Rockburst is a common engineering geological hazard. In order to evaluate rockburst liability in kimberlite at an underground diamond mine, a method combining generalized regression neural networks (GRNN) and fruit fly optimization algorithm (FOA) is employed. Based on two fundamental premises of rockburst occurrence, depth, σθ, σc, σt, B1, B2, SCF, Wet are determined as indicators of rockburst, which are also input vectors of GRNN model. 132 groups of data obtained from rockburst cases from all over the world are chosen as training samples to train the GRNN model; FOA is used to seek the optimal parameter σ that generates the most accurate GRNN model. The trained GRNN model is adopted to evaluate burst liability in kimberlite pipes. The same eight rockburst indicators are acquired from lab tests, mine site and FEM model as test sample features. Evaluation results made by GRNN can be confirmed by a rockburst case at this mine. GRNN do not require any prior knowledge about the nature of the relationship between the input and output variables and avoid analyzing the mechanism of rockburst, which has a bright prospect for engineering rockburst potential evaluation.
Based on data collected during an UCG pilot-scale experiment that took place during 2014 at Wieczorek mine, an active mine located in Upper Silesia (Poland), this research focuses on developing a dynamic fire risk prevention strategy addressing underground coal gasification processes (UCG) within active mines, preventing economic and physical losses derived from fires. To achieve this goal, the forecasting performance of two different kinds of artificial neural network models (generalized regression and multi-layer feedforward) are studied, in order to forecast the syngas temperature at the georeactor outlet with one hour of anticipation, thus giving enough time to UCG operators to adjust the amount and characteristics of the gasifying agents if necessary. The same model could be used to avoid undesired drops in the syngas temperature, as low temperature increases precipitation of contaminants reducing the inner diameter of the return pipeline. As a consequence the whole process of UGC might be stopped. Moreover, it could allow maintaining a high temperature that will lead to an increased efficiency, as UCG is a very exothermic process. Results of this research were compared with the ones obtained by means of Multivariate Adaptative Regression Splines (MARS), a non-parametric regression technique able to model non-linearities that cannot be adequately modelled using other regression methods. Syngas temperature forecast with one hour of anticipation at the georeactor outlet was achieved successfully, and conclusions clearly state that generalized regression neural networks (GRNN) achieve better forecasts than multi-layer feedforward networks (MLFN) and MARS models.