中文摘要：

对目前广泛使用的矿井突水水源判别方法进行了综合评述.从整体而言,除了水化学分析方法外,其他方法都是以一定理论为基础,或构造最优函数,根据判别目标达到最优时的状态进行水源识别;或构造适当的区间,根据一定的法则使判别目标进入不同的区间,进行水源识别.样本较多时采用BP神经网络法,样本较少时采用SVM法会取得更好得预测效果.针对研究区实际状况,选择基于MATLAB的BP神经网络法进行突水预测,准确率达到91.67%,训练样本的选择和数量对预测结果影响较大.

英文摘要：

The various methods for identifying the sources of mine water inrush were comprehensively re- viewed. It was found that in addition to the method for water chemical analysis,other methods were based on a certain theory. Some methods use different optimal structure functions to distinguish the source of mine water bursting. Other methods construct a suitable interval to determine a water inrush source. Generally,if more wa- ter samples were obtained, BP neural network method was used, and if fewer water samples were attained, SVM method was used. Based on 167 original water samples of Hebi Mine, a BP neural network model to distinguish sources of mine water bursting was established. The prediction accuracy of this model was 91.67% ,wherefore the predicted results can provide a reference for mine safety production.