Search

Abstract: This paper proposes the use of a coupled fault tree analysis (FTA) and artificial neural network (ANN) model to improve the prediction of the potential risk of coal and gas outburst events during the underground mining of thick and deep Chinese coal seams. The model developed has been used to investigate the gas emission characteristics and the geological conditions that exist within the Huaibei coal mining region, Anhui province, China. The coal seams in this region exhibit a high incidence of coal and gas outbursts. An analysis of the results obtained from an initial application of an FTA model, identified eight dominant model parameters related to the gas content or geological conditions of the coal seams, which characterize the potential risk of in situ coal and gas outbursts. The eight dominant model parameters identified by the FTA method were subsequently used as input variables to an ANN model. The results produced by the ANN model were used to develop a qualitative risk index to characterize the potential risk level of occurrence of coal and gas outburst events. Four different potential risk alarm levels were defined: SAFE, POTENTIAL, HIGH and STRONG. Solutions to the prediction model were obtained using a combination of quantitative and qualitative data including the gas content or gas pressure and the geological and geotechnical conditions of coal seams. The application of this combined solution method identified more explicit and accurate model relationships between the in situ geological conditions and the potential risk of coal and gas outbursts. An analysis of the model solutions concluded that the coupled FTA and ANN model may offer a reliable alternative method to forecast the potential risk of coal and gas outbursts. [ABSTRACT FROM AUTHOR]

Here we present our workflow and the final results of the 3D prospectivity modeling around the Zhuchong deposit undertaken during this study shown in terms of degree of connection values. The workflow is the core guiding ideology of our work. During the ore formation prediction, we tried to use weights of evidence method and artificial neural networks method to deal with the 3D cells data. And the results indicate these two methods can well predict the location of the known ore bodies. The reason may be that the copper deposits in Zhuchong Fe-Cu deposit are not concentrated but scattered, which makes making a certain influence to training samples in the neural network method. Which cause the defective result of neural network the result of neural network method is defective. What's more, the weight of evidence is also faster than the neural network method in speed. The results of the two prediction methods are roughly consistent. The similarity between these two prediction results indirectly reflects that two methods are feasible in Zhuchong Fe-Cu deposit. In the meanwhile, the prediction results are proved reliable and useful. The final prediction zone of metallogentic can be used as a target area for further exploration in Zhuchong mining area. • We use 3D ore formation prospectivity modeling and weights of evidence method and artificial neural networks method on the Zhuchong Fe-Cu deposit. • Result indicates that areas of known mineralization can be used to train 3D datasets to identify areas that are highly prospective for future exploration. • Compared the prediction results of the weight of evidence and the neural network, synthetically considering, for Zhuchong skarn Fe-Cu-Au deposit, the weight of evidence is more appropriate for the metallogenic prediction work. Our work has guiding significance for future deep prospecting and exploitation for other similar Fe-Cu skarn deposits. Three-dimensional (3D) metallogenic prospectivity modeling has become increasingly important in deep Fe-Cu skarn mineralization prospecting. The selection of prediction models and their comparison are of great significance. In this paper, we integrated the geological and geophysical data from the Zhuchong Fe-Cu deposit in the Yueshan orefield (Anhui, eastern China) to establish the local 3D geological framework. We extracted the 3D ore-controlling factors quantitatively by means of 3D spatial analysis methods, and compared the 3D mineralization locations predicted by the weight of evidence method and the artificial neural network method. The results show that both methods can locate most of the known orebodies, suggesting that both are relatively reliable. In addition, the weight of evidence method performed distinctly better than the artificial neural network method in terms of both prediction effectiveness and efficiency. Therefore, the weight of evidence method may be more applicable in 3D prospectivity modelling for Fe-Cu skarn deposits in the region. Finally, the composite area predicted by the two methods is taken as the final prediction result of the mineralization area of the deposit. And it is suggested that in the quantitative 3D prediction, multiple selection of different methods with good adaptability should be involved in the prediction at the same time, and compound prediction method should be used to delineate the target area. Our work has guiding significance for future deep prospecting and exploitation for other similar Fe-Cu skarn deposits. [ABSTRACT FROM AUTHOR]