摘要
在地质勘探过程中,由于技术、设备的客观条件限制,造成了部分基础地质数据的缺失,这使得矿床建模时地质数据不够完整准确,直接影响了矿体形态及储量估值的精度。为了向矿床模型的构建环节提供完整且可信的基础地质数据,首先研究了地质缺失数据的产生机制,并通过对比分析期望—极大化算法(EM算法)、马尔可夫—蒙特卡洛方法(MCMC方法)以及BP神经网络等数据插补方法的特点及适用条件,提出了基于BP神经网络的地质缺失数据处理方法,构建了地质缺失数据处理的BP神经网络模型,并在山东某金矿进行了实际应用与模型验证。结果表明,模型可以实现地质缺失数据的部分插补,插补结果可信,可以有效解决因基础资料缺失所带来的地质数据不完整问题。
In the process of geological exploration,due to the limitation of technical and equipment objective conditions,there are lots of basic geological data missing.It causes that the geological data is not complete and accurate as building the deposit model,and has a direct impact on the accuracy of the orebody shape and reserves estimation.In order to provide the complete and believable data,so that the deposit model will be more realistic.Firstly the generation mechanism of geological missing data is studied to find out the method which geological missing data obeys.By means of comparing and analyzing the features and applicable conditions of Expectation Maximization（EM）algorithm,Markov Chain Monte Carlo（MCMC）method and Back Propagation（BP）Neural Network,then an interpolation method of geological missing data which based on BP neural network is selected and introduced,and the relative model of processing geological missing data is built up.Finally the whole method is applied in a certain gold mine in Shandong.It has been proved that the model can achieve interpolation of most of the geological missing data,and the results are reliable.In short,it is feasible and effective using the model to solve the integrity problem of geological data caused by basic data missing.
出处
《黄金科学技术》
CSCD
2015年第5期53-59,共7页
Gold Science and Technology
基金
国家自然科学基金项目"金属地下矿山低品位资源动态评估理论与方法研究"(编号:51104010)
中央高校基本科研业务费专项资金资助项目"矿山多源异构数据融合与挖掘技术研究"(编号:FRF-SD-12-001A)联合资助
