Sandification degree classification of sandy dolomite base on convolutional neural networks

Area of Science:

• Geotechnical Engineering
• Artificial Intelligence
• Geology

Background:

• Sandification degrades dolomite strength, impacting tunnel stability.
• Classifying sandification degree in sandy dolomite is challenging for geotechnical projects.
• Traditional methods for classification are often time-consuming and inaccurate.

Purpose of the Study:

• To introduce Convolutional Neural Network (CNN)-based image classification for sandy dolomite sandification.
• To establish a large-scale dataset for sandy dolomite sandification classification.
• To evaluate the effectiveness of CNN models in this classification task.

Main Methods:

• Development of a large-scale dataset with 5729 images of sandy dolomite.
• Classification of images into four distinct sandification degrees.
• Application and experimentation with CNN-based image classification models.

Main Results:

• CNN models achieved a high accuracy rate of up to 91.4% in sandification degree classification.
• The established dataset provides a valuable resource for research and development.
• Demonstrated the potential of AI in addressing complex geological challenges.

Conclusions:

• CNN-based image classification is a pioneering and effective method for sandy dolomite sandification.
• The developed dataset and models offer significant advancements for geotechnical engineering.
• This approach has broad implications for complex geographical analyses and infrastructure stability.