BP神经网络磁性体顶面埋深预测方法

基于磁力异常确定磁性体埋深是磁法勘探的研究目的之一，对磁性矿藏、火成岩及磁性基底的研究具有重要意义。为此，通过提取大量理论长方体模型的磁力异常以及垂向一阶、二阶、三阶导数的极大值，获得三个极值比值并形成序列。利用这些极值比序列与相应模型的顶面深度构成训练样本，对构建的BP神经网络进行训练，并存储训练成功的BP神经网络，用于磁性体顶深预测。模型计算结果表明，91%以上训练样本和非训练样本的预测误差都小于10%。将该方法用于中国YX地区火山岩顶面埋深的预测，结果与钻井数据吻合较好，说明了方法的实用性和有效性。

Prediction of magnetic body top based on BP neural network

Prediction of magnetic body top based on magnetic anomaly is one of the major objectives of magnetic exploration.It is of great significance for the investigation of magnetic minerals,igneous rocks and magnetic basement.By extracting the magnetic anomalies from a large number of theoretical cuboid models and the maximum values of the first,second,or third derivatives in the vertical direction,three extremum ratios were obtained and a sequence dataset was formed.The constructed BP neural network was trained by the extremum ratio sequences and the depth of the model top to build training samples,and the trained BP neural network was stored for predicting the magnetic body top.Modeling results show that the prediction errors of more than 91% of the training samples are less than 10%,and the prediction of the samples not participating in the training also matches well with the results.The method has been applied for the prediction of the depth of the igneous rock in the YX area,China.The predicted depth highly agrees with the depth of the igneous rocks from drilling data,indicating the effectiveness of the method.