深部开采岩爆预测的神经网络方法

Artificial Neural Networks for Predicting Rockburst in Deep Mining

岩爆是深部高地应力岩石地下工程中的一种常见灾害,其影响因素之间存在着极其复杂的非线性关系。在综合分析基础上,选取开采深度、围岩最大切向应力与岩石单轴抗压强度比值、岩石单轴抗压强度和抗拉强度比值、岩石冲击性倾向指数作为岩爆预测的评判指标。应用人工神经网络方法,建立了岩爆预测的计算模型,利用国内外一些深部开采、岩石地下工程资料作为学习样本和测试样本对模型进行训练。该模型成功应用于某矿巷道的岩爆预测,预测结果与实际情况一致,此研究为深部开采岩爆预测提供了新的途径。

Rockburst is a kind of dynamic unstable phenomenon for the surrounding rock mass in deep mining. There exists a complicated nonlinear relationship between rockburst＇ s factors. We selected as the blasting indexes based on an analysis of the mechanism of rockburst： the mining depth, the ratio of maximal tangential stress and uniaxial compressive strength of the surrounding rocks, the ratio of uniaxial compressive strength and uniaxial tensile strength, and the elastic energy index. A rockburst prediction model is proposed by use of artificial neural network （ANN）. The model is trained to converge by use of some samples from rock underground projects at home and abroad. The results show that it is feasible and appropriate to predict rockburst in deep mining.