BP神经网络在花岗岩型铀矿预测中的应用

Application of BP neural network in the prediction of granite type uranium deposit

在已知矿床的外围或深部发现更多的成矿有利区是我国南方地区新一轮热液型铀矿勘查工作的重要方向之一。采用BP网络对某花岗岩型铀矿区以航空放射性信息为主的找矿标志进行综合。15个检验样本的预测有利度与期望有利度基本吻合,表明采用该预测模型对研究区内所有样本进行综合预测是合适的。从全区的预测结果可以看出,14处已知铀矿床均位于成矿有利度大于0.9的区域,且在已知铀矿床的外围发现了具有较好成矿远景的多片有利区。

In the new circle exploration for hydrothermal uranium deposits in the south of China,one of the important tasks is to locate more potential mineralization zones around or beneath currently existing uranium deposits. In this paper,BP network is adopted to integrate all kinds of prospecting indicators,mainly including aero-radioactive information acquired in some granite-type uranium orefields. Expected mineralization probabilities of fifteen testing samples are basically accordant with the predicting mineralization probabilities,which indicates that it is suitable to process all the samples in the study area by using this prediction model. Fourteen existing uranium deposits are found to be located in the zones with probability of mineralization more than 0.9,and several favorable zones around the existing deposits are delineated.