中文摘要：

为了提高识别矿山水害水源（即,判定水的类型）的正确率,利用免疫算法设计并优化了反向传播神经元网络（BPNN）的结构并求得BPNN的各层权系数和阈值的初值,用该初值训练BPNN,获得最佳的BPNN各层的权系数（权重）和阈值,使其适合识别矿山水害水源。用训练好的BPNN识别待判定的水源是哪一种类型的水源,判定水源的危害程度。实验和潞安集团所属煤矿区的矿井和钻孔水样检验结果说明用该方法是有效可行的,识别矿井水的水源的准确率可达到93%。

英文摘要：

In order to improve the recognition correct rate of flood water（ that is, recognition correct rate of type of water）, the structure and the weight coefficient （weight） and threshold of each layer of back -propagation neural network（BPNN） are designed and optimized with Immune Algorithm. The BPNN is trained, to get the optimal weight coefficient （weight） and threshold of each layer of BPNN. Finally, the trained BPNN is used to identify the type of water source to be judged, and the degree of water hazard is determined. Lu＇an Group coal mine and drilling water and experimental results show that the method is feasible and efficient, and the detec- tion right rate of flood waters was above 93 %.