基于免疫原理的径向基神经网络品位插值研究

Research on Grade Interpolation Based on Radial Basis Function Neural Network Combined with Immune Principle

针对地质统计学方法的应用缺陷,基于径向基函数神经网络(RBF网络),结合免疫算法,开展矿石品位插值研究。利用RBF网络对样本数据进行分类及训练,通过免疫算法进行数据聚类分析,确定RBF网络的隐含层节点数、径向基函数中心向量及其宽度等参数。在此基础上,选取某典型矿山品位数据进行插值计算,将插值结果与品位真实值及克立格插值进行比较分析。研究结果表明:所给出的插值模型计算效率高,算法可以覆盖更多的训练数据,全局寻优能力强;插值结果具有较高的精度。当矿床无法满足地质统计学使用条件时,可考虑采用神经网络方法对矿石品位进行估值计算。

For the application defects of geostatistical methods,the study of ore grade interpolation based on radial basis function neural network(RBF network)and combined with immune algorithm was carried out.The RBF network was used to classify and train the sample data,and the data clustering analysis was performed through the immune algorithm to determine the number of hidden layer nodes,the center vector and width of the radial basis function.On this basis,the grade data of a typical mine was selected for interpolation calculation and the interpolation results were compared with the real grade value and Kriging interpolation.The results showed that the interpolation model presented in this paper was efficient,the algorithm could cover more training data,and the ability of global optimization was strong.The interpolation result was of high precision.When the ore deposit cannot meet the use conditions of geological statistics,the neural network method could be used to evaluate the ore grade.