Real-time lithology identification while drilling based on drilling parameters analysis with machine learning
Abstract Accurate formation lithology information is crucial for addressing post-mining issues. Artificial intelligence is increasingly vital for lithology identification but faces challenges in underground coal mines, especially in accurately interpreting lithology from drilling parameters. These c...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-04-01
|
| Series: | Geomechanics and Geophysics for Geo-Energy and Geo-Resources |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s40948-025-00951-5 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Accurate formation lithology information is crucial for addressing post-mining issues. Artificial intelligence is increasingly vital for lithology identification but faces challenges in underground coal mines, especially in accurately interpreting lithology from drilling parameters. These challenges include: the influence of drill string friction, difficulties in extracting valuable data from large datasets, insufficient real-time performance to guide drilling operations, and the limited adaptability of individual machine learning algorithm. These issues hinder the practical application of machine learning-based lithology identification methods in coal mines. To address these challenges, this study developed a smart drilling rig and established an automatic data acquisition system. A novel method for data acquisition, retrieval, and cleaning was proposed to enable real-time data collection, rapid retrieval, and standardized data processing workflows. An automatic drill string friction sensing method was introduced to acquire real-time friction data during drilling, which corrected the drilling parameters and enhanced their correlation with stratigraphic information. Furthermore, an ensemble learning lithology identification model based on soft voting was constructed, integrating SVM, decision tree, KNN, and neural network classifiers. Field tests were conducted in an underground coal mine in China. During the tests, the system completed four boreholes and collected over 100,000 sets of drilling data. The lithology identification model was established, achieving an accuracy of 98.79%. The results demonstrate that the drilling parameters, eliminating the influence of drill string friction, showed a stronger correlation with stratigraphic lithology. The proposed lithology identification model exhibited high accuracy and generalization capability. |
|---|---|
| ISSN: | 2363-8419 2363-8427 |