the understanding of the associated rockburst mechanisms and is also of guiding importance for certain tunnelling designs and constructions in practice. It is also expected that the proposed method can be used to evaluate and predict the rockbursts in many other scenarios in future studies. [ABSTRACT FROM AUTHOR]

• Rock blasting was simulated by a unique self-developed GPGPU-parallelized FDEM. • Dynamic interactions of in-situ stresses, blast loading and fracturing were modelled. • Rock fracture and fragmentation process was properly replicated. • Effects of in-situ stress field and lateral pressure coefficient were clarified. • Effects of blasting decay time ratio and decoupling ratio were elucidated. One highly effective approach for tunnelling in rock at deep depth is blasting. The damage and in-situ stress redistribution in the surrounding rock mass induced by the blast loading during this process is, however, unavoidable. In this study, a powerful GPGPU-parallelized combined finite-discrete element method is implemented to study the damage evolution during controlled contour blasting in the bench of a deep-buried tunnel. The proposed method is characterized by the simulation of the blasting-induced pressure variation via the pressure-gas volume history curve and the modelling of the transition from continuum to discontinuum behaviour in the surrounding rock mass. The in-situ stress distribution, the blasting-induced stress wave propagation, and the corresponding rock fracture and fragmentation process are modelled and analysed. The numerical simulation results indicate that the in-situ stresses play a key role in the damage evolution around the tunnel, strongly influencing the stress redistribution pattern and thus the fracture initiation and propagation around the tunnel during blasting. Several different in-situ stress regimes are considered and discussed, revealing the dominating effects of the major principal stress and lateral pressure coefficient on the rock damage behaviour. Moreover, the blasting results obtained from employing the pressure–time history curve with varying decay time ratios and decoupling ratios are also studied. Longer decay time ratios and higher decoupling ratios induce additional rock fracture and fragmentation, indicating that the model can enable the selection of stemming and explosive type for effective breakage with limited damage around a tunnel. [ABSTRACT FROM AUTHOR]