Analysis and modelling of mining induced seismicity

[摘要] Earthquakes and other seismic events are known to have catastrophic effects onpeople and property. These large-scale events are almost always preceded by smallerscaleseismic events called precursors, such as tremors or other vibrations. The use ofprecursor data to predict the realization of seismic hazards has been a long-standingtechnical problem in different disciplines. For example, blasting or other miningactivities have the potential to induce the collapse of rock surfaces, or the occurrenceof other dangerous seismic events in large volumes of rock. In this study, seismicdata (T4) obtained from a mining concern in South Africa were considered usinga nonlinear time series approach. In particular, the method of surrogate analysiswas used to characterize the deterministic structure in the data, prior to fitting apredictive model.The seismic data set (T4) is a set of seismic events for a small volume of rock in amine observed over a period of 12 days. The surrogate data were generated to havestructure similar to that of T4 according to some basic seismic laws. In particular,the surrogate data sets were generated to have the same autocorrelation structureand amplitude distributions of the underlying data set T4. The surrogate dataderived from T4 allow for the assessment of some basic hypotheses regarding bothtypes of data sets.The structure in both types of data (i.e. the relationship between the past behaviorand the future realization of components) was investigated by means of three teststatistics, each of which provided partial information on the structure in the data.The first is the average mutual information between the reconstructed past and futuresstates of T4. The second is a correlation dimension estimate, Dc which gives anindication of the deterministic structure (predictability) of the reconstructed statesof T4. The final statistic is the correlation coefficients which gives an indicationof the predictability of the future behavior of T4 based on the past states of T4. The past states of T4 was reconstructed by reducing the dimension of a delay coordinateembedding of the components of T4. The map from past states to futurerealization of T4 values was estimated using Long Short-Term Recurrent Memory(LSTM) neural networks. The application of LSTM Recurrent Neural Networks onpoint processes has not been reported before in literature.Comparison of the stochastic surrogate data with the measured structure in theT4 data set showed that the structure in T4 differed significantly from that of thesurrogate data sets. However, the relationship between the past states and thefuture realization of components for both T4 and surrogate data did not appear tobe deterministic. The application of LSTM in the modeling of T4 shows that theapproach could model point processes at least as well or even better than previouslyreported applications on time series data.

[发布机构] Stellenbosch University