Lithology identification method and system based on improved radial basis function neural network

Abstract

The invention provides a lithology identification method and system based on an improved radial basis function neural network. The method comprises the following steps: in combination with samples collected in a mining area, preprocessing data obtained by geophysical joint inversion; performing feature extraction on the preprocessed data by adopting K-L conversion to realize dimension reduction processing; processing the data set by adopting a K-fold cross validation method, and disorganizing the data set and equally dividing the data set into K parts; adopting a fuzzy clustering algorithm to complete clustering of the training set, and obtaining the center of a hidden layer; building a radial basis function neural network, and solving parameters of the radial basis function neural network according to the hidden layer center; verifying the radial basis function neural network by using the test set, and recording various types of recognition accuracy; repeating model training and testing, solving the overall recognition accuracy, and storing the optimal radial basis function neural network. According to the lithology identification method and system based on the improved radial basis function neural network, comprehensive, accurate and efficient lithology identification can be carried out.

Description

Technical field [0001] The present invention relates to the field of geophysical techniques, more particularly to a method and system for lithology identification based on improved radial basis function neural network. Background technique [0002] It refers to reflect the attributes lithologic characteristics of the rock, including color, composition, structure, type of cement and cement, special minerals. Understanding the formation lithology identification and reservoir parameters is a fundamental task solving process. With the exploitation of mineral resources, surface resources are depleted, exploration and mining resources has become a top priority in deep geological work today. Fine characterization of reservoir lithology and structural relationships underground, it can provide important basic information for the deep resource exploration, reservoir construction knowledge. Therefore, for the deep formation lithology identification has become the focus of current research g...

AI Technical Summary

Problems solved by technology

[0007] There may be more overlaps in the feature space of physical property samples obtained by joint geophysical inversion of different types of rocks. Some physical property parameters of various rocks have no obvious boundaries, and the relationship between lithology and physical properties is often ambiguous. The objects in the data set cannot be divided into clearly separated clusters. Using the K-means clustering algorithm to divide a sample into specified clusters may be relatively rigid or even wrong, and it is easy to misclassify lithology with a small sample size; at the same time, the inversion There is a lot of redundant information among the obtained physical properties (velocity, density, magnetic susceptibility, resistivity, etc.), resulting in a high correlation between each other, and the accuracy and efficiency are affected

Therefore, for the physical property data obtained by geophysical joint inversion, it is difficult to carry out comprehensive, accurate and efficient lithology identification only by using the radial basis function neural network based on the K-means clustering algorithm.

Images