U-Net model for multi-component digital rock modeling of shales based on CT and QEMSCAN images

The establishment of digital images of cores with multi-mineral components holds the premise of analyzing the spatial distribution of shale minerals and carrying out numerical simulations of their physical properties. However, it is a challenge to directly obtain 3D multi-mineral-component digital core images by the existing technical means. The image multi-threshold segmentation of grayscale core images is a common method to obtain multi-component core images, but the thresholds need to be manually adjusted, which is time-consuming and may cause large errors. In this paper, to establish shale digital core images with multi-mineral components accurately and quickly, with the shale CT image as the original image and the corresponding QEMSCAN image as the label image, a U-Net model is trained and used to perform automatic segmentation of the CT images to convert grayscale images into multi-mineral-component images. Then, the manually segmented micropores and fractures are added to form the final model. The image segmentation effect of the constructed U-Net model is analyzed, and the results are satisfactory in terms of mineral content, morphological characteristics, and spatial structure. Compared with the traditional method, our method avoids repeated artificial adjustment, makes full use of the information in the CT images and QEMSCAN images, realizes the rapid automatic segmentation, and provides a feasible way for establishing images of a core with multi-mineral components. • A U-Net model is trained for automatic segmentation of shale CT grayscale images. • An accurate and fast 3D multi-component digital core modeling method is presented. • Compared with traditional multi-threshold methods, our method avoids uncertainties and errors caused by manual adjustments.