Automated, real-time geohazard assessment in deep underground mines

Abstract

We introduce an automated, real-time geohazard assessment system designed specifically for underground mining. The system, which we call Geoscience INTEGRATOR, is based on the quantitative 4D geohazard computational strategy that we previously developed and reported on at the Seventh International Conference on Deep and High Stress Mining in Sudbury, Canada, in September 2014. The computational strategy relies on modelling multiple, independent hazard criteria related to geology, structure, rock mass condition, stress, seismicity, mine geometry, and support on the rock interface where the hazards occur, and building a statistical model that relates past hazard occurrence to the state of those criteria at the time of the hazard (McGaughey 2014). Further research and development has been carried out since then with the objective of developing an operational computational infrastructure that enables practical geohazard assessment using this approach. We have developed a system for automatically assembling and managing the required inputs, managing the geohazard assessment computational processes, and providing meaningful reports to mine operators. This has been accomplished within three recent Canadian research programs managed by the Centre for Excellence in Mining Innovation (CEMI): 'Smart Underground Monitoring and Integrated Technologies' (SUMIT), 'Mining Observatory Data Control Centre' (MODCC), and 'Ultra-Deep Mining Network' (UDMN). The initial system design objective under the SUMIT program was a capability to "provide researchers with user-friendly access to mine datasets and their contextual information to facilitate and optimise research efforts." The system was further developed in the ensuing projects to provide a powerful 3D visualisation interface and to support, track, and automatically report on 4D dynamic mine models and associated geohazards in a deep operating mine context. It is now being deployed to manage multi-institution, multi-site data from both SUMIT and other major research projects in addition to implementations at operating mine sites to improve rockburst and other types of geotechnical hazard assessment.

Land Subsidence Susceptibility Mapping Using Bayesian, Functional, and Meta-Ensemble Machine Learning Models

“…Recently, there has been great interest within the hazard prediction community toward improving the performance of hazard susceptibility models. In various fields, machine learning techniques have been shown to be effective in terms of performance [58][59][60][61][62]. In particular, ensemble learning has improved machine learning results by combining several models [17,63,64].…”

Data-driven geotechnical hazard assessment: practice and pitfalls

“…After the modelling engine updates the values on the mine model as required, the rules developed in the data-driven training are applied and new values of hazard are computed and reported. The system (Figure 8) for operational deployment was discussed at greater length in McGaughey et al (2017).…”

“…The phrase 'data-driven geotechnical hazard assessment' in the title of this paper means using the history of geotechnical hazard incidents at a site to predict the likelihood in location and time of future incidents of the same class. We have undertaken such studies at many sites for many types of geotechnical hazard, and incorporated the methods into a working system for automated hazard assessment and reporting now implemented at several operations (McGaughey 2014;McGaughey et al 2017). The subject of this paper is not the operational systems or methods of automated hazard assessment and reporting, but rather what we have learned in terms of the practices and pitfalls of applying data-driven methods to the problem of geotechnical hazard assessment.…”

“…A 'black box' prediction will not provide the required engineering confidence in the output. Furthermore, the system must provide a foundation for a routine, automated computation and reporting system to update hazard assessment in response to new data as described in McGaughey et al (2017).…”

Exploring the Fusion Potentials of Data Visualization and Data Analytics in the Process of Mining Digitalization

“…For hazard prediction and forecasting, time-based prediction is crucial to understanding, assessing, and acting upon mining geomechanical risks [78]. Researchers have studied 4D data analytics for hazard prediction in the mining industry and proposed a system that covers the 4D data management module and the Weight of Evidence (WoE) algorithm-based data analytics module [16], [79], [214].…”