基于变分自编码器利用元素录井数据确定火成岩矿物含量的方法OACSTPCD

Igneous rocks exhibit significant variations in mineral content due to differences in magma types and the environment in which they solidify,and the skeleton parameters of different lithology are obviously different.The determination of mineral content of the rock matrix is an important task in evaluating reservoirs,which is of great significance in stratigraphic lithology division,calculation of matrix parameters and study of depositional environments.In this study,a predictive model for mineral content in igneous rocks is proposed.The model utilizes data from 17 elements obtained through element logging.It employs a VAE(Variational AutoEncoder)approach to predict mineral content and reconstruct the elemental weight content.The model validation reveals that the proposed model has a smaller mean absolute error and mean square error compared to three typical methods:BP(Back Propagation)neural networks,ridge regression and support vector machines.Furthermore,the model is applied to a section of buried hill igneous rock well in the South China Sea.The results demonstrate the superiority of the proposed model over the typical algorithms while maintaining good applicability.