A new method to identify impending failure in rock slopes

Assessing when an unstable slope is at the point of critical equilibrium is one of the main points of research and discussion in the field of rock mechanics. The topic has great relevance especially in the mining industry, as significant economic benefits derive from the ability to safely prolong works in areas where rock deformation is underway. Mining near an unstable slope requires strong confidence that a failure of the excavated area will not happen in the relatively immediate future; achieving this goal determines a more efficient and profitable extraction of the mineral resources.1 An effective monitoring program, able to provide notice of slope instability through the accurate and timely measurement of precursors to failure, clearly represents an essential benefit for the safety and productivity of the mine operation. Adequate anticipation of events of slope failure allows mine operators to plan and implement response actions with sufficient advance to minimize the effects of the failure on personnel safety and mine productivity. As a consequence, in most of the large surface mine operations around the world, extensive slope monitoring programs are undertaken nowadays as part of the mine performance monitoring system, by integrating various instruments such as slope stability radar, robotic total stations and geotechnical sensors.2 Detailed datasets of surface and underground displacements are thus collected and their analysis can provide valuable information for the understanding of the behavior of rock slopes approaching failure. Once accurate monitoring data are acquired in near-real time, the most challenging task for the site staff in charge of risk management is the set-up of suitable alarms representing when slope failure is impending.3 Without entering into their details, a number of “phenomenological” failure criteria (i.e. based solely on datasets of displacement measurements versus time)4 have been proposed in the past to forecast the time of slope failure5, 6, 7, 8, 9 and 10; among these the inverse velocity method, derived from the accelerating creep theory, is the most common tool used to predict the time of failure of progressively accelerating slopes. Failure criteria often provide very useful descriptions of the risk associated with the ongoing deformation, but are also characterized by several limitations. Most notably, universal laws used to describe the displacements of failing slopes do not take into account the specific physical aspects of the phenomenon under investigation, such as the mechanical properties of the material and the influence that these have on the development of the landslide.4 With reference to the inverse velocity method another important limitation is that this assumes that velocity at failure is infinite, whereas the velocity of slopes is evidently never infinite. It follows that failure-time predictions must be regarded just as general estimations and that the inverse velocity method (and failure criteria in general) should be used with caution9 and 11; the margin of error (i.e. the time difference between actual and predicted failure) can in fact range from few hours up to several days.12 and 13 In other cases predictions cannot be performed with adequate confidence. As a result, the issue of determining when slope failure may be impending is still of great concern. According to a different approach, other methods are instead based on the review of databases of failure case histories in order to identify characteristic conditions for slope failure occurrence.1, 14 and 15 Rather than providing failure-time predictions, the aim is to define recurrent correlations between certain variables in close proximity to the instant of failure. In the framework of the ACARP (Australian Coal Association Research Program) C17023 project, Cabrejo and Harries16 analyzed a large database of deformation data acquired by Slope Stability Radar (SSR) devices in several undisclosed Australian open-cut coal mines and reviewed 78 case histories of mine slope failure, which were all anticipated by progressive accelerations. Parameters associated to both displacement and velocity at different stages of the failure process were considered by the authors, but reliable mathematical expressions able to comprehensively characterize the observed events could not be found. In this work further in-depth analysis of this database is presented. In particular, the average accelerations during different sub-sets of time prior to the instant of failure have been studied and highlighted the presence of a common behavior of the slope failures in the database.