Intelligent identification of logging cuttings based on deep learning

Abstract

In oil and gas exploration, rock sample identification is a basic and important work. At present, the methods of rock sample identification mainly include gravity and magnetism, well logging, earthquake, remote sensing, electromagnetism, geochemistry, hand sample, and thin section analysis. Most of these methods for lithology identification are based on manual identification methods, which require a certain professional background and rich identification experience. Based on the deep learning method, this paper constructs the intelligent recognition model of cuttings lithology in logging work, which can realize the automatic recognition of rock images. The pre-training model is VGG16, and the parameters in the pre-training model are frozen and fine-tuned by the method of transfer learning, and finally the rock image identification based on vgg16 is realized. The accuracy of the trained model was 98% in the training set and 86% in the validation set. The experimental results show that the deep learning model based on transfer learning and vgg16 proposed in this paper has strong applicability to cuttings data and can distinguish rock types well. The intelligent lithology identification method proposed in this paper has good generalization performance and can be used for rapid intelligent identification of rock lithology in geology, well logging, power fault inspection, water conservancy and other projects.

Keywords

• Cuttings logging
• Lithology identification
• Deep learning
• Convolutional neural network
• Transfer learning