一种基于深度学习模型的砂岩CT图像渗透率预测方法
By combining deep learning models and sliding window sampling with the LBM method, the problems of inconsistent resolution and training of large-size units in sandstone permeability prediction were solved, achieving fast and reliable permeability prediction, adapting to different resolutions and improving the model's generalization ability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANHUA UNIV
- Filing Date
- 2022-07-25
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies for predicting sandstone permeability suffer from unreliable prediction results due to inconsistent resolution, and the inability to directly train 3D convolutional neural network models on large-size units due to memory limitations, resulting in unrepresentative prediction results.
By employing a deep learning model and establishing a diverse dataset of permeability from three-dimensional binary images of sandstone, we utilize a sliding window sampling method to process large-size images in blocks, combine this with the LBM method to calculate permeability from small-size images, and train a ResNet model. The permeability unit is then converted to pixel squares to adapt to different resolutions, enabling rapid permeability prediction for large-size sandstone CT images.
It enables rapid and reliable permeability prediction of large-size sandstone CT images, and can make predictions flexibly at different resolutions, avoiding the need for retraining the model and improving the accuracy and generalization ability of the prediction results.
Smart Images
Figure CN115393279B_ABST