Rita L. Sousa

Mathematical algorithm development and parametric studies with the GEOFRAC three-dimensional stochastic model of natural rock fracture systems

This paper presents results from research conducted at MIT during 2010-2012 on modeling of natural rock fracture systems with the GEOFRAC three-dimensional stochastic model. Following a background summary of discrete fracture network models and a brief introduction of GEOFRAC, the paper provides a thorough description of the newly developed mathematical and computer algorithms for fracture intensity, aperture, and intersection representation, which have been implemented in MATLAB. The new methods optimize, in particular, the representation of fracture intensity in terms of cumulative fracture area per unit volume, P32, via the Poisson-Voronoi Tessellation of planes into polygonal fracture shapes. In addition, fracture apertures now can be represented probabilistically or deterministically whereas the newly implemented intersection algorithms allow for computing discrete pathways of interconnected fractures. In conclusion, results from a statistical parametric study, which was conducted with the enhanced GEOFRAC model and the new MATLAB-based Monte Carlo simulation program FRACSIM, demonstrate how fracture intensity, size, and orientations influence fracture connectivity. Display Omitted The GEOFRAC 3D model applies sequential stochastic processes to represent rock fracture systems.We propose new algorithms for modeling fracture intensity, size, aperture, and connectivity.Poisson-Voronoi Tessellation of planes into polygons optimizes fracture intensity modeling.A statistical parametric study shows how fracture intensity and size might affect connectivity.

Warning Systems for Natural Threats

This paper is a somewhat revised version of the one previously given as a keynote lecture in the ECI Conference, ‘Geohazards: Technical, Economical and Social Consequences’. Some changes have been made to the original paper and the oral presentation: several of the figures and pictures illustrating the recent events have been eliminated, only one decision analysis example is shown and the review of literature on other warning systems has been expanded. Readers interested in the conference paper and oral presentation are referred to the website: http://services.bepress.com/eci/geohazards/

Computational Modelling of Fracture Propagation in Rocks Using a Coupled Elastic-Plasticity-Damage Model

A coupled elastic-plasticity-damage constitutive model, AK Model, is applied to predict fracture propagation in rocks. The quasi-brittle material model captures anisotropic effects and the distinct behavior of rocks in tension and compression. Calibration of the constitutive model is realized using experimental data for Carrara marble. Through the Weibull distribution function, heterogeneity effect is captured by spatially varying the elastic properties of the rock. Favorable comparison between model predictions and experiments for single-flawed specimens reveal that the AK Model is reliable and accurate for modelling fracture propagation in rocks.

Exploration analysis for landslide risk management

Risk management for landslides involves, in addition to active and passive countermeasures, the collection of information through exploration. With such information, it is possible to reduce uncertainties, make more reliable decisions and therefore reduce risk. This paper addresses two types of exploration, namely (1) exploration (information collection) after the decision to obtain additional information is made, which is the standard type of exploration, and where one uses the new information to update risk, and (2) exploration (information collection) before the decision to obtain additional information is made, in which one conducts ‘virtual exploration’ to establish if exploration is worthwhile. The paper shows that both types of exploration can be assessed using either decision trees or Bayesian networks. Both approaches were applied to the Waltons Wood Landslide in England using infinite slope analysis and produced consistent results. The interested user can, based on what is documented in the paper, select and apply either one or both of the approaches.

Probabilistic identification of subsurface gypsum geohazards using artificial neural networks

Subsurface gypsum dissolution hazards imply risks to the construction and operation of new transport infrastructure including subsidence, cavity collapse and cavity flooding. This is a concern in Abu Dhabi, United Arab Emirates, where gypsum geohazards are observed and an extensive transportation network is planned. This paper proposes an artificial neural network (ANN)-based approach for the prediction of underground gypsum. Moreover, the approach is developed to provide the expected probability of gypsum presence and to generate gypsum hazard maps. Such maps provide both a general planning instrument and an input for the decision support systems. An application to Masdar City, Abu Dhabi, is discussed at the site of a planned metro line. Twenty-one boreholes are used to train and validate the ANN that is used to produce a 3D geological model identifying the expected presence of gypsum. Most significantly, the application illustrates how gypsum hazard maps can be obtained at any required depth providing planners and designers with essential information for risk assessment and management.

Experimental Study of the Mechanics of Gypsum Seam Hazard for Abu Dhabi

Abu Dhabi is predicting a huge growth in population over the next 20 years (Plan Abu Dhabi 2030); further, it seeks to become an international destination for tourists, businesses, and investment while protecting its cultural heritage. A crucial aspect of achieving this goal is the development of large integrated transportation system, underground, and above ground, to ensure Abu Dhabi becomes a sustainable city on a global scale. The presence of gypsum rocks that occurs within Abu Dhabi’s bedrock is a major threat to underground construction and understanding the phenomena is of paramount importance. They are persistent quasi-horizontal bands, at different levels (top level between 10 and 15 m and bottom level between 15 and 25 m), prone to volume change by dissolution or swelling, due to changes in the stress regime and water chemistry and flow. The dissolution of gypsum is also a cause for cavities that can be found within this formation in greater Abu Dhabi. In this paper, the description and results of an experimental study aimed at obtaining a better understanding of the gypsum dissolution process, as well determine factors affecting it, are presented. Tests on the dissolution process of gypsum rock were performed using artificially created intact and fractured gypsum samples which are a representative of the collected in situ fractured gypsum rock samples obtained from Abu Dhabi. The samples are subjected to flow-through tests. Results obtained show that for an initially saturated gypsum specimen, there is a sharp decline in concentration with time (Stage I), followed by a constant concentration (Stage II) before a slight gradual increase is observed (Stage III) with time. This is a fundamental study—part of a larger set of experiments studying the gypsum dissolution process in Abu Dhabi. Using the data collected from the field and the experiments mentioned above, gypsum geohazard risk-related maps which reflect subsidence, swelling, cavity collapse, and cavity flooding associated with Gypsum Karst shall be developed using Geographic Information Systems.

Exploration and decision-making in geotechnical engineering – a case study

ABSTRACT The paper presents methodologies for exploration planning under uncertain conditions based on virtual exploration and Bayesian updating. The process starts with site characterization using existing information to produce geologic profiles. Initial distributions of cost and time are obtained with a Bayesian network model that optimizes the construction strategy for particular geologic conditions. This is followed by the unique step to determine with a “virtual exploration” if additional exploration (e.g. borings) is warranted, and if so, where it should be best done. All this is then applied to the planned Abu Dhabi subway tunnels providing the transportation planners with necessary information for planning and design.

The Use of Data Mining Techniques in Rockburst Risk Assessment

Abstract Rockburst is an important phenomenon that has affected many deep underground mines around the world. An understanding of this phenomenon is relevant to the management of such events, which can lead to saving both costs and lives. Laboratory experiments are one way to obtain a deeper and better understanding of the mechanisms of rockburst. In a previous study by these authors, a database of rockburst laboratory tests was created; in addition, with the use of data mining (DM) techniques, models to predict rockburst maximum stress and rockburst risk indexes were developed. In this paper, we focus on the analysis of a database of in situ cases of rockburst in order to build influence diagrams, list the factors that interact in the occurrence of rockburst, and understand the relationships between these factors. The in situ rockburst database was further analyzed using different DM techniques ranging from artificial neural networks (ANNs) to naive Bayesian classifiers. The aim was to predict the type of rockburst—that is, the rockburst level—based on geologic and construction characteristics of the mine or tunnel. Conclusions are drawn at the end of the paper.